DEV Community: Evgenii Konkin

Harmonic Distortion: The THD Number That Can Hide a Power Quality Problem

Evgenii Konkin — Fri, 24 Jul 2026 05:11:51 +0000

Harmonic distortion is one of those power quality numbers that looks simple until someone uses it as a diagnosis.

A power analyzer gives a THD percentage.

The number looks clean.

Maybe it says:

Voltage THD = 4.8%

or:

Current THD = 13%

Then the mistake begins.

Someone treats that one number as the whole power quality story.

But THD is not the whole story.

It is a ratio. It tells you how much harmonic RMS content exists compared with the fundamental RMS component. It does not tell you every harmonic order, resonance risk, neutral loading, transformer heating, capacitor bank stress, or whether the site complies with every applicable limit.

The formula is useful.

The interpretation needs care.

The basic THD formula

Total Harmonic Distortion compares the RMS value of harmonic components with the RMS value of the fundamental component.

For voltage THD:

THD (%) = [sqrt(H2² + H3² + H4² + ... + H13²) / V1] × 100

For current THD:

THD (%) = [sqrt(I2² + I3² + I4² + ... + I13²) / I1] × 100

Where:

V1 or I1 = fundamental RMS value
H2...H13 = harmonic RMS values above the fundamental
THD = total harmonic distortion, percent

The same RMS structure applies to voltage and current.

But the values must not be mixed.

For voltage THD, use volts RMS for the fundamental and all harmonic components.

For current THD, use amps RMS for the fundamental and all harmonic components.

Do not mix volts and amps.

Do not mix peak values and RMS values.

Do not enter harmonic percentages when the calculator expects RMS magnitudes.

That sounds obvious, but it is one of the most common ways to get a wrong THD result.

THD is an RMS calculation, not an arithmetic sum

A common mistake is adding the harmonic values directly.

That is not how THD is calculated.

Harmonic components are combined as root-sum-square:

Harmonic_RMS = sqrt(H2² + H3² + H4² + ... + H13²)

Then that combined harmonic RMS value is compared with the fundamental.

Example:

Fundamental voltage V1 = 230 V RMS
3rd harmonic H3 = 6 V RMS
5th harmonic H5 = 9 V RMS
7th harmonic H7 = 4 V RMS

Wrong approach:

6 + 9 + 4 = 19 V
THD = 19 / 230 × 100
THD = 8.26%

Correct RMS approach:

Harmonic_RMS = sqrt(6² + 9² + 4²)
Harmonic_RMS = sqrt(36 + 81 + 16)
Harmonic_RMS = sqrt(133)
Harmonic_RMS ≈ 11.53 V

Now calculate THD:

THD = 11.53 / 230 × 100
THD ≈ 5.01%

The wrong arithmetic-sum method gave 8.26%.

The correct RMS method gives about 5.01%.

That is a big difference.

The harmonic values were the same.

The calculation method changed the conclusion.

Worked example: current THD from a nonlinear load

Suppose a power quality check on a panel with VFDs and switching power supplies shows:

Fundamental current I1 = 120 A RMS
3rd harmonic I3 = 8 A RMS
5th harmonic I5 = 12 A RMS
7th harmonic I7 = 5 A RMS

Use the current THD formula:

THD (%) = [sqrt(I3² + I5² + I7²) / I1] × 100

Square the harmonic currents:

I3² = 8² = 64
I5² = 12² = 144
I7² = 5² = 25

Add them:

64 + 144 + 25 = 233

Take the square root:

sqrt(233) ≈ 15.26 A

Divide by the fundamental current:

THD = 15.26 / 120 × 100
THD ≈ 12.72%

So the current THD is about:

Current THD ≈ 12.7%

That is not a clean sinusoidal current waveform.

It does not automatically prove equipment failure, but it is high enough to justify a closer review of nonlinear loads, transformer loading, neutral currents, cable heating, harmonic filters, and possible resonance conditions.

Voltage THD and current THD are not the same problem

Voltage THD and current THD both use the same mathematical structure.

But they do not mean the same thing.

Voltage THD describes distortion in the voltage waveform.

Current THD describes distortion in the load current waveform.

A site can have high current THD but acceptable voltage THD if the upstream system is stiff enough.

For example:

Current THD = high
Voltage THD = moderate or low

That may mean the nonlinear loads are drawing distorted current, but the source impedance is low enough that the voltage waveform is not heavily distorted.

On the other hand, if current harmonics flow through system impedance, they can create voltage distortion:

Harmonic current × system impedance = harmonic voltage distortion

That distorted voltage can then affect other equipment connected to the same system.

So the mistake is saying:

Current THD is high, therefore voltage quality is definitely bad.

Or the opposite:

Voltage THD is low, therefore harmonics are not a problem.

Both can be wrong.

You need to understand which waveform was measured and where.

The common mistake: confusing THD with TDD

THD and TDD are not the same.

THD compares harmonic RMS content with the fundamental RMS component:

THD = harmonic RMS content / fundamental RMS component

TDD compares current harmonic content with maximum demand load current:

TDD = harmonic current content / maximum demand load current

That difference matters in electrical power systems.

Current THD can look high at light load because the fundamental current is small.

Example:

Harmonic current RMS = 10 A
Fundamental current = 40 A

Then:

Current THD = 10 / 40 × 100
Current THD = 25%

Now suppose the same harmonic current exists when the fundamental current is higher:

Harmonic current RMS = 10 A
Fundamental current = 200 A

Then:

Current THD = 10 / 200 × 100
Current THD = 5%

The harmonic current stayed the same.

The THD percentage changed because the fundamental current changed.

That is one reason current distortion compliance work often needs more than a simple current THD number.

THD is useful.

But for standards review, utility requirements, and point-of-common-coupling studies, TDD and individual harmonic limits may be required.

A single THD number can hide the dominant harmonic

Another mistake is looking only at total THD and ignoring which harmonic order is causing it.

For example, two systems may both show:

THD = 8%

But the harmonic spectrum may be very different.

System A:

Mostly 5th harmonic

System B:

Mostly 3rd harmonic

Those are not the same engineering problem.

A strong 5th harmonic may be associated with six-pulse rectifiers, VFDs, and certain power electronic loads.

A strong 3rd harmonic is a triplen harmonic and can be especially important in three-phase four-wire systems because triplen harmonics can add in the neutral conductor instead of cancelling like balanced fundamental currents.

So the total THD value is useful, but the spectrum matters.

A total number tells you distortion exists.

The individual harmonic orders help explain what kind of distortion you have.

Why harmonics matter

Harmonics are not just waveform aesthetics.

They can create real system effects:

Extra transformer heating
Extra conductor heating
Neutral conductor loading
Motor heating and torque pulsation
Capacitor bank stress
Nuisance trips
Generator voltage distortion
Poor power factor behavior
Control and sensor interference
Resonance risk with power factor correction capacitors

A waveform can deliver power and still be distorted.

A panel can operate and still run hotter than expected.

A transformer can be within kVA rating and still be stressed by harmonic heating.

That is why THD is often used as an early warning indicator.

It does not complete the diagnosis, but it tells you whether the waveform is close to sinusoidal or noticeably distorted.

Example: why low voltage THD may not mean “no harmonic issue”

Suppose a facility has several VFDs.

Measured values:

Voltage THD = 3.5%
Current THD = 38%

Someone may say:

Voltage THD is only 3.5%, so harmonics are fine.

That may be too optimistic.

The voltage waveform may still be acceptable because the source is strong, but the current waveform is heavily distorted. That distorted current may still contribute to:

Transformer heating
Cable heating
Generator loading issues
Harmonic filter requirements
Neutral conductor concerns
Distorted voltage under weaker source conditions

The correct conclusion is more careful:

Voltage distortion is currently moderate, but current distortion is high and should be reviewed at the load and distribution level.

That is a better engineering interpretation.

Example: why high current THD at light load may be misleading

Now consider a UPS or VFD at very light load.

Measured values:

Fundamental current = 10 A
Combined harmonic RMS current = 5 A

Then:

Current THD = 5 / 10 × 100
Current THD = 50%

That looks terrible.

But the absolute harmonic current is only 5 A.

At higher load, the same equipment may show:

Fundamental current = 100 A
Combined harmonic RMS current = 10 A

Then:

Current THD = 10 / 100 × 100
Current THD = 10%

The THD percentage dropped a lot.

But the harmonic current doubled from 5 A to 10 A.

So current THD must be interpreted with load level.

High THD at light load can look dramatic because the denominator is small.

That does not mean it should be ignored, but it should not be interpreted without load context.

RMS values matter

The THD formula requires RMS values.

If someone enters peak values for harmonics but RMS for the fundamental, the result is wrong.

For a sine component:

Vpeak = Vrms × sqrt(2)

If a harmonic component is accidentally entered as peak instead of RMS, it will be about 1.414 times too large.

That error propagates into the root-sum-square calculation and inflates the THD result.

The safest workflow is:

Confirm the analyzer reports RMS harmonic magnitudes.
Use the same unit for all components.
Do not mix voltage and current.
Do not mix RMS and peak.
Do not enter percentages as RMS values.

The formula is only as good as the input definitions.

THD does not prove compliance by itself

THD is a screening metric.

It is not a full compliance report.

A final power quality review may require:

Measured harmonic spectrum
Individual harmonic order limits
Point of common coupling definition
Short-circuit ratio
Maximum demand load current
TDD calculation
Voltage distortion limits
Current distortion limits
Measurement duration
Operating scenarios
Utility interconnection requirements

This is especially important for large facilities, industrial plants, data centers, solar inverters, generator systems, and sites with major nonlinear loads.

A calculator can give a consistent first-pass THD estimate.

It cannot replace a full power quality study when standards compliance or utility approval is involved.

Practical design takeaway

Use THD as a power quality screening number.

It helps answer:

How distorted is the waveform compared with the fundamental?
Which harmonic RMS values dominate the result?
Is the waveform close to sinusoidal or heavily distorted?
Is the distortion voltage THD or current THD?
Is a detailed harmonic review needed?

But do not use THD as the only conclusion.

It does not show the full harmonic spectrum by itself.

It does not calculate TDD.

It does not check IEEE 519 compliance by itself.

It does not model resonance.

It does not calculate transformer derating.

It does not prove that neutral current, capacitor stress, or motor heating is acceptable.

It is a first-pass indicator.

A useful one.

But still only one part of the power quality picture.

Final thought

Total Harmonic Distortion is a compact number:

THD (%) = [sqrt(sum of harmonic RMS values squared) / fundamental RMS] × 100

That compact number can be very useful.

It can show when a waveform is close to sinusoidal.

It can warn that nonlinear loads are producing significant distortion.

It can help compare voltage and current quality between systems.

But it can also mislead if the engineer forgets the context.

Current THD changes with load level.

Voltage THD and current THD describe different things.

A single THD value can hide the dominant harmonic order.

And THD is not the same as TDD or full standards compliance.

For quick THD checks from fundamental and harmonic RMS values, use the Harmonic Distortion Calculator on CalcEngineer.

Signal Propagation Delay: The Small Cable Delay That Can Break a Timing Budget

Evgenii Konkin — Thu, 16 Jul 2026 16:30:24 +0000

Signal propagation delay is easy to underestimate.

A cable looks passive.
A fiber run looks like a simple connection.
A PCB trace looks short.

But a signal still needs time to travel through that physical path. It does not arrive instantly.

The basic formula is simple:

Delay = Length / Propagation Velocity

That formula gives the one-way geometric delay.

But the engineering mistake is treating that number as either “too small to matter” or as the full end-to-end system latency. Both assumptions can be wrong.

In fast digital systems, protection relays, synchronization links, trigger circuits, and long communication paths, propagation delay can become a real timing-budget item.

The basic formula

The one-way propagation delay is:

Delay_s = Length_m / Velocity_mps

Where:

Delay_s = one-way propagation delay, seconds
Length_m = signal travel distance, meters
Velocity_mps = propagation velocity, meters per second

If the velocity is entered as a fraction of the speed of light:

Velocity_mps = Velocity_factor × 299,792,458

For example:

Velocity_factor = 0.66

Velocity_mps = 0.66 × 299,792,458
Velocity_mps ≈ 197,863,022 m/s

Then the delay can be converted into more readable units:

Delay_ns = Delay_s × 1,000,000,000
Delay_µs = Delay_s × 1,000,000
Delay_ms = Delay_s × 1,000

That is the whole calculation.

But the interpretation is where the problems start.

One-way delay is not round-trip latency

The calculator gives one-way propagation delay.

That means the signal travels from point A to point B once.

If the system needs a response back from point B to point A, the propagation portion of round-trip delay is roughly:

Round-trip propagation delay ≈ 2 × one-way delay

That still does not include:

Device delay
Connector delay
Repeater delay
Switching delay
Serialization delay
Protocol overhead
Software processing time
Queueing delay
Measurement instrument delay

So propagation delay is one term in the timing budget.

It is not the entire timing budget.

That distinction matters a lot.

A propagation delay estimate can tell you the minimum physical delay caused by path length and signal velocity. It cannot tell you the total system response time.

Worked example: 100 m cable at 0.66c

Suppose an engineer is checking a 100 m copper cable run.

Use:

Path Length = 100 m
Velocity Factor = 0.66c

First calculate signal velocity:

Velocity_mps = 0.66 × 299,792,458
Velocity_mps ≈ 197,863,022 m/s

Now calculate one-way delay:

Delay_s = 100 / 197,863,022
Delay_s ≈ 0.000000505 s

Convert to nanoseconds and microseconds:

Delay_ns ≈ 505 ns
Delay_µs ≈ 0.505 µs
Delay_ms ≈ 0.000505 ms

So the cable adds about:

One-way delay ≈ 505 ns

That sounds tiny.

For a human interface, it is tiny.

For some industrial controls, it may still be tiny.

But for high-speed timing, precision triggering, clock distribution, or protection coordination, 505 ns may be large enough to matter.

The same number can be irrelevant in one system and critical in another.

That is why propagation delay must always be compared against the actual timing budget.

Length matters linearly

Propagation delay is directly proportional to path length.

If the same cable velocity is used:

Delay ∝ Length

So if 100 m gives 505 ns, then:

200 m gives about 1.01 µs
500 m gives about 2.53 µs
1000 m gives about 5.05 µs

There is no hidden magic here.

Double the distance, double the delay.

That is why long fiber and cable paths can become important even when each meter feels harmless.

In many systems, a few meters do not matter.

In some systems, a few meters absolutely matter.

The circuit or protocol decides which case you are in.

