Power supply ripple hall of shame

[Per Hansson]

While our PSU hall of shame thread is fun I thought It'd be fun to have a ripple hall of shame thread too

I'll go first out, it's an industrial PSU that offers +15v -15v -5v +5v rails.
However it has been mounted on a frame with DC/DC converters for producing +12v and -12v as well.
The caps for this have gone a bit high ESR
The measured ripple is around 1700mV, or 1.7v!
After recapping the ripple is gone, also note that the scale on the scope is 50mV/div instead of 500mV per div in this shot
-And that's how you know you have a qualifying PSU: the ripple scale needs to be in the volts range or thereabout

[tom66]

Re: Power supply ripple hall of shame

That ripple is so high and repetitive I'd be tempted to say it was down to oscillation in the controller (caused by bad caps) rather than the caps themselves contributing to it.

My entry. I don't have a scope shot because this was a while ago. But I had a OneHunLo Chinese PSU for an old desktop PC. It had been sitting in this PC unused for a few years so I decided to power up the PC before scrapping it for anything useful.

When I hooked up the PC to my CRT monitor, I noticed that there were strange "hum bars" on the image, and a kind of diagonal saw-tooth ripple through the horizontal axis of the image (imagine the sides of the image have a kind of saw-tooth effect added to them.) This seemed to vary somewhat with load, when the hard drive was accessing things or if the CPU was under a bit more stress the hum bars got stronger. How odd!

I scoped the 12V output and found it had 100Hz ripple on it...about 3Vp-p! 5V and 3.3V rail had <0.5Vp-p on them. Remarkably the HDD, CPU VRM and the other parts of the system seemed otherwise content... I can only guess that the sync pulses or some signal produced by the VGA port was a bit fucked causing the hum bars but otherwise there were no obvious ill effects.

The cause was that the main filter cap was almost open circuit (ESR > 10 ohms, capacitance was about a tenth of the rating), so the output caps were basically handling all of the 100Hz ripple. I'm amazed the supply was able to hold up during the dead periods of the mains cycle. I can only guess that there was enough capacitance on the PSU's 12V output and the CPU VRM bus to keep everything going for the few ms of low switching. The 5V/3.3V rails must have had little load on them or the PSU was more able to regulate them (I don't think it was a DC-DC type though).

So I guess that if this PC was left running for much longer, those VRM caps would eventually cook and the whole thing would die as it wouldn't have the necessary hold-up to keep going.

[eccerr0r]

Re: Power supply ripple hall of shame

I have an on-motherboard 1.2V SMPS that was going about a half volt P-P of ripple due to bad capacitors and going unstable...

Needless to say the machine isn't very happy, the onboard graphics crashes constantly.

---

An actual PSU, an old Antec PSU whose 3v3 output caps fuhjyyu'ed. Was getting about 1V P-P ripple in it. Machine also was not happy with this and refuses to operate properly. It did allow boot oddly enough.

[CapLeaker]

Re: Power supply ripple hall of shame

I like Per's scope meter!

Other than that I had seen PSU's in the 1V to 1.3V ripple territory.

[dmill89]

Re: Power supply ripple hall of shame

This hunk of garbage VIO KY-550ATX, I originally posted in the gutless hall of shame.



(note: I replaced the two visibly bulged caps with used spares I had laying around)

I added a little more load 191W total 12V@ 8.5A, 5V@10A, 3.3V @12A


The scale is set to 200mv (.2V) per division:


12V (around 1.2V ripple):


5V (also around 1.2V ripple):


3.3V (slightly better at around 1V ripple):



I imagine this thing is about on the verge of blowing up with this load and it is much worse than the already bad ripple with the 137W load in the gutless hall of shame.

[kaboom]

Re: Power supply ripple hall of shame

VIO for VIOlent!

A really bad Pow-to-the-Max copy, complete with "bell wire" common mode choke and "two-big, two-small" line rectifiers.
Refer to my user name and sig...

You've got ripple (the sawtooth) as well as hunting in the feedback loop (the spikes). What does the "550" even mean in the fake model number?

How much it draws when tuned for max smoke?

[dmill89]

Re: Power supply ripple hall of shame

How much it draws when tuned for max smoke?

