Power Supply Tests

Here are tests of some power supplies I've built, along with results from other commercial supplies for comparison.

Ripple Rejection Test

In this test, I put a 100 Ω dummy load on the output of the power supply. Since all of the power supples in this test are set for 24V (with one minor exception, noted below), this puts the same load on each, 240 mA. I ran the loaded power supply output to my measurement preamp, set to amplify only AC noise by 100×. I measured this noise with a Fluke 189 DMM to get the mVRMS number, and with a Protek 6502 analog oscilloscope to get the noise characterization and oscillogram. Most of the oscillograms are taken with the same scope settings so they're easy to visually compare, though it sacrifices resolution in the lower-noise shots. See the characterization text for a short description of what one sees at higher magnification.

You may be wondering why I used a low-end analog scope instead of one of the more expensive Tek DSOs I have access to. It's because dollar-for-dollar, analog scopes have less noise than digitals. In previous versions of this test, I observed the noise with my M-Audio Delta 44 sound card which is quieter than even high-end DSOs, but you sacrifice bandwidth, important for some of the test results below. Also, feedback made it clear that the spectrograms captured from the sound card were difficult to interpret correctly.

mVRMS Characterization Osc. YJPS v0.4* 0.007 very fuzzy broadband noise, no fundamentals STEPS** 0.058 some 120Hz ripple visible, but pretty fuzzy TREAD 0.060 essentially identical to STEPS, on this test   Velleman K1823 0.10 unstable waveform, giving large semi-random fluctuations beyond basic noise B+K Precision 1710 0.17 HF noise nearly smothers 120Hz ripple Creek OBH-2 0.25 primarily 120Hz ripple, very little HF noise atop that Elpac WM080 1.8 strong HF content atop 120Hz fundamental Elpac MW1224 7.8 little else but 67kHz sawtooth switching noise Radio Shack 273-1668*** 32 standard 120Hz sawtooth ripple waveform

* The YJPS is a PCB I prototyped for a Jung 2000 super-regulator circuit. After seeing its performance in these tests, I solicited requests to see these boards made, but didn't get enough replies to justify that.

** This is a STEPS v1.1, set for 24V, with a National Semiconductor LM317 and 4000 µF of filter capacitance in it. STEPS v1.0 and v1.2 both test all but identically when configured equivalently.

*** The Radio Shack supply was set on the 13.5V setting, but being an unregulated supply it actually put out 15V under the test load. You would expect 60% higher ripple with a 24V unregulated supply, all else being equal.

Load-Rejection Test

I removed this test because I was unable to get useful results. Either all supplies tested equally good, or equally bad, or it ended up just re-stating the ripple rejection results. I think it comes down to designing a tougher dynamic load than the one I used previously, a PPA driving a headphone dummy load with sine waves. The problem with the PPA is that its supply current doesn't vary all that much in normal use. I suspect you could get better results making the power supply drive a pure class B amplifier with very little rail capacitance. Since that would be a pretty bad sounding amp, I'm not really interested in building it just for these tests.