[SOLVED] ADC offset error measurement

[shifter2013]

Hi,

I have question about how to measure offset error if we have only FFT characteristic of output signal (output signal is data between 0 and for an example 1V)?
SO how to extract offset error from output signal ?

 

[barry]

Offset error is the difference between the measured voltage versus the expected voltage for zero input. If you apply zero input and take the fft of that, the level at zero frequency is your offset error. Otherwise, there's no way to extract that information from a non-zero signal. For example, if you have a single-frequency sine wave centered around zero volts, the amplitude in your zero-frequency bin should be zero. But, what if that sine wave is not centered around zero? You'll see a non-zero value in the DC bin; but that's not offset error, that's an actual component of your signal.

 

[KlausST]

Hi,

FFT result bin0 = DC value.
All other bins are irrelevant.

If your ADC input DC voltage is +20mV, but bin0 shows +25mV, then the ADC offset is +5mV.

Klaus

 

[barry]

The main point here is that you have to apply a known input to determine your offset.

 

[shifter2013]

One more question.
Why is not the same value when we measure tone in dB after fft and when we just do calculation like 20log(Signal_Pik_voltage/votageFS) of the same signal?

 

[KlausST]

Hi,

dB is a relative value.
* maybe it relates to full scale
* maybe it relates to 1V RMS
* maybe it relates to 1V amplitude.
..or other value.

Klaus

Added:
Just recognized

Signal_Pik_voltage

I think you mean "peak". Which is the amplitude. With pure sine the peak voltage is 1.414 of the RMS value. peak-tp-peak is 2.828 of the RMS value.

a factor of..
0.707 means -3dB
1.414 means +3dB
2.828 means +9dB

Klaus

 

[shifter2013]

The FFT in matlab relates to full scale.
Pik is a value of amplitude, so for an example 1V.
So if we put the equation 20log(Signal_amplitude/votageFS) in calculator, the results are not the same.
For me is not clear why?

 

[KlausST]

Hi,

If both show "amplitude" then obviously "full scale" is not "1V".

Klaus

 

[shifter2013]

No, the full scale has another value and it is the same defined in calculations