Velocity factor matters too

Delay is inversely proportional to signal velocity:

Delay ∝ 1 / Velocity

If the propagation velocity is lower, delay is higher.

A common mistake is assuming the signal travels at the full speed of light.

It does not in real cable or fiber.

A velocity factor of 0.66 means the signal travels at 66% of the speed of light.

For 100 m:

At 1.00c:

Velocity = 299,792,458 m/s
Delay = 100 / 299,792,458
Delay ≈ 333 ns

At 0.66c:

Velocity ≈ 197,863,022 m/s
Delay ≈ 505 ns

The length did not change.

Only the propagation velocity assumption changed.

The delay increased from about 333 ns to about 505 ns.

That is about 52% higher.

So using the wrong velocity factor can make the timing estimate look much better than the real system.

The common engineering mistake: treating cable delay as total latency

One common mistake is calculating propagation delay and then saying:

The system latency is 505 ns.

That is usually wrong.

The correct statement is:

The one-way geometric propagation delay through this cable is about 505 ns.

The full system may include many additional delays.

For example:

Cable propagation delay = 505 ns
Input circuit delay = 80 ns
Isolation device delay = 120 ns
FPGA input synchronization = 20 ns
Processing delay = 300 ns
Output driver delay = 60 ns
Second cable path = 505 ns

Total delay:

Total delay = 505 + 80 + 120 + 20 + 300 + 60 + 505
Total delay = 1,590 ns
Total delay = 1.59 µs

In this example, the cable is important, but it is not the only delay.

If the engineer only reports the 505 ns propagation delay, the timing budget is incomplete.

If the engineer ignores the cable entirely, the timing budget is also incomplete.

Both errors are common.

Short cables can still matter

Another mistake is assuming short distance means zero delay.

Take a 10 m cable at 0.66c:

Velocity_mps ≈ 197,863,022 m/s

Delay_s = 10 / 197,863,022
Delay_s ≈ 50.5 ns

So:

One-way delay ≈ 50 ns

For slow control logic, that may be irrelevant.

But in a 10 Gbit/s system, one bit period is:

Bit period = 1 / 10,000,000,000
Bit period = 0.1 ns

A 50 ns propagation delay is not “small” compared with that bit period.

That does not automatically mean the link fails. High-speed systems are designed around serialization, equalization, clocking, encoding, and protocol behavior.

But it does mean that the physical path delay is real and must be handled in the system design.

The same 10 m cable can be invisible to a relay input and very relevant to a high-speed timing path.

Long paths become microseconds or milliseconds

For long links, propagation delay can become easier to visualize.

Suppose a fiber path is:

Path Length = 50 km
Velocity Factor = 0.67c

Convert length:

Length_m = 50 × 1000
Length_m = 50,000 m

Velocity:

Velocity_mps = 0.67 × 299,792,458
Velocity_mps ≈ 200,860,947 m/s

One-way delay:

Delay_s = 50,000 / 200,860,947
Delay_s ≈ 0.000249 s

Convert:

Delay_ms ≈ 0.249 ms
Delay_µs ≈ 249 µs

Round-trip propagation delay would be approximately:

Round-trip propagation delay ≈ 2 × 0.249 ms
Round-trip propagation delay ≈ 0.498 ms

That is only the physical travel time.

Real network latency will be higher once transceivers, switches, routing, buffering, and protocol overhead are included.

This is exactly why propagation delay should be separated from full end-to-end latency.

The timing-budget problem

A useful timing budget should separate delay terms.

For example:

Propagation delay
Transceiver delay
Connector or interface delay
Logic delay
Switch or relay delay
Serialization delay
Protocol delay
Software delay
Safety margin

If a system has a maximum allowable timing error of 2 µs, and the cable alone contributes 0.5 µs, then cable delay is already 25% of the budget.

That may be acceptable.

But it is not negligible.

If several cable paths exist, or if one path is longer than another, skew can appear between channels.

For multi-channel trigger systems, protection relay links, clock distribution, and measurement setups, the difference between path delays may matter more than the absolute delay.

Example:

Channel A cable length = 20 m
Channel B cable length = 80 m
Velocity factor = 0.66c

Length difference:

ΔLength = 80 − 20
ΔLength = 60 m

Delay skew:

ΔDelay = 60 / 197,863,022
ΔDelay ≈ 303 ns

So even if both channels “work,” they do not arrive at the same time.

That skew may need compensation.

Unit mistakes can be severe

Propagation delay calculations are sensitive to length units.

The calculator supports common distance units, but the engineer must select the correct one.

A path length of:

100 ft

is not the same as:

100 m

Convert 100 ft:

100 ft × 0.3048 = 30.48 m

At 0.66c:

Delay = 30.48 / 197,863,022
Delay ≈ 154 ns

For 100 m:

Delay ≈ 505 ns

That is more than 3× different.

The same issue applies to miles and kilometers.

A wrong unit selection can make a link look much faster or slower than it really is.

Propagation delay is not signal integrity

A cable can have the right delay and still have poor signal quality.

Propagation delay tells you when the signal arrives.

It does not tell you whether the signal arrives cleanly.

It does not check:

Attenuation
Reflections
Impedance mismatch
Rise-time degradation
Crosstalk
Jitter
Dispersion
Return loss
Connector quality
Shielding
Ground reference problems
Receiver threshold margin

This is an important distinction.

Timing and signal integrity are related, but they are not the same calculation.

A propagation delay estimate may say the signal should arrive in 505 ns.

It does not prove the receiver sees a clean digital edge at 505 ns.

For high-speed or precision systems, propagation delay is only one part of the design review.

Practical design takeaway

Use propagation delay as a timing-budget term.

It helps answer:

How much delay does this cable, trace, or fiber path add?
Is the delay in ns, µs, or ms?
How much path mismatch exists between channels?
Is one-way delay being confused with round-trip latency?
Is the assumed velocity factor realistic?
Is the path delay small compared with the system timing budget?

But do not use it as a full system latency calculation.

It does not include electronics delay.
It does not include switch or repeater delay.
It does not include protocol overhead.
It does not prove signal integrity.
It does not replace measurement or detailed timing analysis.

It gives you the geometric delay term — the part set by distance and propagation speed.

That is often the right first number to calculate before building the full timing budget.

Final thought

Signal propagation delay is simple:

Delay = Length / Velocity

But the simplicity is the trap.

A 100 m cable at 0.66c adds about 505 ns of one-way delay.

That may be irrelevant for a slow control input, but meaningful for synchronization, protection, trigger, or high-speed timing work.

The important habit is not to call the delay “small” or “large” in isolation.

Compare it with the timing budget.

Check whether you need one-way or round-trip delay.

Use a realistic velocity factor.

And remember that propagation delay is not total latency and not signal integrity.

For quick one-way delay checks from path length and propagation velocity, use the Signal Propagation Delay Calculator on CalcEngineer.

Power Loss: When Small Efficiency Losses Become Real Heat Loads

Evgenii Konkin — Sun, 12 Jul 2026 16:08:31 +0000

Power loss looks like one of the simplest calculations in electrical engineering.

Measure the power going into a device.

Measure the useful power coming out.

Subtract one from the other.

Power Loss = Input Power − Output Power

That gives the missing power.

Then calculate efficiency:

Efficiency (%) = Output Power / Input Power × 100

Simple.

But in real systems, this small calculation can reveal something much bigger: hidden heat load, poor operating point, transformer or UPS losses, inverter inefficiency, motor waste, undersized ventilation, or a measurement boundary that was defined badly.

The formula is easy.

The interpretation is where engineers get into trouble.

The basic power loss formula

The core calculation is:

Power_Loss = Input_Power − Output_Power

Where:

Input_Power = power entering the device or system
Output_Power = useful delivered power leaving the device or system
Power_Loss = power not delivered as useful output

If the input and output are both entered in kW:

Power_Loss_kW = Input_Power_kW − Output_Power_kW

If the values are entered in watts:

Power_kW = Power_W / 1000

If the values are entered in megawatts:

Power_kW = Power_MW × 1000

The important rule is that input and output must be in the same unit before comparison.

You cannot subtract 96 kW from 100 MW and call the result meaningful.

Loss percentage

Absolute power loss is useful, but it does not tell the whole story.

A 4 kW loss may be small in one system and huge in another.

That is why loss percentage is calculated as:

Loss_Percent = Power_Loss / Input_Power × 100

For example:

Input Power = 100 kW
Output Power = 96 kW

Power loss:

Power_Loss = 100 − 96
Power_Loss = 4 kW

Loss percentage:

Loss_Percent = 4 / 100 × 100
Loss_Percent = 4%

Efficiency:

Efficiency = 96 / 100 × 100
Efficiency = 96%

So the same result can be described three ways:

Power Loss = 4 kW
Loss Percentage = 4%
Efficiency = 96%

These are not separate realities. They are different views of the same input-output relationship.

Efficiency and loss always connect

For this simple model:

Loss (%) = 100% − Efficiency (%)

So if efficiency is 96%:

Loss = 100 − 96
Loss = 4%

If loss is 12%:

Efficiency = 100 − 12
Efficiency = 88%

This is useful during design reviews because it avoids double counting.

If someone says a converter has 95% efficiency, then the loss is 5% of input power.

If someone says a system has 8% loss, then the useful efficiency is 92%.

The two numbers must agree.

Worked example: UPS loss is also room heat

Suppose a UPS system has:

Input Power = 250 kW
Output Power = 235 kW

Calculate power loss:

Power_Loss = 250 − 235
Power_Loss = 15 kW

Calculate loss percentage:

Loss_Percent = 15 / 250 × 100
Loss_Percent = 6%

Calculate efficiency:

Efficiency = 235 / 250 × 100
Efficiency = 94%

At first glance, 94% efficiency may sound acceptable.

But the 15 kW loss is not just an abstract efficiency number.

In most electrical equipment rooms, that missing power becomes heat.

So the thermal load from UPS losses is approximately:

Heat Load = 15 kW

Convert to BTU/hr:

1 kW ≈ 3412 BTU/hr

Heat Load = 15 × 3412
Heat Load ≈ 51,180 BTU/hr

That is about:

51,180 / 12,000 ≈ 4.3 tons of cooling

So a “small” 6% loss may still create more than 4 tons of heat load that must be removed from the room.

This is the practical engineering value of the calculation.

Efficiency is not only a performance number.

It is also a thermal design input.

The common mistake: looking only at the percentage

One of the most common mistakes is saying:

Only 4% loss. That is fine.

Maybe it is fine.

Maybe it is not.

The percentage must be tied to the system size.

Example A:

Input Power = 10 kW
Loss = 4%
Power Loss = 0.4 kW

That is 400 W of heat.

Example B:

Input Power = 1,000 kW
Loss = 4%
Power Loss = 40 kW

That is 40 kW of heat.

The percentage is the same.

The equipment-room cooling burden is not.

This is why engineers should always check both:

Loss percentage
Absolute kW loss

A low percentage loss in a large system can still be a serious heat-removal problem.

The other mistake: judging partial-load loss too harshly

Power loss percentage can look worse at light load.

That does not always mean the equipment is faulty.

Many devices have fixed or semi-fixed losses:

Control power
Magnetizing loss
Core loss
Fan power
Standby electronics
Auxiliary power
No-load loss

At low output load, these fixed losses are divided by a smaller useful output.

That can make the loss percentage look high.

Example:

Input Power = 12 kW
Output Power = 10 kW

Power loss:

Power_Loss = 12 − 10
Power_Loss = 2 kW

Loss percentage:

Loss_Percent = 2 / 12 × 100
Loss_Percent = 16.7%

Efficiency:

Efficiency = 10 / 12 × 100
Efficiency = 83.3%

That looks poor.

But now imagine the same equipment near rated load:

Input Power = 105 kW
Output Power = 100 kW

Power loss:

Power_Loss = 105 − 100
Power_Loss = 5 kW

Loss percentage:

Loss_Percent = 5 / 105 × 100
Loss_Percent = 4.8%

Efficiency:

Efficiency = 100 / 105 × 100
Efficiency = 95.2%

The absolute loss increased from 2 kW to 5 kW.

But the loss percentage improved from 16.7% to 4.8%.

That is not a contradiction.

It means the operating point matters.

A light-load efficiency result should not be judged the same way as a full-load efficiency result.

Power loss does not tell you where the loss happens

The formula shows total missing power:

Input Power − Output Power

It does not identify the cause.

The missing power may be caused by:

Copper loss
Core loss
Switching loss
Conduction loss
Friction loss
Fan power
Transformer loss
UPS conversion loss
Motor winding loss
Harmonic heating
Poor power quality
Bad operating point

So if a system has high loss, the calculation tells you:

There is an efficiency problem or a large overhead.

It does not tell you:

The transformer is bad.
The motor is failing.
The inverter is defective.
The UPS needs replacement.

That conclusion requires more evidence.

You need metering, temperature readings, load profile, waveform quality, equipment data, and sometimes manufacturer test curves.

Power loss is the alarm bell.

It is not the root-cause report.

Measurement boundary matters

Another mistake is defining input and output points inconsistently.

For example, consider a UPS system.

Input power could be measured:

At the utility service
At the UPS input terminals
At the rectifier input
At the upstream panel

Output power could be measured:

At the UPS output terminals
At the downstream distribution panel
At the PDU output
At the rack load

Those measurement points are not the same.

If the input is measured upstream of a transformer and the output is measured at the rack, then transformer loss, cable loss, UPS loss, PDU loss, and maybe branch-circuit loss are all included.

That may be exactly what you want.

Or it may be completely wrong for the question being asked.

A good power-loss statement should include the boundary:

Input measured at UPS input terminals
Output measured at UPS output terminals
Load condition: 70% rated load
Measurement period: 15-minute average

That is much better than saying:

UPS loss is 6%

Without a boundary, the number is hard to trust.

Output power cannot exceed input power in this model

In a simple input-output efficiency model:

Output Power ≤ Input Power

If output power appears greater than input power, something is wrong with the measurement or the boundary.

Possible causes include:

Different measurement units
Metering on different time intervals
Incorrect power factor handling
Using apparent power on one side and real power on the other
Regeneration or stored-energy effects
Instrument error
Wrong CT or PT ratio
Different phase boundaries

For example:

Input Power = 95 kW
Output Power = 100 kW

The simple calculation would imply:

Power_Loss = −5 kW
Efficiency = 105.3%

That is not physically valid for a normal passive or conversion device in this simplified model.

The correct response is not to accept the efficiency.

The correct response is to check the measurement method.

Real power vs apparent power

Another common mistake is mixing kW and kVA.

