Mysterious noise from ICs

[Potato Cakes]

Hello, everyone,

For about a year now I have been working on a project that integrates what is essentially a 1073 with additional BA283AV cards, IC differential amps, and a headphone amp. I have written about it here, mostly asking for power supply advice. I thought I had everything thing sorted but now I have a new issue where the ICs (both in the headphone amp and the differential amps) are generating what I can only describe as the sound of the inside of a commercial airliner in flight bandpassed between 1kHz and 3kHz. It isn't pink/white noise nor is it a high noise floor. I've never heard this sound before when working on electronics. When signal is run through the unit there is plenty of headroom and everything sounds as it should, and the mystery noise disappears (perceivably) and then comes back when the inputted signal is stopped. I am using SMPS's (supposedly medical grade), one for the 1073 circuit and one for the ICs. I have exhausted the combinations of who is powering what and how, but I cannot seem to get the noise to go away. I have an earlier version prototype that doesn't have a headphone amp but is powered with the same kind of SMPS and does not have this issue.

On a side note both units are using the same PCBs, but the one without the headphone amp uses a 3x24 gain switch and the other uses a 3x12. The one without the headphone amp buzzes if 0V is connected to ground and the other buzzes without. Both are using the same audio transformers. So I'm fighting not only the PSU/noise issue but also trying to figure out why the grounding is the opposite.

The worst part is that I did have this all sorted, no mystery noise, no buzz, and great sounding audio, but when I put the lid on it and attempted to put it use the noise returned and I haven't been able make it quiet again. This was probably too long winded of an explanation, but my question is has anyone else experienced this type of noise with ICs? If so, what seemed to be culprit?

Thanks!

Paul

 

[abbey road d enfer]

I am using SMPS's (supposedly medical grade), one for the 1073 circuit and one for the ICs.

Are you sure both smps deliver enough current? I believe you are aware of the "hiccup" mode that happens when an smps is too lightly loaded.

 

[Bo Deadly]

generating what I can only describe as the sound of the inside of a commercial airliner in flight bandpassed between 1kHz and 3kHz. It isn't pink/white noise nor is it a high noise floor.... I am using SMPS's (supposedly medical grade), one for the 1073 circuit and one for the ICs.

As Abbey has already said, it's not the ICs. A whistling noise in 1-3kHz is almost certainly the SMPS modulating the output because it is not loaded enough (also known as "hiccup" mode).

Post specifics about the SMPSs. You need to load an SMPS with at least 20% of it's rated current. This is why it's important to use an SMPS that's not too big. If it's not too much larger than it needs to be, you can just add load resistors directly on the outputs of the SMPS.

More specifically crank the pre with the input terminated and listen for the noise on your monitors (or better use spectrum analyzer software). Then momentarily try different load resistors until you can find the smallest value that makes the noise go away. Whatever that value is (should be like 500 or 100 ohms or so depending on the max current of the SMPS), decrease that by another 20% just to be safe, compute the power it's dissipating and then get the right part and find a way to mount it and integrate it. If it turns out that the resistor would have to be too big, you should just buy a new SMPS that is appropriate for your project. Repeat for the headphone supply.

 

[abbey road d enfer]

"Then momentarily try different load resistors until you can find the smallest highest (minimum current)value that makes the noise go away. Whatever that value is (should be like 500 or 100 ohms or so depending on the max current of the SMPS), decrease that by another 20% just to be safe, compute the power it's dissipating and then get the right part and find a way to mount it and integrate it. "

 

[Bo Deadly]

"Then momentarily try different load resistors until you can find the smallest highest (minimum current)value that makes the noise go away.

Sorry for not being clear. Right the smallest resistor value that is just small enough to make the noise go away. So start with 1K and then 470 and then 220 and increasingly smaller until the noise goes away.

But of course the power of each resistor must be considered. You wouldn't want to put a 100 ohm half watt resistor across a 24V supply. It would probably pop instantly.