With the 191W load above it draws 286W so around 66% efficiency.

[momaka]

Re: Power supply ripple hall of shame

The measured ripple is around 1700mV, or 1.7v!
...
https://www.badcaps.net/forum/attach...1&d=1527188421

Nice signal generator, that power supply!

Unfortunately, I don't have a scope to check the ripple out on some of the crappy gutless PSUs in my collection, but this hunk-o-junk L&C LC-B300ATX probably takes the cake. It was used in a Pentium II computer way back when I didn't know any better about crappy power supplies. Essentially, that Pentium II PC kept crashing over and over in certain heavy-load scenarios (mostly in games). But the PC wasn't power-hungry at all, so it never even crossed my mind that this "300W" PSU could be the reason why (that Pentium II PC was very basic, with only a single HDD and a slow S3 Savage 4 video card). After nearly pulling all of my hair out and swapping various components, I eventually "upgraded" back to the ANS L&C LC-235ATX (also posted in the above link). This made the crashing problems disappear once and for all. ... or at least that's the case until I was given a "new" mobo to "upgrade" with (a Pentium III 933 MHz with 384 MB of RAM and Radeon 7000 video card).

So for that Pentium II PC to have been crashing under load like it did, my guess would be that this LC-B300ATX PSU probably couldn't provide even 30 Watts of clean power on the 5V rail. I eventually "improved" the LC-B300ATX (mainly added all of the missing caps, removed some load resistors and added some input filtering.) But by that time, I had a much better looking (well, relatively speaking ) Deer DR-250ATX PSU, which I partially recapped with Japanese caps. So eventually the 250W Deer PSU is what I left with the newly-upgraded Pentium III system and still is with it today, working fine. Now even with 3 HDDs and an upgraded GPU (Radeon 9600), the system is rock-stable. The JEE primary input caps (470 uF) in the 250W Deer still have decent hold-up time too. JEE, I am actually surprised, given their reputation.

[eccerr0r]

Re: Power supply ripple hall of shame

It's bad when you can see the ripple when the scope is DC coupled...

[PeteS in CA]

Re: Power supply ripple hall of shame

That ripple is so high and repetitive I'd be tempted to say it was down to oscillation in the controller (caused by bad caps) rather than the caps themselves contributing to it.

...

The time scale is 1uS/div, so probably not. Were it 100uS/Div that would be a reasonable guess. And you are correct in suggesting that output caps having gone high ESR could cause oscillation.

[PeteS in CA]

Re: Power supply ripple hall of shame

It is necessary when measuring ripple & noise to be sure that spike noise is "real". If you use a ground wire with your probe it may be acting as an antenna, picking up radiated noise. To see if that is happening, connect both your probe ground and your probe tip to PSU output ground. If you see nothing, you're golden; if you see spikes, then part of what you measured was not real.

I didn't learn this the easy way, BTW.

[Per Hansson]

Re: Power supply ripple hall of shame

I like Per's scope meter!

Me too, thanks goes out to Fluke Sweden for it

The time scale is 1uS/div, so probably not. Were it 100uS/Div that would be a reasonable guess. And you are correct in suggesting that output caps having gone high ESR could cause oscillation.

Yea, here is a shot after recapping when it's connected up to the equipment it supports. (So not running idle like in the first post).
Here the operating frequency of the controller can be derived, do note that the time scale is 5µS here instead of 1µS as in the first photos.
The originals where Nichicon VX 85°C, I replaced them with UCC KMG 105°C caps.
The controller, a LM2940OCT could have used low ESR capacitors according to the datasheet.
But I did not look at the datasheet until after the job was done, anyway, system seems to run fine with that amount of ripple, it is after all around 10x less than was there before

It is necessary when measuring ripple & noise to be sure that spike noise is "real". If you use a ground wire with your probe it may be acting as an antenna, picking up radiated noise. To see if that is happening, connect both your probe ground and your probe tip to PSU output ground. If you see nothing, you're golden; if you see spikes, then part of what you measured was not real.

I didn't learn this the easy way, BTW.