Power loss and efficiency should usually be based on real power:

kW

not apparent power:

kVA

For example, if input is measured as kVA and output is measured as kW, the calculation becomes misleading.

A load may show:

Apparent Power = 100 kVA
Power Factor = 0.80

The real power is:

Real Power = kVA × PF
Real Power = 100 × 0.80
Real Power = 80 kW

If someone compares 100 kVA input against 80 kW output, they may incorrectly call the difference “20 kW loss.”

But that is not the same thing.

Some of that difference is reactive power relationship, not necessarily real power converted to heat.

For efficiency calculations, make sure both input and output are real power on a consistent basis.

Why the lost power usually becomes heat

In many practical electrical systems, most power loss ultimately becomes heat.

That matters for:

Electrical room ventilation
Transformer room cooling
UPS room cooling
Inverter enclosure thermal design
Motor control center temperature rise
Panel derating
Component lifetime

If a device loses 8 kW continuously, the room or enclosure must remove approximately 8 kW of heat.

Annual energy waste can also be estimated:

Annual Loss Energy = Power Loss × Operating Hours

For example:

Power Loss = 8 kW
Operating Hours = 6000 h/year

Then:

Annual Loss Energy = 8 × 6000
Annual Loss Energy = 48,000 kWh/year

At an electricity price of $0.12/kWh:

Annual Loss Cost = 48,000 × 0.12
Annual Loss Cost = $5,760/year

That is why small efficiency differences can matter over long operating hours.

A one-time loss percentage may look boring.

A full-year loss cost may not.

Practical interpretation bands

A simple classification can help during early review:

Loss < 1% = very low loss
1% to 5% = low loss
5% to 10% = moderate loss
10% to 20% = high loss
Loss ≥ 20% = very high loss

These bands are not a substitute for equipment-specific judgment.

A 7% loss may be acceptable for one device and poor for another.

A 2% transformer loss might be high in one context, while a 6% inverter loss may be normal depending on type and load.

The classification is useful because it helps decide what to do next:

Very low / low loss: probably normal, verify against equipment data.
Moderate loss: review load point, heat load, and manufacturer efficiency.
High / very high loss: investigate measurement boundary, operating point, thermal impact, and equipment condition.

The band is a screening result.

It is not a final pass/fail certificate.

A useful design review workflow

When checking power loss, do not stop at the arithmetic.

A better workflow is:

1. Confirm input and output are both real power.
2. Confirm both values use the same unit.
3. Confirm the measurement boundary.
4. Calculate absolute power loss.
5. Calculate loss percentage and efficiency.
6. Convert lost kW into heat load if the equipment is indoors.
7. Compare the result with manufacturer data at the same load point.
8. Check whether the result is snapshot, average, or annual.
9. If loss is high, investigate where the loss occurs.

This keeps the calculation grounded in engineering reality.

Final thought

Power loss is not complicated mathematically.

Power Loss = Input Power − Output Power

But it is easy to misread.

A small percentage loss can still be a large heat load.

A high percentage loss at light load may not mean the equipment is broken.

A clean efficiency number may be meaningless if input and output were measured at different boundaries.

And the formula does not identify the cause of the loss.

The best use of the calculation is as a first-pass screening tool. It tells you how much power is missing, how large that loss is relative to input, and whether the result deserves a deeper thermal, efficiency, or troubleshooting review.

For quick checks of power loss, loss percentage, efficiency, and loss severity classification, use the Power Loss Calculator on CalcEngineer.

RC Time Constant: The Simple Timing Formula That Can Hide Inrush and Stored Energy Problems

Evgenii Konkin — Fri, 10 Jul 2026 07:43:02 +0000

RC timing looks easy.

Pick a resistor.

Pick a capacitor.

Multiply them.

τ = R × C

That gives the time constant.

After one time constant, the capacitor has moved about 63.2% toward its final voltage. After five time constants, it is about 99.3% settled.

Simple.

But in real circuits, RC timing is one of those calculations that gets used too casually. The time constant may be correct, while the circuit still has a dangerous inrush current, a capacitor that stores hazardous energy, a slow bleed-down path, or a timing value that changes because the real capacitor is not the value printed on the schematic.

The formula is short.

The engineering consequences are not.

The basic RC time constant formula

For a simple series resistor-capacitor circuit:

τ = R × C

Where:

τ = time constant, seconds
R = effective resistance, ohms
C = capacitance, farads

The word “effective” matters.

The resistance is not always just the resistor shown on the schematic.

It can include:

Series resistor
Source impedance
Switch on-resistance
Capacitor ESR
Trace resistance
Connector resistance
Bleed path, if analyzing discharge

For a rough first pass, the external resistor may dominate.

For power electronics, snubbers, bulk capacitors, and fast transients, the hidden resistance can matter a lot.

Unit shortcuts are useful, but dangerous if misread

The formula uses ohms and farads.

But practical circuits usually use kΩ, MΩ, nF, μF, and sometimes mF.

Some useful shortcuts:

Ω × F = seconds
kΩ × μF = milliseconds
kΩ × nF = microseconds
MΩ × μF = seconds
MΩ × nF = milliseconds

Example:

R = 10 kΩ
C = 1 μF

Using the shortcut:

τ = 10 × 1
τ = 10 ms

Using base units:

R = 10,000 Ω
C = 0.000001 F

τ = 10,000 × 0.000001
τ = 0.01 s
τ = 10 ms

Both are correct.

The danger is entering the wrong unit.

A 1 μF capacitor is:

1 μF = 0.000001 F

A 1 mF capacitor is:

1 mF = 0.001 F

That is 1000× larger.

A timing circuit intended to settle in milliseconds can accidentally become seconds just from a unit mistake.

Charging is exponential, not linear

When a capacitor charges from zero toward a DC supply voltage, the voltage follows:

vC(t) = Vsupply × (1 − e^(−t/τ))

That means the capacitor does not charge at a constant rate.

It rises quickly at first, then slows down as it approaches the final voltage.

Useful checkpoints:

1τ = 63.2% of final voltage
2τ = 86.5%
3τ = 95.0%
4τ = 98.2%
5τ = 99.3%

This is why engineers often use:

t_settle = 5τ

as a practical settling estimate.

It does not mean the capacitor is mathematically “finished” charging. It means the remaining error is small enough for many practical circuits.

Discharge uses the same time constant

When a charged capacitor discharges through a resistor, the voltage follows:

vC(t) = Vinitial × e^(−t/τ)

After one time constant, the capacitor still has about 36.8% of its initial voltage.

After five time constants, it has about 0.7% remaining.

That is why a capacitor can still be energized after the power is off.

The circuit may look dead.

The capacitor may not be dead.

This is especially important in high-voltage DC links, power supplies, motor drives, inverters, test equipment, and capacitor banks.

Worked example: MCU reset delay

Suppose a simple RC delay is used for a microcontroller reset input.

The design uses:

R = 100 kΩ
C = 1 μF
Vsupply = 3.3 V

The time constant is:

τ = R × C
τ = 100,000 × 0.000001
τ = 0.1 s
τ = 100 ms

The 5τ settling time is:

t_settle = 5 × 100 ms
t_settle = 500 ms

At first glance, the circuit gives a nice slow reset ramp.

But many reset inputs do not switch at 99.3% of supply.

They switch at some threshold.

Suppose the reset input releases at about 70% of supply.

The threshold voltage is:

Vthreshold = 0.70 × 3.3
Vthreshold = 2.31 V

The time to reach a threshold while charging from zero is:

t = −τ × ln(1 − Vthreshold / Vsupply)

Since Vthreshold / Vsupply = 0.70:

t = −100 ms × ln(1 − 0.70)
t = −100 ms × ln(0.30)
t ≈ 120 ms

So the reset may release after about 120 ms, not 500 ms.

That is a common interpretation mistake.

5τ means “nearly settled.”

It does not mean “the circuit event happens at 5τ.”

The actual event depends on the threshold.

The engineering mistake: treating supply voltage as part of τ

A common mistake is thinking that a higher supply voltage makes an RC circuit charge “faster.”

It does not change the time constant.

The time constant is:

τ = R × C

There is no voltage term in that formula.

If R and C stay the same, τ stays the same.

For example:

R = 10 kΩ
C = 1 μF

At 5 V:

τ = 10 ms

At 24 V:

τ = 10 ms

At 400 V:

τ = 10 ms

The timing scale is the same.

But supply voltage changes other things:

Final capacitor voltage
Peak inrush current
Stored energy
Shock hazard
Voltage rating margin
Resistor pulse energy

That is the trap.

Voltage may not change the time constant, but it can completely change the safety and stress level of the circuit.

Peak inrush current can be the real problem

When a discharged capacitor is connected to a supply through a resistance, the initial charging current is:

Ipeak = Vsupply / R

This is the maximum current at the instant of switch closure, assuming the capacitor starts at zero volts.

Example:

Vsupply = 24 V
R = 2 Ω
C = 1000 μF

The time constant is:

τ = 2 × 0.001
τ = 0.002 s
τ = 2 ms

That looks fast.

But the peak inrush current is:

Ipeak = 24 / 2
Ipeak = 12 A

The capacitor charges quickly, but the switch, diode, MOSFET, relay contact, connector, or PCB trace may see a large current pulse.

If the engineer only checks τ, the design may look fine.

If the engineer checks Ipeak, the design may need a current limiter, soft-start circuit, NTC thermistor, precharge resistor, or a different switching device.

For many real designs, the question is not only:

How long does it take to charge?

It is also:

What current flows at the first instant?

Stored energy grows with voltage squared

A capacitor stores energy:

WC = 0.5 × C × V²

Voltage is squared.

That makes high-voltage capacitors very different from low-voltage capacitors even when the capacitance looks modest.

Example 1:

C = 1000 μF
V = 12 V

Stored energy:

WC = 0.5 × 0.001 × 12²
WC = 0.072 J

That is 72 mJ.

Now use the same capacitance at 400 V:

C = 1000 μF
V = 400 V

Stored energy:

WC = 0.5 × 0.001 × 400²
WC = 80 J

Same capacitance.

Much higher voltage.

The stored energy increased from 0.072 J to 80 J.

That is more than 1000× higher.

This is why high-voltage bulk capacitors need serious discharge planning. The circuit may be powered off, but the capacitor can still store enough energy to be hazardous.

Bleed resistor design is a tradeoff

A bleed resistor gives the capacitor a controlled discharge path after power-off.

For a bleed resistor:

τbleed = Rbleed × C

The time to discharge from a starting voltage to a target voltage is:

t = τbleed × ln(Vstart / Vtarget)

Suppose a DC link capacitor has:

C = 1000 μF
Vstart = 400 V
Target voltage = 50 V

The design goal is to discharge below 50 V in 60 seconds.

Rearrange the equation:

τbleed = t / ln(Vstart / Vtarget)

Calculate:

τbleed = 60 / ln(400 / 50)
τbleed = 60 / ln(8)
τbleed ≈ 60 / 2.079
τbleed ≈ 28.9 s

Now find the bleed resistor:

Rbleed = τbleed / C
Rbleed = 28.9 / 0.001
Rbleed = 28,900 Ω

So a resistor around 29 kΩ would meet the timing target in a first-pass calculation.

But now check continuous power while the supply is on:

Pbleed = V² / Rbleed
Pbleed = 400² / 28,900
Pbleed ≈ 5.5 W

That is not a tiny resistor.

The engineer may need a higher power rating, multiple resistors in series, voltage-rating checks, thermal spacing, and derating.

The mistake is sizing a bleed resistor only by discharge time.

Faster discharge requires lower resistance.

Lower resistance increases continuous power loss.

There is always a tradeoff.

RC cutoff frequency is related, but not the same question

The first-order RC cutoff frequency is:

fc = 1 / (2π × R × C)

Since:

τ = R × C

then:

fc = 1 / (2π × τ)

This connects time-domain behavior and frequency-domain behavior.

Example:

R = 10 kΩ
C = 1 μF
τ = 10 ms

Cutoff frequency:

fc = 1 / (2π × 0.01)
fc ≈ 15.9 Hz

That number is useful for first-order filter intuition.

But it is not a magic boundary where signals suddenly stop.

At the cutoff frequency, the output is down by about 3 dB. Above that, the signal is increasingly attenuated. Below that, it is not perfectly unchanged either, depending on the exact topology.

The mistake is treating fc as an on/off limit.

A first-order RC filter is a slope, not a brick wall.

Real capacitors are not ideal

The formula assumes a simple capacitor.

Real capacitors bring extra behavior:

ESR
ESL
Leakage current
Voltage rating
Temperature coefficient
Tolerance
DC bias loss
Dielectric absorption
Ripple current limit
Polarity limits for electrolytic and tantalum capacitors

For timing circuits, tolerance can dominate.

A nominal design may use:

R = 100 kΩ ±1%
C = 1 μF ±20%

Nominal time constant:

τ = 100 ms

But the capacitor tolerance alone can push the actual value roughly between:

80 ms and 120 ms

Add resistor tolerance and temperature effects, and the real spread can be wider.

This matters for reset timing, watchdog delays, debounce circuits, one-shot timers, and analog filtering.

For stable timing, capacitor type matters. A Class II ceramic capacitor may lose effective capacitance under DC bias. An electrolytic capacitor may have wide tolerance and leakage. A C0G/NP0 capacitor is stable, but usually available only in much smaller capacitance values.

The formula uses the capacitance value you enter.

The circuit uses the capacitance value the part actually has under voltage, temperature, tolerance, and aging.

The common mistake: ignoring the actual threshold

Many RC timing mistakes come from assuming the circuit switches at 1τ or 5τ.

But real inputs switch at thresholds.

A CMOS input may switch around a logic threshold.

A Schmitt trigger may switch at a defined upper or lower threshold.

A comparator may use a reference voltage.

A 555 timer has internal thresholds.

That means the correct question is not always:

What is τ?

The correct question may be:

When does vC(t) cross the actual threshold?

For charging from zero:

t = −τ × ln(1 − Vthreshold / Vsupply)

For discharge:

t = τ × ln(Vinitial / Vtarget)

This is where engineering judgment enters.

The RC formula gives the time scale.

The receiving circuit defines the event.

Practical design takeaway

The RC time constant is a great first-pass tool.

It helps answer:

How fast does the capacitor charge?
How fast does it discharge?
When is the circuit practically settled?
What is the peak inrush current?
How much energy is stored?
What is the first-order cutoff frequency?
How long does a bleed resistor take to reduce the voltage?
How much continuous power does the bleed resistor dissipate?

But it does not answer everything.

