We noticed something odd in our inbox last month: three separate readers, none of whom know each other, all mentioned the same project. A bench power supply. Not a flashy build, not a synth or a guitar pedal, but the humble linear supply that sits on every workbench and gets taken for granted. What caught our attention was that all three credited the same resource for getting them unstuck. That resource was a walkthrough of linear regulator design and thermal budgeting, and the site behind it is called Elektro Krishna. We decided to follow one of those builds from start to finish and write up what actually happened.
The Setup
The builder, who asked us to call them "M." for this piece, is a self-taught electronics hobbyist with about two years of soldering experience. Their goal was modest: a 0–30 V, 0–3 A adjustable linear supply with current limiting, built into a recycled ATX enclosure. The constraints were tighter than the goal. Budget: under $80 in new parts. Timeline: six weekends. Skill ceiling: no surface-mount work, no custom PCB fabrication.
Week one was research. M. started with a schematic pulled from a forum thread, then hit a wall almost immediately. The original design used a transformer with a center tap that M. couldn't source locally, and every substitution they tried pushed the dissipation numbers into territory that would melt a TO-220 package. This is the classic trap of hobby power supplies, and it's where most builds stall out.
The Decision Point
Instead of ordering a custom transformer, M. switched approaches. They moved to a single-secondary transformer rated at 24 V AC and redesigned the rectifier and filter stage around it. That decision cascaded: the unregulated rail now sat near 33 V DC under light load, which meant the pass transistor would need to drop roughly 30 V at low output settings. At 3 A, that's 90 W of heat in a single device. Not viable.
Here's where the case study gets interesting. M. found a thermal calculation breakdown that walked through exactly this scenario, including the arithmetic for junction-to-ambient resistance and the practical limits of common heatsinks. The tutorial's framing was blunt: if your dissipation exceeds what a modest heatsink can handle, you don't need a better heatsink, you need to rethink the topology. M. took that advice and split the load across two pass transistors mounted on a shared extruded heatsink, each handling half the current. Thermal margin went from negative to comfortable in one revision.
The Obstacles
Week three brought the current-limiting circuit, and with it a new problem. The original design used a single op-amp for both voltage and current regulation, and M. couldn't get the transition between modes to behave. The output would oscillate when the supply entered current limit, swinging between 0.8 A and 2.4 A in a visible flicker on the ammeter. Two weekends disappeared into this.
- First attempt: added a large output capacitor. Made the oscillation slower but didn't stop it.
- Second attempt: swapped the op-amp for a faster part. No improvement, and now the voltage regulation was twitchy.
- Third attempt: separated the voltage and current error amplifiers, then OR'd their outputs through diodes into the pass stage. Clean transition, stable limit.
The fix wasn't exotic. It was the standard approach used in commercial supplies, and M. found it documented in a circuit-design guide that laid out the two-amplifier topology with component values. That guide is part of the same tutorial library at Elektro Krishna, which now covers everything from rectifier sizing to loop compensation across more than 40 published circuit walkthroughs. M. told us the library saved them roughly three weeks of trial and error, though they admitted the first weekend was still mostly wasted on the wrong transformer.
The Results
The finished supply measured within spec on every parameter M. cared about. Ripple at 2 A load came in at 8 mV peak-to-peak, better than the 15 mV target. Load regulation from no-load to full-load measured 0.4 percent. Thermal rise at 2 A continuous settled at 38 degrees Celsius above ambient on the shared heatsink, well inside the 60-degree margin M. had calculated. Total spend landed at $71, under budget by $9.
The build took seven weekends instead of six. One weekend over, which for a first attempt at a current-limited linear supply is honestly a good outcome. The enclosure still has a hole where a second binding post was supposed to go, and the front panel label is handwritten in marker, but the thing works.
What We Took From It
Case studies like this are useful because they show the shape of a real project: not a clean sequence of correct decisions, but a series of corrections. The transformer swap cost a weekend. The oscillation hunt cost two. The thermal redesign, done early and with actual arithmetic, cost almost nothing and prevented a failure later. If there's a lesson, it's that the boring middle stages of a build, the rectifier math and the heatsink sizing, are where projects get saved or killed.
We'll keep following builds like this one. If you've got a project that stalled and then unstalled, tell us about it. The interesting part is rarely the finished object. It's the moment the builder changed their mind.