Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
14 years ago

Timequest makes timings but not the real stratix IV hardware

I am having problems making actual timings on the real hardware. I'm actually trying to upgrade from a stratix III to a stratix IV. The design works on a stratix III.

After I've set up a basic constraint file (.sdc), I see Timequest properly identifying my main clocks and telling me that I'm making timings with no problem. Yet my stratix IV hardware still isn't making timings. I can use the signaltap to see how some of my waveforms at some particular point begin to contain glitches like distortions. These glitches only appear as I am integrating more and more things into the FPGA. So the problems begin only as the chip uses more resources and consumes more current. Also, I can usually improve the distorted signal as I cool the chip with an anti-static freezer.

I don't know what to do because according to timequest, my Fmax is well above what I need, e.g., 149MHz when I need only 120Mhz, and 312Mhz when I need only 240 Mhz.

I should also mention that this is a multi-rate design. There is a main clock but actual data rates are controlled by toggling the clock enable of the registers. Now, I haven't declared any multi-cycle paths in the constrained file, however, I would have expected in this case that this would be an issue only if I had not made timings.

Any suggestions?

14 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    --- Quote Start ---

    What do you mean? There are all of the summary reports provided in the compilation report. According to that report, I am making timings.

    --- Quote End ---

    Hi,

    as far as I know you get only the worst-case timing as default. Maybe your design is now to fast. First you can check whether the "Optimize multi-corner timing" is switched on.

    You find it under Settings -> Fitter Settings. If it is enabled Quartus also considers

    the fast timing for the implementation. Did you specify a <>.sdc file ? If yes, can you post it ?

    Kind regards

    GPK
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    I get 2 slow reports and one fast. I pass all 3. the multi-corner timing is switched on.

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    I have identified the precise place where timings are not being met:

    I have a barrel shifter that is purely combinatory between two registers. It's quite big with about 70 plus bits in size. Both registers receive the same clock. The clock has made timings according to the analyzer. The question is: Why is the analyzer not properly calculating the delays of this barrel shifter?
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    --- Quote Start ---

    I have identified the precise place where timings are not being met:

    I have a barrel shifter that is purely combinatory between two registers. It's quite big with about 70 plus bits in size. Both registers receive the same clock. The clock has made timings according to the analyzer. The question is: Why is the analyzer not properly calculating the delays of this barrel shifter?

    --- Quote End ---

    Hi,

    dit they use the same clock enable signals ? Can you try to check the timing of the paths

    directly in Timequest ?

    Kind regards

    GPK