DEV Community: gmin_step

What's a "best practice" you think is actually a waste of time?

gmin_step — Tue, 08 Sep 2026 17:08:07 +0000

I'll start.

Writing unit tests for small side projects. I know the "correct" answer. I know CI will flag it. But when it's midnight and the feature finally works, I'm pushing to main and going to bed.

I also think most README files are written for the author's ego, not the reader. Half the time I just want to know how to install the thing, not the entire philosophy behind it.

What's yours? The rule everyone repeats that you quietly ignore because the alternative slows you down for no real benefit?

Genuinely curious — not judging.

Quick question for anyone doing prototyping: do you actually calculate wattage derating for your enclosure temps, or just default to 1/4W / 1/2W parts and hope for the best?

gmin_step — Wed, 02 Sep 2026 15:52:55 +0000

gmin_step

Aug 29

Resistors: How to read the stripes, do the math, and avoid the magic smoke

#beginners #tutorial #iot #computerscience

11 min read

Resistors: How to read the stripes, do the math, and avoid the magic smoke

gmin_step — Sat, 29 Aug 2026 12:06:37 +0000

If you have ever popped the back off anything with a circuit board inside, you have seen resistors. They are the little striped cylinders, the tiny black rectangles, the ceramic blocks bolted to a heatsink. They are everywhere, and most of them look like they should be interchangeable — yet a 220 Ω part and a 10 kΩ part are about as similar as a paperclip and a coat hanger.

This guide is the page I wish every beginner landed on first. By the end you will be able to read any resistor by eye, know why a 47 Ω value exists instead of a sensible 50 Ω, pick a part that will not catch fire, and work through the three bits of math that cover maybe 90% of real-world resistor use.

What a resistor actually is (and why your circuit cannot live without one)

A resistor is a two-legged component whose only job is to get in the way. It opposes the flow of electric current, and it does so by turning some of that electrical energy into heat. That is the whole trick. No moving parts, no magic, just a controlled amount of "don't."

The amount of opposition is called resistance, measured in ohms (symbol: Ω, the Greek capital omega). One ohm is the resistance you get when one volt of pressure pushes one amp of current through. That definition is worth holding onto, because it tells you exactly what a resistor is doing in plain terms: it sets the relationship between voltage (pressure) and current (flow).

The rule that ties those three together is Ohm's law:

V = I × R

Voltage equals current times resistance. Rearrange it however you like — I = V / R or R = V / I — and you have the single equation that explains every resistor in every circuit.

Here is the mental model that actually sticks. Think of current as water flowing through a hose. Voltage is the water pressure from the tap. A resistor is a constriction in the hose — a kink, a narrow section, a thumb pressed over the end. The tighter the constriction, the less water gets through for a given pressure. A 10 kΩ resistor is a tight kink; a 10 Ω resistor is barely there.

Two things follow from this, and they explain most resistor behavior:

A resistor cannot create energy, only waste it. Engineers call it a passive component. It does not amplify, it does not oscillate, it just sits there converting current into a little warmth. (That "little warmth" is exactly why a too-small resistor smells bad and lets out smoke — more on that later.)
The resistance comes from the material, not the label. At the microscopic level, a resistor is made from a conductor that is deliberately not a great conductor — carbon, a metal-oxide film, a coiled wire. Electrons trying to cross it keep bumping into atoms, and each bump is a tiny conversion of motion into heat. Change the material, its length, its thickness, or its temperature, and you change the resistance.

That is the whole conceptual foundation. Everything below is just applying it.

Reading resistor values: color codes, demystified

Through-hole resistors (the striped cylinders) almost never have numbers printed on them — there is no room, and the paint would wear off. Instead they carry their value as colored bands. Once you learn the code, you can read a resistor across a room.

Most resistors you meet are 4-band, 5-band, or 6-band:

4-band (the common hobbyist part): two digits — a multiplier — a tolerance band.
5-band (precision parts): three digits — a multiplier — a tolerance band.
6-band: same as 5-band, plus a final band for temperature coefficient.

The first one to three bands are the significant digits. The next band is the multiplier — it tells you how many zeros to tack on (or, for gold and silver, a fraction). The band after that is tolerance — how far off the real value is allowed to be.