It does not automatically verify capacitor voltage rating.

It does not account for DC bias derating unless you do that separately.

It does not guarantee timing accuracy with wide-tolerance capacitors.

It does not make a high-voltage capacitor safe after power-off.

It does not replace checking switch, diode, MOSFET, fuse, and resistor pulse ratings.

The formula is the starting point, not the whole design.

Final thought

The RC time constant is one of the most useful equations in practical electronics:

τ = R × C

It explains debounce circuits, reset delays, filters, snubbers, bulk capacitor charging, discharge paths, soft-start behavior, and timing networks.

But the same simple formula can hide important problems.

Supply voltage does not change τ, but it changes stored energy and inrush current.

A capacitor may look charged after 5τ, but a logic input may switch much earlier.

A bleed resistor may discharge the capacitor quickly, but it may also burn continuous power while the equipment is running.

And a high-voltage capacitor can remain dangerous long after the circuit is switched off.

For quick RC transient checks, 5τ settling time, peak inrush current, stored capacitor energy, cutoff frequency, and bleed resistor discharge calculations, use the RC Time Constant Calculator on CalcEngineer.

Op-Amp Gain: The Feedback Ratio That Fails When Bandwidth Runs Out

Evgenii Konkin — Tue, 07 Jul 2026 07:51:59 +0000

Op-amp gain looks like one of the cleanest calculations in electronics.

For an inverting amplifier:

Av = −Rf / Rin

For a non-inverting amplifier:

Av = 1 + Rf / Rg

Pick the resistors, calculate the ratio, and the gain is done.

At least, that is the way it looks in a textbook.

In real analogue design, the resistor ratio is only the first layer. The circuit can have the correct gain on paper and still fail because of bandwidth, noise gain, output swing, input common-mode range, resistor value selection, bias current, or a simple misunderstanding of what the op-amp is actually being asked to do.

The trap is that op-amp gain is easy to calculate but easy to over-trust.

The basic inverting amplifier formula

The classic inverting op-amp gain is:

Av = −Rf / Rin

Where:

Av = closed-loop signal gain
Rf = feedback resistor
Rin = input resistor

The negative sign matters.

It tells you that the output is inverted relative to the input.

If:

Rf = 100 kΩ
Rin = 10 kΩ

Then:

Av = −100,000 / 10,000
Av = −10

So a 100 mV input signal becomes:

Vout = −10 × 0.1 V
Vout = −1.0 V

That is simple and useful.

But it is not the full design check.

Gain in dB

Voltage gain is often expressed in decibels:

Gain dB = 20 × log10(|Av|)

For a gain magnitude of 10:

Gain dB = 20 × log10(10)
Gain dB = 20 dB

For a gain magnitude of 100:

Gain dB = 20 × log10(100)
Gain dB = 40 dB

For a gain magnitude of 1000:

Gain dB = 20 × log10(1000)
Gain dB = 60 dB

This is a good quick sanity check.

A single-stage gain of 10 or 20 may be routine.

A gain of 100 may be possible, but bandwidth, noise, and offset start to matter much more.

A gain of 1000 in one op-amp stage is usually a warning sign. It may work in a narrow low-frequency case, but it often deserves a two-stage design or an instrumentation amplifier instead.

Non-inverting amplifier gain

For a non-inverting amplifier, the gain is:

Av = 1 + Rf / Rg

Where:

Rf = feedback resistor
Rg = resistor from inverting input to ground/reference

If:

Rf = 99 kΩ
Rg = 1 kΩ

Then:

Av = 1 + 99,000 / 1,000
Av = 100

The gain is positive, so the output is not inverted.

The important detail is that the minimum gain of the standard non-inverting configuration is 1.

You cannot get a gain of 0.5 from the basic non-inverting formula. If you need attenuation, you usually need a divider, an inverting attenuator, a difference stage, or another topology.

This is a common mistake when someone tries to use a non-inverting op-amp as a “buffer with adjustable gain below unity.”

A voltage follower can buffer at unity gain.

It cannot attenuate by itself.

Signal gain vs noise gain

The most important op-amp gain mistake is confusing signal gain with noise gain.

For an inverting amplifier, signal gain is:

Av = −Rf / Rin

But the noise gain is:

Av_noise = 1 + Rf / Rin

These two numbers are not the same.

Example:

Rf = 100 kΩ
Rin = 10 kΩ

Signal gain:

Av = −100,000 / 10,000
Av = −10

Noise gain:

Av_noise = 1 + 100,000 / 10,000
Av_noise = 11

The signal is amplified by −10.

But the op-amp’s input-referred noise, offset, and bandwidth behavior are governed by a noise gain of 11.

That small difference may not matter in a rough low-frequency circuit.

But it matters when checking bandwidth and stability.

Gain-bandwidth product

For many voltage-feedback op-amps, a first-pass closed-loop bandwidth estimate is:

BW = GBW / Av_noise

Where:

BW = estimated closed-loop bandwidth
GBW = op-amp gain-bandwidth product
Av_noise = noise gain

This is a screening calculation, not a complete stability analysis. But it is extremely useful.

Suppose an op-amp has:

GBW = 1 MHz

And the inverting stage uses:

Rf = 100 kΩ
Rin = 10 kΩ

We already calculated:

Signal gain = −10
Noise gain = 11

So the estimated closed-loop bandwidth is:

BW = 1,000,000 / 11
BW ≈ 90,909 Hz
BW ≈ 90.9 kHz

That may be acceptable for a 10 kHz signal.

It may be marginal for a 50 kHz signal depending on amplitude accuracy and phase shift requirements.

It is not acceptable if someone expected a clean 100 kHz amplifier with no margin.

This is where the resistor-ratio calculation can mislead the design.

The gain can be mathematically correct while the selected op-amp is too slow.

Worked example: a sensor amplifier that looks right but fails at frequency

Suppose an engineer needs to amplify a sensor signal.

The requirement is:

Input signal amplitude = 50 mV
Target output amplitude = 2.5 V
Signal frequency = 20 kHz
Op-amp GBW = 1 MHz
Topology = non-inverting

The required gain is:

Av = 2.5 / 0.05
Av = 50

For a non-inverting amplifier:

Av = 1 + Rf / Rg

Choose:

Rg = 2 kΩ
Rf = 98 kΩ

Then:

Av = 1 + 98,000 / 2,000
Av = 50

So far, the resistor calculation looks perfect.

Now check bandwidth.

For a non-inverting amplifier:

Av_noise = Av
Av_noise = 50

Estimated closed-loop bandwidth:

BW = GBW / Av_noise
BW = 1,000,000 / 50
BW = 20,000 Hz
BW = 20 kHz

The required signal frequency is also 20 kHz.

That means the amplifier is being used right at the estimated closed-loop bandwidth.

That is not a comfortable design.

At this point, the gain may be down, the phase shift may be significant, and the waveform may not match the intended signal-chain requirement.

The resistor ratio gave the right gain.

The op-amp selection did not give enough bandwidth margin.

A better design might use:

A higher-GBW op-amp
Lower gain per stage across two stages
A different signal-chain architecture
A reduced target bandwidth

The engineering mistake was not the gain formula.

The mistake was stopping at the gain formula.

The output swing check

Another simple failure mode is ignoring the supply rails.

Suppose the same amplifier has:

Input amplitude = 100 mV
Gain = 50

The expected output amplitude is:

Vout = 0.1 × 50
Vout = 5 V

If the op-amp is powered from a single 5 V supply, that output is probably not possible.

Even a rail-to-rail op-amp cannot usually swing perfectly to both rails under load. A non-rail-to-rail op-amp may need much more headroom.

So a gain calculation can say:

Av = 50

But the real circuit says:

The output saturates.

This is why every gain check should be followed by a voltage swing check:

Expected output min/max
Supply voltage
Output swing limit from datasheet
Load resistance
Required signal headroom

The op-amp does not care that the resistor formula is correct if the output stage cannot physically deliver the voltage.

Resistor values matter

The ratio sets gain, but the absolute resistor values still matter.

For an inverting gain of −10, all of these ratios work mathematically:

Rf = 1 kΩ, Rin = 100 Ω
Rf = 10 kΩ, Rin = 1 kΩ
Rf = 100 kΩ, Rin = 10 kΩ
Rf = 1 MΩ, Rin = 100 kΩ

All give:

Av = −10

But they are not equivalent designs.

Very small resistor values can load the op-amp output and waste current.

Very large resistor values increase sensitivity to input bias current, leakage, noise, parasitic capacitance, and PCB contamination.

For many general-purpose analogue stages, the practical resistor range is often somewhere around:

1 kΩ to 100 kΩ

That is not a universal rule. Precision, low-noise, high-speed, low-power, and sensor front-end designs may need different choices.

But it is a useful first-pass sanity check.

If a gain stage uses 10 Ω feedback resistors, ask why.

If it uses 50 MΩ resistors, ask why.

The ratio alone is not the whole design.

Difference amplifier: resistor matching matters more than the ratio alone

For a simplified matched difference amplifier:

Av_diff = Rf / R1

This assumes matched resistor pairs:

R1 = R3
Rf = R2

The ideal formula is clean.

But real common-mode rejection depends strongly on resistor matching.

That means two difference amplifiers can have the same nominal gain and very different real-world performance.

For example, a difference amplifier may be intended to reject a large common-mode voltage while amplifying a small differential signal.

If the resistor ratios are not matched closely, part of the common-mode signal appears at the output as an error.

So the mistake is thinking:

The gain is correct, so the difference amplifier is correct.

A better check is:

Is the resistor network matched tightly enough for the required CMRR?
Is the input common-mode voltage inside the op-amp limit?
Is the output swing still inside the rails?
Is an instrumentation amplifier more appropriate?

A discrete four-resistor difference amplifier can work well.

But for precision measurement, matched resistor networks or instrumentation amplifier ICs are often the better choice.

Summing amplifier: each input has its own gain

For an inverting summing amplifier:

Av_n = −Rf / Rn

Each input channel has its own resistor and its own gain.

For example:

Rf = 10 kΩ
R1 = 10 kΩ
R2 = 5 kΩ
R3 = 2 kΩ

Per-channel gains:

Av1 = −10,000 / 10,000 = −1
Av2 = −10,000 / 5,000 = −2
Av3 = −10,000 / 2,000 = −5

The output is:

Vout = −(1 × V1 + 2 × V2 + 5 × V3)

The common mistake is checking only the highest per-channel gain and forgetting the summed output magnitude.

Even if each input is inside a reasonable range, several channels can add together and drive the output into saturation.

For example:

V1 = 1 V
V2 = 1 V
V3 = 1 V

Then:

Vout = −(1 + 2 + 5)
Vout = −8 V

That may be fine on a ±12 V supply.

It will not be fine on a single 5 V supply.

Summing amplifiers need both gain checks and worst-case output swing checks.

The common engineering mistake: using gain as the only design criterion

A very common op-amp design review looks like this:

Required gain = 100
Rf = 100 kΩ
Rin = 1 kΩ
Av = −100
Done

But a useful review should continue:

What is the noise gain?
What is the closed-loop bandwidth?
What is the input signal frequency?
What output swing is required?
Can the op-amp output reach that voltage on the selected supply?
Are the inputs inside the common-mode range?
Are the resistor values practical?
Will bias current create an offset error?
Is the single-stage gain too high?
Is the op-amp stable in this configuration?

The resistor ratio is the beginning of the design.

It is not the design.

Practical design takeaway

Use the gain formula to get the first answer.

Then immediately check the things the formula does not cover.

For an inverting amplifier:

Av = −Rf / Rin
Av_noise = 1 + Rf / Rin
BW = GBW / Av_noise

For a non-inverting amplifier:

Av = 1 + Rf / Rg
Av_noise = Av
BW = GBW / Av_noise

For gain in decibels:

Gain dB = 20 × log10(|Av|)

Then ask the practical questions:

Is the bandwidth high enough?
Is the output swing possible?
Are the resistor values reasonable?
Is the input common-mode range valid?
Is the op-amp stable at this noise gain?
Is the single-stage gain realistic?

That workflow catches many problems before simulation, PCB layout, or bench testing.

Final thought

Op-amp gain is simple when it is only a resistor ratio.

But real circuits do not fail only because the ratio is wrong.

They fail because the gain was correct at DC but not at the required frequency.

They fail because the output saturated.

They fail because the designer confused signal gain with noise gain.

They fail because a 1000× stage was forced into one op-amp when two moderate-gain stages would have worked better.

The resistor formula is still the right starting point.

It just should not be the stopping point.

For quick checks of inverting, non-inverting, voltage follower, difference, and summing amplifier gain — including gain in dB, noise gain, and closed-loop bandwidth screening — use the Op-Amp Gain Calculator on CalcEngineer.

RL Time Constant: The Coil Delay and Back-EMF Calculation Engineers Underestimate

Evgenii Konkin — Sun, 05 Jul 2026 16:10:59 +0000

An RL circuit looks harmless.

A resistor, an inductor, a DC supply, and a switch.

But that simple circuit explains a lot of real engineering problems: relay pickup delay, solenoid response, brake coil release time, MOSFET failures, contact arcing, and inductive kickback.

The key number is the RL time constant:

τ = L / R

It tells you how fast current rises or decays in a series resistor-inductor circuit.

That sounds simple.

But the mistake is treating τ as just a classroom formula. In real circuits, τ affects timing, stored energy, heat, and switching stress. And when the circuit is turned off, the inductor does not politely stop conducting current. The stored magnetic energy has to go somewhere.

That is where many failures begin.

The basic RL time constant formula

For a series RL circuit:

τ = L / R

Where:

τ = time constant, seconds
L = inductance, henries
R = total series resistance, ohms

The resistance is not only the external resistor.

It should include:

Inductor winding resistance
External series resistance
Switch on-resistance, if relevant
Wiring or trace resistance, if significant

This matters because τ depends directly on the total resistance.

If resistance is underestimated, the calculated time constant becomes too large and the current prediction becomes misleading.

Current does not rise instantly

When a DC voltage is applied to an RL circuit, the final steady-state current is:

I_ss = V / R

But the inductor prevents current from jumping instantly to that value.

The current rise is exponential:

i(t) = I_ss × (1 − e^(−t/τ))

After one time constant, the current reaches about 63.2% of its final value.

After five time constants, it reaches about 99.3%.

That is why engineers often use:

t_settle = 5τ

as a practical settling-time estimate.

It does not mean the circuit is mathematically finished changing. It means the remaining error is small enough for many engineering checks.

