Teensy 3.6 noise

Status
Not open for further replies.

bob1984

Member
Hello,

I get a lot of noise even with a simple project

Code:
#include <Audio.h>

AudioInputAnalogStereo   adcs1;          
AudioOutputAnalog        dac1;           
AudioConnection          patchCord1(adcs1, 0, dac1, 0);

void setup() {
  AudioMemory(15);
}

void loop() {
 
}

first I made my own pcb with smd components so I thought the problem came from there
... so I tested the teensy on a breadboard

Untitled2.jpg

I could remove a lot of noise using recommended components

Untitled.png

but it still makes more noise than when bypassing the teensy

like, I have to crank the volume half way not to hear the noise anymore while at full volume, I have zero noise bypassing the teensy

- the input is a humbucker guitar pickup
- the output is a marshal MS-2 mini amp

let me know if you need more info, I really need to solve this

have a nice day
 
as I said, bypassing the teensy, the noise is completely gone, so the cables are no culprit here

the noise sounds like high freq hiss mixed with slight aliasing audio artifacts

I'll try recording it...but it really makes the teensy unable to do decent DSP, I just hope it's something I missed
 
as I said, bypassing the teensy, the noise is completely gone, so the cables are no culprit here

the noise sounds like high freq hiss mixed with slight aliasing audio artifacts

I'll try recording it...but it really makes the teensy unable to do decent DSP, I just hope it's something I missed

Hi Bob.

Some ideas:
Is there a reason to use AudioInputAnalogStereo with a mono input?

By the way your components legs are way too log.
That doesn't matter a lot in the output (amplfied signals), but in the input (line signals) you can pick and amplify a lot of noise that way.

What is the impedance of the source you are using?
If impedances don't roughly match you will end most of the times with more noise than signal.
And that will explain too the absence of noise if you bypass the Teensy.
You have a little less that 47K ohms in Teensy input line.
If the impedance of your source it's 400K or superior and the Marshal have close values, you are totally killing your signal but when you bypass the Teensy there will be no problem.
 
https://vocaroo.com/i/s0rY0RymFrxr

hard to find an imgur equivalent

so the first segment is me going through the strings with the teensy bypassed

then the pickup goes through the teensy and into the amp

the recording emphasis a lot more the problem than in real life, yet the difference is enormous
 
Try in a breadboard just one 47k parallel resistor with a series 10uF capacitor.
Just to check if the problem is the impedance.
 
I dont understand what you mean, can you mod the image I uploaded ?

also do I need the 2 GND pins to be connected ? I only soldered a couple of pins, easier to remove that way
 
Hi Bob.
Forget about Teensy now.
Just put one resistor 47K in parallel with the input source, now a 10 uF capacitor in series with the live wire of the signal, and from the capacitor output to the Marshall amplifier connecting the 2 grounds.
It's to mimic the impedance of your input circuit in the breadboard you have with the Teensy.
 
from the doc :

"Noise due to high source impedance, which allows rapidly switching digital signals to capacitively couple... avoiding higher analog impedance is the solution."

how could I do that ? using a preamp ? (opamp or something) ?

buff8.gif
 

Attachments

  • QPkt2.png
    QPkt2.png
    127.8 KB · Views: 64
Last edited:
I been told that jfet were less stable than opamp , besides this opamp is jfet based
anyway both solutions should reduces the impedance
 
using a buffer does not help at all, the teensy is just humming as f@#..
this is so frustrating...

anyone else knows another board that would be more suitable for audio DSP ?? (preferably with the same footprint ?)
 
using a buffer does not help at all, the teensy is just humming as f@#..
this is so frustrating...

anyone else knows another board that would be more suitable for audio DSP ?? (preferably with the same footprint ?)

My TGA Pro audio shield has everything on board, ready to go, no soldering needed. Properly designed power and layout to isolate the analog from the digital stuff as much as possible. It recently sold out but I'll have stock again in about a week. Ordering options make it easy to integrate into a custom project.
https://www.tindie.com/products/Blackaddr/arduino-teensy-guitar-audio-shield/?pt=ac_prod_search
 
that thing is massive, too big to fit in a small compartment that is the size of a teensy 3.6
 
Last edited:
that thing is massive, too big to fit in a small compartment that is the size of a teensy 3.6

The schematic is available for free in the datasheet found on the product page. You are welcome to take a look at the preamp and analog power design for use in your own circuit.
 
