aaron_do
|
Hi all,
so this is what i eventually did. Please comment on whether this method is good or bad. I used a transient noise simulation which ran for 40 clock cycles with a square wave input. I did altogether 250 runs which results in a total of 10000 samples. Of the 10000 samples, I found there to be 2 bit errors when the input voltage was 1 mVpk, so for 1 Vpk maximum signal swing, I can say that the BER is approximately 1/5000 with log(1/0.001)/log2 = 10 bit accuracy. The actual input-referred noise is of secondary importance since we are more concerned with the BER. For accuracy I believe I should have enough runs for about 30 errors or so...
Does this all sound right? I have a feeling I didn't calculate the bit accuracy properly.
cheers, Aaron
|