Actually Pink/White Noise at the output

DoomGuy

Hello everyone/giuliomoro,

There is something, that really annoys me and also stops me from going on:

I wanted to use the Bela as a guitar pedal, but unfortunately I'm getting a crazy amount
of pink/white noise at the input, when I later amplify the output signal.

To make something clear: When I talk about noise, I mean the "hiss" and not any interference.

I mean there is some white noise at the output, but it's not that significant.
However when I use the input, there is lots of pink/white noise.

I've got the standard BelaBoard (BeagleBoard Black + Bela Cape) and I also added
the pins to the unpopulated pads for LineOut.
I'm powering the Bela over 5V USB (5V Barrel connector gives me digital noise)

So to make it more clear, I've created a simple setup, in order to explain this:
Electronically I did this:

-Left output connected to OutputJack(Tip)
-Audio ground pin connected to OutputJack(Sleeve)

With left output I mean LineOut(L+) xor Headphone Output, I tried both, and they give me the same result.
And with audio ground I mean the ground pins at the BelaCape audio interface.
(The DGND pins are just adding digital noise...)

The Output goes over to my Amp over a simple 6,35mm mono cable (I also tried different lengths of cables, but that
didn't change anything)

Programmatically:
A simple project that just copies the left input to the left output.

Settings:
Block size: 16
Sample Rate: 44100
Headphone Level: 0
DAC level: 0
ADC level: 0
PGA gains: 0

What I get (the volume of my amp is at 1.5):

This is how it sounds like:

I also tried to connect my guitar to the Bela audio input, but the noise wasn't even attenuating.
The right input channel also gives that terrible pink noise.

So my question is: how do i get rid of it?
Is it an impedance problem? Are the parts on the BelaCape PCB correct?
Is it the normal behaviour of the TLV320AIC3104 Audio Codec?
I have no clue.

I love the idea behind Bela and I love the fact, that it provides low latency audio.

I would appreciate any help in getting this fixed!

Thanks.

Greetings from Germany

giuliomoro

DoomGuy However when I use the input, there is lots of pink/white noise.

Are you applying any gain to the input signal either in software or using the PGA gain? For noise-sensitive signals such as guitar, one would normally set the codec's PGA gain to 0dB (no gain) and use an external buffer with a high input impedance, a low output impedance and 1.8Vpp output before going into the TLV320AIC3104's input. Most non true-bypass guitar pedals can work as a buffer even when they are in bypass (e.g.: most BOSS pedals would), however I am not sure about whether their output level is reasonably well matched the 1.8Vpp full scale input of the codec. When we used this codec with guitars we tend to put a little single-opamp or single-JFET buffer in front of it.

giuliomoro

@DoomGuy did you delete your latest post on purpose? Is there anything wrong with it or shall I restore it and read it?

DoomGuy

giuliomoro No it's absolutely fine xD, my last post was embarrasing. I applied 1.8Vp to the input, wondering why it was clamping, but unfortunately, it was my stupidity that prevented me of multiplying 2 x 1.8V, which means 3.6Vpp and not 1.8Vpp. My signal generator wasn't showing me Peak to Peak. Then I was thinking: "oh there has to be a faulty component on the PCB, because I'm not getting 1.8Vpp". Facepalm

Anyway...
I was not applying any gain neither software-wise or hardware-wise.
The PGA Gain is set to 0dB and the corresponding test code just looks like this:

void render(BelaContext *ctx, void *userData)
{
	for(uint32_t n = 0; n < ctx->audioFrames; n++) 
		audioWrite(ctx, n, 0, audioRead(ctx, n, 0));
}

When using an external buffer at the codec input, the noise remains the same.
But if it's connected to the codec output, the noise gets attenuated (Boss PS-6 Harmonist)
Having two pedals connected, one at the input, the other at the output of the codec, gives the same attenuated noise.

So no buffered Output (WARNING IT'S VERY LOUD):
10 Watt Transistor amp
Amp Volume goes from 0 to 10

Buffered Output:
Still 10 Watt Transistor amp
Amp Volume goes from 0 to 10

Also on my oscilloscope I get roughly 60mVpp of noise.

The pure noise is arround 170mVpp without any amplification (PGA at 0db, all levels at 0dB) and when the input isn't connected, not even wired.

So, in conclusion buffering the output actually helps!
But when amplifying that buffered output at an amplifier, it is getting much more noticeable, than the noise from the amplifier itself. But I guees I have to live with it. Thank you very much giuliomoro!

Out of curiosity:
How high is the noise on the Bela mini (in mV), when running the test code above,
having all gain levels at 0dB in the settings and nothing connected to the input?

giuliomoro

DoomGuy Also on my oscilloscope I get roughly 60mVpp of noise.

A time-domain observation indicates very little unfortunately: most of the noise will be outside the audio range (above 20kHz) and it's the leftover from the reconstruction filter of the sigma-delta converter. Buffering the output is not supposed to reduce the noise (by definition it can only make the SNR worse), I guess what you are observing is filtering through the PS6's filters (unless there is some full-spectrum attenuation, but that shouldn't happen in a guitar pedal in bypass mode!).

I'd suggest you do your tests alternating sections of full-scale signal generator output and pass-through: this will give you an idea of the usable dynamic range.

I also didn't explain the purpose of the input buffer: I expected that you had evaluated the noise floor in terms of dynamic range with respect to the guitar signal and that this could have been made worse by plugging a passive pickup straight into the codec's 1k input. That's where a buffer would have helped: by buffering the signal it would prevent the signal degradation going into the codec's low impedance input. Have you tested this?

DoomGuy How high is the noise on the Bela mini (in mV), when running the test code above,
having all gain levels at 0dB in the settings and nothing connected to the input?

I think the noise floor on BelaMini is about -80dB (you can measure it yourself with examples/Extras/measure-noisefloor). The codec's spec are a bit better than that (SNR 92dB typical, 80dB min), but things are a bit more complicated in real life when you have an audio circuit sitting a few mm in front of a 1GHz CPU ...

DoomGuy

giuliomoro

giuliomoro A time-domain observation indicates very little unfortunately: most of the noise will be outside the audio range (above 20kHz) and it's the leftover from the reconstruction filter of the sigma-delta converter

You are absolutely right, I wasn't thinking about that. There is some crazy low-pass filtering going on in the PS6, that's actually a thing.

giuliomoro this will give you an idea of the usable dynamic range.

Like applying a 440Hz sinewave at 1.8Vpp at the input (all buffered) and adjust the gain to 0dB at an audio interface? When I do that, the noise is creeping arround -75 dB. So the usable dynamic range is actually fine.

giuliomoro Have you tested this?

Yes indeed, in fact I used a buffered boost pedal with gain control, to achieve a maximum guitar signal voltage
of roughly 1.8Vpp, when I strum a chord. While playing gently, the noise can be a little bit in the way,
when playing on a 60W Amp with volume at 3. And there is no tone sucking anymore.

giuliomoro the noise floor on BelaMini is about -80dB

Yeah I just wanted to know if the normal Bela (which I have) has the same noise floor as the BelaMini.
When running *measure-noisefloor it is roughly -85dB to -80dB on both channels, so it's basically the same.

giuliomoro an audio circuit sitting a few mm in front of a 1GHz CPU ...

Yeah this is impressive.