Worked example: relay coil timing

Suppose a relay coil has:

L = 200 mH
R = 400 Ω
V = 24 V DC

First convert inductance:

L = 200 mH = 0.200 H

Calculate the time constant:

τ = L / R
τ = 0.200 / 400
τ = 0.0005 s
τ = 0.5 ms

The five-time-constant settling estimate is:

t_settle = 5τ
t_settle = 5 × 0.5 ms
t_settle = 2.5 ms

The steady-state current is:

I_ss = V / R
I_ss = 24 / 400
I_ss = 0.060 A
I_ss = 60 mA

So electrically, the coil current approaches 60 mA with a 0.5 ms time constant and is nearly settled after about 2.5 ms.

That is useful, but it does not mean the relay mechanically switches in 2.5 ms.

This is one of the most common interpretation mistakes.

Electrical settling is not mechanical pickup time

A relay or solenoid is not only an RL circuit.

It also has:

Magnetic force buildup
Armature movement
Spring force
Contact bounce
Friction
Mechanical travel distance
Manufacturing tolerance
Temperature effects

The RL time constant describes the electrical current response.

It does not fully predict the mechanical pickup or release time.

A relay may have an electrical current rise that is mostly complete in a few milliseconds, while the mechanical pickup time listed in the datasheet is much longer.

So the correct interpretation is:

τ tells you how fast coil current changes.
The datasheet tells you how fast the relay or actuator actually moves.

Both matter.

Using τ alone as the switching time can make a control circuit look faster than it really is.

Supply voltage does not change τ

Another common mistake is assuming that increasing supply voltage makes the RL time constant smaller.

It does not.

The time constant is:

τ = L / R

There is no voltage term in that equation.

If L and R stay the same, τ stays the same.

What voltage changes is the final current:

I_ss = V / R

Example with the same coil:

L = 0.200 H
R = 400 Ω

At 24 V:

I_ss = 24 / 400 = 60 mA
τ = 0.200 / 400 = 0.5 ms

At 48 V:

I_ss = 48 / 400 = 120 mA
τ = 0.200 / 400 = 0.5 ms

The time constant is unchanged.

But because the final current is higher, the current may cross a required pickup threshold sooner. That can make the device appear to respond faster, but the exponential time scale itself did not change.

This distinction matters when debugging relay drivers, solenoid pull-in circuits, and coil overdrive schemes.

Stored energy matters when the switch opens

At steady state, the inductor stores energy in its magnetic field:

W_L = 0.5 × L × I²

For the relay coil example:

L = 0.200 H
I = 0.060 A

Stored energy is:

W_L = 0.5 × 0.200 × 0.060²
W_L = 0.00036 J
W_L = 360 µJ

That is small, but it still has to go somewhere when the coil is switched off.

For a small relay, that energy may be handled easily with a flyback diode.

For a large solenoid, brake coil, contactor, or actuator, stored energy can be much higher. Then the discharge path becomes an actual design item, not an afterthought.

Back-EMF: the dangerous part of turning off an inductor

An inductor opposes a change in current.

When a switch opens, the current path is suddenly interrupted. The inductor responds by generating whatever voltage is needed to keep current flowing.

A simple screening estimate for the voltage spike is:

V_kickback ≈ L × I / t_switch_open

Where:

V_kickback = estimated inductive voltage spike
L = inductance, H
I = current before opening, A
t_switch_open = switch opening time, s

This is only a first-pass estimate. Real voltage is limited by parasitic capacitance, arcing, avalanche breakdown, clamp devices, and insulation limits.

But as a warning signal, it is very useful.

Example: small coil, fast switch, big voltage estimate

Suppose a coil has:

L = 100 mH = 0.100 H
I = 100 mA = 0.100 A

A MOSFET turns it off very quickly:

t_switch_open = 100 ns = 0.0000001 s

Estimate the kickback voltage:

V_kickback ≈ L × I / t_switch_open
V_kickback ≈ 0.100 × 0.100 / 0.0000001
V_kickback ≈ 100,000 V

That does not mean you will actually measure 100 kV on the board.

Something will clamp or break down first.

Maybe the MOSFET avalanches.

Maybe the relay contact arcs.

Maybe the insulation flashes over.

Maybe the parasitic capacitance absorbs part of the transient.

But the calculation tells the engineer the important message:

This coil cannot be switched off safely without a controlled discharge path.

That is the engineering value of the back-EMF estimate.

It is not a precise oscilloscope prediction.

It is a warning that the circuit needs protection.

The common mistake: omitting the flyback path

A frequent failure mode is driving a relay, solenoid, or small coil with a transistor or MOSFET and forgetting the flyback diode, TVS, Zener clamp, MOV, or snubber.

The circuit may work once.

It may work for a day.

Then the switching device fails.

The reason is not mysterious. Every time the switch opens, the inductor forces current to continue. If there is no safe path, the voltage rises until some unintended path conducts.

That path may be the MOSFET avalanche rating.

It may be the transistor junction.

It may be a contact arc.

It may be a nearby insulation weak point.

A proper protection device gives the inductor a safe place to discharge energy.

Flyback diode vs faster release

A flyback diode is simple and common.

It clamps the coil voltage to roughly a diode drop and protects the switch.

But it also slows current decay.

Because the clamp voltage is low, the rate of current decay is low.

That can delay relay release or solenoid drop-out.

For many relay coils, that is acceptable.

For fast release applications, a higher-voltage clamp may be better:

TVS diode
Zener clamp
RC snubber
Active clamp
MOV for larger industrial coils

Higher clamp voltage usually means faster current decay, but more voltage stress on the switching device.

So the design question is not only:

How do I protect the switch?

It is also:

How fast does the coil need to release?

Protection and timing are connected.

Resistance changes more than one thing

Increasing resistance reduces the time constant:

τ = L / R

So higher R means faster electrical response in terms of τ.

But resistance also reduces steady-state current:

I_ss = V / R

And it changes dissipation:

P_R = V² / R

That means adding resistance is not a free fix.

For example:

L = 0.100 H
V = 24 V

Case 1:

R = 100 Ω

τ = 0.100 / 100 = 1 ms
I_ss = 24 / 100 = 0.24 A
P_R = 24² / 100 = 5.76 W

Case 2:

R = 200 Ω

τ = 0.100 / 200 = 0.5 ms
I_ss = 24 / 200 = 0.12 A
P_R = 24² / 200 = 2.88 W

The time constant became smaller, but the final current was cut in half.

If the coil or actuator needs a certain current to operate, that change may not be acceptable.

This is why RL calculations should be tied to the real device requirement, not done as isolated math.

Characteristic frequency

The first-order characteristic frequency of an RL circuit is:

f_c = R / (2π × L)

This is useful for understanding the frequency scale of the first-order response.

But it should not be confused with a full EMI filter design.

Real inductors have:

Parasitic capacitance
Self-resonant frequency
Core loss
Saturation effects
Winding resistance
Temperature dependence
Coupling to nearby conductors

So f_c is a useful first-order intuition, not a complete high-frequency model.

Unit mistakes are easy

RL calculations are sensitive to unit selection.

Inductance may be listed as:

H
mH
µH
nH

Resistance may be:

Ω
mΩ
kΩ

A 1000× inductance error becomes a 1000× time constant error.

For example:

100 µH = 0.0001 H
100 mH = 0.100 H

Those values look similar in text, but the second is 1000× larger.

If the resistance is 10 Ω:

τ with 100 µH = 0.0001 / 10 = 10 µs
τ with 100 mH = 0.100 / 10 = 10 ms

That is the difference between a fast power electronics transient and a much slower relay or solenoid-type response.

The formula is simple, but the unit dropdown matters.

Practical engineering takeaway

Use the RL time constant as a first-pass design and troubleshooting tool.

It helps answer:

How fast does coil current rise?
How long before the circuit is practically settled?
What is the steady-state current?
How much magnetic energy is stored?
How much resistor power is dissipated?
How large could the switch-opening voltage spike be?
Does the circuit need a flyback diode, TVS, snubber, or active clamp?

But do not use it as a complete device model.

It does not prove relay pickup time.

It does not model saturation.

It does not calculate contact bounce.

It does not fully predict EMI.

It does not replace the datasheet or protection-device rating check.

It gives the engineer the first warning signs.

And that is often exactly what is needed before a small coil becomes a failed transistor, welded contact, or delayed actuator.

Final thought

The RL time constant formula is short:

τ = L / R

But it connects timing, current, energy, heat, and switching stress.

That is why it is more than a textbook circuit equation.

If you are driving a relay, solenoid, brake coil, contactor, or inductor with a switching device, the time constant and the back-EMF path should be checked early.

A circuit that looks fine in steady state can still fail during switching.

For quick RL transient checks, settling time, stored energy, dissipation, and back-EMF screening, use the RL Time Constant Calculator on CalcEngineer.

Inductor Stored Energy: The Small Formula That Can Hide a Big Switching Problem

Evgenii Konkin — Fri, 03 Jul 2026 16:22:39 +0000

Inductor energy storage looks like a small detail.

The formula is short:

W = 0.5 × L × I²

That makes it easy to treat stored magnetic energy as a quick side calculation.

But in real electrical and power electronics work, this small formula can expose a much bigger design issue: switching stress, fault energy, saturation risk, clamp requirements, and the difference between a harmless filter inductor and a component that deserves a deeper protection review.

The dangerous part is not the inductance.

It is usually the current.

Because current is squared.

The basic formula

Stored magnetic energy in an inductor is calculated as:

W = 0.5 × L × I²

Where:

W = stored magnetic energy, J
L = inductance, H
I = current magnitude, A

The formula uses inductance in henries.

That means unit conversion matters:

1 H = 1 H
1 mH = 0.001 H
1 µH = 0.000001 H

So before calculating energy, the inductance must be converted into henries.

For example:

2.5 mH = 2.5 / 1000 = 0.0025 H
220 µH = 220 / 1,000,000 = 0.00022 H

Current direction does not change stored energy because the formula uses current squared. A current of -3 A and +3 A stores the same energy if the inductance is the same.

Current dominates faster than many engineers expect

Stored energy increases linearly with inductance.

If inductance doubles, stored energy doubles.

But stored energy increases with the square of current.

If current doubles, stored energy increases by four times.

Example:

L = 2.5 mH
I = 3 A

Convert inductance:

L = 2.5 / 1000
L = 0.0025 H

Calculate stored energy:

W = 0.5 × 0.0025 × 3²
W = 0.5 × 0.0025 × 9
W = 0.01125 J

So the stored energy is:

W = 0.01125 J
W = 11.25 mJ

Now keep the same inductor, but increase the current to 6 A:

W = 0.5 × 0.0025 × 6²
W = 0.5 × 0.0025 × 36
W = 0.045 J

The current doubled.

The stored energy became four times larger:

11.25 mJ → 45 mJ

This is the first practical lesson.

A design that looks harmless at normal operating current may become much more serious at peak current, startup current, short-circuit current, or switching transient current.

Worked example: buck converter inductor check

Suppose an engineer is reviewing an inductor in a DC-DC converter.

The inductor value is:

L = 220 µH

The expected peak current is:

I = 8 A

First convert inductance to henries:

L = 220 / 1,000,000
L = 0.00022 H

Then calculate stored energy:

W = 0.5 × 0.00022 × 8²
W = 0.5 × 0.00022 × 64
W = 0.00704 J

So:

W = 7.04 mJ

That is not a huge energy level, but it is not zero either. It is enough to matter in switching behavior, snubber selection, MOSFET stress review, and fault interruption analysis depending on the circuit.

Now imagine the same converter has a fault or transient condition where current rises to 20 A before protection reacts:

W = 0.5 × 0.00022 × 20²
W = 0.5 × 0.00022 × 400
W = 0.044 J

Now the inductor stores:

W = 44 mJ

The current increased from 8 A to 20 A.

The energy increased from 7.04 mJ to 44 mJ.

That is more than six times higher.

This is why using normal operating current can understate the stored energy that switching devices and protection components may actually experience.

The engineering mistake: using average current when peak current matters

A common mistake is entering the average current instead of the peak or worst-case current.

For example, in a switching converter, the inductor current may ripple around an average value.

Suppose the average inductor current is:

I_avg = 8 A

But the peak current is:

I_peak = 10.5 A

Using the average current:

W_avg = 0.5 × 0.00022 × 8²
W_avg = 0.00704 J

Using the peak current:

W_peak = 0.5 × 0.00022 × 10.5²
W_peak = 0.01213 J

The difference is significant:

7.04 mJ vs 12.13 mJ

That is about 72% higher stored energy when peak current is used.

The inductor did not change.

The formula did not change.

Only the current assumption changed.

This matters because switching stress, clamp energy, current-limit behavior, and fault energy are usually tied to peak or worst-case current, not the average current shown in a simple load table.

Unit mistakes can be catastrophic

The second common mistake is mixing up µH, mH, and H.

This is easy to do because inductor values often look visually similar:

220 µH
220 mH
220 H

But those are not close.

They are separated by factors of 1,000 and 1,000,000.

Take the same current:

I = 8 A

Case 1 — correct value:

L = 220 µH = 0.00022 H

W = 0.5 × 0.00022 × 8²
W = 0.00704 J

Case 2 — wrong unit entered as mH:

L = 220 mH = 0.22 H

W = 0.5 × 0.22 × 8²
W = 7.04 J

That is a 1,000× error.

The result changes from:

7.04 mJ

to:

7.04 J

Those are completely different engineering situations.

A few millijoules may be a routine power electronics check.

Several joules may require serious review of switching devices, clamps, discharge paths, insulation, thermal behavior, and fault handling.

The calculator can do the unit conversion, but the engineer still has to select the correct unit.

Stored energy is not a saturation check

Another trap is assuming that a stored-energy number proves the inductor is safe.

It does not.

The formula tells you how much magnetic energy is associated with the inductance and current.

It does not confirm:

Core saturation margin
Copper loss
Core loss
Temperature rise
Ripple current rating
Insulation stress
Switching-device stress
Clamp or snubber adequacy
Fault interruption behavior

An inductor can show a modest stored-energy value and still saturate if the core is not suitable for the DC bias current.

Or it can avoid saturation but still overheat due to winding loss or core loss.

Stored energy is a screening calculation, not a complete magnetic design.

Practical interpretation

A useful way to think about inductor stored energy is not simply “low” or “high,” but “what happens if this energy has to go somewhere quickly?”

During normal operation, the inductor stores and releases energy every switching cycle.

During a fault, shutdown, open circuit, or rapid current interruption, the magnetic field collapses and the circuit needs a safe energy path.

