SOLVED: How can I reduce ADC jitter

[richard]

stick a 2200uF low esr electrolytic in every power rail on each bread-board and be done with it. if it still noisy add another set.

if that fails isolate noisy bits behind rfc's [rf chokes]. ditch the hi switching speed [noise inducing] mosfets for pwm, use bjt instead, its only a backlight

as ioannis says, breadboards and low noise are very close to mutually exclusive concepts

[richard]

i tried the burst mode averaging method. very nice if you don't need the adc for anything else, it would be ideal for monitoring a lifepo4 battery during all phases of discharge/charging/float. no great advantage for multiple inputs that i can see

[Demon]

i tried the burst mode averaging method. very nice if you don't need the adc for anything else, it would be ideal for monitoring a lifepo4 battery during all phases of discharge/charging/float. no great advantage for multiple inputs that i can see

I'm going to see about using something like your MEAN average example. I had something like that a while ago (inspired by Melanie), but I had noisy power supplies. The TPS56637 works nicely, so I was hoping ADC2 could make my life easier.

[richard]

I only have a few 1000uF electrolytics, so I put 2 on each rail, same ADC flutter of mostly 1, sometimes 2.

+- 2 counts is as good as it will ever get for raw adc reads on a breadboard, if you right shift 2 bits that's a 256 count resolution that should be very stable

the scope is your friend here , measure the AC noise. every 5mV of noise on the rail and/or the pot wiper is +- 1 more count

[richard]

note some of the variations you are calling flutter are actually "Quantization noise" and can never be eliminated from the adc process,
the best adc result is always +-1 count. at the expense of some resolution[level bandwidth etc] you can filter it out "mostly" people with "hifi" ears reckon they can hear it on the best of cd recoding's , yet scratched vinyl records are pure, go figure. as a true cloth eared pariah i call wtf

[Ioannis]

On a PCB of a irrelevant project I tested the simple ADCin and display the result on the LCD.

As Richard says the +/-1 LSB is what I get with no process. Occasionaly 2 LSBs.

I just sample every sec. and display it.

Of course the PCB is double sided with lot of ground, a cap on the ADC input pin and linear 5Volt regulator. So any external noise is minimized.

Breadboards are just a noise antennas and never give clean results.

Ioannis

[Demon]

Ultimately, the system has to be able to read ADC while managing to 2-3 LED strips and 4 LCDs with PWM.

I'm trying a double PSU approach to see if the input PIC (right) can keep line noise to a minimum, and have all output functions on the output PIC (left).

Input functions will consume very little current, it's the output functions that will consume a lot more, and have multiple PWMs.




The only unknown is the rotary encoders; no idea if their rapid ON-OFF will create noise. I can always move them over to the output PIC.

At worse, I'd end up with a "clean" PIC, and a "noisy" PIC instead of input/output.

[Demon]

stick a 2200uF low esr electrolytic in every power rail on each bread-board and be done with it. if it still noisy add another set....

I only have a few 1000uF electrolytics, so I put 2 on each rail, same ADC flutter of mostly 1, sometimes 2.

...if that fails isolate noisy bits behind rfc's [rf chokes]....

I only have these 2 on hand:
RL-1284 100uH, https://www.electronicsurplus.com/re...uh-irms-0-4amp
79F101K-TR-RC 100uH, https://www.mouser.com/datasheet/2/5...te-1371239.pdf

...ditch the hi switching speed [noise inducing] mosfets for pwm, use bjt instead, its only a backlight...

I was having difficulties turning the backlight all the way off using 2N2907A PNP. The BS250P P-chan worked like a charm from full OFF to full ON.

...as ioannis says, breadboards and low noise are very close to mutually exclusive concepts

Yeah, that's why I moved away from my larger main prototype board to only 2 PCB boards to get ADC working perfectly.