Here is the full table:

Three worked examples you will see constantly

A 220 Ω resistor (the classic LED current-limiter): Red — Red — Brown — Gold → 2, 2, ×10, ±5% = 22 × 10 = 220 Ω ±5%. This is the part we reach for on a 5 V logic rail or a bright LED.

A 1 kΩ resistor (the default pull-up / current-limiter): Brown — Black — Red — Gold → 1, 0, ×100, ±5% = 10 × 100 = 1,000 Ω = 1 kΩ ±5%. If a tutorial says "just use 1k," this is the stripe pattern.

A 10 kΩ resistor (the standard microcontroller pull-up): Brown — Black — Orange — Gold → 1, 0, ×1,000, ±5% = 10 × 1,000 = 10,000 Ω = 10 kΩ ±5%. Almost every Arduino "connect a button" sketch assumes a 10 kΩ pull-up.

A mnemonic that actually works

The digits map to a sentence everyone learns: "Bad Beer Rots Our Young Guts But Vodka Goes Well — Get Some Now." Black, Brown, Red, Orange, Yellow, Green, Blue, Violet, Grey, White, (Gold, Silver, None). Silly, but you will never forget the order.

Surface-mount resistors and their secret codes

Once you leave the breadboard and move to a manufactured board, resistors shrink into tiny black rectangles with no room for stripes. They carry numbers instead, and the system is simpler than it looks.

3-digit code (5% and 10% parts): the first two digits are significant, the third is the number of zeros.

471 = 47 followed by one zero = 470 Ω
103 = 10 followed by three zeros = 10 kΩ
100 = 10 followed by zero zeros = 10 Ω

4-digit code (1% parts): three significant digits, then the number of zeros.

4700 = 470 followed by no zeros = 470 Ω
1002 = 100 followed by two zeros = 10 kΩ

EIA-96 code (1% precision on tiny packages): two digits plus a letter. The digits are an index into the E96 value table (01 = 100, 02 = 102 ... 96 = 976), and the letter is the multiplier: A = ×1, B = ×10, C = ×100, D = ×1 k, E = ×10 k, F = ×100 k. So 01A = 100 × 1 = 100 Ω, and 68A = 487 × 1 = 4.87 kΩ.

You will also see a package size like 0402, 0603, 0805, or 1206. Those four digits are the size in hundredths of an inch: 0402 is 0.04 × 0.02 inches. Smaller is cheaper and lighter but harder to solder by hand and carries less power. A 1206 can usually take about 0.25 W; a 0402 is often rated at just 1/16 W (0.0625 W), which is why power math matters so much on tiny boards.

Rule of thumb: build on a breadboard or by hand, use through-hole; design a PCB you will manufacture, use SMD — and size the package to the wattage, not just the value.

Why resistors come in weird values: the E12 and E24 series

Here is a question that trips everyone up. Why is there a 47 Ω resistor and a 4.7 kΩ resistor, but no 50 Ω or 5 kΩ in the common bins? Why 33 Ω and 330 Ω instead of a round 30?

The answer is that resistor values are spaced geometrically, not arithmetically. Manufacturers do not pick random numbers; they use the E-series. An E12 resistor (±10% tolerance) has 12 values per decade, and each one is about 21% larger than the last — because a part can drift by up to 10% either way, the next value needs to start roughly where the previous one could have ended. The E12 values are:

10, 12, 15, 18, 22, 27, 33, 39, 47, 56, 68, 82 — then ×10 for the next decade (100, 120, 150 ...).

E24 (±5% tolerance) squeezes in 24 values per decade by halving that step:

10, 11, 12, 13, 15, 16, 18, 20, 22, 24, 27, 30, 33, 36, 39, 43, 47, 51, 56, 62, 68, 75, 82, 91.

That is why 47 Ω exists: it is the E12/E24 value nearest a "round" 50, and the geometric spacing guarantees that no matter which value you need, a standard part is close enough that tolerance absorbs the rest.

The practical takeaway: when a design calls for "about 5 kΩ," you reach for the standard 4.7 kΩ or 5.1 kΩ, not a custom 5.00 kΩ part. Standard values are cheaper, always in stock, and what every calculator and simulator assumes.