That path might be:

A diode
A MOSFET body diode
An active clamp
A TVS device
An RC snubber
A flyback winding
A controlled current decay path

If that path is not designed properly, the inductor will force voltage to rise until current can continue flowing somewhere.

That is where switching failures often come from.

The inductor is not “trying” to create a problem.

It is simply obeying the energy equation.

Quick comparison table

Here is how strongly current changes stored energy for a 1 mH inductor:

L = 1 mH = 0.001 H

At 1 A:

W = 0.5 × 0.001 × 1²
W = 0.0005 J = 0.5 mJ

At 5 A:

W = 0.5 × 0.001 × 5²
W = 0.0125 J = 12.5 mJ

At 10 A:

W = 0.5 × 0.001 × 10²
W = 0.05 J = 50 mJ

At 20 A:

W = 0.5 × 0.001 × 20²
W = 0.2 J = 200 mJ

The current increased by 20× from 1 A to 20 A.

The stored energy increased by 400×.

That is the quadratic effect.

Final thought

Inductor stored energy is one of those calculations that looks too simple to be dangerous.

But the formula carries two important engineering warnings.

First, current matters more than intuition suggests because it is squared.

Second, unit selection matters because µH, mH, and H are separated by very large factors.

The stored-energy calculation will not design the full magnetic component for you. It will not verify saturation, thermal behavior, ripple current, or switching protection. But it is an excellent first-pass check for understanding whether an inductor is storing a trivial amount of energy or enough energy to deserve deeper review.

For quick stored-energy checks, unit conversion, and energy-range screening, use the Inductor Energy Storage Calculator on CalcEngineer.

Data Center PUE: The Efficiency Metric That Gets Misread When IT Load Changes

Evgenii Konkin — Mon, 29 Jun 2026 16:59:05 +0000

Power Usage Effectiveness looks like one of the simplest data center metrics.

Take the total facility power, divide it by the IT equipment power, and you get PUE.

PUE = Total Facility Power / IT Equipment Power

A PUE of 1.5 means the facility uses 1.5 watts at the building level for every 1 watt delivered to IT equipment.

Simple enough.

But in real data center engineering, PUE is often misread, especially when people compare different sites, use a short power snapshot, or forget that IT load sits in the denominator.

That last point is where many confusing conversations start.

A data center can reduce total energy and still show a worse PUE.

That does not always mean the design got worse. Sometimes it means the IT load dropped faster than the supporting infrastructure load.

The basic PUE formula

The core definition is:

PUE = Total Facility Power / IT Equipment Power

For energy over a period:

PUE = Total Facility Energy / IT Equipment Energy

The idea is the same in both cases.

Total facility power or energy includes the entire data center boundary: IT load, cooling, UPS and transformer losses, lighting, and other support loads.

IT equipment power or energy includes the computing, storage, and network equipment.

Because total facility energy includes IT energy, PUE should not be below 1.0.

Minimum theoretical PUE = 1.0

A PUE of 1.0 would mean every watt entering the facility reaches IT equipment, with zero cooling, power-distribution, lighting, or support losses. That is an ideal floor, not a normal real-world result.

DCiE is the reciprocal, not the same metric

The reciprocal of PUE is DCiE:

DCiE (%) = (1 / PUE) × 100

If a data center has:

Total Facility Power = 1,200 kW
IT Equipment Power = 750 kW

Then:

PUE = 1,200 / 750
PUE = 1.60

And:

DCiE = (1 / 1.60) × 100
DCiE = 62.5%

That means 62.5% of the incoming facility power reaches IT equipment.

The remaining 37.5% is infrastructure overhead.

This is where language matters. PUE is not a percentage. DCiE is a percentage. Saying “PUE is 62.5%” is mixing two different views of the same energy split.

Overhead ratio: the number engineers should not ignore

A useful way to interpret PUE is overhead ratio:

Overhead ratio = PUE − 1

For a PUE of 1.60:

Overhead ratio = 1.60 − 1
Overhead ratio = 0.60

That means the non-IT overhead equals 60% of the IT load.

So if IT load is 750 kW:

Overhead = 750 × 0.60
Overhead = 450 kW

And total facility power is:

Total Facility Power = IT Load + Overhead
Total Facility Power = 750 + 450
Total Facility Power = 1,200 kW

This is often easier to explain to project teams than the raw PUE number.

A PUE of 1.60 does not mean “60% efficient.”

It means the facility uses 0.60 kW of support power for every 1.00 kW of IT power.

Worked example: component build-up

Suppose a data center has the following power breakdown:

IT Load = 1,000 kW
Cooling / Mechanical = 300 kW
Power Distribution Losses = 120 kW
Lighting = 30 kW
Other Loads = 50 kW

First calculate the total facility power:

Total Facility Power = IT + Cooling + Power Distribution + Lighting + Other
Total Facility Power = 1,000 + 300 + 120 + 30 + 50
Total Facility Power = 1,500 kW

Now calculate PUE:

PUE = 1,500 / 1,000
PUE = 1.50

Now calculate DCiE:

DCiE = (1 / 1.50) × 100
DCiE = 66.7%

So 66.7% of the facility power reaches IT equipment.

The overhead is:

Overhead = Total Facility Power − IT Load
Overhead = 1,500 − 1,000
Overhead = 500 kW

Now break down the overhead:

Cooling share of overhead = 300 / 500 = 60%
Power distribution share = 120 / 500 = 24%
Lighting share = 30 / 500 = 6%
Other share = 50 / 500 = 10%

This is more useful than simply saying “PUE is 1.50.”

It tells the engineer where the overhead actually lives.

In this case, cooling dominates the non-IT load. If the facility team wants to improve efficiency, cooling is likely the first place to investigate.

The common mistake: using a snapshot as a reporting-grade number

One of the biggest PUE mistakes is taking a single power reading and treating it as the data center’s annual PUE.

A snapshot is useful for a quick engineering check.

But it changes with:

IT load
Outdoor temperature
Cooling mode
Chiller staging
Economizer operation
UPS loading
Humidity control
Part-load equipment efficiency

A power snapshot at 2 PM on a hot day can look very different from a nighttime or winter reading.

For reporting and comparison, energy over a longer period is much more meaningful:

Annual PUE = Annual Total Facility Energy / Annual IT Equipment Energy

For example:

Annual Total Facility Energy = 10,500,000 kWh
Annual IT Equipment Energy = 7,000,000 kWh

Then:

Annual PUE = 10,500,000 / 7,000,000
Annual PUE = 1.50

That number carries more weight than a five-minute power snapshot because it includes weather, operating profile, load variation, and seasonal cooling behavior.

The mistake is not calculating snapshot PUE.

The mistake is quoting it as if it were an annualized performance metric.

Why PUE can get worse after IT load drops

This is the part that surprises many people.

Imagine a facility before a server consolidation project:

IT Load = 1,000 kW
Facility Overhead = 500 kW
Total Facility Power = 1,500 kW

PUE is:

PUE = 1,500 / 1,000
PUE = 1.50

Now the IT team virtualizes workloads and removes old servers. IT load drops:

IT Load = 700 kW
Facility Overhead = 430 kW
Total Facility Power = 1,130 kW

The total facility power improved:

1,500 kW → 1,130 kW

That is a real reduction.

But the new PUE is:

PUE = 1,130 / 700
PUE = 1.61

The PUE got worse.

Does that mean the project failed?

Not necessarily.

The total energy dropped, but the denominator dropped more sharply than the support load. Cooling, power distribution, controls, lighting, and fixed losses often do not scale perfectly with IT load.

This is called the denominator effect.

At low utilization, PUE can look worse even when the building is using less total energy.

That is why PUE should not be the only metric used to judge a data center efficiency project.

PUE does not measure server efficiency

Another common mistake is using PUE as if it measures useful computing work.

It does not.

PUE measures infrastructure overhead around the IT load.

Two data centers can both have a PUE of 1.30, but one may run efficient modern servers at high utilization while the other runs old underutilized hardware.

The PUE would not tell you that.

For IT efficiency, you need different metrics: workload per watt, server utilization, compute efficiency, storage efficiency, or application-level performance per energy unit.

PUE answers one specific question:

How much total facility energy is required for each unit of IT energy?

It does not answer:

How useful is the IT work being done with that energy?

Measurement boundary matters

A PUE number is not very useful unless the measurement boundary is clear.

Where is total facility power measured?

Where is IT power measured?

At the UPS output?

At the PDU?

At the rack?

At the server input?

Those choices change the result.

For example, measuring IT power upstream may include some downstream distribution losses inside the IT number. That makes overhead look smaller. Measuring closer to the rack may produce a more conservative PUE.

Neither number is automatically “wrong,” but they are not directly comparable unless the boundary is stated.

A good PUE statement should say something like:

PUE = 1.48 annualized
Total facility energy measured at utility service entrance
IT energy measured at PDU output
Period: Jan 1 – Dec 31

That is much more useful than simply saying:

Our data center PUE is 1.48

Practical engineering takeaway

PUE is useful when it is used for the right job.

It is good for:

Tracking one facility over time
Comparing before and after retrofit projects
Estimating infrastructure overhead
Identifying whether cooling or power losses dominate
Communicating facility efficiency in a simple ratio

It is weak for:

Comparing unrelated data centers without context
Measuring IT productivity
Measuring carbon impact
Measuring water use
Judging low-load facilities
Replacing detailed energy modeling

The formula is simple, but the interpretation is not.

A PUE of 1.50 may be acceptable for one facility and disappointing for another, depending on climate, redundancy, utilization, measurement boundary, and age of infrastructure.

Final thought

PUE is one of the most useful data center efficiency metrics because it turns a complex facility into a simple ratio.

But the simplicity is also the trap.

A snapshot is not the same as annual performance.

PUE is not the same as DCiE.

A lower IT load can make PUE look worse even while total energy improves.

And a good PUE does not mean the servers are doing useful work efficiently.

The best way to use PUE is as a screening and tracking metric, not as a complete judgment of data center performance.

For quick checks, component build-up, DCiE, overhead ratio, and efficiency banding, use the Data Center PUE Calculator on CalcEngineer.

Hospital Operating Room Airflow: The Simple ACH Calculation That Still Gets Misused

Evgenii Konkin — Sat, 27 Jun 2026 08:26:58 +0000

Operating room airflow looks simple on paper.

Take the room volume, multiply it by the target air changes per hour, convert it to CFM, and the result becomes the required supply airflow.

But in real hospital HVAC design, that simple number is easy to misuse.

The most common mistake is treating operating room airflow as a generic ventilation calculation. An operating room is not just another room with a high air-change rate. The airflow has to support dilution, pressure relationship, thermal control, filtration strategy, and clean air delivery over the surgical zone.

That is why the basic ACH calculation is useful — but only if the engineer understands what it does and what it does not prove.

The basic formula

For an operating room in imperial units:

Room Volume = Length × Width × Height

Required Supply Airflow = Room Volume × ACH / 60

Where:

Room Volume = ft³
ACH = air changes per hour
Required Supply Airflow = CFM

The division by 60 converts cubic feet per hour into cubic feet per minute.

For metric units:

Room Volume = Length × Width × Height

Required Supply Airflow = Room Volume × ACH

Where:

Room Volume = m³
ACH = air changes per hour
Required Supply Airflow = m³/h

So the math itself is not complicated. The problem is usually the interpretation.

ACH is not the same thing as “good OR design”

ACH tells you how many times the room air volume is supplied per hour.

If an operating room has a volume of 4,800 ft³ and is designed for 20 ACH, the supply airflow is:

Required Supply Airflow = 4,800 × 20 / 60
Required Supply Airflow = 1,600 CFM

That looks like a clean answer.

But 1,600 CFM does not automatically prove that the operating room is correctly designed.

It does not prove that the room is positive to the corridor.

It does not prove that the outdoor air quantity is correct.

It does not prove that the diffuser array covers the surgical field properly.

It does not prove that the filtration, return location, door leakage, or balancing strategy is acceptable.

It only proves one thing:

For that room volume and that target ACH, the total supply airflow is 1,600 CFM.

That is an important starting point, not the final design sign-off.

Worked example

Suppose an operating room has the following dimensions:

Length = 24 ft
Width = 20 ft
Height = 10 ft
Target ACH = 20

First calculate the room volume:

Room Volume = 24 × 20 × 10
Room Volume = 4,800 ft³

Then calculate the required supply airflow:

Required Supply Airflow = 4,800 × 20 / 60
Required Supply Airflow = 1,600 CFM

So the operating room needs 1,600 CFM of total supply air to achieve 20 ACH.

Now compare that with a lower airflow case.

If someone only supplied 1,200 CFM to the same room:

ACH = CFM × 60 / Volume
ACH = 1,200 × 60 / 4,800
ACH = 15 ACH

That is a very different ventilation basis. The room may still look like it has “a lot of air” because 1,200 CFM is not a small number, but normalized against the room volume it is only 15 ACH.

This is why CFM by itself can be misleading. ACH gives the airflow number context.

The engineering mistake: mixing up total supply air and outdoor air

A common design mistake is confusing total supply airflow with outdoor air.

For an operating room, the total supply airflow may be based around 20 ACH, while the outdoor air component may be a smaller part of that total. For example, if the same 4,800 ft³ operating room needs 4 ACH of outdoor air:

Outdoor Airflow = 4,800 × 4 / 60
Outdoor Airflow = 320 CFM

So the room may need:

Total supply airflow = 1,600 CFM
Outdoor air component = 320 CFM

Those are not the same number.

If an engineer treats all 1,600 CFM as outdoor air, the air handling unit may be oversized for heating, cooling, humidification, and dehumidification loads. The system becomes more expensive to operate and harder to control.

If an engineer goes the other way and only supplies 320 CFM because that is the outdoor air quantity, the room gets only:

ACH = 320 × 60 / 4,800
ACH = 4 ACH

That misses the total operating room ventilation intent completely.

This is one of the most important distinctions in OR HVAC design:

Outdoor air is a ventilation component.

Total supply air is the airflow used to achieve the room ACH, air distribution, pressure control, and thermal performance.

They overlap, but they are not interchangeable.

Room size changes everything

Another mistake is using a fixed CFM rule without checking room volume.

Two operating rooms can both be designed for 20 ACH but require very different airflow.

Example 1:

Room A = 20 ft × 20 ft × 10 ft
Volume = 4,000 ft³
Airflow at 20 ACH = 4,000 × 20 / 60
Airflow = 1,333 CFM

Example 2:

Room B = 30 ft × 24 ft × 10 ft
Volume = 7,200 ft³
Airflow at 20 ACH = 7,200 × 20 / 60
Airflow = 2,400 CFM

Both rooms are designed for the same ACH.

