Power supply ripple

[JuanGg]

Hello everyone,
I have been "putting together" a lab power supply with one of those DPS3005 modules and a 24 V 10 A power supply that goes with the model number S-240-24. I wanted a floating one so I don´t blow anything up probing around with the oscilloscope. I tested it with the multimeter, all looked as expected. Wired it all up the best I could and made case for it (see attached photo).

Then I hooked it up to the scope to measure its noise and what was my surprise when I found 180 Vpp (volts, not milivolts) ripple on the output. I used the "poor-man's diferential probe" and I don't really know how to interpret the results (see screenshot) Is this normal on a floating output?

Later, I referenced the supply to mains earth and measured the noise, this time with a single probe, and I found almost 0.5 Vpp noise! I knew some was to be expected, it being a switching power supply, but half a volt seems like a lot. (see other screenshot). I tried with the low inductance springy thing, but it made no difference.

Is there any way I can fix this, maybe a filter of some sort? Or is it just a faulty power supply?

Thanks in advance!

[RobertHolcombe]

Curious; have you tried measuring the output ripple with a load connected?

[ArthurDent]

It sounds like from what you were saying that initially you were using the 24V/10A SMPS without the necessary ground (earth) connected, which is not safe, and that probably accounted for the 180 volt capacitor coupling from the mains to the output. If you had probed the SMPS metal chassis with the scope with SMPS earth disconnected you'd probably have gotten about the same reading. 

If you try measuring the ripple or noise with a handheld DMM set to the most sensitive A.C. range you will probably find that the output from either the SMPS or the P.S. module will be much lower than 0.5 volts. Most good DMMs will have a wide enough frequency response to read higher frequency noise as well as ripple. As suggested, see if there is a difference if you put a load on the supply. It is sometimes hard to prevent induced loop currents in scope probe ground leads from giving false indications when the scope and the supply are both using a common ground. That is not to suggest that you should run either the supply or scope ungrounded which could represent a safety hazard.     

Oh, nice job on making the case, but I prefer metal to help keep the noise at a minimum. A thin layer of metal foil glued and tacked to the inside of the wood case wouldn't hurt. Just a thought.

[JuanGg]

First of all, thank you for the fast responses!

Curious; have you tried measuring the output ripple with a load connected?

Yes, I have tried with various resistors, and it did not seem to make much of a difference, which makes me think of noise pickup on the scope probes maybe?

It sounds like from what you were saying that initially you were using the 24V/10A SMPS without the necessary ground (earth) connected, which is not safe, and that probably accounted for the 180 volt capacitor coupling from the mains to the output. If you had probed the SMPS metal chassis with the scope with SMPS earth disconnected you'd probably have gotten about the same reading. 

The SMPS earth was connected all the time, and its case was grounded (the multimeter reads a direct short between mains earth and the metal case, as expected) . Probing the case with the scope results in 200 mVpp noise around ground. Both the scope and the power supply are conected to the same power strip.

The 24 V output of the SMPS is suposed to be floating (no direct short to mains earth) and goes directly to the P.S. module. The output of the module is then suposed to be floating, so if I want it to be mains earth referenced, I would join the negative output of the module to mains earth (black binding post and green binding post). I cannot do that before the module as it has a low side current shunt.

If you try measuring the ripple or noise with a handheld DMM set to the most sensitive A.C. range you will probably find that the output from either the SMPS or the P.S. module will be much lower than 0.5 volts. Most good DMMs will have a wide enough frequency response to read higher frequency noise as well as ripple. As suggested, see if there is a difference if you put a load on the supply. It is sometimes hard to prevent induced loop currents in scope probe ground leads from giving false indications when the scope and the supply are both using a common ground. That is not to suggest that you should run either the supply or scope ungrounded which could represent a safety hazard.

Measuring the ripple with my UniT 61E multimeter on the ac range, it reads about 10 mVpp on both the SMPS and the module's output. Maybe it is just those loop currents on the scope probes. Both the supply and the scope are conected to the same power strip. I understand the importance of the ground conection.

Oh, nice job on making the case, but I prefer metal to help keep the noise at a minimum. A thin layer of metal foil glued and tacked to the inside of the wood case wouldn't hurt. Just a thought.

I know, wood was all I had at hand. I thought about using some copper tape or similar inside the case and ground it. I will try it as soon as I can.

I also found that putting a 100nF ceramic cap between earth and the negative output of the SMPS cuts the 180 V swings to about a volt and decreases the noise from 500 mVpp to about 40 mVpp. Is that a valid solution?

Thanks again and sorry for the lengthy post.

[xavier60]

It looks to me that the probe's ground lead is open circuited somehow.

[JuanGg]

It looks to me that the probe's ground lead is open circuited somehow.