Yes, this is very good advice.
I've been bitten also by the fact that the Intel design spec for ATX PSU's specify to use a 0.1uF ceramic and 10uF eletrolytic capacitor at the probing location.
I never bothered to and usually that works fine, if you measure with a load.
But in one case it was without a load, and the scope was seeing some HF ripple due to that which would have disappeared with any sort of equipment hooked up, so that's worth keeping in mind as well...

[stj]

Re: Power supply ripple hall of shame

does that series of scopemeter still use 4 C cells??

[Per Hansson]

Re: Power supply ripple hall of shame

No, it uses a NiMH battery pack called BP190

[CapLeaker]

Re: Power supply ripple hall of shame

Here is a link to a KB Power Supply 120-12-8 that I had recently on my bench. It will rock your socks, promise.
https://www.badcaps.net/forum/showthread.php?t=71608

[Per Hansson]

Re: Power supply ripple hall of shame

This is from a 12v to 5v DC-DC PSU that was installed on an I/O board for a machine.
I had some fun with it today, it's sensitive to both heat but then I also noticed physical shocks.
I can make it do all kinds of weird things by just poking it
The nicest ripple is the one in the big attached picture, about 0.7v
Interestingly the 5v output is only at 2v or thereabout when this happens.

It's bad when you can see the ripple when the scope is DC coupled...

That's a great idea for this thread, when the ripple shows up DC coupeled you know you have a problem that is worthy

[Per Hansson]

Re: Power supply ripple hall of shame

So I got an identical PSU to the one in the first post above.
This one had fared far worse though: the caps either read in the picofarad range or ESR was 0.14K which is the highest my ESR meter can display.
The ripple is around 11.6v on the +12v line!
Yup, I think we have a winner!
I forgot to DC couple however I think you can imagine the result!
With new caps the ripple is completely gone!



With the above I for sure thought I had found the worst possible ripple, I mean how could it be any closer to the actual rail voltage?
But the mainboard this PSU was powering had failed, now my plan is to replace it if I can find something in time.
But customers always want the machines fixed yesterday, so I thought I have nothing to loose: I'll give it a shot.
An N-Channel MOSFET was shorted & popcorned, as was 2x ceramic caps.
I replaced the Si4410DY with a ME4410A that I scavenged from an old laptop mainboard, along with 2x 100nF caps from the same board.
I then started it up but there was no display.
So I started checking things, of course the silly 30mm CPU fan had seized so the board had gotten quite hot, I checked the caps and they seemed to read ESR fine in circuit.
However a cursory ohms check gave very low readings across C40 so I decided to desolder it and check it on my meter: sure enough it reads as a 40ohm resistor on my ESR tester so it is shorted out!
So in its place I put a Samxon GC 2200uF 6.3v cap, mainly because I'm a cheapskate and don't want to sacrifice a nicer cap on a board that might never work (these Samxon caps are around 10 years old now)
I fired it up and to my surprise it went through POST with no issues!



I took some cursory readings with my DMM and everything looked normal except on C36, the meter showed something like -160VDC
I thought obviously that is completely infeasible, must be a bug in my UT61E?
So I was about to bring out my nice Fluke but then decided for the old trusty scope as I anyway had it setup for checking the RGB output on an Amiga 600 I'm fixing.
To my amazement the UT61E was not wrong, the ripple is in the range of 50v!!!



Now obviously the UT61E has some problems to understand such a waveform so it shows a higher value, but still damn!
So I desoldered that cap too and put in another Samxon GC 2200uF 6.3v and the ripple then was an order of magnitude lower!
The system also drew 0.3A less power I could see by coincidence since the bench PSU was in the frame of both pictures.
The crazy thing is that the CPU survived this, it is an Intel Tillamook 266Mhz, they sure don't make em today like they used to!

[momaka]

Re: Power supply ripple hall of shame

The ripple is around 11.6v on the +12v line!

Ripple?!? More like tsunami waves!!

You know how they say don't connect a speaker to a DC source for a long time, because it will burn out? Well, with this PSU, I think your speaker will be just fine, if not a little "buzzy".

The crazy thing is that the CPU survived this, it is an Intel Tillamook 266Mhz, they sure don't make em today like they used to!

Nice!!