Simple Power Supply with Noise Problems

[Swend]

Hi Friends,

I've got this simple power supply with clean sinusoidal input, but after the bridge rectifier some noise is introduced and I have no idea why, maybe you have some suggestions?

As you can see on the left there is a bog standard ZVS driver running at about 40KHz and two secondary windings with bridge rectifier and a smoothing cap on the output..



Here is the output of the secondary winding



Here is the output noise of the positive side



Here is a closer look at the peak noise

[andre_teprom]

It seems a perfectly plausible artifact considering the presence of inductors and capacitors at the output stage performing a tank circuit capable of sustaining an oscillation during the recharge of the capacitors. By the way, it would be more clarifying if you added some resistive load to the output, so you would see how much that ringing 'noise' would remain in the face of the peak current in the capacitor recharge (ripple) in the secondary side of transformer.

[Swend]

Hi Andre!

Thank you for your reply.

It seems a perfectly plausible artifact considering the presence of inductors and capacitors at the output stage performing a tank circuit capable of sustaining an oscillation during the recharge of the capacitors.

So you are saying this is normal? Is there anything I could try to diminish it? e.g. larger caps, other diodes?

By the way, it would be more clarifying if you added some resistive load to the output, so you would see how much that ringing 'noise' would remain in the face of the peak current in the capacitor recharge (ripple) in the secondary side of transformer.

Yes you are right, I used a resistive voltage divider (1Meg/1K) to measure the output voltage, would that qualify as a resistve load in this context?

[andre_teprom]

Even for 600V_dc, 1MΩ is almost no load in the context of a power supply (360mW) for that magnitude. Add something not greater than 10KΩ at the output stage, or even smaller; this would make waveforms more realistic during circuit working under normal operation. Anyway, I would agree with you that increasing capacitances would somehow attenuate these spikes.

[asdf44]

The noise is at 40khz and your switching is at 40khz so it's switching noise correct?

First figure out of its 'real' by tying your scope probe to ground and see if its scope noise or common mode pickup.

If not then try more/better capcitors, diode snubbers, and/or a post LC filter.

[Easy peasy]

Yup the noise is at 40kHz intervals - so it is switching noise - most likely from the o/p diodes - put snubbers across all.

- - - Updated - - -

yes - those diodes are 300nS recovery - so they carry reverse current when they turn off - they are really designed for lower frequencies than 40kHz, I see you have tried to mitigate this with a sine wave source - however they will need snubbers - or use faster & better 600V diodes - now with smaller snubbers ...

[Swend]

The noise is at 40khz and your switching is at 40khz so it's switching noise correct?

First figure out of its 'real' by tying your scope probe to ground and see if its scope noise or common mode pickup.

If not then try more/better capcitors, diode snubbers, and/or a post LC filter.

Hi asdf44!

The noise is 'real' - grounding does nothing.

I tried doubling up on the caps - does nothing special.

Changed diodes to UF1007 which have 75nS Trr - that reduced noise amplitude with about 60% at 1Mohm load.


I will try with some diode snubbers now, would a couple of nF be fine?

- - - Updated - - -

Yup the noise is at 40kHz intervals - so it is switching noise - most likely from the o/p diodes - put snubbers across all.

- - - Updated - - -

yes - those diodes are 300nS recovery - so they carry reverse current when they turn off - they are really designed for lower frequencies than 40kHz, I see you have tried to mitigate this with a sine wave source - however they will need snubbers - or use faster & better 600V diodes - now with smaller snubbers ...

Hi Easy peasy!

I did as you and asdf44 suggested, please see my post #7.

[Easy peasy]

at a guess 330ohm and 220pF ( 1 or 2kV ) across each diode.

a small LC post filter will do a lot to remove noise too ... esp for a low C choke ...

[Swend]

at a guess 330ohm and 220pF ( 1 or 2kV ) across each diode.

a small LC post filter will do a lot to remove noise too ... esp for a low C choke ...

Hi Easy peasy!

On the diodes, tried with 1nf up to 22nf - but could not register any significant change.

LC filter, in the uF HV range I only had a 50uF/800V so I tried that in combination with several different coils just to see if any change could be registered e.g. 50uF/10mH I get Fc=7KHz, but no significant change could be registered.

[asdf44]

Wait, let's be really clear. When you say "The noise is 'real' - grounding does nothing" does that mean you still saw the noise after grounding the probe tip?