You mean it is not making a proper conection? I am using brand new probes, that would be strange.
I checked it: 0.02  from the ground clip to the BNC.

[xavier60]

At what places have you tried connecting the ground lead?

[ArthurDent]

Your previous post wasn't too long, it was complete and helped answer some questions I had. The steps you have taken so far seems very logical to me and The final bit about adding a 100nf cap from neg to ground is a good idea and I would probably add a second from pos to ground as well. Sometimes chasing an elusive problem like this is as much magic as it is science - or maybe they are the same and we just don't recognize it. 

I had a couple of other thoughts after I posted last night. The wooden case will also insulate everything thermally. Both the module (with the fins) and the supply will be generating heat and it is important to be able to get rid of that heat in some way and the top vent slots probably won't be enough. You may find that a metal case will be needed after you do some more testing but in the meantime you may want to add a fan and add more slots on top and sides. A metal case might help with noise problems as well

Also It looks like the A.C. line goes directly to the 24V supply input terminals and I would add a switch, fuse, and LED. Even though the module will be displaying when powered I still like lots of LEDs. There is generally a fuse inside the supply but it isn't that easy to get to so a smaller value external fuse that will blow first would be a good idea.   

[JuanGg]

At what places have you tried connecting the ground lead?

First, I removed ground leads from two probes, attached them to the outputs of the power supply and substracted them on the scope (emulating a diferential probe). Later I attached the ground clip of one probe to the negative output and the probe to the positive, earth referencing the supply.

Your previous post wasn't too long, it was complete and helped answer some questions I had. The steps you have taken so far seems very logical to me and The final bit about adding a 100nf cap from neg to ground is a good idea and I would probably add a second from pos to ground as well. Sometimes chasing an elusive problem like this is as much magic as it is science - or maybe they are the same and we just don't recognize it. 

I am happy to hear that the cap is a good idea. I was a bit worried it would not leave the power supply completely floating, but it should only kill the noise a bit I suspect. Will try adding another one to the positive rail. For me half of it seems like magic, still learning the fundamentals...

I had a couple of other thoughts after I posted last night. The wooden case will also insulate everything thermally. Both the module (with the fins) and the supply will be generating heat and it is important to be able to get rid of that heat in some way and the top vent slots probably won't be enough. You may find that a metal case will be needed after you do some more testing but in the meantime you may want to add a fan and add more slots on top and sides. A metal case might help with noise problems as well

Sure, I just wanted to do some testing on it so I could see how much heat it produced. Also, the SMPS is rated for 10 A and I am only going to use a maximum of 5 so it should not heat up much. I will add some move vents and maybe a temperature controlled fan on the back panel. Will see if I can get my hands on some aluminum sheets...

Also It looks like the A.C. line goes directly to the 24V supply input terminals and I would add a switch, fuse, and LED. Even though the module will be displaying when powered I still like lots of LEDs. There is generally a fuse inside the supply but it isn't that easy to get to so a smaller value external fuse that will blow first would be a good idea.   

I know, it is just provisional. I ordered one of those panel-mount mains conectors that come with a switch and a fuse, it just arrived this morning, will mount it later.

[JuanGg]

Just added a proper mains conector (see photo). I think I can live with 40-60 mVpp on the output for a general purpose power supply. The module is just a bit noisy I supose, is there any filtering I could add to the output to further reduce the noise?

[JS]

  At what frequency the pulses do appear?
  There is a modulation of the pulses at about 100Hz but I'm asking about between pulses, I'd expect the output ripple of a SMPS have a triangular wave shape with spikes on the corners, plus the modulation seen in your captures.

  Also, check the noise between the AD-DC supply and the converter, you could need some filtering there as well.

  All of those problems should be dealt separately, is strange a 100Hz modulation after two switching stages, so for there I'd add some filtering in between, which could be a big cap, no problem there.
  At the output you want a small cap so it doesn't store too much energy to dump in the load before going into current limiting mode, low ESR preferably there, so MLCC could be a good choice. For it to work properly you will want to add a choke before.

  As you have screw terminals you could place the caps on those to test the setup, before committing to a PCB with fixed caps if that looks better for you. The choke should be added to the brown wire going to the front panel, the cap you could start testing it at the binding posts, and use the banana output to run the measurements. Then for the 100Hz I'd try to place a big-ish electro between black and blue in the front panel module terminals.

  To better know the proper values for the application the freq is needed for the math, and better at a few different loads, try open, 10%, 50% and 90% and document any differences between them.

JS

[ArthurDent]

I don't know if you've seen this video Dave posted but it is interesting.

[JuanGg]

  At what frequency the pulses do appear?
  There is a modulation of the pulses at about 100Hz but I'm asking about between pulses, I'd expect the output ripple of a SMPS have a triangular wave shape with spikes on the corners, plus the modulation seen in your captures.

