The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> spur question of fft spectrum in ADC
https://designers-guide.org/forum/YaBB.pl?num=1401359646

Message started by cchen on May 29^th, 2014, 3:34am

Title: spur question of fft spectrum in ADC
Post by cchen on May 29^th, 2014, 3:34am

It's a 50MHz 14bit pipeline ADC. In the chip measurement, a spur (blue line) shows the largest power except signal. I'm not sure the spur is noise or distortion. Different window methods are tried in fft program and the results are similar.
Any comments about this spur? What's the possible root cause?

Title: Re: spur question of fft spectrum in ADC
Post by cchen on May 29^th, 2014, 3:35am

here's the whole fft spectrum.

Title: Re: spur question of fft spectrum in ADC
Post by sheldon on May 29^th, 2014, 4:48am

CChen,

Not sure what it could be, have you looked at the clock to see
how clean it is and at the input source to see how clean it is?

Sheldon

Title: Re: spur question of fft spectrum in ADC
Post by cchen on May 29^th, 2014, 10:41pm

Hi Sheldon,
Thanks for your replying. The analog input source should be ok. The clock may have a risk to be not so clean.
Single-ended clock from equipment is converted to differential one by a balun, and is converted to single-ended clock again using a comparator on PCB, and then goes into the chip through analog IO. Actually, the on-chip clock is treated as pure digital signal with analog power supply driving standard cells.
If the clock is the problem, why does the spur always locate besides the signal for inputs with different frequency?

Regards,
cchen

sheldon wrote on May 29^th, 2014, 4:48am:

CChen,

Not sure what it could be, have you looked at the clock to see
how clean it is and at the input source to see how clean it is?

Sheldon

Title: Re: spur question of fft spectrum in ADC
Post by RobG on May 31^st, 2014, 8:55pm

Can't say if it is the cause, but you will get skirting like that if your sampling is not coherent. In addition to having the ratio correct you should sync the clock and input sources using the trigger in/out connections on the back of the generators.

Title: Re: spur question of fft spectrum in ADC
Post by cchen on Jun 3^rd, 2014, 12:54am

If the sampling point for fft analysis is increased, for example 16 times, the spur disappears and the HD3 turns to have the largest power. As a result, SFDR increases a lot.

Title: Re: spur question of fft spectrum in ADC
Post by BackerShu on Jul 18^th, 2014, 1:50pm

Congratulations!
But I think the fft points only makes the spectrum resolution better. It should not kill the spur if it really exists.

I guess RobG may actually have the point about coherent sampling. How did you increase the fft point? The way you increase the fft points may actually get close to coherent sampling relationship.

Appreciate if you can enlighten me why the spur is removed.

Title: Re: spur question of fft spectrum in ADC
Post by eternity on Jul 22^nd, 2014, 8:05am

Hi there,

In order to have a promising coherent sampling situation. try letting the tool set the input frequency by setting the other variables needed to compute it (fin = (n/N)*fs) {n-number of integer periods of signal; N-FFT points;fs-sampling frequency}. This is to be done this way not to approximate the input frequency to two or three decimal points accuracy.

Like the other colleague mentioned here, increasing FFT points should bascially distribute the noise bins wider. If increasing FFT points resolves your issue, I suspect only the coherent sampling condition setting.

Eternity