lagos.jl
Joined: 13 Apr 2006
Posts: 47
Helped: 1
|
 28 Mar 2007 17:55 vcd tetramax |
|
|
|
|
Hi all. I am experiencing a very annoying problem while trying to run a sequential fault simulation with Tetramax, using external patterns saved (via a VCD dump) from a back-annotated simulation done in ModelSim. The problem is that when I load the VCD dump file in Tetramax and simulate it, the tool produces wrong simulation results, exhibiting an akward behavior. The following caption explains the whole situation:
h**p://images.elektroda.net/37_1175093373.jpg
****[Upper part of image: Modelsim 5.8b simulation - Results: OK, as expected]****
Notice the timing of the clocks PAD_EMTCK and PAD_CLK. At the beginning they have periods of 40 and 20 ns, respectively. Then, @ 27,240ns, signal PAD_SCAN_EN goes up and PAD_CLK assumes also a 40 ns period (the other clock remains unaffected). Notice also the 40ns-wide pulses appearing on signals PAD_SCAN_IN1...3.
****[Lower part of image: Tetramax Z-2007.03 simulation - Results: WRONG]****
This is Tetramax's simulation as obtained through the reading of modelsim's vcd dump through the following command:"set patterns external my_dump.vcd -sensitive -strobe period 40 ns -strobe offset 37 ns" (note that this command implies the sampling of the test patterns at periodic intervals of 40ns, with a suitable offset of 37ns that should allow for an appropriate sampling of stable values in all waveforms).
Notice now the awful part: at the beginning Tetramax seems to get it right, and defines the test patterns at regular intervals 40ns apart. But then misfortune happens: after the edge of PAD_SCAN_EN, Tetramax incredibily does two wrong things:
1. Contrary of what was specified in the "set patterns" command, it starts to sample the test patterns EVERY 20ns!!!!
2. It starts mis-simulating everything: the periods of both clocks switch to 20ns, the pulses appearing in PAD_SCAN_IN1...3 are 20ns-wide ,etc.
Therefore, since the dumped file is not appropiately simulated by Tetramax, I cannot perform the desired fault simulation, which I desperately need.
I would be the most grateful if anybody could please give a clue to what is going wrong here. Why doesn't Tmax respect the 40ns-period samping? Am I missing something in the "set patterns" command? I have struggled for 2 weeks with this problem without success.  I have even tried using other versions of Tmax, and also giving different values to the "offset" parameter, without any luck. I have also verified the vcd dump file's correctness with a third tool, with positive results.
Any suggestions are really, really welcome!!!
Thanks in advance!!!
Regards, JL.
|
|
lagos.jl
Joined: 13 Apr 2006
Posts: 47
Helped: 1
|
13 Nov 2007 6:22 tetramax fault simulation |
|
|
|
|
Just for the sake of completeness, I would like to inform that this issue was finally circumvented long time ago.
It turned out that, well, when simulating the dumped vcd Tetramax basically did what it wanted instead of what it was supposed to do according to the documentation. Even the (auto-generated) command-line instructions it inferred from what we entered using the dialog windows had errors on them. The moral to this story would just to be VERY carefull with this piece of (largerly bugged) software, and try, try, try and try again until, maybe with a lot of luck, the darn thing starts working. 
...I still recall earlier times when I naively thought serious EDA were unbugged and user-friendly; oh boy can't believe how wrong I was!
Jorge.
|
|
