DEV Community: Evgenii Konkin

Harmonic Filter Design: The LC Tuning Mistake That Can Miss the Target Harmonic

Evgenii Konkin — Sun, 09 Aug 2026 07:41:12 +0000

A passive harmonic filter can look correct on a drawing.
A reactor.
A capacitor.
A target harmonic order.

But if the LC values tune the filter to the wrong frequency, the filter may not absorb the harmonic it was supposed to handle.

That is the basic risk in harmonic filter design.

The key formula is:

f_t = 1 / (2π√(LC))

Where:

f_t = filter tuning frequency, Hz
L = reactor inductance, H
C = capacitor capacitance, F

Once the tuning frequency is known, the actual tuning harmonic order is:

h_t = f_t / f_1

And the target harmonic frequency is:

f_target = h_target × f_1

For a 60 Hz system, the 5th harmonic is:

f_target = 5 × 60
f_target = 300 Hz

So if a filter intended for the 5th harmonic is actually tuned near 300 Hz, it is in the right region.

If it is tuned far away from 300 Hz, the filter may miss the problem.

Example: a filter intended for the 5th harmonic

Suppose a passive LC filter has:

L = 5 mH
C = 100 µF
f_1 = 60 Hz
Target harmonic = 5th

Convert units first:

L = 5 / 1000 = 0.005 H
C = 100 / 1,000,000 = 0.0001 F

Calculate the LC product:

LC = 0.005 × 0.0001
LC = 0.0000005

Now calculate tuning frequency:

f_t = 1 / (2π√0.0000005)
f_t ≈ 225 Hz

The actual tuning harmonic order is:

h_t = 225 / 60
h_t ≈ 3.75

But the target 5th harmonic frequency is:

f_target = 5 × 60
f_target = 300 Hz

Now calculate detuning:

Detuning = [(225 − 300) / 300] × 100
Detuning ≈ −25%

The filter is tuned about 25% below the target harmonic.

That is not a small difference.

It means the selected reactor and capacitor values do not align well with the intended 5th harmonic target.

The common mistake: checking the THD but not the tuning

A common workflow mistake is:

Measure high THD.
Decide the 5th harmonic is the problem.
Install a passive filter.
Assume the filter is tuned correctly.

But the LC values decide where the filter actually resonates.

The label or design intent does not tune the filter.

The formula does.

A filter intended for the 5th harmonic can end up closer to the 4th harmonic region if the inductance or capacitance is wrong.

That can reduce attenuation at the intended harmonic and may create unwanted interaction with the system impedance.

Detuning is not always bad

A filter does not always need to be tuned exactly on the target harmonic.

In many capacitor bank applications, engineers intentionally tune slightly below the lowest dominant harmonic. For example, a system with strong 5th harmonic distortion may use a detuned filter around order 4.7 or 4.85.

That is done to reduce resonance risk.

So the issue is not simply:

Exact target = good
Any detuning = bad

The better question is:

Is the detuning intentional and within the expected design range?

A small negative detuning may be part of a good design.

A large accidental detuning is a warning sign.

Unit mistakes can destroy the result

The LC formula is very sensitive to units.

The calculator expects:

Inductance in mH
Capacitance in µF

and converts them internally to:

H
F

If someone enters 5 mH as 5 H, the tuning frequency shifts by a huge amount.

If someone enters 100 µF as 100 F, the result becomes unrealistic.

This is why the first review step should always be:

Check mH vs H.
Check µF vs F.
Check 50 Hz vs 60 Hz.
Check the target harmonic order.

The formula itself is simple.

The input units are where the mistake often starts.

What the calculation does not prove

A correct tuning frequency does not mean the harmonic filter is fully designed.

It does not verify:

Harmonic attenuation
System impedance scan
Quality factor
Damping
Capacitor kvar
Reactor current rating
Thermal loading
Switching transients
Resonance amplification
IEEE 519 or utility compliance

The LC tuning check is a first-pass design screen.

It tells you whether the filter is aimed at the right frequency region.

It does not prove that the full power quality problem is solved.

Practical takeaway

Use the harmonic filter tuning calculation early.

