ADC problem in Zigbit module from Meshnetics

[anujaa]

Hello Friends,

I am using Zigbit module of Meshnetics which consists of ATMega 1281.
Supply : 3V battery.
ADC Specifications:
Resolution: 10 Bit
Reference Voltage : 1.25V
Accuracy :±2 digit LSB.

I have applied constant input(1V) to ADC1 pin.
For this input, ADC count should be 819.2
But the count value I am getting is between 812 to 832.

Why I am not getting count between 818 to 820?

I want a constant count value.How to compensate ADC error?

Any help would be appreciated.

Thanks in advance

 

[joby16]

adc to meshnetics

Hi Anuja,

May be some noise is getting added with the input signal.. You can try adding an RC filter circuit in the input..

 

[xulfee]

adc in atmega noise

which ADC(ic number) u r using?

 

[anujaa]

ADC problem

ADC is inside the MCU

 

[FvM]

Re: ADC problem

ADC noise depends on many parameters, e. g. ADC clock, possible usage of the ATMega "noise canceler" (AD conversion in CPU sleep mode), reference voltage and hardware design. Does Meshnetics give the +/- 2LSB specification? I doubt, if it is realistic with 1.2 Vref anyway. If it's a trustworthy manufacture spec, the problem should be expected in a noisy input signal. The ADC is sampling the input voltage, it's aware a lot of noise you won't see with a multimeter! In this case, the suggested RC filter should help.

Usual software noise reduction technique is digital filtering by low pass or averaging of samples.

 

[nandhu015]

Re: ADC problem

by FvM

Usual software noise reduction technique is digital filtering by low pass or averaging of samples.

Can you explain a little bit more?
and how can it be applied?

Nandhu

 

[FvM]

Re: ADC problem

1st order digital filter can be done like this:https://www.radiolocman.com/shem/schematics.html?di=36513

Averaging is achieved by adding n samples and dividing by n, for n being a power of two, divison can be replaced by rightshift.

 

[Siniša]

Re: ADC problem

LSB value with 1.25V reference would be 1.22mV. You can exceed that in noise level fairly easy.
It looks that you have small ammount of offset too.

±2LSB is refering to nonlinearity errors. Offset can be adjusted in software fairly easy. For noise, you should start first in reducing and if possible eliminating sources of noise, like already suggested putting MCU to sleep while sampling, making sure you have nice clean power supply for MCU, filtering input suignal as much as practicable, conditioning signal with buffer/filter opamp,...
Further filtering could be done in software where simple averaging will do wonders. Once you have noise under controll you can attack offset and span errors by doing software compensation.

 

[FvM]

Re: ADC problem

I didn't see a Meshnetics specification for expectable effective resolution or accuracy of ATMega1281 under specific operating conditions. I also noticed, that all Atmel datasheet specification are for 4V reference. So I think, there is no assured resolution or accuray specification for 1.25 reference operation at all. But better results should be achievable as Sinisa explained, although the sleep mode acquisition is may be not possible due to device firmware framework requirements.