Designer’s Guide Community

Mahdi3999

Junior Member

Offline

Posts: 12

"Phase" parameter of the input source in a PAC analysis
Jan 02^nd, 2016, 7:14am

Hi,
I am simulating a switched-capacitor filter, I know that the circuit's frequency response (in my particular circuit) depends on the phase of input source. In fact, the phase difference between input source and the clock is important. The transient simulations verify this dependence. However, changing the phase of the input source in PAC simulations, does not make any difference in my results.
I would be really grateful if anybody can help me with this.

regards,
mahdi

Back to top

IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: "Phase" parameter of the input source in a PAC analysis Reply #1 - Jan 2^nd, 2016, 8:08am What do you mean when you say the phase difference is important? -Ken
Back to top	IP Logged

Mahdi3999

Junior Member

Offline

Posts: 12

Re: "Phase" parameter of the input source in a PAC analysis
Reply #2 - Jan 2^nd, 2016, 12:38pm

Thanks for your time dear Ken,
In fact I am simulating an N-path bandpass filter with center frequency of 1GHz, i.e the filter is clocked at 1GHz. I am testing a 2-path filter. Since, in a 2-path filter, each capacitor integrates half of the input sine wave, depending on the phase difference of input sine and the utilized clock the capacitors may have zero or non-zero voltage. Since the frequency response of such an structure depends on the phase of input nobody practically uses a 2-path filter. As a matter of fact such a circuit may pass the input Vin=sin(2*pi*1GHz+0) but suppress the same input with a phase of 90 degree (Vin=sin(2*pi*1GHz+90degrees)).
I want to verify this effect by PAC simulations but as I change the "PAC phase" parameter the frequency response of the circuit does not change.

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: "Phase" parameter of the input source in a PAC analysis
Reply #3 - Jan 2^nd, 2016, 11:27pm

Quote:

Since the frequency response of such an structure depends on the phase of input

I believe you are misunderstanding the nature of N-path filters. The frequency response does not depend on the phase of the input. The filter would have to be nonlinear for that to be true. N-path filters are linear with respect to the input. However N-path filters are also time-varying, which may be what you are thinking.

Since the filter is linear with respect to its input, it is easiest to consider the input as being asynchronous from the clock. In that case the phase is constantly shifting and has no meaning.

To simulate an N-path filter with SpectreRF you need to include all the paths, run a PAC analysis, and you would sweep the input over a range of frequencies.. You should leave the PACphase at 0. Generally the only time you set PACphase is when you have two sources that are generating a differential signal. In that case one would be set to 0 and the other set to 180.

-Ken

Back to top

IP Logged

Mahdi3999

Junior Member

Offline

Posts: 12

Re: "Phase" parameter of the input source in a PAC analysis
Reply #4 - Jan 2^nd, 2016, 11:54pm

thanks for your reply,

the frequency response of an N-path filter for N>=3, is independent of the input phase, but if you consider a 2-path filter, each capacitor sees the input for half the period of the input (because there is only 2 capacitors and in one period, either of these two capacitors must connect to the input). We know that the capacitor will integrate the input signal over the time it is connected to the input. Then a phase difference between input and the clock may lead to a zero integral. In fact this is equivalent to integrating a sine over half its period and we know that this integral may become zero.

I can verify this with transient response, i.e if I check the response in time domain changing the input phase results in suppressed signal in the output. However, PAC does not predict this.

Back to top

IP Logged

Mahdi3999 Junior Member Offline Posts: 12	Re: "Phase" parameter of the input source in a PAC analysis Reply #5 - Jan 3^rd, 2016, 12:06am The dependence I am discussing is particularly clear when we apply a sine at the center frequency of the filter, Vin=sin(2pif_clock).
Back to top	IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: "Phase" parameter of the input source in a PAC analysis
Reply #6 - Jan 3^rd, 2016, 12:22am

In the case you describe, PAC is effectively equivalent to a transient analysis followed by a Fourier analysis. So if you are seeing this effect in transient, you should also see it in PAC. Perhaps your setup is wrong.

If you sweep the frequency, then you should see a peak as you pass through the clock frequency. Are you seeing that?

-Ken

Back to top

IP Logged

Mahdi3999

Junior Member

Offline

Posts: 12

Re: "Phase" parameter of the input source in a PAC analysis
Reply #7 - Jan 3^rd, 2016, 12:35am

Yes, I see the expected bandpass response, but the input phase has no effect on this response.

I don't know the method used by the simulator in detail, but it is interesting that when I change the "delay" parameter in clock signals, the PSS does not show it(when I plot the clock in time domain under PSS form). As if I never have modified this parameter. I was wondering maybe this is related to my observations.

Back to top

IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: "Phase" parameter of the input source in a PAC analysis Reply #8 - Jan 3^rd, 2016, 9:36pm The phase cannot play a role when you are sweeping frequency because the input signal and the clock signal are not synchronized. The only time the phase of the input signal could play a role is if the input signal and the clock signal have commensurate frequencies (are synchronized). -Ken
Back to top	IP Logged

Mahdi3999

Junior Member

Offline

Posts: 12

Re: "Phase" parameter of the input source in a PAC analysis
Reply #9 - Jan 4^th, 2016, 3:07am

Based on your suggestion that the phase does not play a role when the input and the clock are not at the same frequency, it occurred to me that a "a single-point PAC" simulation (at the center frequency of the filter) may give the expected result. However, this does not make a difference.

I conclude that a PAC simulation won't work in my case, am I right? If so, how can I plot the frequency response of this circuit?

A simple way would be to perform Fourier analysis on the results of transient analysis but I would rather another approach to be sure about the final results I get.

Back to top

IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: "Phase" parameter of the input source in a PAC analysis Reply #10 - Jan 4^th, 2016, 4:31pm Is the center frequency of the filter equal to the clock frequency? -Ken
Back to top	IP Logged

Mahdi3999 Junior Member Offline Posts: 12	Re: "Phase" parameter of the input source in a PAC analysis Reply #11 - Jan 5^th, 2016, 1:51am yes, the center frequency of the filter is equal to the clock frequency (f_clock=1GHz) and I did a PSS+PAC simulation in which the PAC was single-point at 1GHz.
Back to top	IP Logged

Mahdi3999 Junior Member Offline Posts: 12	Re: "Phase" parameter of the input source in a PAC analysis Reply #12 - Jan 7^th, 2016, 10:14am Thanks for your time and your helpful comments dear Dr. ken. -Mahdi
Back to top	IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: "Phase" parameter of the input source in a PAC analysis Reply #13 - Jan 7^th, 2016, 9:16pm Sorry I could not be more help. The results from PAC and transient analysis should be consistent. I cannot imagine what is causing the difference. -Ken
Back to top	IP Logged