It helps answer:

What frequency is this LC filter actually tuned to?
Which harmonic order does that represent?
How far is it from the intended target harmonic?
Is the detuning intentional or accidental?
Are the L and C units entered correctly?

That is enough to catch many obvious filter selection mistakes before going deeper into manufacturer review, harmonic measurements, or a full impedance study.

Final thought

A passive harmonic filter is not tuned by its name.

It is tuned by its inductance and capacitance.

f_t = 1 / (2π√(LC))

If the LC values put the filter near the intended harmonic, the design may be moving in the right direction.

If the tuning point is far from the target harmonic, the filter may miss the harmonic it was meant to absorb.

For quick checks of passive single-tuned LC harmonic filters, tuning harmonic order, target frequency, and detuning percentage, use the Harmonic Filter Design Calculator on CalcEngineer.

Ground Ring Resistance: The Soil Resistivity Mistake That Can Break a Grounding Estimate

Evgenii Konkin — Wed, 05 Aug 2026 17:05:05 +0000

Ground ring resistance looks like a simple grounding calculation.

Enter soil resistivity.

Enter ring size.

Enter conductor diameter.

Get resistance in ohms.

But the result can change dramatically depending on one input: soil resistivity.

That is where many early grounding estimates become misleading.

A ground ring may look acceptable in moist clay and completely weak in dry sand or rocky soil. The same conductor. The same ring diameter. Very different resistance to earth.

The formula is useful, but it should not be treated as a final grounding design.

It is a screening tool.

The basic ground ring resistance formula

For a single circular buried ground ring, the resistance can be estimated with:

R = (ρ / 2π²r) × ln(8r / d)

Where:

R = ground ring resistance, ohms
ρ = soil resistivity, Ω·m
r = ring radius, m
d = conductor diameter, m
ln = natural logarithm

The formula shows three important things:

Higher soil resistivity increases resistance.
Larger ring radius reduces resistance.
Larger conductor diameter reduces resistance, but usually less strongly.

That last point matters.

Increasing conductor diameter may help a little, but increasing the ring radius or improving the grounding system layout usually has a much stronger effect.

Soil resistivity is the big driver

Soil resistivity is often the most important input.

For example, use:

ρ = 100 Ω·m
r = 5 m
d = 20 mm = 0.020 m

First check the geometry ratio:

8r / d = (8 × 5) / 0.020
8r / d = 2000

Natural logarithm:

ln(2000) ≈ 7.601

Denominator:

2π²r = 2 × π² × 5
2π²r ≈ 98.696

Now calculate resistance:

R = (100 / 98.696) × 7.601
R ≈ 7.70 Ω

So the estimated ground ring resistance is:

R ≈ 7.70 Ω

That may be a moderate result depending on the project target.

Now keep the same ring geometry, but change only the soil resistivity.

If the site has dry or rocky soil:

ρ = 500 Ω·m

Because soil resistivity is directly proportional in the formula, the result becomes about five times higher:

R ≈ 38.5 Ω

Same ring.

Same conductor.

Different soil.

Very different grounding performance.

That is the engineering lesson.

The soil assumption can dominate the result.

The common mistake: using a guessed soil resistivity

A frequent early-design mistake is entering a generic soil resistivity value without checking whether it matches the actual site.

For example:

Assumed soil resistivity = 100 Ω·m
Actual dry-site resistivity = 500 Ω·m

The estimate can look five times better than reality.

That is not a small rounding error.

It can change the design decision completely.

With 100 Ω·m, the engineer may think:

The ring is probably acceptable.

With 500 Ω·m, the better conclusion may be:

The ring alone is likely not enough.
Additional grounding measures or field testing are needed.

So the ground ring calculation should always be tied to a realistic soil resistivity assumption.

For serious projects, that means soil resistivity testing, not just a table value.

Ring radius matters more than conductor diameter

The formula also explains why ring size is powerful.

Resistance is divided by ring radius:

R ∝ 1 / r

There is also a logarithmic term, but the main practical effect is clear: a larger ring usually reduces resistance significantly.