But Room B needs more than 1,000 CFM additional supply airflow because the volume is larger.

This matters for AHU capacity, duct sizing, diffuser selection, terminal unit sizing, sound control, and balancing.

A large OR at “standard” ACH can easily become a high-airflow room from a ductwork and equipment standpoint.

The formula does not design the air pattern

The ACH formula assumes the room is well mixed.

Real operating rooms are more sensitive than that.

The location and type of supply diffusers matter. The return or exhaust locations matter. Door openings matter. Ceiling obstructions matter. Surgical lights and equipment booms can disturb airflow. A room may technically have enough CFM but still perform poorly if the air pattern does not support the surgical zone.

That is why operating room airflow should not be checked only as a spreadsheet number.

A better early workflow is:

1. Calculate required total supply airflow from room volume and ACH.
2. Calculate outdoor air separately.
3. Check supply/exhaust or supply/return balance for positive pressure intent.
4. Confirm diffuser coverage over the surgical area.
5. Coordinate thermal load, humidity control, filtration, and controls.
6. Verify final requirements against the project’s healthcare standard and local authority.

The calculator helps with step one. The engineering review completes the rest.

Practical design takeaway

The ACH calculation is useful because it quickly exposes under-sizing.

If the room is 4,800 ft³ and the design only provides 900 CFM, the actual ACH is:

ACH = 900 × 60 / 4,800
ACH = 11.25 ACH

That is not a small rounding error. It is a different design basis.

But the same calculation can also expose over-simplification. If the room gets 1,600 CFM, the ACH may look right, but the design still needs pressure control, outdoor air verification, diffuser layout, filtration review, and commissioning.

The simple formula answers:

“How much supply airflow is required to hit the target ACH?”

It does not answer:

“Is this operating room fully compliant and ready for healthcare use?”

That difference is where many design mistakes happen.

Final thought

Operating room airflow is one of those engineering topics where the formula is easy, but the consequences of using it casually are serious.

The volume-based ACH calculation should be the starting point for the design conversation. It gives the engineer a fast, defensible airflow target. After that, the design still has to prove pressure relationship, outdoor air basis, clean-air distribution, filtration, temperature control, humidity control, and commissioning performance.

For quick preliminary sizing, use the Hospital Operating Room Airflow Calculator on CalcEngineer.

Return Air Ratio: The HVAC Shortcut That Can Hide an Outdoor Air Problem

Evgenii Konkin — Sun, 21 Jun 2026 17:50:15 +0000

Return air ratio looks like a simple HVAC percentage.

An air-handling unit supplies air to the building, some of that air returns from the occupied space, and some fresh outdoor air is added at the mixing box.

That sounds straightforward.

But this ratio can easily be misunderstood.

A system with 80% return air and 20% outdoor air may look normal. It may even look efficient.

But that does not automatically mean the space receives enough outdoor air for ventilation.

That is the main engineering trap:

Return air ratio is an airflow balance metric.
It is not a ventilation compliance calculation by itself.

A high return air ratio can reduce heating and cooling energy because more conditioned air is being reused.

But if the outdoor air volume is too low, the system may still fail the actual ventilation requirement.

The core calculation

The calculator uses a fixed air-mixing model.

The total supply airflow is split into two parts:

Total Supply Airflow = Return Airflow + Outdoor Airflow

If outdoor air is known, return air can be calculated as:

Q_return = Q_supply − Q_outdoor

If return air is known, outdoor air can be calculated as:

Q_outdoor = Q_supply − Q_return

Then return air ratio is:

RA_ratio = (Q_return / Q_supply) × 100

Outdoor air fraction is:

OA_fraction = (Q_outdoor / Q_supply) × 100

And the verification check is:

RA_ratio + OA_fraction = 100%
Q_return + Q_outdoor = Q_supply

This is simple arithmetic, but it gives a clear picture of what the air-handling unit is actually doing.

What the ratio really means

Return air ratio tells you how much of the supply airflow is recirculated indoor air.

Outdoor air fraction tells you how much of the supply airflow is fresh outdoor air.

For example:

80% return air = 20% outdoor air
70% return air = 30% outdoor air
0% return air = 100% outdoor air

Those pairs always move together.

If return air ratio goes up, outdoor air fraction goes down.

If outdoor air fraction goes up, return air ratio goes down.

That relationship affects both energy and indoor air quality.

Higher return air ratio usually means lower conditioning load because the system reuses air that is already close to indoor temperature.

Lower return air ratio usually means more fresh air, better dilution, and higher outdoor air conditioning load.

Example: AHU mixing box calculation

Suppose an air-handling unit has:

Total Supply Airflow = 8,000 CFM
Return Airflow = 6,400 CFM
Outdoor Airflow = unknown

Step 1: Calculate outdoor airflow.

Q_outdoor = Q_supply − Q_return
Q_outdoor = 8,000 − 6,400
Q_outdoor = 1,600 CFM

Step 2: Calculate return air ratio.

RA_ratio = (Q_return / Q_supply) × 100
RA_ratio = (6,400 / 8,000) × 100
RA_ratio = 80%

Step 3: Calculate outdoor air fraction.

OA_fraction = (Q_outdoor / Q_supply) × 100
OA_fraction = (1,600 / 8,000) × 100
OA_fraction = 20%

Step 4: Verify the airflow balance.

6,400 + 1,600 = 8,000 CFM
80% + 20% = 100%

So the AHU operating point is:

Return Air Ratio = 80%
Outdoor Air Fraction = 20%
Outdoor Air Volume = 1,600 CFM

At first glance, that looks like a typical commercial HVAC mixing condition.

But the calculation is not finished.

The engineer still needs to ask:

Is 1,600 CFM of outdoor air enough for the actual occupancy and floor area served?

That is the difference between an airflow ratio and a ventilation requirement.

The common engineering mistake

The most common mistake is assuming that a “normal” return air ratio automatically means the system meets ventilation requirements.

For example:

80% return air / 20% outdoor air

may be perfectly reasonable for one office zone.

But it may be too little outdoor air for a dense conference area.

It may be completely inappropriate for a laboratory, healthcare procedure room, kitchen exhaust makeup system, or other space where recirculation is restricted.

The ratio tells you the mix.

It does not tell you whether the outdoor air volume satisfies the code basis.

ASHRAE 62.1 ventilation checks are based on outdoor air volume, people, floor area, zone effectiveness, and system ventilation efficiency.

They are not based only on a generic outdoor air percentage.

Why percentage can be misleading

Outdoor air fraction is a percentage of total supply airflow.

But ventilation requirements are often based on absolute outdoor air volume.

That means the same outdoor air fraction can mean very different outdoor air quantities.

For example:

20% outdoor air of 8,000 CFM = 1,600 CFM
20% outdoor air of 3,000 CFM = 600 CFM

Same percentage.

Very different outdoor air volume.

This is why “20% outdoor air” is not a complete ventilation statement.

The actual outdoor air CFM must be checked against the required outdoor air CFM.

What happens when supply airflow changes?

Suppose the AHU originally supplies:

Q_supply = 8,000 CFM
OA_fraction = 20%

Then:

Q_outdoor = 8,000 × 0.20
Q_outdoor = 1,600 CFM

Now imagine the supply airflow is reduced during part-load operation:

Q_supply = 5,000 CFM
OA_fraction = 20%

If the outdoor air fraction stays at 20%:

Q_outdoor = 5,000 × 0.20
Q_outdoor = 1,000 CFM

The outdoor air volume dropped by:

1,600 − 1,000 = 600 CFM

The outdoor air fraction did not change.

But the actual outdoor air volume changed a lot.

That is why VAV systems, minimum damper positions, demand-controlled ventilation, and outdoor air measurement deserve careful review.

A fixed percentage may not guarantee the required outdoor air volume at all operating conditions.

Another mistake: confusing return air ratio with supply air effectiveness

A high return air ratio does not mean the supply air is delivered well.

It only means a high percentage of the supply airflow is recirculated return air.

It says nothing about:

Air distribution effectiveness
Short-circuiting
Diffuser performance
Dead zones
Ventilation effectiveness
Outdoor air delivery to the breathing zone

A system can have the correct return air ratio and still distribute air poorly.

That is why return air ratio is a mixing-box calculation, not a complete indoor air quality analysis.

Another mistake: applying standard ratios to special spaces

Standard commercial HVAC systems often use a high percentage of return air because it saves energy.

That can be fine for normal offices, retail, and similar occupancy types when outdoor air requirements are met.

But some spaces should not be treated like standard recirculating systems.

Examples include:

Laboratories
Commercial kitchens
Paint booths
Isolation rooms
Operating rooms
Certain healthcare procedure spaces
Hazardous exhaust applications

In those cases, return air may be prohibited or heavily restricted.

A normal-looking return air ratio can be the wrong design concept entirely.

Energy recovery becomes important at high outdoor air fractions

When outdoor air fraction increases, the HVAC system must condition more outdoor air.

That can increase:

Cooling load in hot climates
Heating load in cold climates
Dehumidification load in humid climates
Coil capacity
Plant load
Energy use
Operating cost

If the outdoor air fraction is high, energy recovery may become worth evaluating.

This is especially true for systems with large exhaust volumes, high outdoor air requirements, or long operating hours.

A low return air ratio may improve ventilation, but it can also create a major conditioning penalty.

The right design is not simply “more outdoor air” or “more return air.”

The right design balances ventilation, energy, code requirements, contaminant control, and operating mode.

Practical design checks

Before accepting a return air ratio result, ask:

1. Is the total supply airflow correct?
2. Is return airflow measured, balanced, or assumed?
3. Is outdoor airflow measured directly or only inferred?
4. Does outdoor air volume meet ASHRAE 62.1 requirements?
5. Does the ratio change during VAV part-load operation?
6. Does economizer mode change the outdoor air fraction?
7. Is the space allowed to use return air recirculation?
8. Is energy recovery needed for high outdoor air fractions?
9. Is the system being checked at design load, minimum ventilation, or another operating mode?

These checks matter because the ratio is only one operating point.

The real system may move through many operating states during the day.

Practical engineering takeaway

Return air ratio is calculated from a simple relationship:

RA_ratio = (Q_return / Q_supply) × 100

Outdoor air fraction is the complementary value:

OA_fraction = (Q_outdoor / Q_supply) × 100

And the airflow balance should always close:

Q_return + Q_outdoor = Q_supply
RA_ratio + OA_fraction = 100%

But the engineering interpretation is the important part.

A return air ratio can tell you how much air is being recirculated.

It cannot, by itself, prove that the occupied space receives enough outdoor air.

That requires checking the absolute outdoor air volume against the ventilation requirement for the actual occupancy, area, and system design.

For a quick first-pass review, you can use the Return Air Ratio Calculator.

It calculates return air ratio, outdoor air fraction, return air volume, and outdoor air volume from AHU supply, return, and outdoor airflow inputs, helping you verify the mixing-box balance before deeper ventilation compliance review.

School Classroom Ventilation: The Mistake of Counting Supply Air as Outdoor Air

Evgenii Konkin — Thu, 18 Jun 2026 16:28:46 +0000

Classroom ventilation is easy to overestimate.

A school may have a rooftop unit, supply diffusers, return grilles, and a balancing report showing hundreds of CFM delivered to the room.

At first glance, that can look acceptable.

But classroom ventilation is not checked from total supply airflow alone.

The important value is outdoor air.

That distinction matters because a large part of the supply air may be recirculated return air. Recirculated air can help with heating and cooling, but it does not count as new outdoor air for ASHRAE 62.1 ventilation compliance.

That is why the better question is not:

“How much supply air enters the classroom?”

The better question is:

“How much outdoor air reaches the classroom breathing zone?”

The core ventilation formula

The calculator uses the ASHRAE 62.1 Ventilation Rate Procedure for a single classroom zone.

The breathing zone outdoor airflow is calculated from two components:

Vbz = (Rp × Pz) + (Ra × Az)

Where:

Vbz = breathing zone outdoor airflow
Rp = people outdoor air rate
Pz = number of occupants
Ra = area outdoor air rate
Az = floor area

For classrooms in Imperial units, the calculator uses:

Rp = 10 cfm/person
Ra = 0.12 cfm/ft²

For Metric units:

Rp = 5 L/s per person
Ra = 0.6 L/s per m²

The reason for the two-part formula is simple.

People create bioeffluents, CO₂, odors, and airborne contaminant load. That is the people component.

The room itself also contributes pollutants from materials, furnishings, finishes, and general background sources. That is the area component.

A classroom needs both components.

Distribution effectiveness changes the required supply outdoor air

After calculating breathing zone outdoor airflow, the calculator applies zone air distribution effectiveness:

Voz = Vbz / Ez

Where:

Voz = zone outdoor airflow required
Ez = zone air distribution effectiveness

If Ez is 1.0, the system is assumed to deliver outdoor air effectively to the breathing zone.

If Ez is 0.8, the system receives less credit for the supplied outdoor air, so the required zone outdoor airflow increases.

That matters because some diffuser and return configurations do not distribute outdoor air as effectively during certain operating modes.

A classroom with the wrong Ez assumption can look compliant on paper while still needing more outdoor air.

The supporting checks

The calculator also reports air changes per hour:

ACH = (Supplied Outdoor Airflow CFM × 60) / Room Volume

For Imperial units:

Room Volume = Floor Area × Ceiling Height

It also checks ventilation adequacy:

Adequacy (%) = Supplied Outdoor Airflow / Voz × 100

And per-person ventilation rate:

Per-Person Rate = Supplied Outdoor Airflow / Pz

These outputs answer different questions.

Voz tells you the required outdoor air.

Adequacy tells you whether the supplied outdoor air meets that requirement.

ACH gives a room-volume perspective.

Per-person rate gives a useful occupancy-based sanity check.

Example: classroom ventilation check

Suppose a classroom has:

28 students + 1 teacher = 29 occupants
Floor Area = 850 ft²
Ceiling Height = 9 ft
Supplied Outdoor Airflow = 380 CFM
Ez = 1.0

Step 1: Calculate the people component.

People Component = Rp × Pz
People Component = 10 × 29
People Component = 290 CFM

Step 2: Calculate the area component.

Area Component = Ra × Az
Area Component = 0.12 × 850
Area Component = 102 CFM

Step 3: Calculate breathing zone outdoor airflow.

Vbz = 290 + 102
Vbz = 392 CFM

Step 4: Apply distribution effectiveness.

