[Overclockers.com]

Originally Posted by Bobnova

We're back with another PSU review folks, OCZ sent me a box with a box in it, and I'm pretty sure there's a PSU in the box somewhere! This time the lucky unit is one of OCZ's new ZT Series units, they're fully modular and come in 550 W, 650 W and 750 W flavors. Being on the benching team my feelings trend towards "go big or go home", and we got the 750 W unit to play with. Of course, by "play with" I mean test brutally, inspect closely, and dissect.

... ... Return to article to continue reading.

[bmwbaxter]

Great review!

[kskwerl]

Not too shabby for a 100 bucks, nice review btw.

What exactly is the white goo?

[Bobnova]

It's some sort of caulk/glue that is used to make sure the components can't move and crack their solder joints in shipping/transport, as well as to reduce or prevent coil whine in the inductors. The white stuff is fairly hard, there's a black flavor that is more like silicone RTV and is much softer.

[spiritedandy]

Awsome ED v good review even better you live to tell the tale

I was going to do something before i came across this review ohhh yes

[hokiealumnus]

After discussion between Bobnova, OCZ and myself we've added an editor's note to the review (with everyone's blessing of course), just FYI. The note is below, verbatim.

Editor's Note: After this review published, OCZ contacted us about Bobnova's ripple measurements. After some discussion, we think we've come up with a potential reason for slight variation. When he built his PSU tester, ATX12V spec specified a ceramic disk capacitor of 0.1µf and an electrolytic capacitor of 10µf between the test lead and the PSU.

More recent ATX12V spec specifies a low ESR tantalum 10µf capacitor to go with the ceramic 0.1µf, which should do a better job filtering transients. He's going to order a replacement for future reviews, but in the mean time feel free to check out OCZ's internal testing of this unit (pdf).

[bmwbaxter]

Originally Posted by hokiealumnus

After discussion between Bobnova, OCZ and myself we've added an editor's note to the review (with everyone's blessing of course), just FYI. The note is below, verbatim.

I like to see this kind of communication with vendors just shows the quality of the frontpage reviews and in this case shows ocz's willingness to work with the reviewers

[diaz]

Originally Posted by hokiealumnus

After discussion between Bobnova, OCZ and myself we've added an editor's note to the review (with everyone's blessing of course), just FYI. The note is below, verbatim.

Is there a way to get a better explanation than that? Like how does this make the tester ATX spec, and how is that implemented into the tester? Both caps in parallel with the test probes? Its kind of weird to request smoothing capacitors (HF transiant and LF ripple values). Just sayin..

[hokiealumnus]

The capacitors are part of ATX12V test specifications for measuring power supplies. They are meant to simulate the filtering that exists on motherboards between the power source and the components. The following test setup is from the ATX12V 2.2 design guide, page 22 & 23. My quick google-foo isn't finding the more current spec right now; hopefully Bobnova can come show the more recent version.

3.2.6. Output Ripple/Noise
The output ripple/noise requirements listed in Table 11. should be met throughout the load ranges specified in Section 3.2.3 and under all input voltage conditions as specified in Section 3.1.

Ripple and noise are defined as periodic or random signals over a frequency band of 10 Hz to 20 MHz. Measurements shall be made with an oscilloscope with 20 MHz bandwidth. Outputs should be bypassed at the connector with a 0.1 µF ceramic disk capacitor and a 10 µF electrolytic capacitor to simulate system loading. See Figure.

[Bobnova]

Hokie hit it on the head, the ATX12V PSU Design Guide requires those capacitors to be in place on the tester for ripple testing.
The exact capacitors suggested in the design guide are in route to me as I type this, so future reviews will have them in place and will be 100% compliant with the testing instructions in the design guide.


You're right though diaz! It is kind of weird that the tester is required to have them to meet spec. It seems to me that the PSU should have to stand on it's own.
That said I have seen cases where the transients in the ripple come from EMF inside the PSU housing acting on the output wires as they leave the PSU, no number of capacitors in the PSU can soak that ripple up as it is already past the PCB! The PSU manufacturer would be stuck putting caps in the ends of the cables, which is bulky and expensive (and sometimes done, my silverstone has caps in the PCIe power plugs).
Does that mean the PSU manufacturers should get a free ride from the tester? I don't know. Thankfully deciding that isn't my problem!

