Do you do copper fill on your PCB?

Andy Belov

Well-known member
I make a Teensy 4.1 based sound effects processor and have my guitar preamp very close to the Teensy and audio processor.
I can hear some digital noise on the guitar input and want to minimize it.

I try to separate all analog stuff from the digital and put all the good capacitors, but wonder what else I can do.
One thing could be to fill all empty space on the board with grounded copper, like some people do. Others say it acts as a capacitor and makes things worse.

Please share your experience.
In general, what do you do to lower digital noise in analog inputs?
Thanks.
 
Good question.

In all of my audio devices - digital or not - I put in a large sheet of copper connected to ground. This does reduce noise. It doesn't eliminate it.

I have tried combining sub- miniature tube preamp with a Teensy and as long as there's no display technology involved it works well. On devices with an spi driven display Iget a cacophony of digital noise...copper sheet or not.

I have looked at the schematic of a successful unit ( digitech gsp 2101 ) that combines digital logic with tubes and in that case the tube circuits are on a virtual ground. As for applying that information to my own designs no real success, but to be specific I've only tried a simple resistor divider circuit to do this and haven't yet ventured off into using opamps and buffer chips like buf634a. For best results as far as I can tell, audio IO would need to be a differential pair and best I can tell no *readily available* audio device for teensy does this. Ground for the audio is ground for the digital logic. If anyone knows otherwise, I'm listening.

But to circle back to the specific idea in your question, yes, a denser ground plane does more good than harm * in my observation *
 
For low noise audio with a teensy 4, I'd use an external, fully differential ADC with a clean supply and reference voltage. With a differential pair to it.
 
Thanks @ETMoody3

I figured out the screen noise, it's quite "obvious": the noise is clearly related with what's happening on the screen.
I'll try to find a way to turn all SPI communications with it until necessary, only for changing settings.
I'll keep it off at other times, to minimize the noise.

And I will fill all the empty space with copper, just to see what happens, if you say you didn't notice any negative sides to it.
 
@jonr.
My thingy is battery-powered. Would really hate to have 2 separate batteries, but I'll give it a try, just for the heck of it. The ground will still be common, though.
 
I might have to try that as well, at the moment I put screen refreshing in kinda wherever, mostly event driven but there's the odd redraw that isn't gated by an event, depends on where the code is at. Never bothered to tighten that up.

I've tried shielded cable to connect the spi display and got rewarded with bizarre screen behavior. Cable was probably too long...

All of my work is hand soldered point to point, some on perfboard, some on busboard. I don't know how effective pcb fill is.

I've also noticed radically different results with the tube driver *after* the digital audio as opposed to before it. For guitar effects you generally want time effects and modulation after your drive/distortion. In those cases I've heard the most noise. The most successful hybrid I've done is a headphone preamp that uses a T3.2, a pt8211, two 6418 pentode tubes and an OPA2134 opamp. When attached to some computers, I get a little digital noise. Attached to an iPad the signal is pristine so I kinda don't suspect the Teensy as a noise source.

To jonr : I'd like to know a good fully differential ADC to try out. Recommend any?
 
@jonr.
My thingy is battery-powered. Would really hate to have 2 separate batteries, but I'll give it a try, just for the heck of it. The ground will still be common, though.

The important part to remember is to not pull any current through the AC ground. Your digital and your analog grounds should be completely separate except for ONE PLACE where they meet directly with the ground from the battery or power supply. Google "star grounding."

And yeah, at a certain point differential is the only way you're going to get great noise performance, especially if you have a very sensitive high impedance audio circuit like a tube amp as a buffer. There's a differential codec supported directly in the audio library, but so far as I know you'll have to make your own board. The CS42448 boards I have seen were made to fit like a shield and not at all suited for low noise performance (irony).
 
Thanks boxxofrobots,

I'll try the star grounding in my next version of PCB. Just ordered this one from JLCPCB yesterday.
On this version, I'll try adding components one by one and see which one contributes to the noise the most.
I have tons of things on a 8x10 cm board. Everything is very close to each other. Hard to isolate things.

Although, looks like the copper fill and the star grounding can be mutually exclusive. I think I'll try the copper fill first.
I have Teensy on one side of the board and the audio things on the other. Maybe having a copper shield between them will help somewhat?
 
Last edited:
Thanks boxxofrobots,

I'll try the star grounding in my next version of PCB. Just ordered this one from JLCPCB yesterday.
On this version, I'll try adding components one by one and see which one contributes to the noise the most.
I have tons of things on a 8x10 cm board. Everything is very close to each other. Hard to isolate things.

Although, looks like the copper fill and the star grounding can be mutually exclusive. I think I'll try the copper fill first.
I have Teensy on one side of the board and the audio things on the other. Maybe having a copper shield between them will help somewhat?

You can do multiple ground planes. One for digital, one (or more) for analog. Sometimes one for the power supply. Then connect them together in one spot via a star.
 
This is my board, with the ground highlighted. Just don't see how I can improve much there, other than do some copper fill here and there, probably on the green side.
The whole thing is 6x8 cm.
All audio stuff is at the upper left corner.

ground.jpg
 
No, star grounding and copper fill are NOT exclusive in any way. They work together.

Every connection is a signal from the pin, through the input, back via ground. Every switching signal between the psu ground and analog ground is going to introduce a path for those spikes to travel to your analog circuitry.

Analog circuits belong on a ground plane, with that plane connected at a single point directly at the power ground. You don't have to do it that way, but you'll never get great s/n without it.

The color selection in your post makes it impossible to actually evaluate the work you've done; all I see is white lines on black.
 
If you use differential signal/inputs, ground can be quite noisy or offset and it doesn't matter. Nor does other noise that gets induced in both lines.
 
Thanks for the video, boxxofrobots.

I do have a bunch of those capacitors, maybe not enough or not in the right places.
Will experiment with that. Just not sure about the capacity and the placements.
 
Should have added ground bounce and overshoot as search terms. My oversight.

IMXRT1060CEC_rev0_1.pdf, page 36 Table 25 indicates that Output Pad Transition Times, rise/fall could be anywhere between 1.06 and 5.57 nS depending how the pin is configured.

Translating this to frequency component, we see 180 - 943 Mhz.

A physical connection between eg. a T4 output to an input pin of some other device may be several centimeters long and meander around the board along the way and for steady state signals is fine however for above frequency component, needs to be seen as a complicated string of inductances with various stray capacitances hanging off so it's pretty easy to create some unhelpful resonances. One would expect to see harmonics of the frequency component of our rise/fall edges.

Any resonant circuit will serve as an antenna radiating its energy and we have close neighboring traces.

From the perspective of the device on the receiving end and thinking of overshoot let's ponder where it goes. One scenario would be the ESD protection between V+ and an input pin. Typically we see a diode junction here which will try to shunt overshoot energy to the V+ rail, in other words modulate it with more noise.

This thread has some discussion of overshoot and has some nice 'scope pics.

https://forum.pjrc.com/threads/4187...(ring-oscillator-example)?highlight=overshoot
 
Thanks everybody for your help and advises.
Well, turns out that all I needed was to turn up the amplification on my guitar preamp.

I put a couple of potentiometers for that and made sure that my oscilloscope shows around 3V amplitude on the preamp output.
Then I could turn down the audio input level on Teensy and the noise went down to nothing.

I still can hear some high pitch "music" in the right channel only when I turn the volume way up, but I can live with that.

But I thing I'll still try the copper fill. 8)
 
Back
Top