Designer’s Guide Community

Yutao Liu

Community Member

Offline

Posts: 76
Guangzhou, China

How to verify the quadrature clock signal?
Mar 19^th, 2012, 7:29am

Hello everyone,
I want to check whether the quadrature clock signal is 90 degree phase shift between each other. The clock is square waveform.

The old way I used is to measure the time difference between rising edge or falling edge of two waveform t_d, and then use the equation: phase shift = �t_d/T_clk*360, where T_clk is clock period. However, I suspect the precision of this way, because it is difficult to define the measuring point of t_d.

So I want to use the correlation critique (1/T)*integral(I(t)*Q(t)*dt), where I(t) and Q(t) are the quadrature clock waveforms. Is this a effective and reliable way?

Welcome any other approaches. Thanks for advanced.

Best
Yutao

Back to top

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: How to verify the quadrature clock signal? Reply #1 - Mar 19^th, 2012, 10:55am hello, how you are planing to evaluate the integral, r u using any numerical technique? i guess finding the phase difference is much better and accurate, i donno why you said measuring point Td is different, get the max+min/2 point and find the time corresponding to that. Thanks, raj.
Back to top	IP Logged

RFICDUDE Community Fellow Offline Posts: 323	Re: How to verify the quadrature clock signal? Reply #2 - Mar 20^th, 2012, 6:31pm The approach that I use is to take the FFT of each of the clock signals then compare the amplitude and phase of the fundamental signals. This works well for linear circuits that depend on the the quadrature clocks (like mixers), but I do not know if it is adequate for clocking/timing circuits.
Back to top	IP Logged

Yutao Liu

Community Member

Offline

Posts: 76
Guangzhou, China

Re: How to verify the quadrature clock signal?
Reply #3 - Mar 21^st, 2012, 6:25am

Hello raja,
Thanks for your suggestion. I also use (max+min)/2 to define the measuring point in my old method. A tiny overshoot in the waveform would deviate (max+min)/2 from the ideal point. So I am afraid of that this would affect the measurement accuracy.

In correlation critique, I am planing to use the calculator in Spectre to do the integral. I(t) and Q(t) are the transient simulation result in Spectre as well. Therefore, I am confused the transient simulation precision has effect on the integral result.

Is my concern unnecessary?

I think the method proposed by RFICDUDE is similar with correlation critique method.

And my quadrature clocks drives a direct up-conversion mixer.

Thanks

Yutao Liu

Back to top

IP Logged

Yutao Liu

Community Member

Offline

Posts: 76
Guangzhou, China

Re: How to verify the quadrature clock signal?
Reply #4 - Jul 26^th, 2012, 8:33am

RFICDUDE wrote on Mar 20^th, 2012, 6:31pm:

The approach that I use is to take the FFT of each of the clock signals then compare the amplitude and phase of the fundamental signals.

This works well for linear circuits that depend on the the quadrature clocks (like mixers), but I do not know if it is adequate for clocking/timing circuits.

How do you get the phase of the fundamental signals? Do you run a transient analysis and then DFT to get the real part and image part of the fundamental frequency? This simulation consumes lots of memory and hardly get result in parametric sweep. Is there any better method?

Thanks.
Yutao

Back to top

IP Logged

RFICDUDE

Community Fellow

Offline

Posts: 323

Re: How to verify the quadrature clock signal?
Reply #5 - Jul 26^th, 2012, 5:29pm

Yes, I run transient analysis and obtain the amplitude and phase from a DFT.

I am confused why you are having trouble with memory using transient simulation unless you are saving all nodes and currents in the circuit. If you are running many parametric sweeps then consider saving only the nodes of interest in the circuit.

Other simulation methods consume even more memory to find the steady-state solution you are interested in finding.

Back to top

IP Logged

Yutao Liu

Community Member

Offline

Posts: 76
Guangzhou, China

Re: How to verify the quadrature clock signal?
Reply #6 - Jul 29^th, 2012, 1:09am

RFICDUDE wrote on Jul 26^th, 2012, 5:29pm:

Yes, I run transient analysis and obtain the amplitude and phase from a DFT.

I am confused why you are having trouble with memory using transient simulation unless you are saving all nodes and currents in the circuit. If you are running many parametric sweeps then consider saving only the nodes of interest in the circuit.

Other simulation methods consume even more memory to find the steady-state solution you are interested in finding.

Thanks for your suggestion! It's a good idea to save only the nodes of interest.
But I have another question on computing the amplitude and phase from the DFT.
I tried to use the "harmonic" function provided by the calculator in spectre to get the level of the fundamental frequency from DFT, but it does not work. And there is the warning that "harmonic" function can not handle "scaler " data from DFT. So which function in the calculator do you use to compute the amplitude and phase? Or you mark the fundamental frequency by yourself?

Thank you again!
Yutao Liu

Back to top

IP Logged

RFICDUDE

Community Fellow

Offline

Posts: 323

Re: How to verify the quadrature clock signal?
Reply #7 - Jul 29^th, 2012, 5:19am

You can use the "value" function to select a dft result at a particular frequency. The dft function returns complex values (rectangular I think).

Here is an example from the calculator that takes the dft and returns a value at 1GHz.

value(dft(v("vtest" ?result "tran") 50e-9 250e-9 4096 "Rectangular" 1 "default" ) 1e9 �?period nil ?xName "time" �)

Another option is to use the Fourier analysis component that you place in the schematic and it will perform a Fourier analysis of the signal at that node and at the fundamental frequency you specify. This might be the easiest way for you to get the complex harmonics you are looking for.

The fourier analysis schematic elements are in analogLib under category Analysis.

Back to top

IP Logged