Seemingly random number of repetition in stepped sine mode - measurements take very long

[Ratterbass]

When using the stepped sine mode, the software sometimes does dozens (I counted) repetitions before actually starting with the averaging.

Sometimes it does just one or two passes of capturing the signal before starting the averaging and moving onto the next step, but sometimes it takes 20 or more passes with no obvious (at least not to me) reason. I understand taking one or two passes before starting the actual measurement makes sure the DUT has settled and that there are no errors, especially when using rectangular window but surely 20+ seems a bit excessive.

This is with 0 settling time and 0 waiting time set up in the stepped sine dialog.

Is this a bug? Am I missing something?

Using latest build (but also happened on earlier ones) on MacOS 14.

 

[John Mulcahy]

Please provide information about the measurement setup and test configuration. Do the replay and capture devices share the same clock?

 

[Ratterbass]

Please provide information about the measurement setup and test configuration. Do the replay and capture devices share the same clock?

this happens with multiple tested devices (several audio interfaces of varying quality such as RME ADI 2/4 pro, arturia audiofuse, zoom u44). Playback and recording device is the same (audio interface).

This even happens with direct loopback (line out -> cable to line input), so there is no DUT other than the interface itself.

 

[John Mulcahy]

Not something I've come across on any of the interfaces I've tested. There is a stability check for the fundamental on the first input in THD vs frequency tests if that's the test you are referring to. Worth including a screenshot of the stepped test dialog so I can see all the settings and perhaps an example mdat file for a test result.

 

[John Mulcahy]

Please also include a screenshot of your REW soundcard preferences.

 

[Ratterbass]

Here's some testing with the ADI 2/4, measurement taken with XLR loopback from channel 1 output to channel 1 input:

Behavior is the same for THD vs. Level and THD vs. Frequency

I've recorded two short-ish screen captures of the process (please note the spaces in the link as I am not yet allowed to post links):

https:// we. tl/t-2novqbSgxJ


REW preferences:

MacOS preferences:

RME preferences (although other interfaces show the same behavior as mentioned and devices like the U44 don't even have their own preference editor)

The resulting measurements are attached as well. Note: I forgot to disable the -1dB treble rolloff which I had active for my speakers, but this does not affect the behavior.

 

[John Mulcahy]

First issue will be the buffer sizes selected in the Soundcard preferences. When the stimulus changes REW has to wait long enough for the buffers to flush through, the stability check can shorten that wait if the signal stabilises sooner. I normally use 16k buffers on macOS so try that.

 

[Ratterbass]

Thank you, that seems to take care of the issue. I guess I never really noticed the "k" in the buffer settings in REW. Makes sense that 512k samples at 48kHz SR take a while to flush.

However, I do wonder why sometimes the capturing already started after just two passes?

I also ran into trouble with the input data not being detected, which I wanted to open yet another thread about but I'll do some further testing to see if that was caused by the buffer as well.

Thank you so far for the very quick response!

 

[John Mulcahy]

No input data on macOS has been reported when the generator is started before input capture with some interfaces.