Help with noise in Sample/Hold project

[Ripthorn]

So I am working on a digitally controlled sample/hold project. It is sort of like a DIY Proteus Subdecay, but with a couple differences. I have all the digital voltage generation done and working well and have hooked it up to the VCF portion. For reference, I am using the VCF portion of the Madbean Sharkfin project. However, I seem to have noise getting into the signal path such that I can hear the filter shifting frequencies even when not playing. I am looking for some help as to why this is happening and what potential mitigation strategies there are. Here is a quick and dirty video I made for reference. Sorry it's a little quiet, just a cell phone video. I'm open to any suggestions. Thanks!

[Kevin Mitchell]

I'm pretty sure that you'll get that noise no matter what with the FSH circuit - as long as the filter is being engaged by your LFO/sample hold.

You could come up with a way to null or pause your supply voltage when you're not playing. Like a gate. I'm sure you can find or design a circuit like that easily. That's the direction I would lean towards anyways.

Alternatively you could just bypass the effect when you're not using it 
I know! Such an original thought.

[Ripthorn]

I will bypass it when not in use, I was just curious as to the noise, what is causing, and how I might be able to control it. A gate is an interesting thought, but it might be too many added components. I'm trying to fit this all inside a 125B. However, if this is unavoidable, that's fine, I just haven't played a sample/hold before. Thanks!

[Kevin Mitchell]

I don't see why it wouldn't fit into a 125B enclosure.
For perspective... I'm almost done with a unvibe layout the size of a credit card - fitting into a 125B 

This page might be of interest to you.
Minimal S/H circuit

You've reminded me that I have an unfinished layout for one of these in a 1590A 

[Rob Strand]

To me it sounds like something *is* actually wrong with it.
Without a circuit it's hard to speculate.
Maybe it's a wiring or part value error.

At first I thought there was a dog barking in the background then I realized that was the noise problem.

[Ripthorn]

So I have made a little headway, but don't have it all solved. The ground connection on my voltage generating IC was a little dodgy, so that helped a little. I also found that the LED blinking on my control chip is contributing some high frequency noise, but I still have the main filter whooshing. I'm working my way through it, but don't understand exactly how the VCF section works quite well enough to be able to make intelligent debugging steps.

[ElectricDruid]

Noise getting to the filter isn't the only problem. Any DC getting to the filter will also be modulated by the changing cutoff, so you need to avoid DC offsets too. And then the filter itself with allow some of the CV to leak into the audio output ("control rejection"). Even for decent quality synth filters designed to avoid this, typical figures are in the low tens of dBs.

I was working on a LM13700 state-variable filter last year. In the end I abandoned it because of this control feedthrough problem. I finished up with a vactrol filter instead, which seems to be much quieter.

Rob's right though - make sure that what you've got is working correctly before you try and evaluate how well it works. And yes, I realise how hard that is to determine for circuits you've not played with before!

HTH,
Tom

[Rob Strand]

So I have made a little headway, but don't have it all solved. The ground connection on my voltage generating IC was a little dodgy, so that helped a little. I also found that the LED blinking on my control chip is contributing some high frequency noise, but I still have the main filter whooshing.

Now you mention that it's go me thinking there's a loose connection.  Perhaps around where the output of the detector connects to the control inputs of the OTA's.   It's kind of jumping about which is what you get with a loose connection.

A good practice is to use separate ground and supply for the audio and the non-audio parts of the ckt then connect the power and grounds back to one point on the breadboard perhaps putting a 100uF cap right at that point.    That will stop things like LED currents modulating the supply/ground lines.

[Ripthorn]

I think what I will do first is build up the classic noise generation circuit and tie it in to the vcf and see what it sounds like. That should att least help me determine if the issue I'd in the noise or filter section.

[Ripthorn]

Alright, I tried a couple other things and I think I'm slowly boxing the problem in. I breadboarded the noise generation section of the schematic in the link above. When I hooked it to my main board, as soon as I connected the +9V and -9V connections (even before the ground and control voltage) I got a huge drop in the noise and everything else is working fine. I have double and triple checked my power section and it looks fine, so it seems like maybe there is some loading or something that is coming in to play. I really don't know, but the difference in noise is dramatic. There is a very high pitched whine that gets introduced.

However, I did not get the analog noise generation circuit working. I'll have to make it wait until tomorrow. Ugh.

Also, regarding the LED, it seems like the noise is actually loud when the LED is NOT lit and is attenuated when it is. I don't understand that at all.

[Rob Strand]

Not sure exactly which noise ckt you are using.  Is it the one with the opamp and the 2M7 feedback resistor?

Maybe the noise opamp is clipping?  In that case reduce the 2M7 resistor.

Another weird thing that could make it clip earlier is the fact full noise bandwidth is getting through.   Adding a small cap across the 2M7 will decrease the o/p voltage swings due to noise above 20kHz.

