Odd Impulse Response

[mortod]

First of all I'm a newbie so please bear with me.

Secondly I am only using REW to analyse results from another measurement/filter generation system, namely InguzAudio and DRC (I'm implementing room correction for a squeezebox based system).

I've imported the impulse respone generated by the InguzAudio app, and have the following questions:

1. The documentation for REW states that 32bit WAV impulse responses can be imported. I can only import as 16bit (which is OK as I have used Audacity to convery from 32bit-float). Is this a known feature?

2. My impulse response (in the %FS view) looks to be much as expected, although the graph tails off after just 50ms, rather than, say, 500ms in the documented examples. Should that concern me?

3. With db FS for the y axis, the impulse response looks very odd - a) the background nose seems quite high at -30db before the inital spike, and b) the shape continues to look odd afterwards.

The frequency response curve does however look OK, and the filters generated by DRC do make a positive change to the sound of my system. But in trying to use REW to better understand my system's behaviour I seem to be getting somewhat confused. Any help appreciated.

[brucek]

My impulse response (in the %FS view) looks to be much as expected, although the graph tails off after just 50ms, rather than, say, 500ms in the documented examples. Should that concern me?

To me, it looks like you don't have sufficient signal above the noise (-30dB) to produce normal looking results.

You're dBFS plot looks like it should, given the horizontal scale you've chosen.

With your response graph, your scale is being kind to you. Change it to a more realistic vertical of 45dB-105dB. You could also turn on smoothing to get a better look at it.

brucek

[JohnM]

The 32 bit import is integer rather than float. If you use 16 bit you should normalise your responses first to make the most of the resolution, but the main problem looks to be poor signal/noise in the initial measurement you are importing. You should try using a longer impulse response window nonetheless, perhaps 150 - 200ms, to get some resolution in the low frequency part of the measurement.

[mortod]

Yes, I managed to get a better signal to noise ratio by using 'Stereo Mix' rather than 'Line' input and that did the trick- see below.

One further thing - I don't seem to be able to make the RS SPL cal file change the shape of my frequency response (by loading or clearing in settings). Am I missing something?

[brucek]

One further thing - I don't seem to be able to make the RS SPL cal file change the shape of my frequency response (by loading or clearing in settings). Am I missing something?

Yeah, both the soundcard and meter calibration data aren't included in the impulse response. The mic/meter and soundcard calibrations are only applied when calculating the frequency response.

To apply or remove a mic/meter or soundcard calibration, simply load or clear the cal data as required and press the Apply Windows button (invoked when you click the IR Windows ICON) to recalculate the frequency response with a new calibration file.

Man, you have one huge bass peak at ~43Hz eh?

brucek

[mortod]

That's what I'm doing - I import the impulse response, view it in 'Filter Adjust' tab. Load in the SPL cal file, and press Apply Windows. Same again when clearing the calibration. The response never changes the shape, though the cal curve is displayed.

[brucek]

Huh, very interesting, I've used this feature quite often with measurements I've taken and it works fine.. Perhaps there's a small bug in this area as a result of the import. I'll send JohnM (author of REW) a PM and see what he thinks.

brucek

[JohnM]

That's actually intentional, though undocumented, behaviour. Imported impulse responses will have been generated by some other measurement system so it would usually not be appropriate to apply the REW soundcard cal or mic cal to the data being imported, as REW cannot know what calibrations or corrections had been applied when the impulse response was captured. I'll add a feature request to have the option of applying either or both cal files to imported responses.