 

[jokeramik]

Is ist possible for you to post a oscilloscope image?
Best regards!
jokeramik

 

[Potato Cakes]

I would say that I am definitely having issue with note loading it down to at least 20%. The 48V SMPS is 2.4A and the 36V (for the headphone amp) is 1.67A. The reason I went so high in the first place is that I had used a 48V 0.57A with this circuit but it kept power cycling, causing me to believe it was loaded down too much but it didn't make sense because the 1073 plus 2 more BA283AM cards is about 200mA. Thinking about it now, there are three separate 1000uF reservoir caps for the two AM cards and the one AV card, but I have mistaken those caps causing too much of a load on the SMPS. I had a similar issue with an 8 channel JLM Baby Animal preamp that I built and Joe Malone helped sort me out by having me reduce the reservoir cap values on each of the cards. This is why I went with a much more higher rated than necessary SMPS so that I wouldn't run into this trouble but it seems that I inadvertently caused a new problem.

The 1073 part of the circuit does not have this whistling issue. It actually sounds really incredible. Regardless, I will swap out for some lower current rated power supplies and or load down the V+/- out and report back.

And thanks for the reminder on paying attention to heat dissipation. I can almost guarantee that I would not have remembered and would have definitely blown something up.


Thanks!

Paul

 

[Audio1Man]

Run an FFT on the power supply voltages with and without a test tone signal. Then again run an FFT on the output jack with and without the test tone. Post the graphs
Duke

 

[Bo Deadly]

The reason I went so high in the first place is that I had used a 48V 0.57A with this circuit but it kept power cycling, ...Thinking about it now, there are three separate 1000uF reservoir caps for the two AM cards and the one AV card, but I have mistaken those caps causing too much of a load on the SMPS.

Yup, the 2000uF of capacitance will draw too much current for too long and make the SMPS short circuit protection kick in. These modern SMPS are pretty fancy gizmos. If you try to draw too much current or too little, it does special things.

There are two solutions to stopping the short circuit protection from kicking in:

A) Reduce the capacitance. You probably don't really need anywhere near as much filtering as a linear supply because the SMPS provide a low impedance source with almost no low frequency noise (there should be literally no measurable mains hum). Any little bit of high frequency switching noise will be snuffed out easily by the simplest of filters.

B) Use a capacitance multiplier. A CM will not only filter very well (and at LF if you really want) but also the RC will turn on the pass transistor relatively slowly and and thereby keep the SMPS short circuit protection from kicking in. Note that you don't need a regulator. An SMPS *is* a regulator. Even the constant current ones put out the specified voltage up to the rated current.

If you already have the whole thing built, then just use option A, get the 0.57A SMPS and drop each 1000uF to 100uF. That should work dandy but I would always verify with a spectrum of the noise floor.

Although I am slightly confused as to why the SMPS is 48V. I thought all the Neve stuff was 24V.

 

[Potato Cakes]

I swapped out for a 1.12A 48V SMPS (https://www.mouser.com/ProductDetail/TRACO-Power/TXM-050-148?qs=%2Fha2pyFadui4o%252BkPTC%252Bq2p1Vil9AsnMB2sK5pdT8L4pZUqMTlAtaEw%3D%3D) but the noise still persisted, so I started adding resistors and got down to 160ohm but the whistling never went away. So then I swapped out for a 0.57A SMPS (https://www.circuitspecialists.com/ps1-25w-48.html) but then ran into the problem with the SMPS power cycling (which I mentioned above), so I wound up having to remove all of the reservoir caps from each board to get it power up properly. However the new issue is that the noise floor (hiss) is higher than the whistling sound. I added back one 1000uF cap but it seemed that that value is too high and the power cycling ensued. Without any reservoir caps, total circuit draw (no headphone amp) is 80mA, so I would figure even with those caps I would be well within the 0.57A rating of the SMPS.