Using the same soil:

ρ = 100 Ω·m
d = 20 mm

At:

r = 5 m

we got:

R ≈ 7.70 Ω

Now increase the radius:

r = 10 m

The new estimate is about:

R ≈ 4.20 Ω

That is a major improvement.

Now compare that with increasing conductor diameter instead.

Keep:

ρ = 100 Ω·m
r = 5 m

Double the conductor diameter:

d = 20 mm → 40 mm

The result improves only from about:

7.70 Ω → 7.00 Ω

That helps, but it is not nearly as strong as increasing the ring radius.

Why?

Because conductor diameter is inside the logarithm:

ln(8r / d)

Logarithmic changes are slower.

This is why oversizing the grounding conductor is not always the most effective way to reduce earth resistance.

The layout and soil conditions usually matter more.

Radius vs diameter: a simple input mistake

Another common mistake is entering ring diameter as ring radius.

Suppose the actual ground ring has:

Diameter = 10 m
Radius = 5 m

The calculator asks for radius.

If the engineer accidentally enters:

r = 10 m

instead of:

r = 5 m

the result may look much better than reality.

With:

ρ = 100 Ω·m
d = 20 mm

Correct radius:

r = 5 m
R ≈ 7.70 Ω

Wrong radius input:

r = 10 m
R ≈ 4.20 Ω

That mistake can make the grounding design look stronger than it really is.

The physical ring did not improve.

Only the input was wrong.

This is why radius and diameter should be checked carefully before trusting the result.

What the calculation does not prove

A single-ring resistance estimate is not a complete grounding design.

It does not model:

Layered soil
Multiple ground rods
Grounding grids or meshes
Ring-plus-rod systems
Touch voltage
Step voltage
Ground potential rise
Fault current distribution
Lightning impulse behavior
Seasonal soil moisture changes
Corrosion or clamp resistance
Field measurement results

That limitation matters.

A low resistance value is useful, but grounding safety is not only about one ohm number.

For substations, towers, generator pads, transformer pads, solar inverters, battery systems, and industrial equipment, the grounding review may need much more than single-ring resistance.

A good result from the formula means:

This single-ring estimate looks promising.

It does not mean:

The grounding system is fully verified.

Practical design takeaway

Use ground ring resistance as an early screening calculation.

It helps answer:

Is the ring likely to be low, moderate, or high resistance?
How sensitive is the result to soil resistivity?
Would a larger ring radius help?
Is conductor diameter actually the best place to improve the design?
Does the project need supplemental electrodes or a grounding grid?
Is field testing required before final acceptance?

But do not use it as the final safety decision.

The most important engineering habit is to separate:

Preliminary calculation
Final grounding design
Field verification

The formula gives the first part.

It does not replace the other two.

Final thought

Ground ring resistance is not just a conductor-size calculation.

It is mainly a soil and geometry problem.

A ring in low-resistivity soil can perform very differently from the same ring in dry, rocky ground.

Increasing the conductor diameter may help a little, but increasing the ring radius or changing the grounding layout often has a much stronger effect.

And if the wrong soil resistivity or wrong radius is entered, the result can look much better than the real installation.

For quick preliminary checks of a single circular buried ground ring using soil resistivity, ring radius, and conductor diameter, use the Ground Ring Resistance Calculator on CalcEngineer.

Residential Electrical Load: The Service Size Mistake Hidden Behind Square Feet

Evgenii Konkin — Thu, 30 Jul 2026 04:59:23 +0000

A house does not need a 200 A service just because it is “modern.”

And it does not automatically fit on an old 100 A service just because the lights turn on today.

Residential service sizing is more structured than that.

The calculation starts with floor area, small-appliance circuits, laundry, major appliances, HVAC, motors, and EV charging. Then demand factors are applied. The result is a calculated load in volt-amperes and amperes.

The mistake is thinking this is just a square-footage problem.

It is not.

A 2,000 ft² house with gas appliances can look very different from a 2,000 ft² all-electric house with electric heat, electric range, dryer, and EV charger.

Same floor area.

Very different service calculation.

The basic idea

A dwelling electrical load calculation tries to answer a practical question:

How much load should the service be sized for?