Voz = Vbz / Ez
Voz = 392 / 1.0
Voz = 392 CFM

So the classroom requires:

Required Outdoor Airflow = 392 CFM

Step 5: Compare supplied outdoor airflow with required outdoor airflow.

Supplied Outdoor Airflow = 380 CFM
Required Outdoor Airflow = 392 CFM

Now calculate adequacy:

Adequacy = 380 / 392 × 100
Adequacy = 96.9%

The classroom is slightly below the ASHRAE 62.1 minimum.

The shortfall is:

392 − 380 = 12 CFM

That may look small, but the result is still important:

Status = Insufficient ventilation

The room needs at least 392 CFM of outdoor air for the stated occupancy, area, and Ez condition.

ACH can tell a different story

Now calculate the room volume:

Room Volume = 850 × 9
Room Volume = 7,650 ft³

Calculate ACH:

ACH = (380 × 60) / 7,650
ACH = 22,800 / 7,650
ACH = 2.98 ACH

So the classroom has:

ACH ≈ 3.0

This is useful context.

The room is close to the code minimum outdoor airflow requirement, but the ACH value is not high. If the school is targeting enhanced dilution, improved IAQ, or post-pandemic ventilation performance, the design conversation may not stop at minimum ASHRAE 62.1 compliance.

This is the practical lesson:

Meeting minimum outdoor air and achieving higher dilution performance are related, but not identical.

Common engineering mistake: entering total supply airflow

The biggest mistake is using total supply airflow instead of outdoor airflow.

For example, a balancing report might show:

Total Supply Air = 900 CFM
Outdoor Air Portion = 380 CFM

If the engineer enters 900 CFM into the ventilation calculator, the result will look excellent.

But that would be wrong.

The ASHRAE 62.1 ventilation check is based on outdoor air delivered to the zone, not total mixed supply air.

The 900 CFM includes recirculated air.

Only the outdoor air portion counts for the ventilation requirement.

This mistake can create a false-compliant classroom.

The room may appear to exceed the requirement, while the actual outdoor air delivery is still below the minimum.

Another mistake: forgetting the teacher

A classroom count should include all occupants in the breathing zone.

That means the teacher counts too.

A room with 28 students is not 28 occupants if the teacher is present.

It is:

28 students + 1 teacher = 29 occupants

That one person changes the people component by:

10 cfm/person × 1 person = 10 CFM

In a single classroom, that may look minor.

Across a school with dozens of classrooms, repeated undercounting can create a real ventilation gap.

Another mistake: ignoring Ez

Zone air distribution effectiveness can change the required outdoor airflow.

If the same classroom has:

Vbz = 392 CFM
Ez = 0.8

Then:

Voz = 392 / 0.8
Voz = 490 CFM

That is a large change.

The required zone outdoor airflow increases from:

392 CFM to 490 CFM

The increase is:

490 − 392 = 98 CFM

That is a 25% increase.

So if the actual air distribution condition requires Ez = 0.8, using Ez = 1.0 will understate the outdoor air requirement.

This is why diffuser and return configuration should not be treated as a minor detail.

Another mistake: treating ACH as the only compliance metric

ACH is useful, but it is not the same thing as the ASHRAE 62.1 classroom ventilation calculation.

A small classroom and a large classroom can have the same ACH but different occupant-based ventilation adequacy.

A low-occupancy room and a crowded classroom can also have the same ACH but very different per-person outdoor air rates.

That is why the calculator reports several outputs:

Vbz
Voz
Adequacy
Per-person rate
ACH

Each output tells part of the story.

The required outdoor air calculation protects the code basis.

ACH gives a dilution and room-volume perspective.

Per-person rate helps check whether the result makes sense for the occupant load.

Practical design checks

Before accepting a classroom ventilation result, ask:

1. Did we enter outdoor airflow only, not total supply airflow?
2. Does the occupant count include the teacher?
3. Is the floor area the correct net classroom area?
4. Is the selected Ez value consistent with the diffuser and return configuration?
5. Is the supplied outdoor airflow measured, balanced, or assumed?
6. Does the classroom meet the required Voz?
7. Is ACH acceptable for the project’s IAQ target?
8. Is this a single-zone check or part of a multi-zone AHU calculation?

These checks matter because classroom ventilation errors are often input errors, not formula errors.

The equation is straightforward.

The interpretation is where mistakes happen.

Practical engineering takeaway

Classroom ventilation should be checked from outdoor air delivered to the zone, not total supply air.

The main calculation is:

Vbz = (Rp × Pz) + (Ra × Az)

Then:

Voz = Vbz / Ez

And the delivered outdoor airflow should be compared against Voz:

Adequacy (%) = Supplied Outdoor Airflow / Voz × 100

If the result is below 100%, the classroom does not meet the minimum outdoor air requirement for the stated inputs.

The most common mistake is simple:

Total supply air is not the same as outdoor air.

For a quick first-pass check, you can use the School Classroom Ventilation Calculator.

It calculates classroom breathing zone outdoor airflow, required zone outdoor airflow, ventilation adequacy, per-person ventilation rate, and ACH using occupant count, floor area, ceiling height, supplied outdoor airflow, and air distribution effectiveness.

Adaptive Comfort Model: When Indoor Temperature Should Follow Outdoor Climate

Evgenii Konkin — Tue, 16 Jun 2026 15:29:42 +0000

Adaptive comfort is easy to misuse.

Many HVAC discussions treat indoor comfort as a fixed temperature target.

For example:

Set the room to 22°C.
Keep the office at 24°C.
Anything above 26°C is too warm.

That kind of thinking may be reasonable for fully air-conditioned buildings.

But it does not always fit naturally ventilated buildings.

In naturally ventilated spaces, occupants can adapt. They may open windows, use fans, adjust clothing, shift expectations, or accept a wider indoor temperature range when the outdoor climate is warmer.

That is the idea behind the adaptive comfort model.

It does not ask only:

“What is the indoor temperature?”

It asks a better question:

“Given the recent outdoor climate, is this indoor temperature still acceptable for occupants in a naturally ventilated building?”

The core formula

The calculator uses the ASHRAE 55 adaptive comfort relationship:

T_comf = 0.31 × T_outdoor + 17.8

Where:

T_comf = neutral indoor comfort temperature, °C
T_outdoor = running mean outdoor temperature, °C

The running mean outdoor temperature is not just a random outdoor temperature from one hour.

It is a recent outdoor temperature average, typically representing the thermal history that occupants have adapted to.

Once the neutral comfort temperature is calculated, the acceptable comfort band is applied.

For 80% acceptability:

Comfort Range = T_comf ± 3.5°C

For 90% acceptability:

Comfort Range = T_comf ± 2.5°C

The 80% band is wider.

The 90% band is stricter.

That means a space may pass the 80% comfort check but fail the 90% comfort check.

Why this matters

Adaptive comfort is useful because it recognizes that comfort is not always fixed.

For example, an indoor temperature of 27°C may feel too warm in a fully air-conditioned office where occupants expect tight temperature control.

But in a naturally ventilated building during warm weather, 27°C may be acceptable if occupants are adapted to the outdoor climate and have some control over their environment.

That does not mean “hot rooms are always fine.”

It means the comfort limit moves with the running mean outdoor temperature.

The model is especially useful for:

Naturally ventilated buildings
Mixed-mode buildings
Passive design studies
Low-energy cooling strategies
Tropical and temperate climates
Early comfort screening
Post-occupancy comfort checks

But it must be applied to the right building type.

Example: naturally ventilated office comfort check

Suppose a naturally ventilated office has:

Running Mean Outdoor Temperature = 28°C
Actual Indoor Temperature = 27°C
Acceptability Level = 80%

Step 1: Calculate the neutral comfort temperature.

T_comf = 0.31 × T_outdoor + 17.8
T_comf = 0.31 × 28 + 17.8
T_comf = 8.68 + 17.8
T_comf = 26.48°C

So the neutral comfort temperature is:

T_comf = 26.48°C

Step 2: Calculate the 80% comfort range.

For 80% acceptability:

Comfort Range = T_comf ± 3.5°C

Lower limit:

Lower Limit = 26.48 − 3.5
Lower Limit = 22.98°C

Upper limit:

Upper Limit = 26.48 + 3.5
Upper Limit = 29.98°C

So the acceptable range is:

22.98°C to 29.98°C

Step 3: Check the actual indoor temperature.

Actual Indoor Temperature = 27°C

Since 27°C is between 22.98°C and 29.98°C, the space is within the 80% adaptive comfort range.

Step 4: Calculate distance from the nearest comfort boundary.

Distance = min(T_indoor − Lower Limit, Upper Limit − T_indoor)

Substitute the values:

Distance = min(27 − 22.98, 29.98 − 27)
Distance = min(4.02, 2.98)
Distance = +2.98°C

The positive result means the indoor temperature is inside the comfort range.

So the result is:

Indoor temperature = 27°C
Status = Within 80% adaptive comfort range
Distance from nearest boundary = +2.98°C

That is a useful result because it prevents an overly rigid comfort judgment.

A fixed-temperature mindset might say:

27°C is too warm.

But the adaptive comfort model says:

For this running mean outdoor temperature and this building type, 27°C can still be acceptable.

What happens with 90% acceptability?

Now keep the same temperatures, but use the stricter 90% acceptability band.

Inputs:

Running Mean Outdoor Temperature = 28°C
Actual Indoor Temperature = 27°C
Acceptability Level = 90%

The neutral comfort temperature is unchanged:

T_comf = 26.48°C

For 90% acceptability:

Comfort Range = T_comf ± 2.5°C

Lower limit:

Lower Limit = 26.48 − 2.5
Lower Limit = 23.98°C

Upper limit:

Upper Limit = 26.48 + 2.5
Upper Limit = 28.98°C

The 90% comfort range is:

23.98°C to 28.98°C

The indoor temperature is still:

27°C

So it remains inside the range.

Distance from the nearest boundary:

Distance = min(27 − 23.98, 28.98 − 27)
Distance = min(3.02, 1.98)
Distance = +1.98°C

The space still passes, but the margin is smaller.

That is the practical difference between 80% and 90% acceptability.

The 90% band is not just a label.

It narrows the acceptable temperature range and leaves less room for drift.

What happens if indoor temperature rises?

Now suppose the indoor temperature rises to 30°C while the running mean outdoor temperature stays at 28°C.

Inputs:

Running Mean Outdoor Temperature = 28°C
Actual Indoor Temperature = 30°C
Acceptability Level = 80%

Neutral comfort temperature:

T_comf = 26.48°C

80% comfort range:

Lower Limit = 22.98°C
Upper Limit = 29.98°C

Now compare:

Actual Indoor Temperature = 30°C
Upper Limit = 29.98°C

The indoor temperature is slightly above the upper comfort limit.

Distance from comfort range:

Distance = Upper Limit − T_indoor
Distance = 29.98 − 30
Distance = -0.02°C

The negative result means the indoor condition is outside the acceptable range.

This is where the model becomes useful for design decisions.

The result is not simply “30°C is always bad” or “30°C is always acceptable.”

The result is:

At this running mean outdoor temperature and acceptability level, 30°C is just outside the 80% adaptive comfort range.

That is a much more precise engineering statement.

Common engineering mistake: using adaptive comfort for fully air-conditioned spaces

The biggest mistake is applying the adaptive comfort model to the wrong building type.

The adaptive model is intended for naturally ventilated buildings where occupants can interact with the environment.

That usually means people can do things like:

Open windows
Use fans
Adjust clothing
Change local airflow
Experience outdoor climate variation
Adapt expectations to seasonal conditions

A sealed, fully air-conditioned office is different.

If occupants have limited control and expect mechanical cooling to maintain a narrow setpoint, adaptive comfort may not be the right model.

For mechanically cooled spaces, PMV/PPD methods or standard HVAC comfort criteria may be more appropriate.

The formula may still produce a number.

But the engineering interpretation can be wrong.

Another mistake: using one outdoor temperature instead of running mean temperature

The input is running mean outdoor temperature.

That matters.

Using a single afternoon peak temperature can distort the result.

Using one random weather value can also distort the result.

The adaptive model is based on recent outdoor thermal history, not one isolated moment.

For example:

Outdoor peak today = 34°C
Running mean outdoor temperature = 28°C

Those are not the same input.

If the engineer uses the peak temperature instead of the running mean value, the comfort temperature shifts too high and the acceptable band may look more permissive than it should.

That can lead to an overly optimistic comfort conclusion.

Another mistake: ignoring humidity and air movement

The basic adaptive comfort formula does not directly include humidity.

It also does not fully describe local air movement, radiant asymmetry, solar exposure, or internal heat gains.

That matters in real buildings.

A room can be inside the calculated adaptive temperature range and still feel uncomfortable if:

Humidity is very high
Air movement is too low
Direct sun hits occupants
Radiant surfaces are hot
Internal equipment gains are high
People cannot open windows
Fans are not available
The space is densely occupied

Adaptive comfort is a powerful screening method.

It is not a full replacement for detailed thermal comfort analysis.

Practical design checks

Before accepting an adaptive comfort result, ask:

1. Is the building naturally ventilated or mixed-mode?
2. Do occupants have real control over windows, fans, or airflow?
3. Is the outdoor input a running mean temperature, not a single peak value?
4. Is the selected acceptability level 80% or 90%?
5. Is the indoor temperature an operative comfort temperature, not only a random air sensor value?
6. Is humidity high enough to create discomfort even inside the temperature band?
7. Are solar exposure and radiant temperature effects important?
8. Is the outdoor running mean temperature within the valid model range?

These checks matter because the adaptive comfort model is simple, but the application is not automatic.

Practical engineering takeaway

The adaptive comfort model changes the way engineers think about comfort.

Instead of treating indoor temperature as one fixed target, it links acceptable indoor temperature to recent outdoor climate.

The main formula is:

T_comf = 0.31 × T_outdoor + 17.8

Then the acceptable range is applied:

80% acceptability: T_comf ± 3.5°C
90% acceptability: T_comf ± 2.5°C

The result helps answer a practical design question:

Can this naturally ventilated space be considered thermally acceptable under these outdoor conditions?

Used correctly, the model can support passive design, mixed-mode operation, wider comfort bands, and lower mechanical cooling energy.

Used incorrectly, it can justify uncomfortable spaces with the wrong formula.

The key is to apply it only where the assumptions match the building.

For a quick first-pass check, you can use the Adaptive Comfort Model Calculator.

It calculates neutral comfort temperature, acceptable comfort range, and distance from the comfort boundary based on running mean outdoor temperature, actual indoor temperature, and the selected ASHRAE 55 acceptability level.