I used a verilogA module in my schematic which measures the period of the output waveform of a block.
The verilogA code is as follows:-
`include "disciplines.vams"(*instrument_module*)module frequency_meter(in); parameter real thresh=0; // threshold (V) parameter integer dir = 1 from [-1:1] exclude 0; // 1 for rising edges, -1 for falling input in; voltage in; integer timing; real t0, t, period, freq; analog begin t = last_crossing(V(in) - thresh, dir); @(cross(V(in) - thresh, dir)) begin if (timing) begin period = t - t0; freq = 1/period; $strobe("period = %rs (measured at %rs).\n", t - t0, $abstime); end t0 = t; timing = 1; end endendmodule
I want to plot freq versus time using OCEAN script.
Can anybody please tell how it can be done.
I did :
save I5:period I5:freqplot(I5:freq).
This gives error in OCEAN script.
But in input.scs file I worte:-
save I5:period I5:freq
I ran the simulation and in the wavescan I was able to plot freq-vs-time.
In the spectre netlist, you can add:
myOpts options saveahdlvars=all
or in OCEAN you could say:
saveOption( ?saveahdlvars "all" )
And then to reference the freq signal, you'd do:
plot(getData("I5:freq" ?result 'tran))
Or you could make the frequency an output pin of the meter instead - but that involves changing it a bit so that you are contributing to a node - for your needs the above is probably sufficient?
In reply to Andrew Beckett:
Thanks a lot.
Now, I am beginner to verilogA coding.
I have some doubt regarding the above verliogA code.
1:- $Strobe prints at what instants
2:- Looks like it is printing even if there is NO thresold cross
In reply to RFStuff:
It should print when there's a threshold crossing and timing==1. That's what I see. Are you sure you don't have threshold crossings?
Theree are thresold crossings. But it is printing ( what is inside $strobe )even between the zero crossings.
To me it looks like it is being printed at the end of each simulation time step ? This I am not able to understand.
Can you put together a simple example (just with voltage sources, analogLib res/cap) which shows the problem? Then post the input.scs (also the version of spectre that you're using).
I misinterpreted the result. It is printing as you have mentioned.
1:- I have some doubt regarding the initial parameter value of "t0" and "timing" in the 1st iteration.
2:- Could you please tell what change I have to make to make freq as an output node in the code ?
3:- Could you please tell how to plot it in GHz unit instead of GV ?
1. The initial values will be 0 - this is in the VerilogAMS language definition.
2 & 3 - see the model below. By specifying the output as a frequency discipline, the graph comes out with the right units too (I'd not seen your third point until I'd updated the model, but it solves 3 too...)
`include "disciplines.vams"// could specify the fout signal as being electrical// but it's nicer to do it as the specific discipline - this// way we can have a custom abstol and blowup, since the// magnitude of the frequency will be much larger than // voltages in the circuitnature Frequency units="Hz"; access=Freq; abstol=1; blowup=100G;endnaturediscipline frequency potential Frequency;enddiscipline (*instrument_module*)module frequency_meter(in,fout); parameter real thresh=0; // threshold (V) parameter integer dir = 1 from [-1:1] exclude 0; // 1 for rising edges, -1 for falling input in; voltage in; output fout; frequency fout; integer timing; real t0, t, period, freq; analog begin t = last_crossing(V(in) - thresh, dir); @(cross(V(in) - thresh, dir)) begin if (timing) begin period = t - t0; freq = 1/period; $strobe("period = %rs (measured at %rs).\n", t - t0, $abstime); end t0 = t; timing = 1; end // use a transition filter to ensure that the frequency doesn't // change discontinuously Freq(fout) <+ transition(freq,0,100p); endendmodule
Whether setting abstol and blowup in this code affects the accuracy.
Actually, I am seeing some differnece in the frequency plots ( i.e. periodic spike ) in this code as compared to the code ( without having a pin)
blowup won't, abstol might. Depends on what frequencies you're dealing with - ideally abstol should be about a millionth of a typical frequency that is being measured.
Also, the transition time should not be excessively sharp if the frequency is low. Or it may be too short - I don't know what frequency ranges you're dealing with.