I meant you should leave your probe ground connected and clip or touch the probe to its own ground lead. This should obviously show a perfect 0V. If you still see noise then its scope pickup and isn't actually on your output. The scope pickup may be worth trying to solve...but its a different problem.

[Swend]

Wait, let's be really clear. When you say "The noise is 'real' - grounding does nothing" does that mean you still saw the noise after grounding the probe tip?

I meant you should leave your probe ground connected and clip or touch the probe to its own ground lead. This should obviously show a perfect 0V. If you still see noise then its scope pickup and isn't actually on your output. The scope pickup may be worth trying to solve...but its a different problem.

Hi asdf44!

I think I misunderstood what you said, I thought you meant 'ground' as in 'earth'. My mind was elsewhere as I have been blowing up half a dozen of scope power supplies while working with HV, the galvanic isolation was non-existent (but now I'm using a separation trafo).

Anyway I did as you said, and the noise is now 'unreal'. I still see noise, bit different wave form but same amplitude, when connecting the the probe to it's ground clip.

[Easy peasy]

LC filter, in the uF HV range I only had a 50uF/800V so I tried that in combination with several different coils just to see if any change could be registered e.g. 50uF/10mH I get Fc=7KHz, but no significant change could be registered.

this likely means that you are picking up only CM noise on your scope - taking noise measurements at low levels is an art, try looping your scope probe as many times as you can thru a high mu toroid - this will reduce CM pickup ...

using caps only as a snubber is counter productive ...

- - - Updated - - -

for a post filter you need a low capacitance choke - say 100uH, and a low ESR/ESL cap ( i.e. no long leads ) to get attenuation at the noise freq, say 220nF, film/foil.

[Swend]

this likely means that you are picking up only CM noise on your scope - taking noise measurements at low levels is an art, try looping your scope probe as many times as you can thru a high mu toroid - this will reduce CM pickup ...

"taking noise measurements at low levels is an art" - yes indeed, and taking measurements at high levels is also an art of a different kind, and performing both at the same time is black magic I have a little hi mu toroid, so I will try your suggestion.

for a post filter you need a low capacitance choke - say 100uH, and a low ESR/ESL cap ( i.e. no long leads ) to get attenuation at the noise freq, say 220nF, film/foil.

Will try that too, thank you very much. This calculator says Fc=34KHz with the suggested values (R=10k) would that be correct?

using caps only as a snubber is counter productive ...

In this context or generally? because I see many technical documents with caps only 'snubbers'.

[Easy peasy]

1) "at low levels of noise" esp in the presence of CM radiation

2) not good to believe everything you stumble across on the internet -

3) generally, and refer to 2) above.

[treez]

i think OP has problem with "pickup" in his scope lead, as many others here have suggested.
You can use a coaxial probe (you have to make it yourself), which doesnt have the ground clip.

When i worked at a major telco company, 4 graduates were given a "tail chasing" job where they were asked to solve the pickup problem on a power supply .......they were led to believe that the pickup was noise that was "infecting" the microcontrollers...and told to solve the pickup...they couldnt solve it, as they kept using the "dangling gnd clip" probe.

After a full 2 days, the gaffer went back to them and told them it was just "pickup" and that he had just wanted to make them look stupid for 2 days.

[Swend]

Hi treez.

i think OP has problem with "pickup" in his scope lead, as many others here have suggested.
You can use a coaxial probe (you have to make it yourself), which doesnt have the ground clip.

That sounds interesting, but I couldn't find any pertinent refs on the interwebs to the "coaxial probe" other than "open ended coaxial probes" is that what you mean?

[treez]

coaxial probe is just a piece of coaxial cable, with a BNC connector on one end, and the other end you splay out to get you the ground and centre conductor to solder to your circuit.

You usually have to buy it as a pice of 1 metre (or so) coax with a BNC at each end...you then cut one BNC off, and then roll the gnd conductor into a wire so you can solder it.

Its a home made kind of thing

[Velkarn]

to remove noise pickup use probe with spring https://electronics.stackexchange.co...f-oscilloscope

[Swend]

to remove noise pickup use probe with spring https://electronics.stackexchange.co...f-oscilloscope

Hi Velkarn

Thank it's nice, perhaps that was also the intention with treez's coax suggestion i.e. minimizing ground loop.