The result is usually expressed as:

Total Calculated Load = VA
Service Current = Total VA / Service Voltage
Recommended Service Size = next standard ampere rating above the sizing basis

For a typical one-family dwelling service:

Service Voltage = 240 V

So:

Service Current = Total Calculated Load / 240

If the calculated load is:

Total Calculated Load = 33,125 VA

Then:

Service Current = 33,125 / 240
Service Current ≈ 138 A

The next common standard service size above that is:

150 A

That is the engineering logic.

But getting the 33,125 VA number is where most of the work happens.

General dwelling loads

The general load starts with floor area.

For the standard dwelling calculation:

Lighting VA = Floor Area × 3 VA/ft²

Then add required small-appliance and laundry circuits:

Small-appliance circuits = 1,500 VA each
Laundry circuits = 1,500 VA each

A common baseline for a dwelling is:

2 small-appliance circuits
1 laundry circuit

That adds:

2 × 1,500 + 1 × 1,500 = 4,500 VA

before the range, dryer, HVAC, fixed appliances, EV charger, or motor loads are even considered.

This is one of the first mistakes engineers and homeowners make.

They start with the visible appliances, but forget that the base dwelling calculation already includes required circuit allowances.

Example: base general load

Suppose a dwelling has:

Floor area = 2,000 ft²
Small-appliance circuits = 2
Laundry circuits = 1

Calculate lighting load:

Lighting VA = 2,000 × 3
Lighting VA = 6,000 VA

Add the small-appliance and laundry circuits:

General VA = 6,000 + 2 × 1,500 + 1 × 1,500
General VA = 6,000 + 3,000 + 1,500
General VA = 10,500 VA

So the general load starts at:

10,500 VA

That is before adding the range, dryer, fixed appliances, HVAC, largest motor, or EV charger.

This is why “just use watts per square foot” is not enough.

The service calculation is not only an area calculation.

Standard method demand factors

Under the standard method, general lighting demand is not simply carried at 100% forever.

A simplified demand structure is:

First 3,000 VA @ 100%
Next portion up to 120,000 VA @ 35%
Remainder @ 25%

For the 10,500 VA general load example:

First 3,000 VA = 3,000 VA
Remainder = 10,500 − 3,000
Remainder = 7,500 VA

Apply 35% to the remainder:

Remainder demand = 7,500 × 0.35
Remainder demand = 2,625 VA

Total lighting/general demand:

General demand = 3,000 + 2,625
General demand = 5,625 VA

Notice what happened.

The connected general load was:

10,500 VA

But the demand load became:

5,625 VA

That is why residential service calculations cannot be done by simply adding all nameplate loads at 100%.

Demand factors matter.

Range demand is not always nameplate

A household range is another common source of mistakes.

For one household range:

Range ≤ 12 kW → 8,000 VA demand

So a 12 kW range does not necessarily enter the standard method as 12,000 VA.

It can enter as:

8,000 VA

That surprises people.

The nameplate is not always the demand value.

For a range above 12 kW, an adjustment is applied, but the key engineering habit is simple:

Do not blindly enter range nameplate as full demand under the standard method.

Use the correct demand rule for the calculation method and range size.

Dryer load has a floor

A household dryer is also easy to misread.

The dryer demand is:

Dryer demand = max(5,000 VA, nameplate VA)

So if a dryer nameplate is:

4,500 VA

the demand value is not 4,500 VA.

It is:

5,000 VA

The mistake is entering the smaller nameplate and thinking the load calculation is finished.

The floor matters.

Fixed appliances and the 75% factor

Other fixed appliances may include:

Water heater
Dishwasher
Garbage disposal
Built-in microwave
Trash compactor
Other fastened-in-place appliances

For the standard method, when there are four or more fixed appliances, a 75% demand factor may apply:

Fixed appliance demand = Total fixed appliance VA × 0.75

But this factor does not apply to everything.

A common mistake is applying the 75% factor to the range, dryer, HVAC, or EV charger.

That is not the same category.

The fixed-appliance count matters.

The load type matters.

A clean load calculation needs categories, not just a pile of wattage numbers.

