Splitting and isolating a signal

[mark2]

I mentioned part of this in the other thread: But I'm using an IC to boost, shift, and sample the signal and write some visualizations to a display.

On its own it works great. Then problem is I'm struggling to find a way to keep the input signal usable. It gets VERY VERY noisy.

Ideally I'd like a way to split and completely isolate it, but is that even possible with a shared ground?  In a simplified form, the ideal result is a pedal that has an input and output, and the input passes purely and unaffected to the output, while the digital circuit inside can sample and use the input without affecting the output.

Here's the schematic if it helps. Any thoughts would be most appreciated. I feel like I'm simply lacking the vocabulary/knowledge to know what to search for or study.

[mark2]

I think I figured out my problem.
I had a JFET buffer which quieted the noise, but I had the split after the buffer.

I moved the split before the buffer (with the buffer on the digital circuit, and the other side of the split going straight to out) and it seems to have fixed it.

[mark2]

Here's what I currently have. The buffer dramatically reduces the noise, but it's still there on the OUT.
Any suggestions?

https://imgur.com/cohBbgR

[antonis]

A BJT buffer should be substantially "quieter"..
or
Use a really noiseless op-amp and scale down respective resistive circuitry..

P.S.
Or just locate signal's noise source and try to eliminate it..

[mark2]

I have an unused side of the LM358. Would I be better off making an input buffer with it (like this) rather than the JFET input buffer I have in my schematic I posted above.

Or am I misunderstanding?

[antonis]

It could save you the extra JFET circuitry components and provide a more linear & noiseless responce..
But I think your issue is more signal noise source rather than buffer/gain stage one..

[mark2]

As in I need to find and mitigate the noise this digital circuit is introducing? And there's no way to simply "wall it off" or isolate so it can't affect the output?

Thanks for your help and patience in explaining.

[antonis]

As in I need to find and mitigate the noise this digital circuit is introducing?

Of course, as mitigating the issue source is elementary "cure"..
(better safe than sorry..)

If it's located and verified as an inherent issue, then proceed to eliminate it by extra circuitry..

[niektb]

what you could try is having 2 buffers such that one buffer feeds into your circuit and the other on is feeding the out. See for example this page from AMZ:
http://www.muzique.com/lab/splitter.htm

[niektb]

Also, displays can be an incredible pain in the ass, i've had issues with oled displays in the past that caused a lot of noise on the 5V supply. Try disconnecting the display and see if that makes any difference.

If that's the cause, try isolating the ground from the microcontroller and display from the audio signal ground as much as possible (Google 'Star Point Grounding')

[PRR]

What is this in the middle? Doesn't it suck your "4.5V" down to a half volt? And cut out all the bass?

[mark2]

What is this in the middle?

This is to shift the AC signal into the positive so I can read it with an AVR's (attiny85) ADC.

Edit: and by the way, C9, C7, C1 are probably redundant. I've been piecing this together (buffer -> DC offset -> amplify) and haven't tried removing some of those  yet.

Doesn't it suck your "4.5V" down to a half volt? And cut out all the bass?

Not that I noticed but I will look again.

Try disconnecting the display and see if that makes any difference.

If that's the cause, try isolating the ground from the microcontroller and display from the audio signal ground

Will test that shortly, thanks. Unfortunately I can't isolate the  grounds any further.

The analog is literally just a straight input-to-output. I'm not doing any audio shaping (and trying really hard to avoid it).

[mark2]

Try disconnecting the display and see if that makes any difference.

Yes, the display is exactly the problem.  It's whisper quiet without it.

I'll try another buffer on the output, like the signal splitter, but otherwise not sure what else I can do here.

I tried dual buffers, like the splitter, and surprisingly it actually made the noise louder. 

i.e. the second is louder

Code Select Expand


                   -> buffer -> digital visualizer
Input -> split --<
                   -> output (has quiet noise from OLED)


                             -> digital visualizer
Input -> buffer splitter --<
                             -> output (has loud noise from OLED)


So far my best option seems to be to simply accept the quiet noise and tweak the component values to reduce noise. e.g. the AMZ buffer is a bit quieter than my original (w/ same JFETs). Or possibly try an opamp buffer.

Doing a bit more research uncovers this to be a common problem with I2C - which this display uses - with no simple solution. Maybe some sort of LC filter.

[PRR]

> to shift the AC signal into the positive

But then you AC-couple the signal! And re-reference it to +4.5V. Which is awful high for a 5V ADC.

Besides the 1k dividers being way-low for the 10k divider. (So it may work less-bad than it seems. But is still just wrong.)

[PRR]

The '358 can be a terrific buffer. Bias it right. Add some isolation in front so if the '358 does make crap (likely from clipping from too much gain) it doesn't get out the front.

[mark2]

Thanks a lot for taking a look at that and making some changes. I didn't have a 270p but tried with a 220 and it seems to have a similar effect, but centers around 2.5V.

Everything I found about DC coupling or shifting has 2 resistors like in my original. Do you have any topics you can point me toward to better understand how this works? And how you chose 270p?

The '358 can be a terrific buffer. Bias it right.

Are you suggesting I replace the JFET buffer in my original with the unused half the 358? Or did I miss your point?

Add some isolation in front so if the '358 does make crap (likely from clipping from too much gain) it doesn't get out the front.

Is this just a buffer? Or if not, any topics I can look up to learn how to isolate it?

Thanks again!

Edit: Also, for what it's worth, this works well for getting a good signal to the ADC, but - unless I inadvertently changed something else - the jfet buffer doesn't seem to be taming the noise at all.  Whereas I was able to get the wrong/broken one very quiet. (not suggesting I should use the wrong one, just noting)

[niektb]

Will test that shortly, thanks. Unfortunately I can't isolate the  grounds any further.

I don't mean disconnecting but keeping the physical distance as large as possible by directly connecting the ground to your LDO's ground. (Star Point Grounding, look it up on the internet! )

You could also try a second LDO that also generates 5V (or 2.5) so your Bias Voltage isn't connected to the supply of your display.

Also, add a couple of 100nF caps close to the supply pins of your Attiny, display and LDO (1 for each device)

[mark2]

Thanks. Yeah I understand about the grounding. It's as separate as possible right now.

A second power supply should probably clear this up, though I haven't tried, but I'd really like to avoid that. Definitely a good idea though.

Also thanks for the recommendation about decoupling caps - I tried them right as close to the devices as possible but it didn't help.

[mark2]

I got a nice big 680u cap and put it between 5v and gnd at the screen and AVR IC, and it made a pretty huge difference in the noise. Anything under 470 seems to barely make a dent.

It's still unacceptably loud, but it's progress.

I'm starting to wonder if I should try an LC filter. Curious to hear if anyone has thoughts on that.

[antonis]

A cap, by its own, can't form any filter, despite its capacitive value..!! 
(even those big electros reservoir caps on rectifier circuits DO need a series resistance to form LPF - which resistance is considered the sum of secondary winding and rectifier diode(s) dynamic ones..)

That resistance is your 5V regulator Output impedance..

An LC filter simply substitutes inductor (choke) for resistance in a LPF..
(of impedance according to 2π*f*L..)
The lower the corner frequency of this LPF the higher the inductor value (more heavy/bulky item) for a given C value..
[ f = 1/2π(LC)^1/2 ]