Welcome to buchanan1.net

J.R. Buchanan


Using a cheap Chinese LM2596 board in a power supply

June 2016

Board being tested

I've read multiple articles on the Internet about building general purpose (bench) power supplies from inexpensive Chinese LM2596 based buck mode switching regulator boards, which are available on Amazon, eBay, etc. at very low prices. These boards take an input voltage of up to 40 volts and output an adjustable voltage, controlled by a 20 turn trimmer potentiometer, from about 1.2 volts up to something a bit under the input voltage. They are rated at 3 amps. They have short-circuit protection built in. In these projects, the 20 turn trimmer is unsoldered and replaced with a potentiometer mounted on the front panel of the unit. The input voltage is often supplied by a used (a thrift store is a good source) laptop power supply.

It seems like an awfully nice solution, simple and inexpensive. By using a switching regulator, one of the biggest problems with variable bench supplies is eliminated, the horrible efficiency and high power dissipated by the pass transistor, which can easily get out of hand. It seemed almost too good to be true. Was it?

To test the idea out, I ordered one of these modules from Amazon. The first thing I noticed was that it didn't match the description. Instead of "All SANYO solid capacitors" (important for low ESR (Equivalent Series Resistance) in a switching supply), it had some pretty generic looking aluminum electrolytic capacitors. The LED power indicator was missing as well, but why would you need this in any case? Would the capacitors matter? According to the LM2596 data sheet, high quality capacitors are needed. If the input capacitor has too high an ESR, it will dissipate too much heat and will have a short lifespan. Solid tantalum capacitors are indeed recommended in the data sheet, although properly spec'd aluminum electrolytics are acceptable. If the ESR of the output capacitor is too high, there will be too much ripple on the output, a subject we'll get back to. Interestingly, if the ESR of the output capacitor is too low, the circuit will become unstable.

Closeup of board

I hooked the unit to a bench supply set to 20 volts and set the trimmer to as close to 12.0 volts as I could get. I then hooked a 100 mA load, the voltage dropped from 12.08 to 12.04 volts, a 0.3% drop. Then I tried 1 amp and it dropped from 12.08 to 11.97 volts, an 0.9% drop. Not really great, but for the price, we could live with this. I then tried dropping the input voltage from 20 volts to 15 volts. At 100 mA the output dropped 0.03 volts, a 0.2% drop. At 1 amp, the output dropped 0.04 volts, or 0.3% Again, not bad. So everything looks great, right?

Not so fast. Looking at a circuit with a multimeter is like looking into a room through a tiny knothole. Let's hook an oscilloscope up to the output and see what we see. Whoa. At 100 mA load current, we have 130mV peak to peak of ripple at about 55 KHz. At 1 amp, we have 278 mV peak to peak ripple at about 55 KHz. At 1 amp, the ripple is equal to more than 2% of the supplied voltage! OK for many things, but in my opinion, not so hot for a general purpose bench supply.

At 1 amp load

Closeup of board

If I'd read the data sheet before running these tests, I would not have been surprised by the results, as the data sheet states, " A typical output ripple voltage can range from approximately 0.5% to 3% of the output voltage." Our unit isn't even at the top of this range despite the generic looking capacitors. Maybe solid tantalum capacitors would take it to the lower end of this range? That might be acceptable if it could be managed across the entire output range.

I had planned on trying the circuit powered by a used computer power brick, and I had considered using a transistor to switch the load on and off to see what the transient response is like, but since the device has already pretty much failed for this use case, I see no point. All hope is not completely lost, the data sheet does discuss using an inductor and a low ESR capacitor as a "post ripple filter" to lower the output ripple voltage. I don't have a suitable capacitor on hand to try that now, I might in the future.