Designer’s Guide Community

fangjh

Junior Member

Offline

Posts: 13
hefei china

About ADC's FFT and crosstalk
Mar 29^th, 2005, 6:27pm

I have two issues:
� 1) When I �calculate ADC's(8bit) �FFT, I should use at least 512 point,but in fact I use 400 point, because the input and sample frequency is not match, and test equipment 's speed is not enough, the result become bad or good? why?
� 2) How to calculate two channel ADC's crosstalk(how to define), and I how to test ???
� I am glad to receive your solution
� �Thanks

Back to top

IP Logged

fangjh Junior Member Offline Posts: 13 hefei china	Re: About ADC's FFT and crosstalk Reply #1 - Apr 6^th, 2005, 12:04am please help me, thanks, I only want to the vary direction.
Back to top	IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: About ADC's FFT and crosstalk
Reply #2 - Apr 7^th, 2005, 2:17am

Hi, Fang,

Are you runing SPICE simulation or lab measurement?

I don't understand why do you have only 400 point. Pls go to maxim's website for applications about ADC test.

About the crosstalk. I think you must make one has zero input, and another has near full scale input. Then sample the first ADC and calcultate the output signal.

Best regards,
Yawei

Back to top

IP Logged

fangjh

Junior Member

Offline

Posts: 13
hefei china

Re: About ADC's FFT and crosstalk
Reply #3 - Apr 7^th, 2005, 4:51am

thank for ywguo's answer

1)I use 400point to calculate is not used in Hspice, is used in matlab in actual test. because the test equipment's speed is not so quick, the data will be errors sometimes. so I select some correct data to calculate, but I don't know the result is godd or bad.
2) I have test crosstalk follow what you said, but how can I know the result is good or bad. Such as 50db in AD9883's result.

Back to top

IP Logged

analogic

Junior Member

Offline

Posts: 11

Re: About ADC's FFT and crosstalk
Reply #4 - Apr 12^th, 2005, 7:25pm

you said ----
1)I use 400point to calculate is not used in Hspice, is used in matlab in actual test. because the test equipment's speed is not so quick, the data will be errors sometimes. so I select some correct data to calculate, but I don't know the result is godd or bad.

-- i m confused by your question, why not to capture more data for FFT analysis. It is better to use a sample number which is power of 2 for FFT. Even 512 is too few. You need 4096 or more to get good frequency resolution. Please read some application notes before you jump into testing.

you said --
2) I have test crosstalk follow what you said, but how can I know the result is good or bad. Such as 50db in AD9883's result.

-- crosstalk is always undesired, so the lower the better. you cannot say it is good or not without referring to a specific application.

hope it helps.

Back to top

IP Logged

fangjh

Junior Member

Offline

Posts: 13
hefei china

Re: About ADC's FFT and crosstalk
Reply #5 - Apr 12^th, 2005, 10:29pm

Thanks analogic's help

1)FFT should to �use a sample number which is power of 2 for FFT. 4096 is more better. I kown this. In fact, I
use 400 point to calculate because of actual equipment's limit ,so I have use 400 point with hann window in matlab, the result is beter or worse than use 4096;
2) ADI�s AD9883 the crosstalk is 55db, I want to know how to calculate the parameter.

Back to top

« Last Edit: Apr 12^th, 2005, 11:44pm by fangjh »

IP Logged

analogic

Junior Member

Offline

Posts: 11

Re: About ADC's FFT and crosstalk
Reply #6 - Apr 13^th, 2005, 8:56pm

1)FFT should to use a sample number which is power of 2 for FFT. 4096 is more better. I kown this. In fact, I
use 400 point to calculate because of actual equipment's limit ,so I have use 400 point with hann window in matlab, the result is beter or worse than use 4096;

-- worse. the more points, the more frequency resolution. for example, you will not see harmonics if the sample number is small.

2) ADI�s AD9883 the crosstalk is 55db, I want to know how to calculate the parameter.

-- i guess, say u have 2 channels, apply signal only at channel 1, measure output of channel 1 (P1) and that of channel 2 (P2). 10log10(P1/P2) is crosstalk. it may not be exact definition of crosstalk, but should help to characterize it.

Back to top

IP Logged

sheldon Community Fellow Offline Posts: 751	Re: About ADC's FFT and crosstalk Reply #7 - Apr 14^th, 2005, 9:32pm Analogic, Basically, the definition is up to you. If you look at the AD9229, there AD defined the crosstalk as the signal on a quiet channel when the adjecent channel is driven by a full scale signal. Best Regards, Sheldon
Back to top	IP Logged

fangjh Junior Member Offline Posts: 13 hefei china	Re: About ADC's FFT and crosstalk Reply #8 - Apr 17^th, 2005, 6:37pm Thank everyboday's help!!! 1）I thank that the harmomic will alias signal when sample and input ratio is not large(eg: 2~4) in little point. That is right or not. 2) I have test two channel. I sent full range(1 Vpp) to 1 channle, and test 2 channel always jump 1 LSB.This crosstalk is ???
Back to top	IP Logged

sheldon

Community Fellow

Offline

Posts: 751

Re: About ADC's FFT and crosstalk
Reply #9 - Apr 17^th, 2005, 7:04pm

Fangjh,

1) When simulating, a 256 point FFT should have enough resolution
provide you all the information you need, at least for 10 bit, Nyquist
type ADCs. I usually see the FFT noise floor at about -100dB for
a 256 point FFT and this is enough resolution.

You do need to be careful about input frequency selection. The correct
simulation setup has been discussed on other threads. The point is to
avoid windowing if you can, that is, when simulating setup to use
the rectangular window. Unfortunately, when using measured data
this may not be possible. However, a 256 point FFT should still
provide useful information even when using measured data.

The frequencies do "fold" back into the FFT spectrum. Not sure what
the concern is, are you somehow post processing the data digitally to
remove the harmonics at high frequency? If not, you should include the
effect of distortion in your effective number of bits calculation.

2) My guess is that 1dB lsb of crosstalk is real bad, the AD9229, had
-95dB of crosstalk.

Best Regards,

Sheldon

Back to top

IP Logged