Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

How pac analysis works in spectre??

Status
Not open for further replies.

gunturikishore

Full Member level 2
Joined
Aug 19, 2004
Messages
146
Helped
13
Reputation
26
Reaction score
6
Trophy points
1,298
Location
INDIA GANDHINAGAR
Activity points
1,363
pac analysis

Hi,
Can anybody tell me how the PAC analysis works?The doubts are as follows


1. When thr are more than one ac sources like 'vsin' or 'port' whr the small stimulus is applied in the circuit? I saw the netlisiting but it does not include any such information.

2.Which operating point is used for the small signal analysis. The documents say it uses previously calculated peridically time varying operting point.If so this small signal itself will have the frequency conversion effcts which will be again convoluted in the Frequncy domain with appripriate side band of previously calculted PSS analysis.The number of calcultions are going to increse then enormously. I have a feeling tht somewhere thr lies my understanding problem.

Thanks in advance
 

spectre pac

1. You can have your small-signal excitation input in the "pacmag" parameters of vsin source. If you have more then one input excitation, the output is just the linear combination of the response by these many excitation by superposition.

2. The pac analysis will include frequency conversion effect, i.e. you can calculate the gain from one side band to the others with different harmonics index (This is useful, for e.g., in SC filters and up down conversion mixer). You can select the amount of sidebands that you needs in simulation, and of course the more the sideband you specified, the longer the simulation time. Hope it helps and pls correct me if I am wrong.
 

spectre pss analysis

I'd like to ask one question too. I'm using PSS to estimate IIP3 of IF filter, but as every people meet, the 3rd-order curve in low-power area isn't correct. Anyway, in my case, this incorrectness occured up to the high power area that make me very hard to find the 3 dB/dB slope of the curve, compare to PSS+PAC simulation. I have to set the number of harmonic up to 500 then that 2 curves (PSS vs. PSS+PAC) can be approximated the same.

My question is why this thing happen? Since in spectreRF's theory, they said the accuracy of PSS doesn't depend on number of harmonic. Anyone has an idea?
 

pac analysis spectre

Hi terryssw,

Your comment is good in the sense tht now I remembered tht we have to use the smallsignal parameters available in the models to work with the pac pxf kind of analysis because they are small signal analysis after the PSS operating point information.

But still I did not get proper answer for my second question tht for the small signal applied I doubt only a single OP is used rather than entire Peridic Operating points of PSS.
 

pac spectre

gunturikishore said:
But still I did not get proper answer for my second question tht for the small signal applied I doubt only a single OP is used rather than entire Peridic Operating points of PSS.

If you needed to use PAC, your circuit must be driven by a large periodic stimuli (e.g. clock in SC circuit or LO in mixer), and your circuit must work under the influence of such periodic stimuli. That's is why you cannot use normal AC analysis, since your circuit does not work probably with single DC operating point. PAC always output over the whole period of PSS operating point.

So whenever you use PSS (and its sons PAC, PXF, PNOISE, PSP, etc....) you must interest in the output over one period, not just a single DC operating point since SC circuit or mixer cannot work under DC.

For example, you simulate an ideal sample-and-hold amplifier with PAC, with the signal frequency of fo and clock of fs. Then you can use PAC to simulate the gain of the sample-and-hold from input to output at the same frequency fo (of course fo will still vary with the AC simulation), or you can simulate the gain from input of fo to output at N*fs + fo, i.e. the modulated gain response with N as the "reference sideband number" (I don't quite remember is it the correct name of variable N).
 

spectre pac analysis

Hi,

I don't mean tht the PSS is not required before the pac or pxf or whatever sons or grandsons tht you call. PSS is definitely required and my doubt is abt pac small signal response in the circuit. More precisely to say if periodic operating point is used the small signal itself has a frequncy conversion effects. Then depending on the reference side band it is mixed with the appropriate band of the Large signal PSS frequency by frqquency domain convolution. My doubt is wheteher this small signal has any frquency conversion effets before tht convolution to the large signal. If Periodic Operating point is used before this large signal convolution the small signal itself will have the frequncy conversion effects ( the time domain Multiplication and hence the frequency domain convolutio are performed LATER over this nonliner small signal response depending on the reference side band tht we specify if it happens so).

Otherwise if it uses no frequency conversion techniques on this small signal (before large signal convolution effect) it should be using a single operating point
which I should say unique and only one DC operating point. I am skeptical abt this frequcny conversion effects whether applied on this small signal or not, before mixing with the PSS large signal. I am also testing with various band selections to conform the same. The problem should be more clear to u by now I suppose.
 

pxf spectre example

I think it would be better to explain in the following few points:
1. There must be only one frequency conversion effect done by the combination of PSS+PAC.

2. In PSS, you define your periodic operating frequency fs, which is also your center of the reference sidebands and also the location of the convolution in freq. domain.

3. In PAC, you define your input signal frequency fo and also the reference sideband number N, which means that from the info of PSS+PAC you will calculate small signal gain from fo+N*fs. So I think there is only one freq conversion effect, performed by PSS+PAC at the same time, since only PSS analysis does not show you any information about freq. conversion.

4. The only nonlinear effect considered in PSS+PAC is the one that caused by the large periodic stimuli nonlinearity (nonlinearity caused by clock or LO). In additions to this, no nonlinearity is considered in PAC (ie, the nonlinearity from the input signal will not be considered).
 

spectre ac è·Ÿ pac

Ya,

I think now you got my point of question.If no frequency conversion or non liner effects are produced by the pac then it should have the unoque operating point which can be only Dc operating point. Otherwise it will have the non linear effects in its response itself. But the documents like the SpectreRf Theory says otherwise that it will use the previous PSS calculated operating point. Thts the point where I was struck.
 

spectre pac pss

terryssw said:
I think it would be better to explain in the following few points:
quote]

Thanks to terryssw, it's really a wonderful explanation for PAC simulation.
Any recommendation to get more accurate result by pac simualtion?
Especially in SC filter.
Thak you,
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top