ICL7660 - Whining w/ Power Supply but NOT with a Battery?

[Paul Marossy]

So I have a question. I resurrected a little function generator circuit I built a loooong time ago that uses an ICL8038 chip. I thought it would be cool to operate the circuit off of a 9V battery and use an ICL7660 to double the voltage. I also thought it would be nifty to have a switching DC jack so I can use my bench power supply too. The voltage doubling works fine, but the problem is that I get that approx. 6 kHz whining sound whenever I am using the power supply (in other words, it's fine with the battery).

I am perplexed as to why it would do the whining when on the power supply but not with the battery. Is it something to do with how the grounding is done? (and I know, just get a MAX1044 but I want to understand what is going on)

[Mark Hammer]

How well-regulated is your bench supply?

[Rob Strand]

You need a big cap across the input supply terminals of the charge-pump.   The cap needs to be close to the charge-pump chip.
Within half to double the caps used in the charge pump (bigger the better).

This confines the large current pulses going to charge pump over some very short tracks.
Without the input cap the current pulses spread through the whole circuit and cause noise problems.

[Paul Marossy]

How well-regulated is your bench supply?

Should be pretty good I would think. It was a little kit that I built years ago, I think it uses an LM-317.

[Paul Marossy]

You need a big cap across the input supply terminals of the charge-pump.   The cap needs to be close to the charge-pump chip.
Within half to double the caps used in the charge pump (bigger the better).

This confines the large current pulses going to charge pump over some very short tracks.
Without the input cap the current pulses spread through the whole circuit and cause noise problems.

How big is big? I currently have 100uF... not large enough? I do have it very close to the chip.

[antonis]

It was a little kit that I built years ago, I think it uses an LM-317.

Any chance for output cap been an aged tantalum..??

[anotherjim]

For that ICL7660, the data says...

The oscillator, when unloaded, oscillates at a nominal frequency of 10kHz for an input supply voltage of 5.0V. This frequency can be lowered by the addition of an external capacitor to the OSC terminal,

10kHz and can be lower! Exact frequency not really possible to predict - depends on load and supply volts. Is this part really suitable for audio? The whine might be in a ground loop when the PSU is used.
Big electro cap has higher self-inductance so less good with high frequencies (don't forget the switching waveform makes high harmonics). You also need ceramic bypass caps on the power rails on everything.

[Paul Marossy]

It was a little kit that I built years ago, I think it uses an LM-317.

Any chance for output cap been an aged tantalum..??

No, I used electrolytic cap.

[Paul Marossy]

For that ICL7660, the data says...

The oscillator, when unloaded, oscillates at a nominal frequency of 10kHz for an input supply voltage of 5.0V. This frequency can be lowered by the addition of an external capacitor to the OSC terminal,

10kHz and can be lower! Exact frequency not really possible to predict - depends on load and supply volts. Is this part really suitable for audio? The whine might be in a ground loop when the PSU is used.
Big electro cap has higher self-inductance so less good with high frequencies (don't forget the switching waveform makes high harmonics). You also need ceramic bypass caps on the power rails on everything.

Ah... the ceramic cap might be worth a try. I suppose it could also be a ground loop, but wouldn't that be at a much lower frequency?

[Rob Strand]

100uF is normally enough.  I could have dried out so it wouldn't hurt soldering another 100uF in parallel as a test.

You could even try temporarily adding an inductor on the input side to see if it helps.
https://www.maximintegrated.com/en/design/technical-documents/tutorials/2/2027.html

Yep, it could be a ground loop.

It could also be your bench power supply is oscillating.   Maybe because some caps have got old.   

You can try adding a 100uF cap across the output  terminals of your power supply.

Another test you can do is keep the long wires to your power supply but
- try connecting the battery at the point where you connect the PSU.  That checks weird layout and wiring base problems.
- then after that, run from battery with the long wires but now connect the -ve lead of the PSU to the negative lead of the battery.
  Don't connect the PSU positive.
  That might show-up any ground loops.

If you don't get the whine in these two tests I'd start to suspect an issue with the PSU.   The aim is to separate out the cause.

-----------------
EDIT:  Something which did occur to me, if you have an oscilloscope measure the switching frequency with the PSU and the battery.

[anotherjim]

So, is the bench PSU a switch mode with a linear (317) regulator feeding the project?
You could be hearing heterodyne whine between the 7660 frequency and a switching PSU.

[italianguy63]

I think you need an ICL7660S.

[anotherjim]

I think you need an ICL7660S.

Yes, that's the one! I fogot it's the same part number, but the "S" is significant.

Changing the ICL7660S and ICL7660A Oscillator Frequency.
It may be desirable in some applications, due to noise or other considerations, to alter the oscillator frequency. This can be achieved simply by one of several methods.By connecting the Boost Pin (Pin 1) to V+, the oscillator charge and discharge current is increased and, hence, the oscillator frequency is increased by approximately 3.5 times.

...which puts it's switching frequency well above audio.

[Rob Strand]

Despite the better chip, the question why there's a difference between PSU and battery still needs an explanation.