For contrast, the other prototype (exact same number of active components) runs on a 24V 1.5A SMPS (https://www.circuitspecialists.com/hengfu_hf35w-lsm-24_single_output_power_supply.html) and is perfectly quite and uses all of the reservoir caps that I removed in the unit which I am having trouble. So I'm certain that removing those caps on the problematic unit is the reason for dramatically raised noise floor. I also realized that I have some caps that are not quite the correct voltage rating, so I'll have to order some more parts before I can resume this mess.

Thanks!

Paul

 

[Bo Deadly]

So I'm certain that removing those caps on the problematic unit is the reason for dramatically raised noise floor.

That seems odd. I don't know why that filter caps would matter so much. But in that case you could use the capacitance multiplier. Like I said, the CM will slowly turn on the pass transistor so the SMPS short circuit protection won't kick in. Then you can put whatever filter caps you want on there. Between the CM and the filter caps, it should be quieter than the dark side of the moon.

 

[Potato Cakes]

Yup, the 2000uF of capacitance will draw too much current for too long and make the SMPS short circuit protection kick in. These modern SMPS are pretty fancy gizmos. If you try to draw too much current or too little, it does special things.

There are two solutions to stopping the short circuit protection from kicking in:

A) Reduce the capacitance. You probably don't really need anywhere near as much filtering as a linear supply because the SMPS provide a low impedance source with almost no low frequency noise (there should be literally no measurable mains hum). Any little bit of high frequency switching noise will be snuffed out easily by the simplest of filters.

B) Use a capacitance multiplier. A CM will not only filter very well (and at LF if you really want) but also the RC will turn on the pass transistor relatively slowly and and thereby keep the SMPS short circuit protection from kicking in. Note that you don't need a regulator. An SMPS *is* a regulator. Even the constant current ones put out the specified voltage up to the rated current.

If you already have the whole thing built, then just use option A, get the 0.57A SMPS and drop each 1000uF to 100uF. That should work dandy but I would always verify with a spectrum of the noise floor.

Although I am slightly confused as to why the SMPS is 48V. I thought all the Neve stuff was 24V.

All of the transistors on the BA boards can handle 60V, so I'm running it at 48V to simplify the PSU setup (that is the hope) since I need another PSU for the headphone amp which needs the VDC to be isolated (the other PSU has V-/0V tied to ground). The INA134's are powered off of the 48V supply with a rail splitter following drop resistors. This is how I am doing it on the other unit (but with no drop resistors) and it works great. The other reason to give slightly more headroom. I did build an CLC filter but it was in conjunction with some regulators for another purpose that didn't work out, so it looks like I just need to build just the filter. On the schematic for the BA boards I am using, it looks like there are RC filters being used, but those resistors may have been intended for circuit protection. Either way, yet another Mouser order is needed before I can proceed.

Since V- is being tied to ground, would a CLC filter be preferred to CM or does it not matter which one is used in this instance?

Thanks!

Paul

 

[Potato Cakes]

That seems odd. I don't know why that filter caps would matter so much. But in that case you could use the capacitance multiplier. Like I said, the CM will slowly turn on the pass transistor so the SMPS short circuit protection won't kick in. Then you can put whatever filter caps you want on there. Between the CM and the filter caps, it should be quieter than the dark side of the moon.

Nor do I. All I know is that I removed them and the mystery noise is no longer there or it is just being masked by the now very high noise floor.

How high of a uH value for the CM choke would be appropriate?

Thanks!

Paul

 

[Potato Cakes]

Run an FFT on the power supply voltages with and without a test tone signal. Then again run an FFT on the output jack with and without the test tone. Post the graphs
Duke

I'll do this when I get the parts in to fix what I broke.

Thanks!

Paul

 

[Bo Deadly]

How high of a uH value for the CM choke would be appropriate?

