Questions about ENOB definition

[lionelgreenstreet]

I have some doubts about enob definition:
1)Is it correct to define enob as signal to (quantization) noise of an adc expressed in db?
2)For example, if a 20 bit adc has enob equal to 18 bit: what does it mean?
Thanks

 

[yxo]

enob standard deviation


from
17 Nov 2008 15:55

 

[Kral]

enob formula

lionelgreenstreet,
Also se the article "Does ENOB Tell the Whole Story" by Bonnie Baker.
www.edn.com/article/CA6549029.html?text=Does+ENOB+tell+the+whole+story? -
Regards,
Kral

 

[lionelgreenstreet]

resolution enob

Thanks for infos and link,
i know its definition, but i try to understand its pratical meaning...after reading, i understand that two definition of enob exist: with and without quantization error (correct me if i'm wrong). I think that enob is equivalent to snr (if all sources of error (quantization noise,kt/c,distortion, ecc) are considered as noise), only expressed in bit. Is this right?

 

[yxo]

bonnie baker 1.76 is quantization error

Thanks for infos and link,
i know its definition, but i try to understand its pratical meaning...after reading, i understand that two definition of enob exist: with and without quantization error (correct me if i'm wrong). I think that enob is equivalent to snr (if all sources of error (quantization noise,kt/c,distortion, ecc) are considered as noise), only expressed in bit. Is this right?

ENOB always includes quantization noise. Actually, after integration of the quantization noise we get the relationship between SNR and ideal NumberOfBit (SNR=6.02*NoB+1.76). THis is ideal case(only math is used) which means, that, for example, SNR of a 10-bit ADC won't be better than 61.94 or even equal. Because in the real world many non-ideal things occur and when we measure real ADC we see much more worse result. Another worlds, we meausure ADC, see on the screen of spectrum analyzer SNR worse than 61.94, place it to the formula and get number of bits less than was in ideal case. ANother example example if we have a 10 bit ADC and ENOB about 9.5 bits and it is sin on the input of ADC, on the output we will see that low bit will always change. But in the datsheets is written SNR that is non -ideal but all harmonic were cuted in the particular bandwidth. SINAD - harmonic were not cuted. ANd SINAD have to be insert to thouse formula. If you insert SNR it won't be honest . But in the case of a static input signal (fin =dc) it have to be all 10 bits. SNR, SINAD,ENOB are dynamic parameters.

 

[lionelgreenstreet]

enob 1.76

Thanks of help...

ENOB always includes quantization noise. Actually, after integration of the quantization noise we get the relationship between SNR and ideal NumberOfBit (SNR=6.02*NoB+1.76).

In link posted by Kral is wrote:
"Use the following calculation to derive ENOB from SINAD: ENOB=(SINAD–1.76)/6.02. "..."if you apply a dc signal to a delta-sigma ADC and record a large number of samples, you can then derive the standard deviation for these codes. The formula for ENOB is: N–log2(σ), where σ is the standard deviation of data and N is the number of converter bits."
Two different definition of enob are present: i think that if i make multiple measurement of the same dc signal, and quantization noise is the only non-ideality present, i obtain always the same output code. So, repeatability of measurement is guaranteed: in this case quantization noise shoud be unimportant in enob measurement..is this right?

ANother example example if we have a 10 bit ADC and ENOB about 9.5 bits and it is sin on the input of ADC, on the output we will see that low bit will always change.

This is the big question:
if i have a 10bit adc and enob equal to 9.5bit, if i measure the same sinusoid, in multiple measurement only the last bit change??For example, if i compare output codes associated to sinusoid peak obtained by multiple measuremnt, only the last bit change?Or i should consider noise free resolution (enob-2.7bit) to obtain the real nuber of unchanged bit?