[bluebunny]

My money's on the heterodyne that Jim pointed out.  But we haven't established the nature of the PSU yet.

[Paul Marossy]

Despite the better chip, the question why there's a difference between PSU and battery still needs an explanation.

Yeah, so here's the weird thing. I use this ICL-8038 based function generator to inject a test tone into D.U.T. and then send the output of that to an LM-386 audio amp (re-housed RadioShack Mini-Amp). Taking the D.U.T. out of the equation, I get the whining only if I have both of them on the same power supply. If I take one or the other off of the power supply and have it on battery operation instead, no whining. 

So in other words, audio amp on power supply and function generator on battery, no whine. Function generator on battery and audio amp on power supply, no whine. Both on same power supply, whining.

I am a little perplexed. I am guessing it's being injected from one circuit into the other via the grounds? I know I could just not have one of them not be on the power supply but that's not really what I want to do.

[Rob Strand]

I am a little perplexed. I am guessing it's being injected from one circuit into the other via the grounds? I know I could just not have one of them not be on the power supply but that's not really what I want to do.

The fact it *can* occur makes 100% sense.  It's very common with single-end audio inputs like the LM386.

I'd basic scenario is:
-  The function generator, which contains the ICL7660, will have current pulses on the wires between the ICL7660 and the power supply.   The current pulses introduce small voltage fluctuation in the ground wires.

- When you connect the LM386 to the oscillator, you will connect signal and ground.   However if you connect both power and ground of LM386 of the the power supply.   The current pulses from the ICL7660 have many paths to follow:
- Function generator power ground back to the PSU
- Through the function generator ground, through the LM386 ground, then back to the PSU.

The last point is where the the problems occur  since the small voltage drops due the ICL7660 current pulses are now flowing through the audio cable between the function generator and that signal adds onto the audio signal at the LM386 inputs via the ground.

If you "cut" the audio ground wire between the LM386 and the function generator  then the LM386 would see see the voltage fluctuations on the ground between the function generator and the PSU.   Here the whole function generation ground potential is moving up and down with the current pulses.

So the only option left is it "cut" the ground between the LM386 and the PSU.  In this case the LM386 gets it's power from via the audio line.   While this isn't great and can case oscillations, it does prevent the ICL7660 current pulses going down the audio.

You should try it.

With the DUTs present they have to get the power via the audio line as well.  If you connect the power back to the PSU then the current pulses are on the audio again!

So that's only one scenario.   There's more possibilities.    Also more evil possibilities like the demodulation case anotherjim mentioned.

The root cause is common supplies + single ended audio.   It can often have issues.   That's why people use differential inputs on a lot of audio test stuff.    Unfortunately you are likely to want to use single-end audio on simple systems but that's exactly when you end-up with problems.     So the only the option is separate power.  I have put differential input front-ends on simple systems in the past because it is not possible to control the grounds and the noise on the grounds.

[Paul Marossy]

So the only option left is it "cut" the ground between the LM386 and the PSU.  In this case the LM386 gets it's power from via the audio line.   While this isn't great and can case oscillations, it does prevent the ICL7660 current pulses going down the audio.

You should try it.

With the DUTs present they have to get the power via the audio line as well.  If you connect the power back to the PSU then the current pulses are on the audio again!

So that's only one scenario.   There's more possibilities.    Also more evil possibilities like the demodulation case anotherjim mentioned.

The root cause is common supplies + single ended audio.   It can often have issues.   That's why people use differential inputs on a lot of audio test stuff.    Unfortunately you are likely to want to use single-end audio on simple systems but that's exactly when you end-up with problems.     So the only the option is separate power.  I have put differential input front-ends on simple systems in the past because it is not possible to control the grounds and the noise on the grounds.

Yeah, I kinda came to the same conclusions about sharing the power supplies thru the process of elimination. Your explanation of the pulsing on the ground makes sense, although it's kind of weird to conceptualize because I think of ground as zero volts... but I guess I could think of it as a power supply ripple on the ground wiring. Before your reply I was fiddling around with different scenarios and found that cutting the ground on the audio was the only thing that really worked. So I just added a ground lift switch on my audio oscillator. That works!

Thanks for taking the time to explain all that, it makes sense. I'm a bit rusty, haven't done too much electronics related stuff in the last 2-3 years.

[Rob Strand]

but I guess I could think of it as a power supply ripple on the ground wiring.

Yep, totally reasonable way to view it.

Before your reply I was fiddling around with different scenarios and found that cutting the ground on the audio was the only thing that really worked. So I just added a ground lift switch on my audio oscillator. That works!

In non-ideal scenarious like this fiddling about is as good or better than theorizing.    There's too many possibilities and usually only one or two fixed end up being better than the rest.   You have to work with what is there.   if you were working in lab you can afford to dump some your test set-up but at home you don't have that luxury.

Thanks for taking the time to explain all that, it makes sense. I'm a bit rusty, haven't done too much electronics related stuff in the last 2-3 years.

Problems like these never get any easier,  you just expect they are there.