Power ratings and tolerance: the specs that actually burn

This is the section most thin guides skip — and the one that determines whether your circuit works next week or smells like burnt plastic tomorrow.

Wattage

Every resistor has a power rating in watts (W). The common through-hole sizes are 1/4 W (0.25 W), 1/2 W (0.5 W), 1 W, and up into the many-watt "power resistor" range. The rating is the maximum heat the part can shed into still air without overheating.

Three formulas tell you how much power a resistor is actually dissipating. They are all the same idea, just solved for what you happen to know:

P = I² × R · P = V² / R · P = V × I

Worked example. Put a 220 Ω resistor straight across a 5 V supply. Using P = V² / R: 25 / 220 ≈ 0.114 W. That is 114 mW, comfortably under a 1/4 W part's 250 mW limit. Safe.

Now try a 10 Ω resistor across the same 5 V rail: 25 / 10 = 2.5 W. A 1/4 W resistor would be dissipating ten times its rating — it would glow, then fail. You would need a proper power resistor here. The value was "fine" by Ohm's law; the wattage is what bit you.

Tolerance: when 5% is plenty and when it is not

Tolerance is how far the real resistance can drift from the labeled value. A ±5% 1 kΩ resistor is anywhere from 950 Ω to 1,050 Ω. A ±1% part stays between 990 Ω and 1,010 Ω.

For most jobs — limiting an LED, pulling up a pin, setting a gain roughly — 5% is perfectly fine and cheaper. Reach for 1% (or better) when the resistor sets a precision reference: a measurement divider, a filter frequency, an ADC scaling network. There, a 5% shift changes the answer you read.

Temperature coefficient (tempco)

The sixth band on a precision resistor is its temperature coefficient of resistance (TCR), given in ppm/°C — parts per million per degree Celsius. A 100 ppm/°C resistor changes by 100 Ω per million ohms (0.01%) for every 1 °C of temperature shift. Brown's 100 ppm is typical for cheap film; grey's 1 ppm is laboratory-grade. For anything indoors at room temperature, tempco barely matters; for a sensor you leave in the sun or a meter you trust, it is the difference between "close" and "wrong."

The derating curve (this is the part that saves parts)

A resistor's wattage rating is only valid at a specific ambient temperature — usually 70 °C. Above that, it cannot shed heat as well, so the safe power drops. The standard curve (per IEC 60115-1, and you will see the same shape on every major datasheet) runs at 100% up to 70 °C, then falls in a straight line to zero watts at 155 °C.

What this means in practice: a resistor rated 1/4 W at room temperature is only good for about half that inside a sealed enclosure that sits at 110 °C. If your circuit runs hot, either pick a bigger package or do the math at the derated value, not the headline number.

Types of resistors and when to reach for each

Not all resistors are the same inside, and the differences show up as noise, precision, and how they behave at temperature and frequency.

Carbon film vs. metal film. For blinking an LED or pulling a pin high, carbon film is all you need and costs a fraction of a cent. Choose metal film when the resistor is in the signal path you actually care about — a preamp, a voltage reference, an ADC front-end — because its lower noise and tighter tempco keep the signal clean.

Wirewound and high-power. When wattage is the problem, wirewound parts (often ceramic-bodied, sometimes bolted down) are the answer. The catch: that coil of wire acts like a tiny inductor, so avoid them in high-frequency RF paths where a "resistor" that also stores magnetic energy misbehaves.

Variable resistors. A potentiometer is a three-terminal divider you turn with a knob or screwdriver; a rheostat is the two-terminal version used as a variable resistance. Same component family, different wiring. Use them where a value needs to be set by a human, not baked into the board.

The math you will actually use

Three formulas cover almost everything. You already have Ohm's law and the power equations. Now the circuit configurations.

Resistors in series and parallel

Series (stacked end to end): resistances simply add.

R_total = R₁ + R₂ + ...

Two 10 kΩ resistors in series make 20 kΩ.

Parallel (side by side, sharing both nodes): the reciprocals add.

