Designer’s Guide Community

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Measure SNR of DAC with spectrum analyzer
Dec 28^th, 2011, 6:07am

Hi Guys,

As well known, we usually measure output spectrum of DAC with spectrum analyzer. It is easy to calculate SFDR or HDs automatically. However, is it a reliable tool to calculate SNR? For example, the following is a picture that is measured with Agilent E4404B spectrum analyzer. The clock frequency is 50 MHz, and the signal frequency is 813 kHz. The SFDR is 69.94 dB. To calculate SNR, I use the formula below.

signal_dB = -5dB
noise_floor_dB = -88dB
signal = 10^(signal_dB/10) = 0.316 mW
RBW = 1000 Hz
Assume Nyquist_band = 25 MHz and noise floor is flat in Nyquist band, although the stop frequency is 2.5MHz only in that spectrum plot.
noise = 10^(noise_floor_dB/10)*(Nyquist_band/RBW) = 39.6 μW
SNR = 10*log10(signal/noise) = 39.02 dB

It seems that SNR is too low for a 10-bit DAC. But I don't know what's wrong with the above derivation. Any comments are appreciated.

Yawei

Back to top

SCREN094.GIF

IP Logged

raja.cedt

Senior Fellow

Offline

Posts: 1516
Germany

Re: Measure SNR of DAC with spectrum analyzer
Reply #1 - Dec 28^th, 2011, 9:00am

hello,
i normally do in matlab this calculations, but seems your calculations are correct. But it's Quite common to get less than 6*n SNR at nyquist rate, due to SNR droop with clock frequency. How much your o/p swing? because higher swing DAC's this is droop is much faster.... This lesser SNR can be come from many sources like data dependent delay of the current switch, tail current impedance �variation with clock frequency.

Can you check SNR with lesser clock frequency, if it's SNR is more at lower frequency means the above factors are impacting.

Thanks,
Raj.

Back to top

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Measure SNR of DAC with spectrum analyzer Reply #2 - Dec 28^th, 2011, 11:18am hello, one more thing, did you measured DNL, because with DNL you can estimate low frequency SNR. Thanks, Raj.
Back to top	IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Measure SNR of DAC with spectrum analyzer
Reply #3 - Dec 29^th, 2011, 7:00am

raja.cedt wrote on Dec 28^th, 2011, 9:00am:

hello,
i normally do in matlab this calculations, but seems your calculations are correct. But it's Quite common to get less than 6*n SNR at nyquist rate, due to SNR droop with clock frequency. How much your o/p swing? because higher swing DAC's this is droop is much faster.... This lesser SNR can be come from many sources like data dependent delay of the current switch, tail current impedance �variation with clock frequency.

Can you check SNR with lesser clock frequency, if it's SNR is more at lower frequency means the above factors are impacting.

Thanks,
Raj.

You are right. I have another spectrum that is measured at 8 MHz clock. The signal frequency is 130 kHz. Obviously it has lower noise floor. Ignore the skirt of the signal, which I think is caused by the phase noise of clock source. The calculated SNR is around 50 dB.

Back to top

SCREN088.GIF

IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Measure SNR of DAC with spectrum analyzer
Reply #4 - Dec 29^th, 2011, 7:06am

raja.cedt wrote on Dec 28^th, 2011, 11:18am:

hello,
one more thing, did you measured DNL, because with DNL you can estimate low frequency SNR.

Thanks,
Raj.

Yes, I did measure DNL. But what's the relationship between DNL and SNR at low frequency?

I find a few literatures about DAC SNR. One is The Data Conversion Handbook from ADI, http://www.analog.com/library/analogDialogue/archives/39-06/Chapter%205%20Testin..., the other is http://grouper.ieee.org/groups/1658/NextMeetingArchive/061030DAPONTE/DAPONTE_P.p....
They clarifies the definition and test procedure for me.

Yawei

Back to top

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Measure SNR of DAC with spectrum analyzer Reply #5 - Dec 29^th, 2011, 7:22am hello, please find the attached low frequency relations.... One more thing did you corrected for noise floor based on the fft points? Thanks, raj.
Back to top	relations.png IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Measure SNR of DAC with spectrum analyzer Reply #6 - Dec 29^th, 2011, 6:18pm Thank you, Raj. The spectrum analyzer does not sample the signal and do FFT. But the noise floor need to be corrected because resolution bandwidth (RBW) is not 1Hz in my spectrum plots. On pages 5.20-5.21, The Data Conversion Handbook, it also corrects the noise with Process Gain Best Regards, Yawei
Back to top	IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Measure SNR of DAC with spectrum analyzer
Reply #7 - Dec 29^th, 2011, 7:03pm

Hi Raj,

One thing I am not sure is attenuation in spectrum analyzer. I read Agilent Spectrum Analysis Basic Application Note 150. But I am not an expert of test and instrument. I don't know if it is correct to compensate it when calculating SNR. If YES, the SNR in my first example would be 49.02 dB (50 MHz clock), and SNR would be around 60 dB for my second example (8MHz clock). Do you have any ideas?

Yawei

Back to top

IP Logged

raja.cedt

Senior Fellow

Offline

Posts: 1516
Germany

Re: Measure SNR of DAC with spectrum analyzer
Reply #8 - Dec 30^th, 2011, 2:44am

hello,
i guess you have to correct for RBW, because noise you have measure in the 1 Hz bandwidth, so you have to subtract 10log(RBW) from the results. Normally when you measure Phase noise it does this correction, let me check in the lab after two days.

I guess after correction, at lower frequency 60db and at higher frequency 50db is fine i guess..what do you say. At least this will say that at higher clock frequency all non-idelaities come into picture. Refer the following pap.
https://lirias.kuleuven.be/bitstream/123456789/171625/1/Gielen_3_Full_Paper.pdf

Thanks,
raj.

Back to top

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Measure SNR of DAC with spectrum analyzer Reply #9 - Dec 30^th, 2011, 7:13am Hi Raj, I have corrected RBW in my first example. In analog's The Data Conversion Handbook, it is corrected too. SNR =Signal/Noise_floor - 10log₁₀(Fs/2/RBW) Now I am not sure whether noise floor is attenuated as the signal. I am going to lab in a few days. Thank you very much. Yawei
Back to top	IP Logged

raja.cedt

Senior Fellow

Offline

Posts: 1516
Germany

Re: Measure SNR of DAC with spectrum analyzer
Reply #10 - Dec 30^th, 2011, 7:50am

hello sir,
what do you mean by noise floor is attenuated as the signal. I am guessing your DAC working fine, because in your previous post you shown you get 60dB at lower frequency. For me it looks no problem. One more things it's hard to maintain constant SNR till fs/2 is difficult, what is your architecture? by choosing carefully segmentation depth some people made this droop lesser.

Thanks,
Raj.

Back to top

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Measure SNR of DAC with spectrum analyzer Reply #11 - Jan 6^th, 2012, 2:16pm hello ywguo, i have verified about this agilent spectrum analyzer, you have to subtract 10log(RB). So your ADC is working. Enjoy Thanks, Raj.
Back to top	IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Measure SNR of DAC with spectrum analyzer Reply #12 - Jan 6^th, 2012, 11:03pm Hi Raj, What is RB? Do you mean Resolution bandwidth (RBW)? Thanks Yawei
Back to top	IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Measure SNR of DAC with spectrum analyzer Reply #13 - Jan 7^th, 2012, 2:34am sorry..yes it's RBW. So is your problems solved regarding this dac. I have checked with people who are in lab, what they is for Phase noise it does this subtraction by default where as for Spectrum we have to do this. Thanks, Raj.
Back to top	IP Logged