[Kevin Mitchell]

However, I did not get the analog noise generation circuit working. I'll have to make it wait until tomorrow. Ugh.

It can be tough finding a good NPN that makes the cut. I've had a bag of 30 where only a few actually did something for the circuit. Some folks had better luck putting the base to the negative supply instead of the collector. I've tried this but didn't like the range as much as my "good" candidates in the original format.

Also to note - I had not listened to your videos. Just killing time in the office 

[Ripthorn]

So my current noise circuit is actually based around an ATTINY85 microcontroller that does things like tap tempo control, divisions, patterns, etc. The output voltage is actually a PWM duty cycle of the 5V supply voltage. To filter the PWM, I was using a Sallen-Key low pass filter with Q of 1/sqrt(2) and corner frequency very low. I was using an LM358P to allow for driving signal very close to 0V at minimum. I tried swapping out the Sallen-Key LPF for a simple RC LPF and much of the noise went away. I still have some, but it is far less offensive. However, I still have a little bit of the PWM frequency audible and still some of the whooshing. I'm gradually trimming away parts that may not be needed to see if I can simplify things a little more. I'll keep working on and keep this updated.

[Ripthorn]

Alright, more data. I changed the physical location of where I was doing the LPF for my PWM control voltage. That eliminated the PWM noise. The last bit of noise I have left I am able to almost completely eliminate by connected +9V to pin 8 of a TL072 and -9V to pin 4. Not having either of those connections results in increased noise. I'm trying to figure out what the root cause is, but I guess I don't understand the nitty gritty of op amp internals well enough. Is there something I could use that would have the same effect and be smaller footprint? Why would just supplying voltage to an opamp result in super reduction of noise? Hmm...

[Rob Strand]

Alright, more data. I changed the physical location of where I was doing the LPF for my PWM control voltage. That eliminated the PWM noise. The last bit of noise I have left I am able to almost completely eliminate by connected +9V to pin 8 of a TL072 and -9V to pin 4. Not having either of those connections results in increased noise. I'm trying to figure out what the root cause is, but I guess I don't understand the nitty gritty of op amp internals well enough. Is there something I could use that would have the same effect and be smaller footprint? Why would just supplying voltage to an opamp result in super reduction of noise? Hmm...

Without a schematic I'm guessing what is going on.  You have the circuit connected up, including the opamp but with the exception of the opamp power pins?   Then, when you connect the power the noise disappears?

In this case it is very simple!  When the opamp is unpowered  the opamp output impedance is high and the signals at the input pass through the R's and C's around the opamp then out through the other side where the opamp output is.  If the input signal is noisy it will pass the noisy signal to the output.

As far as a smaller foot print goes it depends on the gap between the PWM frequency and where you want the filter cut-off.  If there is a large frequency ratio between the PWM frequency and the filter cut-off you can use a second or third order RC network with as low a cut-off as you can stand in your circuit.   However if you need a low impedance output then it is unlikely you can removed the opamp.    A transistor is an option but really the overall footprint is not likely to get smaller.

[Ripthorn]

Alright, problem solved. Using the results of last night's experiments, I decided that the opamp was likely producing some loading on the charge pump. Since my circuit doesn't use the envelope filter section or the noise generation section, the current draw is a lot lower than a stock FSH-1. So I put a 10k trim pot between +9V and -9V with it set at max resistance. I then lowered the resistance just until the noise disappeared. This turned out to be about 3k or so. Using good old Ohm's law, I found that that would result in about 100mW dissipation. I have settled on a 3k3 load resistor. All seems to be working well. Thanks to all who have contributed. I know it's a bit of a ramble. I hope to have a proto all built up shortly and then i will likely do a writeup with complete project files for others to build if they so desire.

[ElectricDruid]

Hang on, I'm missing something. I mean, it's great that you're fixing your problem, but...

If the problem is caused by loading on the charge pump, then why does turning the resistance *down* and thereby *increasing the loading* seem to help?  I'm not getting this.

Thanks,
Tom

[Ripthorn]

I wish I understood charge pumps and what is going on better. I was talking to a friend at work today who has worked with them quite a bit in the past and he mentioned that sometimes charge pumps have to have enough current draw to work well. That, combined with what I had seen on my breadboard, I decided to experiment. I don't know why, but increasing the load on the charge pump helped reduce the noise. I suppose it's possible that there is some other actual mechanism at work and I just found a workaround. I have a board layout almost complete for what I have on my breadboard, so once I get the proto boards back, I can no-pop that resistor and see if it was more of a layout issue or if it was actually a circuit issue.

[Ripthorn]

Just for completeness' sake, here is a video of what it sounds like with all the noise fixed. Extremely little background noise when not playing. I'm pretty excited about it!

[Rob Strand]

Just for completeness' sake, here is a video of what it sounds like with all the noise fixed.

Good to see it all workout out.   

The problem with analog stuff is you can have few small things wrong and all of a sudden you have some weird-ass behaviour on your hands.