S
Svenn Are Bjerkem
Guest
Hi,
I have a well proven flip-flop that I want to run some characterization
runs on. In a previous question I was made aware of SpectreMDL. Fine, I
thought, let's give it a try.
I wrote a netlist of my testbench out of artist and started hacking mdl
files on the command line. I have used the dflop example in the mdlref
pdf as template. I have successfully run the dflop tutorial example,
just to have that mentioned.
Of course, my bench is not working, and I try to debug what happens.
Spectre writes a .raw directory that I try to read with wavescan. That
works, but when I rerun the simulation there is no way that I can update
the signal graphs. I have to close the wave window and double click on
the signals in the list again. Very irritating, indeed.
Next is that the simulation does not compute. I get the error message:
Error found by spectre.
Error in search binary function, check search condition for start
My mdl look something like:
// Control card
alias measurement setup {
export real vddelay, outcross, Tsetup, vcdelay, setdelay, maxq
run tran1(stop=500p)
vddelay = cross( sig=V(D), thresh=0.45, dir='rise, n=1 )
vcdelay = cross( sig=V(CLK), thresh=0.45, dir='rise, n=1 )
outcross = cross( sig=V(Q), thresh=0.45, dir='rise, n=1 )
maxq = max(Q)
setdelay = VDIN:delay
Tsetup = vcdelay - vddelay
}
search VDIN:delay from binary( start=350p, stop=450p, tol=1p ) {
run setup
} until (setup -> maxq < 0.45 )
and the signal drivers look something like:
VDIN (D VSS!) vsource type=pulse mag=0 phase=0 val0=0 val1=vvdd \
period=1 delay=0 rise=50p fall=50p width=1
VCLKN (CLK VSS!) vsource type=pulse mag=0 phase=0 val0=0 val1=vvdd \
period=1 delay=200p rise=tflank fall=tflank \
width=pwratio/fclk-tflank
If somebody has an idea where I should start looking for the error or
how to convince that wavescan in standalone mode to reload the
simulation _and_ update the graphs, then I would be very thankful.
--
Svenn
I have a well proven flip-flop that I want to run some characterization
runs on. In a previous question I was made aware of SpectreMDL. Fine, I
thought, let's give it a try.
I wrote a netlist of my testbench out of artist and started hacking mdl
files on the command line. I have used the dflop example in the mdlref
pdf as template. I have successfully run the dflop tutorial example,
just to have that mentioned.
Of course, my bench is not working, and I try to debug what happens.
Spectre writes a .raw directory that I try to read with wavescan. That
works, but when I rerun the simulation there is no way that I can update
the signal graphs. I have to close the wave window and double click on
the signals in the list again. Very irritating, indeed.
Next is that the simulation does not compute. I get the error message:
Error found by spectre.
Error in search binary function, check search condition for start
My mdl look something like:
// Control card
alias measurement setup {
export real vddelay, outcross, Tsetup, vcdelay, setdelay, maxq
run tran1(stop=500p)
vddelay = cross( sig=V(D), thresh=0.45, dir='rise, n=1 )
vcdelay = cross( sig=V(CLK), thresh=0.45, dir='rise, n=1 )
outcross = cross( sig=V(Q), thresh=0.45, dir='rise, n=1 )
maxq = max(Q)
setdelay = VDIN:delay
Tsetup = vcdelay - vddelay
}
search VDIN:delay from binary( start=350p, stop=450p, tol=1p ) {
run setup
} until (setup -> maxq < 0.45 )
and the signal drivers look something like:
VDIN (D VSS!) vsource type=pulse mag=0 phase=0 val0=0 val1=vvdd \
period=1 delay=0 rise=50p fall=50p width=1
VCLKN (CLK VSS!) vsource type=pulse mag=0 phase=0 val0=0 val1=vvdd \
period=1 delay=200p rise=tflank fall=tflank \
width=pwratio/fclk-tflank
If somebody has an idea where I should start looking for the error or
how to convince that wavescan in standalone mode to reload the
simulation _and_ update the graphs, then I would be very thankful.
--
Svenn