Heating and cooling are usually non-coincident

For many dwellings, heating and cooling are not expected to operate at full load at the same time.

So the calculation uses the larger of the two:

HVAC demand = max(air-conditioning load, electric heat load)

Example:

Air-conditioning load = 5,000 VA
Electric heat load = 10,000 VA

Then:

HVAC demand = max(5,000, 10,000)
HVAC demand = 10,000 VA

The mistake is adding both:

5,000 + 10,000 = 15,000 VA

That can oversize the service calculation.

But the opposite mistake is also possible: ignoring electric heat because “the house has AC.”

If electric heat is installed, it must be evaluated.

The calculation must reflect the actual equipment.

Largest motor allowance

Motor starting and motor load behavior can affect service calculations.

Under the standard method, the largest motor receives an additional allowance:

Largest motor adder = 25% × largest motor VA

If the largest motor is:

Largest motor = 1,800 VA

Then:

Motor adder = 0.25 × 1,800
Motor adder = 450 VA

This is not usually the biggest number in a dwelling service calculation, but it is still part of the method.

Small omissions add up.

EV charger load can change the answer

EV charging is where many existing residential services get stressed.

An EV charger is often treated as a continuous load:

EVSE demand = EV charger nameplate VA × 1.25

If the charger is:

EV charger = 9,600 VA

Then:

EVSE demand = 9,600 × 1.25
EVSE demand = 12,000 VA

At 240 V, that is:

12,000 / 240 = 50 A

So a 9.6 kW charger can add 50 A to the service calculation.

This is why a house that looked comfortable before EV charging may suddenly land near the next service size.

The charger nameplate is not the final load in the calculation.

The 125% factor can matter.

Worked example: 2,000 ft² all-electric house

Suppose a dwelling has:

Floor area = 2,000 ft²
Small-appliance circuits = 2
Laundry circuits = 1
Range = 12 kW
Dryer = 5 kW
Fixed appliances = 4,500 VA
Air-conditioning = 5,000 VA
Electric heat = 10,000 VA
Service voltage = 240 V

Start with general load:

Lighting VA = 2,000 × 3
Lighting VA = 6,000 VA

Add small-appliance and laundry circuits:

General VA = 6,000 + 3,000 + 1,500
General VA = 10,500 VA

Apply standard-method lighting demand:

First 3,000 VA @ 100% = 3,000 VA
Remainder = 10,500 − 3,000 = 7,500 VA
Remainder @ 35% = 7,500 × 0.35 = 2,625 VA

So:

General demand = 3,000 + 2,625
General demand = 5,625 VA

Range demand:

12 kW range → 8,000 VA

Dryer demand:

Dryer demand = max(5,000 VA, nameplate)
Dryer demand = 5,000 VA

Fixed appliances:

Fixed appliances = 4,500 VA

HVAC demand:

HVAC demand = max(5,000, 10,000)
HVAC demand = 10,000 VA

Total calculated load:

Total VA = 5,625 + 8,000 + 5,000 + 4,500 + 10,000
Total VA = 33,125 VA

Convert to service current:

Service Current = 33,125 / 240
Service Current ≈ 138 A

Recommended service size:

150 A

That result is not based only on the 2,000 ft² floor area.

It is based on the actual load categories and demand rules.

What happens when an EV charger is added?

Now add:

EV charger nameplate = 9,600 VA

Apply the 125% factor:

EVSE demand = 9,600 × 1.25
EVSE demand = 12,000 VA

New total load:

New Total VA = 33,125 + 12,000
New Total VA = 45,125 VA

New service current:

Service Current = 45,125 / 240
Service Current ≈ 188 A

Recommended service size:

200 A

That is a major change.

The same house moved from a 150 A result to a 200 A result because of one EV charger.

This is the kind of thing that gets missed when someone says:

The house is only 2,000 square feet.

Square footage alone does not describe the service load.

Optional method can produce a lower result

The optional method groups many general loads together and applies a different demand structure:

General demand = first 10,000 VA @ 100% + remainder @ 40%

Then HVAC is handled separately.