  Also, check the noise between the AD-DC supply and the converter, you could need some filtering there as well.

To better know the proper values for the application the freq is needed for the math, and better at a few different loads, try open, 10%, 50% and 90% and document any differences between them.

I measured both the output of the module and the switch mode power supply with different loads as sugested (100% is 5A). Please find the measurements below. I tried measuring with two probes and substracting (floating supply), but even measuring the exact same point with both probes, noise was seen and the math fuction displayed diferences between both channels. (see first capture). The remaining measurements were made grounding the negative of the supply via the low-inductance spring (instead of the earth lead).

All of those problems should be dealt separately, is strange a 100Hz modulation after two switching stages, so for there I'd add some filtering in between, which could be a big cap, no problem there.
  At the output you want a small cap so it doesn't store too much energy to dump in the load before going into current limiting mode, low ESR preferably there, so MLCC could be a good choice. For it to work properly you will want to add a choke before.

  As you have screw terminals you could place the caps on those to test the setup, before committing to a PCB with fixed caps if that looks better for you. The choke should be added to the brown wire going to the front panel, the cap you could start testing it at the binding posts, and use the banana output to run the measurements. Then for the 100Hz I'd try to place a big-ish electro between black and blue in the front panel module terminals.

I tried with a 330 uF and 1000 uF electrolitics (what I had at hand that could withstand the voltage) on the output of the SMPS and it made no difference. Also tried several ceramic caps on the binding posts with the same result. Will looping the wire a couple times around a ferrite bead do instead of a choke?

I don't know if you've seen this video Dave posted but it is interesting.

Yes, I have seen it. I tried to measure as he recommends (20 MHz bandwidth limit, low inductance paths...within the equipment I have)

Thanks again to everyone for spending time helping others, and sorry for measurements not being taken correctly or anything else for that matter...still learning how to use the scope.

[JS]

  For me looks like the module is switching at about 70kHz and making the SMPS sag a bit at that frequency, can you see how the dips grow in the SMPS as you increase the load? Also, something similar happens to the module, where with no load there are no spikes, then at 10% load the spikes are there but the ripple ignoring the spikes is very low, and once the load increases more there is some ripple to be clearly seen in the 50mV/div scale which was barely showing off the noise in the 10mV/div scale.

  It might be interesting to see how the SMPS behaves on it self, loaded with a easier load than the SMPS.

  To filter the spikes you need a capacitor that has low enough ESR and ESL to deal with them, it needs to provide 5A, now you have a ripple of 200mVpp which suggests 40 to 20 mΩ ESR+ESL on the output cap, the sharp spikes suggest main problem is ESL. You should try using MLCC SMD caps, as they are shorter and thus have lower ESL. Zooming in even further, to check for the spike itself could throw some numbers in the capacitance and inductance shown in there.

  Then you have the ringing between each spike, that's over the 70kHz, it can almost fit two cycles so about 140kHz, but as it's much lower than the spikes I'd leave them as they are till I get the spikes down. As the spikes are so fast, I wouldn't expect much inductance to be needed to deal with them, a few turns into a core in one of the wires coming from the module to the output should help, then dealing with the ringing it could need something different..

  For the low inductance caps, I think the best way would be a double sided PCB, one side untouched as ground plane, the other with a slot in the middle for the cooper, so the caps go mounted side to side. You get into the PCB with the wires from the module and came the other way with the wires to the output or just mount the output directly to the PCB, holes to screw the banana jacks directly (you might want to have ±outputs side by side and the earth to the negative side as is sometimes done). In the PCB I'd install quite a few 0805 caps, or something similar, rated for 50V but with not much capacity, under 1µ for sure. 10 caps of 220nF would make 2.2µF which is a nice number for the output cap of a lab PSU while getting them distributed in a 50x50mm PCB as described will make seriously low ESL. You could even mix the caps, interleaving some 1nF and 10nF to go crazy with the 2.2nF ones. You don't need as many as the small ones as you are not trying to increase the overall capacitance with them but reduce the HF impedance of the solution.

  In any case, even as fun as this is and informative, you shouldn't expect this SMPS to be the low noise PSU from your lab, but the flexible one for the odd load. Having that output range (V and I) will make it's use inside the lab even being this noisy. I still believe some improvement could be made.

JS

[JuanGg]

  To filter the spikes you need a capacitor that has low enough ESR and ESL to deal with them, it needs to provide 5A, now you have a ripple of 200mVpp which suggests 40 to 20 mΩ ESR+ESL on the output cap, the sharp spikes suggest main problem is ESL...

I zoomed into the spikes and they appear to have a frequency of about 190 MHz (see attached capture).
I added 5x 100 nF and 4x 10 nF on a veroboard on the binding posts and 6 turns around a ferrite core on the positive wire. (see photo). I know it is nor even near what you proposed, just used the materials at hand. I will add some more capacitors I am waiting for, or even make a proper PCB.