Inductor capacitor filters need to be designed to handle the required power and to be damped properly. But generally speaking, the larger the inductance, the lower the frequency of the noise being filtered. But for an SMPS the CM choke is really just to snuff out the high frequency switching noise. So if the switching frequency is 100kHz, pick something that is going to give you good attenuation at that frequency. Use one of the online calculators. Or better, LTSpice. Then just find a part that handles the required current given the space available.

It sounds like you have some relatively complicated requirements with some higher voltage stuff and a virtual ground (if you have the BA283 outputs, a virtual ground is highly dubious) so you will need to post a schematic of the supply with blocks for each device being powered. List the voltage(s) and current(s) required of each block. Draw it on a piece of paper and take a picture with your phone. Be detailed. Show how mains enters, SMPS with model numbers, CM chokes, caps, the virtual ground circuit, how the ground lines run, etc. Otherwise, it's just too complex for me to know what you're doing.

 

[Potato Cakes]

I forgot that I hah parts to do CLC filters and I also had 100uF and 220uF with the correct voltage rating. I also swapped back to the 1.12A PSU. I changed out the 1000uF caps for 220uF and added a CLC filter. No more whistling sound and the noise floor is lowered drastically to what one would expect with modern audio gear.

For the headphone amp, I kept the same PSU and it no longer had the whistling noise from before, so that was a little bit confusing but I am just glad that everything is working properly without noises that emanate from unhappy electrons.

I really appreciate the help, squarewave and abbey road. I still have some reading to do as there is more I want to understand about power supplies. But for now I have a fully working prototype from which I make tweaks, and for me that is quite a milestone.

Thanks!

Paul

 

[MisterCMRR]

If you have two switching power supplies whose outputs are connected together in any way (common-ground, stacking, whatever), I'd submit that the tone you're hearing is the beat note (difference) in their switching frequency. This frequency is rarely tightly controlled, but it sets the frequency of the ripple at each output - and also will be the frequency of radiated magnetic fields (where small voltages can be induced in the output side - twisting of all DC outputs and their returns is recommended until the cable is a foot or so away from the supply.

Incidentally, "hospital-grade" rarely has anything to do with the purity of the DC output - the leakage current on the AC input side is what's tightly limited for anything used in a hospital, where tiny leakage currents can become lethal if the flow in the wet "innards" of a patient.

 

[abbey road d enfer]

Incidentally, "hospital-grade" rarely has anything to do with the purity of the DC output - the leakage current on the AC input side is what's tightly limited for anything used in a hospital, where tiny leakage currents can become lethal if the flow in the wet "innards" of a patient.

Yes.
In another thread
https://groupdiy.com/threads/should-phantom-power-be-supplied-by-a-linear-supply.77291/someone wondered why many SMPS were not "quiet".
I believe that audio is about the only activity concerned with noise in the audio band. Most other activities are more concerned with RF/digital noise. To my knowledge, no SMPS manufacturer produces a product clean enough for direct integration in replacement for a linear PSU. That's why we have to add bits and pieces to make it in conformity with our demands. This is not a sign of using Band-Aids to overcome the faults of an inadequate product, just good engineering. Just like using a simple rectifier and reservoir cap*, which used to be the standard in many radio sets, needs to be complemented with active regulators for our builds.
Although not necessary, the use of medical grade SMPS in audio products is not a bad thing, since the low conducted emissions help cleaning the complete system.

*admittedly, often with the help of the inductance of the loudspeaker excitation coil

 

[Matt Syson]

A 'cheap and cheerful' SMPS is a lot less cheap by the time all the sundry parts necessary to make it 'clean and quiet' for both the incoming mains wiring (filter necessary) and DC output (more filters necessary) and the quite often horrible waveform of the (supposedly) DC output. Ainear 18 Volt 12 Amp module I am aquanted with regularly manages less than 2mV of ripple and noise (10 - 30kHz), whereas a switcher of similar rating manages 50mV or more of 'mains related ripple' and spikes at 60kHz plus at way more than half a volt. As Mr CMMR says, using multiple switchers inside a unit (and linked) will give sum and difference noises that will be audible.
Matt S