Spectral leakage in stepped sine mode

Discussion in 'Official REW (Room EQ Wizard) Support Forum' started by keantoken, Apr 12, 2019.

  1. keantoken

    keantoken Member
    Thread Starter

    Joined:
    May 31, 2018
    Messages:
    29
    Likes Received:
    0
    In the normal spectrum mode I can resolve the THD of my setup, but whenever I used a stepped sine measurement, spectral leakage obscures the harmonics as you can see in the screenshot. I think this is due to the absence of windowing in stepped sine mode.
     

    Attached Files:

  2. John Mulcahy

    John Mulcahy REW Author

    Joined:
    Apr 3, 2017
    Messages:
    1,432
    Likes Received:
    240
    The generator is synchronised to the FFT length for stepped sine so that it is periodic within the length, removing the need for windowing. What is your setup? If the generator and input are on different devices their clocks could be sufficiently different to cause a problem.
     
  3. keantoken

    keantoken Member
    Thread Starter

    Joined:
    May 31, 2018
    Messages:
    29
    Likes Received:
    0
    You guessed right, different clocks. Motherboard DAC, USB ADC. But the solution is still to use windowing, no?
     
  4. John Mulcahy

    John Mulcahy REW Author

    Joined:
    Apr 3, 2017
    Messages:
    1,432
    Likes Received:
    240
    Yes, I have added a Window option for the next beta release.
     
  5. keantoken

    keantoken Member
    Thread Starter

    Joined:
    May 31, 2018
    Messages:
    29
    Likes Received:
    0
    How does the silence interval work in stepped sine mode? This seems like a poorly conditioned input as we have to know the measurement period in order to know what silence interval we want, and we shouldn't really need to do that calculation in the first place.

    What makes more sense to me is a duty cycle. So for instance if I have a 300W amp with a 30W heatsink (or a 30W PSU), I need to do testing with a 10% duty cycle in order to not overload.

    Instead of a duty cycle maybe this should be an average power input. So at 0dbFS, if I want 10% max power, the duty cycle would be 10%. At -3dbFS, power is halved, so duty cycle should be 20% to achieve the same average power output. This approaches what is called isothermal conditions, as the average temperature does not change.

    Secondly, if this is used to avoid average power limits, then the measurement interval is a critical number. A unit might survive a 1 second measurement interval but not a 4 second one. So the measurement period MUST be known. Also for this reason, averaged measurements should be broken up into individual test/rest periods, so the test periods don't string together into an unnecessarily long period.
     

Share This Page