Designer’s Guide Community

Visjnoe

Guest

DAC clock jitter
Jun 04^th, 2005, 8:09am

Hi,

I'm currently investigating (@ the architectural level) a 12b delta-sigma (oversampling) CMOS DAC.

One of the main non-idealities associated with CMOS ADCs/ DACs is the clock jitter.

Modeling of clock jitter for ADCs is fairly straightforward: one just assumes 'ideal clock instants' but adds a certain deviation dv to the 'ideal sampled values'. Afterwards, FFT processing of the data shows the effect of the clock jitter.

However, for a DAC, i can't come up with a way to model the clock jitter...of course, generating a clock with jitter is not the problem, but data that is generated in such a way can not be used for FFT processing (requires equidistant points).

Does anyone have an idea and/or experience with modeling of clock jitter effects on DAC performance?

Thanks a lot!

Peter

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: DAC clock jitter
Reply #1 - Jun 4^th, 2005, 8:46am

If you have access to Spectre you can use it. It provides two ways of performing a Fourier analysis. The FFT, built into Artist, is a discrete-time Fourier transform. The Fourier component, built into Spectre itself, computes the true Fourier integral, and so is a continuous-time Fourier transform. It will accurately catch the jitter.

-Ken

Back to top

IP Logged

Visjnoe Guest	Re: DAC clock jitter Reply #2 - Jun 6^th, 2005, 5:11am Hi, Does anyone know an alternative to using the continous FFT (requiring SpectreRF)? Greetings Peter
Back to top	IP Logged

sheldon

Community Fellow

Offline

Posts: 751

Re: DAC clock jitter
Reply #3 - Jun 6^th, 2005, 8:07am

Visjnoe,

Ken described Spectre's Fourier Analysis. This analysis
is a basic capability of Spectre and does not require a
SpectreRF license. However, the fourier element
implemented in analogLib is only a subset of the full
functionality and you should probably run Spectre from
the command line(unless you want to modify the CDF
yourself). BTW, Spectre's Fourier analysis is described
in chapter 5 of Ken's book.

Also had a question for Ken

, when using the fourier
integral for analyzing oversampling ADC/DAC is there
any method to window the data? Since for oversampling
ADC/DAC the data is not periodic outside the time
selected for the Fourier analysis, it seems like the
data needs to be windowed to prevent spectral leakage.

Best Regards,

Sheldon

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: DAC clock jitter
Reply #4 - Jun 6^th, 2005, 9:30am

Sheldon,
The only way I know to apply a window to the Fourier Integral components in Spectre is to add Verilog-A code between the Fourier component and the rest of the circuit that implements the window.

Peter,
The only alternative to Spectre's Fourier Integral that I know of is to build a Fourier analyzer directly into your circuit using multipliers, integrators, etc. Each harmonic requires a sine and cosine source, two multipliers, and two gated integrators. Such an analyzer can be implemented using Verilog-A. You would only want to do this if you were interested in very few harmonics (say no more than 3).

-Ken

Back to top

IP Logged