Help getting rid of ripple from SMPS

[Benedict]

I'm using two RS - 100 -12  switchers as PSU's for my linear speaker amp in series for + and - rails (no need to ground outputs - safety standards good to go = class II = no hum; + high efficiency, industrial grade etc). (http://www.meanwell.com/mw_search/RS-100/RS-100-spec.pdf )
It has a rated 120mV ripple. On my scope it's more like 100 mV both at load and idle. The amp doesn't amplify it like it does ground loops (and it shouldn't).

It's VERY audible. I've analyzed the differential  reading on the scope, listened to the ripple on speakers and compared it to arbitrary frequency generator by ear, and recorded and analyzed the ripple (from speakers) using the microphone and PC. All three findings conclude it's around 4040 Hz and 120 Hz ripples, in around 100 mV range. Next ripple/noise is in Mhz - not an issue.

I have no real idea how to eliminate it. I've tried decoupling it with 15pF caps, but it didn't work... I never really had any need to remove such small ripples before, and I'm an ophthalmologist (electronics enthusiast) so it's not really my field. I do know about safety, I'd say enough - all my equipment is either earthed or class II.

My next bet would be using some sort of a LC circuit to try and suppress the ripple, or maybe linear regulators though I'd very much like to avoid those.

Any suggestions? Pls

[Kalvin]

Capacitance multiplier might do it as the noise frequency is at quite low frequency:
http://sound.westhost.com/project15.htm
Adding an additional LC-filter doesn't hurt as it will filter the higher switching frequency components.

[Benedict]

Thanks! That was a good read and good info. My only real problem is that my amp has a minimum voltage req. of 10 volts +/-, and these darlingtons drop 3 volts (or so they  say), putting me at 9 volts, which is a bit too low. If that can be solved, should I apply heatsinks (just read it - yes, I should. not big though)? Also, would it be smarter to add the circuit to the amp itself or as close to the PSU as possible?

[Kalvin]

If you are lucky, the power supplies may have a trimpot inside which allows adjusting the output voltage.

Ps. Don't do anything stupid, like kill your self. If you do not know what to do, you may get electrocuted, which is not a good thing.

[Benedict]

Hahahah I wont! There is a trimmpot there, I just assumed it's for fine tuning

[nickm]

How much power is your amp? That power supply is 100W and i bet you're listening at  less then 10W so i'm guessing the power supply is skipping cycles to increase efficiency.  It might also be adjusting its frequency instead of its duty cycle to regulate (real popular with the cheap high power supplies).  Either way an lc is going to require a huge L and huge C to be effective at 120Hz.  Good luck.

[Zero999]

Do you have a schematic of the amplifier?

The power supply rejection ratio of the amplifier should be high enough to stop the ripple from passing through the amplifier.

[Benedict]

Amp has a tdp of 125 W. You are certainly correct that it's not above 10w for listening, but I figured I'd give it it's rated power :/

I don't have the exact schematic but its just dual lm3875 and a few components

[Zero999]

The power supply rejection of the LM3875 us above 50dB, over the audio range and above 65dB at the frequencies you're dealing with. 65db is 1778, meaning the noise on the power supply will be reduced by this factor before appearing on the output, so 120mV will become 120×10^-3/1778 = 67.5×10^-6 = 67.5µV which should be inaudible. Either something is wrong with the design or it's the RF generated by the SMPS which is being detected and amplified.

See figure 34 of the datasheet.
http://www.ti.com/lit/ds/symlink/lm3875.pdf

[awallin]

I'm using two RS - 100 -12  switchers as PSU's for my linear speaker amp in series for + and - rails (no need to ground outputs - safety standards good to go = class II = no hum; + high efficiency, industrial grade etc). (http://www.meanwell.com/mw_search/RS-100/RS-100-spec.pdf )
It has a rated 120mV ripple. On my scope it's more like 100 mV both at load and idle. The amp doesn't amplify it like it does ground loops (and it shouldn't).

a scope view of the waveform would probably help.

the usual problem with SMPS is common-mode noise:
http://www.murata.com/~/media/webrenewal/products/emc/emifil/knowhow/26to30.ashx

The filtering solutions are quite different for single-ended vs. common-mode noise. Either common-mode chokes or caps across the isolation barrier to case-ground are suggested to suppress common-mode.

