PAC Simulation: outputperiod option

 Hi Albert,

Which version of the IC tools and simulator are you using, and what are the analysis statements in the netlist? I don't recall having seen such an output file.

Also, pac normally gives you the time-averaged response, so I'm assuming you must be using a "sampled" mode? It's not clear what simulation you're actually doing.

Regards,

Andrew.

Hi Andrew,

I am simulating pac using an option called "outputperiod".  The option supposes to give me time integreated dc output.  Please see below.

pac  pac  sweeptype=relative  relharmnum=0  start=0
+    annotate=status  freqaxis=out  outputperiod=period

With this option set, the simulation output will give me a file called pac.real.pac, in additional to the normal pac output.  The Result Browser is not able to show this additional output result. The file exists (~800k in size), but I cannot look at the result in the file.  

I am using  Virtuoso 6.1.4-64b and I am using MMSIM72.243.

Thank you,

Albert

Recently, I've found that there is an option called "outputperiod" in PAC simulation.  According to the reference manaul, it lets you see the time-domain small-signal response. I am curoius on how this time domain small signal reponse is computed in Cadence. 

This is how I've simulated with the output period option with period equal to 1ns.

pac  pac  sweeptype=relative  relharmnum=0  start=0
+    annotate=status  freqaxis=out  outputperiod=1n

I've tried to  simulate the circuit without the outputperiod option and computed the time response myself.  The settling below gives me 50 harmonic output sidebands.  I then take the reverse fourier transform to obtain the time function.The input frequency is at a single point: 0Hz.

 pac  pac  sweeptype=relative  relharmnum=0  start=0
+    maxsideband=50  annotate=status  freqaxis=out 

I compute the time domain output by the following equation for each time point t: 

H(0; t) = sum_{m = -50}^{m = 50} H_{m}(0) * exp(j * m * wc * t)

where H_{m}(0) represents the mth harmonic relative to the input (0Hz) at DC. wc is the fundamental frequency = 2*pi/(1ns).

For some reason, I did not get the same time domain waveform as if I directly use the Cadence outputperiod option. I've checked my math a few times, but I didn't see any problems.  The outpuperiod gives more correct result.   Could anyone please clarify how Cadence compute this output?

 Many thanks,

Albert

I've found the bug.  In the Cadence simulation output directory, there is a file called "logFile." When doing the PAC simulation with outputperiod option on, it does not write to the logFile correctly, resultting in (Result Browsers/Direct Plot) not able to read the result. 

A fix from Cadence may be needed.

Albert, I merged your two posts.

My understanding has always been that the outputperiod was an option added in the early days of SpectreRF to allow R&D to debug things, and as far as I know it does not  behave sensibly - for example, if you sweep frequency in the pac analysis, you get a combined waveform for all the points. I suspect at some point the logFile has been incorrectly updated, or ViVA not checked to ensure it supports this output file.

In practice, I think the outputperiod option is of no use. As far as I can see, there have been no service requests or CCRs on this option, so my assumption (since there are things about it which don't work), is that nobody has tried to use it either.

If you wish to persue this further, please log a service request via http://support.cadence.com so that this can be discussed with R&D further, but I don't believe it is intended to be a useful option for users. In fact I am tempted to request that the option be removed...

Best Regards,

Andrew.

Hi Andrew,

It is interesting to know that this option is just for debuggin.  Actually, it is not a completely useless function.  It lets you calculate time-varying DC gain.