You're implying one hell of a dangerous assumption here: that copying other people's testing criteria (or to be more precise, not trying different testing methods and criteria) is somehow preferred. I'll explain as I go along.
MikeC wrote:
W/ regard to SPCR PSU testing, the only test mentioned by multiplexer that we don't and can't do is dynamic testing. Crossloading simply refers to the minimum required load on each of the main rails to keep the other rails in spec; we do this only for the 12V line by putting the lower lines at 1A.
This is all you really need to test for computer PSUs specifically. It's kind of marginally interesting to know what the PSU does in other crossload patterns, but this rarely yields useful standard results (i.e. efficiency, regulation, ripple). However, something I have said many times on other related forums: not everything you test needs to be published. You should be playing around and trying weird load patterns if you for instance see that the noise characteristics change dramatically as you 'unbalance' the crossload further. Anyway,
Quote:
Output noise and ripple have been standard in all our tests for years... and I know of no one who measures input noise.
You do

You measure power factor, which for electronic loads is an integral quantification of the input current distortion. Electrically, nowadays, the actual power factor actually isn't that interesting anymore. Also, the value of publishing the power factor without context is a source of confusion for many readers because it is kind of an abstract quantity. What is important though, is the input distortion as a function of frequency, i.e. the FFT of the input current waveform. Basically all power analyzers can display this, and this can reveal a lot about the input filter and PFC quality. That's what I mean about input noise. The same goes for output noise: it's really not that important to know what the peak values are, as is the shape of the FFT.
doveman wrote:
They say Transient Response, not Dynamic Response, which seems to refer to their Transient Load Testing results at the bottom of
this page, briefly how the PSU responds to short duration loads, such as a RAID array spinning up or draw from a video card. They mention the disappointing results are quite common with these ATNG built units.
But as you say, this and the voltage regulation issues may not have any meaningful effect in practice.
They do! Transient response is one of the most important tests a manufacturer does for its power supplies, and something that is very strangely absent from reviews. I think this is because power supply reviewers generally don't come from either the science or engineering, but from the general hardware/journalist genre and as such do not know how manufacturers usually characterize their equipment.
Transient response is not 'dynamic' response, nor is it the response to macroscopically (i.e. disks spinning up) 'short' duration peaks. Transient response is the response to (1) an instantaneous step current increase, (2) a dirac-function (very peaked impulses) load pattern (3) a ramp current increase/decrease (i.e. constant dI/dt) and (4) turn-on/turn-off. Now, what is important to know about this is that when you do such a thing (for instance, suddenly go from 1 to 5A current draw) the output voltage does not stay the same, but it has a tendency to either overshoot (underdamped) or undershoot (overdamped), depending on the attainable current slope. In designing a power supply, it is absolutely critical to design these output filters such that they 'do' this properly. If you don't, and you have for instance a load that turns on and off at 20kHz, the voltage may keep under- and overshooting or even amplify this effect every time over, causing what is called an oscillating power supply. What is most important to note here is that in general, these overshoots and undershoots cause way more output voltage disturbance than the ripple at constant load. Especially when you're dealing with power supplies that already have appreciable ripple, try the transient load patterns. You'll see that they will end up out of spec. Also, good transient response is a marker for a good quality supply. Bad transient response is a good indicator of early failure (bad transient response is usually caused by underrated output capacitors, with high ESR causing a lot of self-heating under dynamic load conditions, causing overheating and early aging of the capacitors).
Even if all of this text doesn't seem that important to you, there is one big lesson in here. Transient response, the quality of filtering, the brand and quality of the components used and the topology used (these last things can usually only be determined by doing a full tear-down, refer to hardwaresecrets) are actually the classic, 'original' if you will, markers for power supply quality and suitability. Back in the day when power supplies weren't reviewed and just built, this is what the engineers would base their decisions upon. These are the areas that actually cost money and that indicate a significant R&D effort. Efficiency, up until recently, as well as acoustics weren't that important. The only thing engineers would look at was the longevity of the fan used and whether cooling was sufficient. I'm not saying that that was a particularly good thing in the industry, but I am saying that just because none of the other reviewers don't do it, those testing criteria aren't important. They are definitely more important to assessing power supply quality and performance than measuring output peak-to-peak ripple and line regulation - things that are actually structurally done wrong*.
I hope it's a bit more clear now why I said what I said. And yes, I am a power electronics engineer who makes and mods power supplies, I am sort of biased
* (1) Vpp is often measured with purely resistive load behind a ginormous low-esr capacitor bank. This hugely dampens ripple both due to the constant nature of the load and the dampening effect of the uncommonly large capacitor bank. When measuring ripple with a real-life load it is seldomly less than 20% higher. (2) line regulation is often measured with too thin wires (e.g. all 12V power goes through 2 EPS12V plugs) and in general measured at the wrong connector. This causes some reviewers to report very steep regulation ramps which are actually just caused by the resistance of the wires used. They also don't take into account the specification which says that the voltage drop at maximum conductor current can sag a certain amount more than the ATX spec itself.Oh, by the way, this is in no way an insult to SPCR or SPCR's testing methodology, I intend what I write to be informative.