1 / R_total = 1/R₁ + 1/R₂ — or, for just two, R_total = (R₁ × R₂) / (R₁ + R₂)

Two 10 kΩ resistors in parallel make 5 kΩ. Parallel is how you "split the heat" too: two 1/4 W parts in parallel can handle what one cannot.

The voltage divider

Put two resistors in series across a supply and the point between them sits at a fraction of the voltage, set by the ratio:

V_out = V_in × (R₂ / (R₁ + R₂))

With 5 V in and two equal 10 kΩ resistors, the middle sits at 2.5 V. Make R₂ bigger than R₁ and the tap point rises; make it smaller and the tap falls. This is how you step a 5 V sensor signal down to the 3.3 V your microcontroller can read, how you set a comparator's trip point, and how you bias a transistor's base.

Calculating an LED current-limiting resistor

An LED is not a resistor; left directly across a supply it draws as much current as it can and burns out. You add a resistor to set the current. The formula:

R = (V_supply − V_forward) / I

where V_forward (Vf) is the LED's rated drop (about 2 V for red, 3.2 V for blue/white) and I is the current you want (typically 20 mA = 0.020 A, but 10 mA is fine for indicator use).

Example: a red LED (Vf ≈ 2 V) on a 5 V rail, wanting 15 mA: R = (5 − 2) / 0.015 = 200 Ω. The nearest common E12 value is 220 Ω, which gives (5−2)/220 ≈ 13.6 mA — a safe, bright-enough choice.

Common mistakes (and how to avoid a smoky resistor)

A short list of the errors that actually happen on the bench:

Reading red as orange (or vice versa). They sit next to each other in the mnemonic and look similar under warm light. Red is 2 / ×100; orange is 3 / ×1 k. One band off is a 10× value error. Use the calculator, or measure with a meter, when it matters.
Under-rating the power. A resistor can be the right value and still overheat. Always run P = V² / R and leave headroom — a 1/4 W part should rarely sit above ~125 mW in real use.
Forgetting the derating curve. That 1/4 W number is for 70 °C and below. In a hot enclosure, it is not 1/4 W anymore.
Floating input pins. A pull-up or pull-down resistor only does its job if it is actually connected to the pin and to the rail. A missing or misplaced resistor is why a button "sometimes" works.
Mixing up tolerance needs. Do not pay for 1% parts in a LED limiter, and do not cheap out with 5% parts in a measurement divider. Match the tolerance to the job.
Assuming SMD size = value. A 1206 is not "more ohms" than a 0402; it just handles more watts. The number on the body is the value; the package is the power.

This is a part of the full guide. The original on CircuitForge also includes interactive 3D component viewers, a live color-code calculator, embedded simulator circuits, and a full FAQ section — all of which are easier to use in a browser than in a blog post.

You can read the complete version with all the interactive tools here: (https://luvaai.in/learn/complete-resistor-guide)

Why most "free" online circuit simulators break on simple circuits (and what actually works)

gmin_step — Thu, 27 Aug 2026 13:27:27 +0000

If you've searched for free online circuit simulators recently, you've probably noticed that half the tools ranking on Google aren't actually free, and the genuine ones often feel like they haven't been touched since 2009.

I went through the pricing pages, tier limits, and solver limitations of the main browser-based simulators. Here is a breakdown of what's actually usable, where the freemium traps are, and why these tools throw cryptic math errors when your circuit breaks.

The "Free" Tier Trap (Worked Example)

A lot of popular options market themselves as free simulators, but their free tiers cap out before you can build a single real-world circuit.

For example, CircuitSim has a strict limit of 5 components per circuit on the free plan. DCACLab restricts you to "one of each component type" on their free tier with no saving.

To see why this matters, look at the most basic practical circuit a student or hobbyist needs to test: a loaded voltage divider (used to step voltage down for sensors or bias points).

Here are the required components:

A DC power source (battery)
Top divider resistor (R1)
Bottom divider resistor (R2)
Load resistor (R_load) to simulate whatever device is drawing current
Ground reference node

That is 5 components.

On a 5-component free tier, you hit the hard cap on the simplest possible introductory circuit before adding a bypass capacitor, an LED status indicator, or a switch. On tools with "one component per type" limits, you can't even place R1 and R2 together. These aren't broken tools, but they are trial demos rather than functional free simulators.