You can’t eat the cow and her butter, as we say in France. There are reasons why in high end audio devices, there are separate power supplies, separate grounds and separate shielded sections for the analog and digital circuit parts inside a common compartment, and why there are either audio transformers and/or additional buffer and amplifier stages to make sure that level, phase, and impedance are optimally matched between stages.
Thus, I’d say that it’s not the best idea to limit yourself too early with a probably too small compartment while you haven’t yet a circuit which fulfills your specs.
 
fyi : the compartment is big enough just for the teensy, the power supply will be external (2x 9v battery based + buck boost)

for the ADC, I was thinking about using this : https://store.digilentinc.com/pmod-i2s2-stereo-audio-input-and-output/ if the teensy ADC are that terrible ....

I mean , for a board that has a dedicated audio library and even demo videos, I really feel I got scammed on that one, but if an external I2S ADC can make it work...
the teensy audio board is too big otherwise I would have used it
 
Last edited:
yes but the problem is that I cant solder such small smd packages, nor CNC such small traces

and building it myself pushes me away from my target...

I thought teensy would be it...appart from these dodgy ADC, I like the board

I am currently looking for small I2S ADC (humanly solderable) chips or teensy like breakout boards
 
Last edited:
I mean , for a board that has a dedicated audio library and even demo videos, I really feel I got scammed on that one, but if an external I2S ADC can make it work...
the teensy audio board is too big otherwise I would have used it

I'm not sure how you could possibly think you got scammed. Mixed signal (analog and digital) design is tough, period. The built-in ADC is not intended for high quality audio applications like guitar effects, it's only 12-bits! 12-bit is for low-fi audio applications. That's why PJRC provides their audio board for higher quality 16-bit audio at line-level., but you say that board is too big, but you're not going find something much smaller than that using easy to solder components and actually meets all the impedance and low-noise requirements for guitar as well.

But since you are doing a guitar application instead of line level, your needs go up quite a bit for a robust design. That's why the TGA Pro linked above has so much circuitry for handling guitars properly. Everything from high impedance inputs, to massive dynamic range needs, not to mention power isolation and voltage protection.

Take a look at the PCB for a Boss DD-3 digital delay. All it does is a simple delay, but look how much circuitry is required to handle a guitar input and output properly!

Boss_DD-3_Top.jpg
 
did'nt know about that

is there any human-solderable I2S stereo ADC out there ?

I been digging a lot last nigh about this, but they are all zillion-tiny-pins packages so far
 
A quick look into the source code of the I2S objects of the audio lib will tell you that one frame contains two 32bit words (one per channel). The audio lib does use 16bit audio which means that the least significant bits will be ignored. Thus, your 24bit ADC will use 24 of the 32 available bits, leaving the remaining 8 least significant bits equal zero. The audio lib will then use the 16 most significant of these 24 bits and truncate the 8 LSBs. So, everything should work for you, if you set the CS5343 up in slave mode (10k resistor from SDOUT to GND) and configure everything for single speed mode with 256x MCLK/LRCK and 64x SCLK/LRCK.

And, for heaven's sake, please respect to avoid another disappointing situation:

"As with any high-resolution converter, designing with the CS5343/4 requires careful attention to power sup- ply and grounding arrangements if its potential performance is to be realized. Figure 3 shows the recommended power arrangements, with VA connected to a clean supply. Decoupling capacitors should be as near to the ADC as possible, with the low value ceramic capacitor being the nearest. All signals, especially clocks, should be kept away from the FILT+ and VQ pins in order to avoid unwanted coupling into the modulators. The FILT+ and VQ decoupling capacitors, particularly the 0.1 μF, must be positioned to minimize the electrical path from FILT+ to GND. The CDB5343 evaluation board demonstrates the optimum layout and power supply arrangements. To minimize digital noise, connect the ADC digital outputs only to CMOS inputs."
 
Status
Not open for further replies.
Back
Top