For the same general group:

General total = 32,000 VA

The optional-method demand is:

First 10,000 VA = 10,000 VA
Remainder = 32,000 − 10,000
Remainder = 22,000 VA
Remainder @ 40% = 22,000 × 0.40
Remainder @ 40% = 8,800 VA

So:

General demand = 10,000 + 8,800
General demand = 18,800 VA

If the HVAC demand comes out as:

HVAC demand = 6,500 VA

Then:

Total VA = 18,800 + 6,500
Total VA = 25,300 VA

Service current:

Service Current = 25,300 / 240
Service Current ≈ 105 A

Recommended service size:

110 A

The same house can produce different calculated loads under the standard and optional methods.

That does not mean one calculation is “fake.”

It means the method matters.

But the optional method is not a casual shortcut. Eligibility must be checked before using it as the sizing basis.

The 100 A minimum trap

Another common mistake is screening old services against the raw calculated amperes only.

For a one-family dwelling, the sizing basis cannot drop below the minimum service requirement.

So the sizing basis is:

Sizing basis = max(calculated amps, 100 A)

Example:

Calculated load = 19,200 VA
Service voltage = 240 V

Calculated amps:

Calculated amps = 19,200 / 240
Calculated amps = 80 A

Existing service:

Existing service = 90 A

If you compare only against the calculated load:

90 A > 80 A

It looks adequate.

But the sizing basis is:

Sizing basis = max(80, 100)
Sizing basis = 100 A

Now compare:

90 A < 100 A

So the service is undersized against the minimum sizing basis.

The mistake is saying:

The calculated load is only 80 A, so a 90 A service is fine.

The correct interpretation is:

The load calculation is 80 A, but the dwelling service minimum governs at 100 A.

That distinction matters during old-service reviews.

Do not use 120 V for total service current

A typical one-family dwelling service is 120/240 V single-phase.

The total calculated VA is divided by 240 V, not 120 V.

If someone uses 120 V by mistake, the calculated service current doubles.

Example:

Total load = 33,125 VA

Correct calculation:

Service Current = 33,125 / 240
Service Current ≈ 138 A

Wrong calculation:

Service Current = 33,125 / 120
Service Current ≈ 276 A

That mistake could turn a 150 A result into something that looks like it needs a 300 A service.

The arithmetic is simple.

The voltage basis matters.

The calculation does not size everything

A dwelling load calculator gives the service load and service-size screening.

It does not complete the entire electrical design.

It does not automatically size:

Service-entrance conductors
Main breaker
Neutral conductor
Grounding electrode conductor
Panelboard layout
Feeder taps
Voltage drop
Short-circuit rating
Load-management controls
Local amendment requirements

Those are separate checks.

This is important because people often treat one calculator result as a complete permit package.

It is not.

A service load result is a major input.

It is not the whole design.

Practical design takeaway

A good residential load calculation should answer:

What method was used?
What NEC edition or reference basis was assumed?
What floor area was included?
Were small-appliance and laundry circuits included?
Was the range treated by demand rules?
Was the dryer floor applied?
Were fixed appliances counted correctly?
Was only the larger of heating and cooling included?
Was the largest motor adder included?
Was the EV charger counted at 125%?
Was the result divided by 240 V, not 120 V?
Did the 100 A minimum govern?
What standard service size is recommended?

That workflow prevents the biggest mistakes.

The final service size should not come from a guess, a square-foot rule, or a nameplate sum.

It should come from a structured load calculation.

Final thought

Residential electrical load sizing is not just:

House size × watts per square foot

And it is not just:

Add every appliance nameplate at 100%

The real calculation is category-based.

Some loads have demand factors.

Some loads have floors.

Some loads are non-coincident.

Some loads, like EV chargers, can push the service to the next standard size.

And sometimes the calculated load is below 100 A, but the minimum service requirement still governs.

That is why the most important part of a dwelling load calculation is not only the final amp value.

It is the breakdown.

The breakdown shows what was counted, what was demand-adjusted, which method was used, and why the service size changed.

For quick residential service load checks using the standard and optional methods, including general loads, appliances, HVAC, EV charging, service current, and service adequacy screening, use the Electrical Load Calculator on CalcEngineer.

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.