Forum Discussion

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

FFT 7.1 streaming mode

Hi folks,

I have two issues with the FFT 7.1 MegaCore in streaming mode. It is configured as 256-point 10-bit FFT.

1. when simulated in ModelSim, the functional model generated by MegaWizard always produces a constant exponent output. The value of the exponent is always the maximum for the selected FFT length. In this case, it's always -11. The post-synthesis model produces the correct exponent, which varies with the input.

2. when debugged in-circuit via SignalTap, the FFT outputs appear to be misaligned. The exponent part always leads I and Q by one frame. No timing violation was reported by Quartus. However they appear to be aligned in simulation using the post-synthesis model and fed with the same data (captured via SignalTap).

I have sent request to my local FAE. But I'm just wondering if any of you have experienced the same.

Francis

4 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    18 years ago

    Did you happen to "Upgrade" from the 6.1 streaming or earlier core?

    I know the streaming interface is different, and although the conversion document states how the signal names are re-mapped, they don't mention that the new core is much more picky on both start and stop of the controls.

    I'm using the 7.2 core in streaming mode now, without any problems. But I've know Altera has had issues with specific version and specific settings in the past.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    18 years ago

    Thanks for the reply anakha. I discovered the second problem was due to that I didn't follow some 'fine print' in the user guide (2-19 Compile the Design):

    ensure the fft_pack_fft_71.vhd file is at the top of the list in the

    file name list window.

    Once I followed the instruction, the frame misalignment problem and some other strange issues were gone! I'm still puzzled why Quartus doesn't produce an error or warning message to stop this from happening.

    My first problem still stands and I'm still waiting for a reply from Altera.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    17 years ago

    Did you ever resolve the FFT v7.1 vho problem?

    I have been having problems with the FFT v2.2.1 vho so I tried updating the FFT Megacore to v7.1. I get the problem you are having with the maximum exponent (-16 in my case) and the SOP/EOPs are not correctly generated.

    Post synth netlist doesn't work properly when I create a netlist for just the FFT itself, and if I create a post synth netlist for full signal processing block ModelSim runs out of memory (I have 16 FFTs). The only way I am currently able to simulate is to generate I/O in ModelSim, export WLF to VCD, perform a Quartus simulation, export the results to VCD, then convert to WLF. It's not a quick or pleasant way to simulate.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    17 years ago

    I'm not sure whether you are referring to the same problem. But I did resolve the problem# 1 in my original post.

    It was caused by a race condition between clock and async reset to the fft block. Once the condition was removed, I got expected simulation result.

    We don't do post-synthesis simulation here so I can't confirm the problem you were having.