A Question About Generating Impulse Responses of Equipment Measurements

[Otavio]

Hello there,

I have recently begun using REW, and today I tried to make some measurements on my Steinbeg UR-RT2 audio interface. The .mdat file is attached in this post.

The microphone input of this interface has a button that activates a transformer which is intended to color the sound in some way. I wanted to know what the changes of activating this transformer were, so I:

1) Calibrated the interface properly, without the transformer being activated, with a connection from the L output into Input 1 as a loopback, no acoustic output involved;
2) Measured the input with the transformer now activated.

The plot below shows the response I obtained from the measurement of the input with the transformer activated.

The results were in agreement with my expectations of the transformer's behavior.

Now, here's where I might be mistaken: I believe that if I export the impulse response of this measurement of the transformer-activated input as a WAV impulse response, I could get the changes in coloration of this transformer by convolving the exported impulse response with another signal such as a voice or guitar recording. This is what I wanted to do, intending to use this within a DAW.

However, I attempted to generate the impulse response of this measurement, and the IR that was exported effectively does not change the sound. In case it is of interest, here are the settings I used to generate the impulse response.

So here are my two questions:

1) If everything is done correctly, is it correct to assume that an exported impulse response of this measurement, when convolved with a signal, will be able to imprint the phase and amplitude behavior of the measured device into the signal?

2) Is there some mistake in the described process that could lead the generated impulse response to be different in its amplitude and phase characteristics from the originating measurements in REW?

In terms of the convolution process and settings of the plugin, I'm sure they were all properly set as I know the plugin and have worked with other impulse responses of rooms and equipment before in it. As I'm new to REW, I imagine I may have missed something important when going through measurement and exporting.

I'm sorry if there were any details left out, I'll inform anything extra that might be needed to properly assess this situation, and thank in advance any help that can be given.

 

[John Mulcahy]

The process is fine, but there is a caveat. The exported impulse response is the linear response of the system, distortion products are removed from it by the sweep measurement process, so you won't get the effect of the transformer's rising low frequency distortion. The response is pretty flat, with just a 1 dB lift around 10 kHz, so the effect would be very subtle.

 

[Otavio]

Thank you very much for the answer, John! I'm glad the process was correct.

What I actually intended on obtaining from the convolution was much more this distortion aspect that you talked about rather than the soft amplification on the higher frequencies.

At least a general distortion behavior seems to be picked up in the plot that resulted from the sweep measurement. Is this distortion plot that was generated by the sweep measurement correctly showing the distortion behavior of the measured transformer?

The exported impulse response is the linear response of the system, distortion products are removed from it by the sweep measurement process, so you won't get the effect of the transformer's rising low frequency distortion.

Is there any functionality of REW that allows me to export this distortion behavior in some way so that I can use it elsewhere for convolution or filtering, recreating this measured frequency-based distortion? Perhaps using a different form of measurement, or exporting it a different format other than a WAV IR? Thank you again for the answers, as I believe these must be basic questions.

 

[John Mulcahy]

The distortion plot is correct. Whilst the distortion can be measured, distortion cannot be reproduced in a convolution engine. Convolution is a linear process, non-linear processing is needed to generate non-linear (distortion) outputs. As a trivial example, if you have a processing block whose output is the input plus some proportion of the square of the input, the output will have second harmonic distortion.

 

[Otavio]

I see! It's a pity that it can't be reproduced by convolution, but being able to properly plot the transformer distortion as above is already very satisfying. Thank you once again for the time answering me.

 

[DanDan]

Nice try Otavio. I recommend the CraneSong Phoenix plug in.