Digital + Analog: Severe noise bleed-thru problem

[dano12]

I have a microcontroller board that generates PWM and other pulses, through a vactrol, to an analog filter. When I run the analog filter on its own 9v supply, and the MCU on its own, it is quiet as can be: no bleedthrough of the digital signal. When the MCU and analog filter share a power supply, the bleed thru is overpowering.

I've learned a little about power supply filtering, but that is centered around cleaning up the power coming in. This one is completely new to me.

Any ideas on where to start looking for a cure?

[Auke Haarsma]

I'm not 100% sure if this applies to your situation, but you could try to keep the digital gnd and the analog gnd separate as long as possible. Connect them together via a 10 Ohm resistor.

[dano12]

I'm not 100% sure if this applies to your situation, but you could try to keep the digital gnd and the analog gnd separate as long as possible. Connect them together via a 10 Ohm resistor.

Thanks for the suggestion. I tried the 10ohm with no luck. The entire apparatus is spread across three breadboard so I'm sure that's adding noise.

I need to do some reading on decoupling. At least I think that's it.

Any other ideas from the night crew?

[Paul Perry (Frostwave)]

You have two problems:
1. sudden current draws in the digital side, put spikes on the + rail.
So as has been suggested you need to run the parts creating the spikes, and the parts picking up the spikes, either via separate regulators or by (as has been suggested) via resistors with heavy filtering to ground from both ends. Depending on current draw of a section, you might be able to get the resistor up to 220 ohm. I have done this successfully.

2. you also get ground return current spikes. Hence the need for separate ground returns for the various sections, meeting at only one point.

3. Don't forget, if the + rail has noise on it, so do all bias points.

If the micro is running off a +5 reg, then the noise probably comes via the combined ground return.

[dano12]

You have two problems:
1. sudden current draws in the digital side, put spikes on the + rail.
So as has been suggested you need to run the parts creating the spikes, and the parts picking up the spikes, either via separate regulators or by (as has been suggested) via resistors with heavy filtering to ground from both ends. Depending on current draw of a section, you might be able to get the resistor up to 220 ohm. I have done this successfully.

2. you also get ground return current spikes. Hence the need for separate ground returns for the various sections, meeting at only one point.

3. Don't forget, if the + rail has noise on it, so do all bias points.

If the micro is running off a +5 reg, then the noise probably comes via the combined ground return.

Hi Paul, thanks for the pointers. I'm trying to understand the issues here and your comments help explain the problem. Given this (very simplified) view:



I understand about filtering the + side of the supply and have included the basics there. The MCU does have a 5v regulator, the wah filter does not. So adding a 9vDC regulator to the wah and a ~100 ohm resistor from the mcu to ground would accomplish what you suggest? I don't know how to do filtering to ground, only filtering on the + rail as I've seen in so many stompbox schematics...

Thanks.

[R.G.]

Imagine your power supply floats, no ground reference.
+ side goes to two 10 ohm resistors. One resistor goes to the analog section, one to the digital section.
Each section, analog and digital, has a large electro cap from + to -, 100uF or possibly larger. Each also has a 0.1uF monolithic ceramic and a 0.01uF mono ceramic right at the power supply entry. Each section has a 10 ohm resistor from its - side back to the power supply. The two - terminals are connected by yet another 10 ohm resistor.

We call the analog - terminal "ground".

The idea is to force AC currents to be contained in the local loop which includes either analog plus its decoupling caps or digital and its decoupling caps, and to have NO current flowing through a shared conductor. The caps are arranged so that low frequency needs are sourced form the local 100uF (or 1000uF or array of 100uF's) and high frequency needs are satisfied by the 0.1 and 0.01's. Only a highly filtered, sluggishly moving DC goes back to the power supply to contaminate the other side.

To the extent that we are successful at separating the power supply flows of current in the opposite side, we will prevent any cross contamination with noise.

[Paul Perry (Frostwave)]

What RG. said.
As for your diagram, the MCU should be on the other side of the 100 ohm resistor, or (better) have its own 100 ohm resistor.
(I say '100 ohm' at a guess, what you should use on each side, depends on current draw, and regulator drop out voltage, if a regulator is used.)
I went through this with my Sonic Alienator, which has a PIC micro doing some simple DSP, plus a D to A chip, and an analog OTA filter section. Separate analog & digital grounds, separate regs. It had a 7662 voltage inverter/doubler as well..

[dano12]

I went through this with my Sonic Alienator, which has a PIC micro doing some simple DSP, plus a D to A chip, and an analog OTA filter section. Separate analog & digital grounds, separate regs. It had a 7662 voltage inverter/doubler as well..

Check out Adrian Belew's blog (one of my favorite guitar players of all time) http://elephant-blog.blogspot.com/2007/12/tale-of-trentsville-part-two.html

and you'll see a pretty awesome white pedal standing out front....

[dano12]

Imagine your power supply floats, no ground reference.
+ side goes to two 10 ohm resistors. One resistor goes to the analog section, one to the digital section.
Each section, analog and digital, has a large electro cap from + to -, 100uF or possibly larger. Each also has a 0.1uF monolithic ceramic and a 0.01uF mono ceramic right at the power supply entry. Each section has a 10 ohm resistor from its - side back to the power supply. The two - terminals are connected by yet another 10 ohm resistor.

We call the analog - terminal "ground".

The idea is to force AC currents to be contained in the local loop which includes either analog plus its decoupling caps or digital and its decoupling caps, and to have NO current flowing through a shared conductor. The caps are arranged so that low frequency needs are sourced form the local 100uF (or 1000uF or array of 100uF's) and high frequency needs are satisfied by the 0.1 and 0.01's. Only a highly filtered, sluggishly moving DC goes back to the power supply to contaminate the other side.

To the extent that we are successful at separating the power supply flows of current in the opposite side, we will prevent any cross contamination with noise.

Very helpful, as always. Here's the schematic representation of what I read. Look right?