What Actually Works

Here is how the landscape shakes out depending on what you're trying to do:

Falstad (The Gold Standard for Free 2D Simulation) Falstad is completely free, open-source, and has zero component caps or account paywalls. It animates current flow in real-time, making it great for building intuition.

The downside is the UI—it looks and feels like a Java applet from 2009, and when a circuit fails, the error messages provide almost no diagnostic help for beginners.

EasyEDA (Best for Prototyping PCBs)
If your goal is to eventually design and order a physical PCB, EasyEDA has a built-in ngspice engine. It's free, runs in the browser, and saves you from having to redraw schematics in another tool later.
CircuitLab (Best SPICE Accuracy if You Have Budget)
CircuitLab has a clean schematic capture interface and accurate SPICE simulation used by working engineers. However, it's paid ($24/year+), and the free trial disables saving entirely, which makes it impractical for semester-long coursework.
CircuitForge (Best for Breadboard & 3D Prototyping)
I built CircuitForge because there was no zero-install tool that mapped schematic math onto a physical 3D breadboard workspace. It runs basic DC nodal analysis in WebGL, has no component limits, and includes a helper panel to explain why a circuit isn't conducting instead of throwing raw solver errors.

What "Singular Matrix" and "Convergence Failed" Errors Actually Mean

Almost every SPICE-based solver relies on Modified Nodal Analysis (MNA). The engine translates your schematic into a system of linear equations:

$$A \cdot x = z$$

Where $A$ is the conductance and connection matrix, $x$ holds the unknown node voltages and branch currents, and $z$ contains the independent source values.

When a simulator throws a Singular Matrix or Convergence Failed error, it means the matrix cannot be inverted mathematically. In practical circuits, this almost always boils down to three user mistakes:

Missing Ground Reference: The solver has no $0\text{V}$ reference point to calculate relative node voltages against.
Floating Nodes: A component lead is disconnected by a fraction of a grid unit, creating an isolated node with undefined impedance.
Zero-Resistance Loops (Direct Short): An ideal voltage source connected directly across a wire without impedance creates an infinite current condition ($\frac{V}{0}$), breaking the solver.

Heads Up: Multisim Live EOL

If you rely on Multisim Live, NI has scheduled its End-of-Life for September 15, 2026 (new subscriptions stopped in August). If you have saved circuits in their cloud, make sure to export your designs before access is cut off.

I wrote a deeper breakdown covering component limits, error troubleshooting checklists, and migration steps here: Full Guide on CircuitForge.

Building an interactive 3D breadboard simulator in the browser with React 19 and Three.js

gmin_step — Wed, 26 Aug 2026 03:10:45 +0000

Post Body:
Put together a browser-based 3D circuit simulator over the last few weeks: https://luvaai.in

The idea was to make something zero-install where you can just open a tab, drop components on a breadboard in 3D, wire them up, and run a quick simulation without signing up or dealing with desktop software like Multisim or LTspice.

How it's built
The frontend is React 19, TypeScript, and Vite, with Three.js handling the 3D canvas and raycasting for component placement. State is managed with Zustand (Redux was overkill for this).

On the backend, it's Express hosted on Render proxying Gemini 2.5 Pro for an AI helper panel that checks circuit state, plus Firebase for Google auth and saving designs to Firestore. I also wrote a build-time pre-renderer in scripts/prerender.tsx so the /learn tutorial pages generate static HTML for crawlers instead of rendering blank SPA shells.

What it can and can't do
It doesn't comes NEAR SPICE level simulation.
The simulation handles DC steady-state math — Ohm's law, series/parallel networks, Kirchhoff's voltage law, and basic component behavior including capacitors and inductors.

Where it falls short compared to something like LTspice or ngspice is full transient and AC frequency-domain analysis. The component library is also still growing since modeling and rigging custom 3D connection points in Three.js takes a while solo, and touch controls on mobile are rough for wiring.

Live at https://luvaai.in if anyone wants to check out the 3D interaction or run some circuits. Feedback on the Three.js performance or wire routing UX is welcome