The spikes haven't disappeared, but are a third of what they were before:

                Spike's Amplitude
  Load       Before          After
    0%       17 mVpp      16 mVpp
  10%       145 mVpp    50 mVpp
  50%       330 mVpp    100 mVpp
  90%       280 mVpp    95 mVpp

  In any case, even as fun as this is and informative, you shouldn't expect this SMPS to be the low noise PSU from your lab, but the flexible one for the odd load. Having that output range (V and I) will make it's use inside the lab even being this noisy. I still believe some improvement could be made.

No doubt I am learning a lot, and that is what it's all about. Thanks again for the contribution. At the moment it is the only PSU in my "lab" (appart from an Atx PSU with binding posts...) I will eventually make a linear one with a couple LM317s if I need to.

[JS]

  Hey, not too shabby, 1/3rd of the amplitude is not a bad improvement! At 190MHz construction is everything, so with that same configuration and a better construction you can be getting closer. Note that the frequency is well over your scope BW so I'd expect them to be a bit higher than they look but the relation should be pretty much the same.

  Also, if you probe at the load (which is what interests you) the amplitude might get even lower as the inductance of the wires between the supply and the load helps with that.

  If you get around to build it with SMD components and a double sided clad let me know, I'd like to know the actual difference in that.

JS

[JuanGg]

For now I think it is good enough, I will make a proper pcb with smd caps eventually. (And post the results here, for sure)

My scope BW is 50 MHz (so less than 1/3 of 190 MHz). Spikes should be really attenuated by the scope I suppose. I was surprised they even showed up.

I will try probing the load when I have time.

Thanks again.

[JS]

  It just has a filter at 50MHz, so some attenuation at 190MHz but you can unlock to move the filter up to 100MHz... No biggie, you should, specially if you are trying to measure such fast events in SMPS.

JS

[JuanGg]

It just has a filter at 50MHz, so some attenuation at 190MHz but you can unlock to move the filter up to 100MHz... No biggie, you should, specially if you are trying to measure such fast events in SMPS.

Will do that eventually.

The SMPS gave up a while ago, oppened it up and there were no clear faulty components, I don't want to mesh much with its insides anyway...
I have been running the DPS3005 module of a 15 V 1 A wall wart, which is not ideal, but it does the job for now.

I am thinking of building a linear supply with some transformers I have salvaged, have to do more research on that. I have build a dc load that will hopefully help me with that. (https://www.eevblog.com/forum/projects/arduino-based-electronic-load/)

[cdev]

You should put a small fan pulling hot air out of your case as the case is likely to impede airflow through both the original power supply's ventilation holes and the DC-DC converter's heatsink somewhat.

Those DC-DC converters likely need every bit of heat sinking they have got to work. I would also put some vent holes on the bottom, and 'feet' on the case so air can flow upward through the fins, not just from the sides up.

[Simon]

Silly suggestion, have you checked your math function settings on your scope? just make sure it does not have a probe divider set that does not correspond to your probes. Else it may be a differential probe.

[JuanGg]

.  You should put a small fan pulling hot air out of your case as the case is likely to impede airflow through both the original power supply's ventilation holes and the DC-DC converter's heatsink somewhat.

Those DC-DC converters likely need every bit of heat sinking they have got to work. I would also put some vent holes on the bottom, and 'feet' on the case so air can flow upward through the fins, not just from the sides up.   

As mentioned above, the 24 V switching PSU I was using gave up. I got hold of a 22 Vac transformer and I will use that to make a raw PSU for the module. I'll add a small fan to it and make some more vent holes.

Silly suggestion, have you checked your math function settings on your scope? just make sure it does not have a probe divider set that does not correspond to your probes. Else it may be a differential probe. 

I am pretty sure I did. I'll check it again anyway. Setup has changed as I am no longer using the initial switching PSU measurements above were made with.

    Juan

[JuanGg]

I have changed the wall-wart with a linear power supply inside the case. Now I have the full 30 V out and more power available. I have added a -rather noisy- 12 V fan. I may make it temperature controlled.
    Juan

[Jwillis]

It looks like one of those cheaper SMPS .
That "ping" is typical of SMPSs.It's the power MOSFET switching on and off.The weird looking sine wave is also typical for that grade of SMPS .I suspect that the incorrect material is used in the transformer.I get the same results testing various ferrite cores at various frequencies .At the correct frequencies for the core material used the sine wave smooths out and looks correct on the scope.At different frequencies I get the same type of distortion to the sine wave.
I've got a few of these cheap SMPS and they all do the same thing.Better quality ones don't exhibit that kind of noise as bad.