With a scope, try probing case-ground vs. +output, vs. -output, and vs. Common-rail. If it's common mode noise from the SMPS the common rail might have noise from both supplies, perhaps at different frequencies and probably not synchronized to eachother.

[Benedict]

You are absolutely right, was just reading the datasheet right now too, that same thing. But if it's noise collected from the power supply via rf, it should be amplified, but it's constant. Really odd. Might have something to do with the wiring inside the case. Since I can't really "rewire the pcb" my only option is to smoothen the dc input

It's not common mode noise I'm afraid, it's there even if I disconnect the ground, and it wasn't there with linear psu. There is ofcourse common noise on all 3 rails referenced to ground but it's not there differentially so it shouldn't matter

[Zero999]

Try powering it from the linear PSU again, then turn on the SMPS nearby and see if the noise appears. If it doesn't, then try connecting one of the power supply rails to the output of the SMPS.

[Benedict]

Powered it from linear with smps running, no noise.

Differential of + and gnd psu  outputs when not plugged in mains:


Powered on, no load


Powered on, loaded with amp


Same thing on speaker inputs

[Benedict]

The wide band on the first picture is noise in Mhz regions - no issue
Anyways I managed to set the voltage with trimpot to 13,8 volts, so I have some room for linear regulation or that capacitive multiplier up there. Any recommendations? I'd need some hefty regulators to withstand around 9 amps, I KNOW there will never be such a draw in reality but I want it done properly for the amp :/

I forgot... A huge thank you to all folks

[MagicSmoker]

The power supply rejection of the LM3875 us above 50dB, over the audio range and above 65dB at the frequencies you're dealing with.
...

This is exactly the point I was going to make... While MeanWell switchers are just one step up from the bottom of the barrel, the LM3875 should handily reject the 120mVpp of noise/ripple from them. There is something else going on here - most likely layout related, but I wouldn't rule out marginal components (particularly capacitors) or incorrect wiring.

...
Powered on, loaded with amp


Same thing on speaker inputs

It's kinda hard to tell with the fuzziness, but it sure looks like the ripple shown in the first scope shot is making it into the second, which implies capacitive or E-field coupling. The actual schematic and pictures of how it was constructed would be helpful here.

[Benedict]

I don't have a schematic, but it's a wiring mess on the inside. It isn't a problem when the rails are clean of ripples, as there is no noise to couple and jump then..

So I've got 3 options, redo the entire amp myself with proper wiring, ditch the nice metal box, or 

a) find LT1083/4/5-12 somewhere, use 2 of those - I'm not ecstatic about their price and availability... also seems that their ripple rejection seems to be bit low from the charts, even as low as 20db or lower? nope
b) do the capacitance multiplier circuit from the 1st post in the PSU itself - should work,  have only one question about it:

Am I correct in thinking that it's only important that TIP35 and TIP36 transistors handle up to 25 amps, and BD139/140 are merely driving them so they don't have to handle the load currents?

[Kalvin]

b) do the capacitance multiplier circuit from the 1st post in the PSU itself - should work,  have only one question about it:

Am I correct in thinking that it's only important that TIP35 and TIP36 transistors handle up to 25 amps, and BD139/140 are merely driving them so they don't have to handle the load currents?

The main current of 9A will pass through TIP35/TIP36, and you will definitely need a heat sink for them. The Low frequency current gain is around 30 - 40, so the base current is approx. 9A / 30 = 300mA or less. This 300mA will flow through the BD139/BD140 so I would prepare to use small heat sinks for those too. Of course you are not pushing the 9A all the time, so you possibly can get away with a easier thermal design requirements. I would just test the concept first and see whether the capacitance multiplier gives good results before buying the actual heat sinks. Ps. The 13.8V should give nice margin.

[Benedict]

Great! Much obliged! Manythanks! Will report results.

[Benedict]

 

It's DEAD SILENT! :O

I did the capacitance multiplier thing and the speakers are dead silent even at 120-ish watts of amplified "no signal".  (playing some music on 120 watts would probably tear my house apart and eject me through the window behind me). I've got no ground loops, even when the case and speakers are connected to earth. Such happiness. Thanks everyone for your help! Some pics incoming:











Speakers, amp, PSU and filterbox are handmade.