Help to obtain pre-ringing measurement

[Leo11]

Hi all,

I'm trying to make an impulse measurement on my sound card's digital loopback to see how the pre-ringing of a room correction filter develops.

I play a full range sweep and then record the signal with digital loopback.

If I'm not mistaken I should see it with the step response.

But the step I get with REW is a symmetrical peak with the comb effect...

What are the correct window settings I should set to make that measurement?

 

[John Mulcahy]

Best posting the mdat.

 

[Leo11]

Sure, attached.

Here how it looks to me...
NOTE: no room correction filter applied in this measurement. I expect to see perfect step...
But... REW sample rate is set to 44.1 and loopback audio device at 192, so upsampling maybe happen somewhere... it's REW that upsample to match driver sample rate or it's the driver that upsample?

 

[John Mulcahy]

If HiFi Cable supports WASAPI exclusive you should select the device entries that start EXCL, then the sample rate of the device should match what you choose in REW. If you don't do that Windows will resample from the device rate to the rate chosen in REW.

The window settings do not affect the captured impulse response, but the integration of the impulse response to produce the step response uses the windowed impulse response and starts at the beginning of the left window. Generally better using windows that are flat within the region of interest rather than tapered across their full extent, Tukey is a good choice.

 

[Leo11]

Ok, thanks for the feedback.

I accidentally removed previous image, but reuploaded it with the same measurement...

However, why that comb effect is displayed? Shouldn't it be a continuous line, at most with a sinusoidal trend?

I applied Tukey window then to perform other measurements, but something is not clear to me.

This is what I get with WASAPI Shared, REW 44.1, HiFi Cable 192 (Windows is resampling)

This is with WASAPI Shared, REW 192, HiFi Cable 192 (Windows in theory is not resampling)

This is with WASAPI Exclusive, REW 192, HiFi Cable 192 (Windows in theory is not resampling)

Shouldn't the last two have the same IR?

To me it's necessary to understand in order to make proper measurement of filter mentioned previously.

 

[John Mulcahy]

Band limited responses have a sinc-like response, that's what you see. The graph uses band-limited interpolation to show what happens between the sample points, per the sampling theorem.

 

[Leo11]

Band limited responses have a sinc-like response, that's what you see. The graph uses band-limited interpolation to show what happens between the sample points, per the sampling theorem.

View attachment 62989

SORRY, I must be ignorant... how do I get that graph with REW?
I can play only sweeps...

 

[John Mulcahy]

Change the axis from dBFS to %, top left of the graph when the mouse is over the graph.

 

[Leo11]

Thanks for the clarification.
I managed to make the measurement that I had to.

If you don't do that Windows will resample from the device rate to the rate chosen in REW.

I am not sure about this.
I tried to monitor what REW transmit to the driver using API Monitor, following this instruction.
With REW set to 44.1kHz and WASAPI Shared output, and output device set to 192kHz from Windows settings, it results that REW passes data at 192kHz sample rate, so performing the upsampling.
In fact, this behavior is also seen in other players that support WASAPI Shared, like Foobar, and meant to avoid the bad Windows resampler.
I believe this is determined with this option called "wasapiAutoConvert", set by the application calling the driver.

In fact, using Exclusive mode the API Monitor shows that REW transmits data at the sample rate specified in REW's preferences.

Could you please confirm this behavior?

 

[John Mulcahy]

If you use the WASAPI exclusive device entries (names starting with EXCL) a WASAPI Exclusive driver is used and the driver will apply the rate set in REW. If you use a Java device entry without EXCL JavaSound will ask the OS for data at the rate set in REW. If the device is configured to run at a different rate the OS will resample the data from the device before passing it to REW at the requested rate. REW (as with all Java applications) has no direct access to hardware.

 

[Leo11]

Ok, I understand what you mean. We were talking about two different things though.
You talk about the capture stream, while I talk about the playback stream.
The behavior I've seen, that REW does the resample, is with playback.
In fact in this case if REW and OS sample rate differ, it turns out with API Monitor that REW passes already upsampled data to match the OS setting.

 

[John Mulcahy]

REW does not resample the audio data. Any resampling happens outside REW and outside the Java runtime.