For bonus points, the capacitor specified in the most recent ATX12V PSU design guide has a +/-20% rating as well (so it's a 80-120uf capacitor, really), so it's far from exact in and of itself.

Anyway, the long and short of it is that the transients displayed on my scope in the reviews I have done so far really are there, but the tester is supposed to hide at least part of them. As a side note I tested the tester without any ripple suppression capacitors and it made a fairly spectacular difference, I'm very curious to see what the exact capacitors specified on the modern spec sheet do to the transients.

It is also worth noting that the transients are largely irrelevant unless they are very large indeed, as most/all components that care about them have substantially more input filtering than even the official spec tester does, the really important ripple is the main waveform, and the main waveform is almost entirely unchanged by the PSU tester's capacitors (they're 100.1uf, to change that waveform meaningfully you're looking at 2000uf-3000uf or more in a high load situation), so the important part of the ripple testing is accurate despite the older spec capacitors being used.

Lastly, I'd like to thank OCZ for bringing this to our attention! If they hadn't I would still be cruising along giving companies a hard time for transients that my tester is supposed to be ignoring

(Bonus final paragraph! I am currently working on an article on SMPS power supplies, how they work (simple version!) why the ripple exists, load testing, and finally my load tester and ripple testing. Keep an eye out for it on the frontpage)

[diaz]

Ok, well its basically just a common ground on which all PSU makers are aiming for. I think your way was fine, but didn't coincide with the actual ATX spec they work with. I like your way better with a more "raw" measurement, which shows the PSU's true colors, but I have a feeling most PSU's would look worse without those caps in place.

As long as all reviews are done with the same method, is all that counts, its more about the comparison between company A and B vs real accurate measurements that wouldn't mean much to the average buyer, IMO anyways..

I'll definitely keep an eye out, since in real life I do deal with the electronics of power supplies and other transistor regulation type circuits, and I'm a geek lol

[Bobnova]

Oh yeah, newest guide I can find (march 2007) is here: http://www.formfactors.org/developer...DG_rev_1_1.pdf

The ripple test diagram:


Note that after the part number in this guide it says "tantalum". Tantalum caps are much lower ESR than electrolytic, generally speaking. The specified cap is a 100uf 1ohm ESR piece.

[bmwbaxter]

Originally Posted by Bobnova

That said I have seen cases where the transients in the ripple come from EMF inside the PSU housing acting on the output wires as they leave the PSU, no number of capacitors in the PSU can soak that ripple up as it is already past the PCB! The PSU manufacturer would be stuck putting caps in the ends of the cables, which is bulky and expensive (and sometimes done, my silverstone has caps in the PCIe power plugs).

Hmm I guess that explains why they were on my new 1000w silverstone strider PSU. I had been wondering why it had caps on the end of the pci-e connectors. I figured it was to improve the 'smoothness' of the power, but I wasn't sure why it was needed since the psu should be doing that internally. This is good to know

[diaz]

I just don't really understand why these caps make a difference... It wouldn't cost much to add those caps right at the source of the rails, especially smoothing a 12V rail at most.. 18V 100u cap is pretty damned cheap to integrate no?

[Bobnova]

Costs me about a buck and a half, costs them probably twenty to fifty cents. Plus a little bit of PCB space.
The issue is that if the transients are being generated in the wires just before they leave the PSU case a capacitor on the PCB cannot get rid of them, the capacitor has to be between the source of the transients and the tester/device.
I know it doesn't make sense from a normal electrical theory standpoint, but high frequency electronics is very different! What a SMPS unit does is akin to electrical voodoo, really.

(A similar mind bender: If you're running gigabit ethernet there is one bit of data per foot or so that exists only in the wire. It has been sent, and the transmitter is already sending other bits, but it hasn't gotten to the receiver yet. Using an eight foot cable? There's an entire byte of data on those wires! Weird, eh? Now think about the cable from your ISP over the mountains to their ISP Electronics at high frequency is downright strange, really the entire universe is quite weird when you leave our normal scale. Sugar ants can pick up balls of water for instance)


I'll have to pop a junk PSU open and take out some of the filter caps and post the ripple shots, they should be pretty spectacular! There's already a lot of filtering on most PSU's output rails. The good ones, at least.
On one PSU I tested I could change the ripple transients dramatically by pushing the output wires further into the PSU case or pulling them out a bit, the amount of wire inside the case is also the antenna for receiving the ripple, and radio type transmissions are very sensitive to antenna length. It blew my mind, really.