Loopback measurement IR difference between 5.20b5 and 5.19

[3ll3d00d]

I was checking something and noticed that 5.20 shows a quite different IR to 5.19 when presented with a digital loopback. Is this expected? If so, why?

Measurement is via a motu 1248 looping back through the driver alone, no DAC/ADC round trip involved, running on windows 8.1. Same REW settings (I think) in either case.

5.19

5.20b5

 

[John Mulcahy]

That's a result of this change from 5.20 beta 2:

The way the inter-sample behaviour of the impulse response is generated has changed. It used to be a simple Hermite interpolation, but now uses band limited oversampling to produce a result that shows the same shape regardless of fractional sample offsets in the sample positions. A side effect of that has been to enable more accurate determination of the value for the IR peak when it lies between samples, improved precision for the estimated IR Delay and more precise acoustic timing reference delay estimates.

In essence, the IR now shows the correct inter-sample response, obtained by windowed SINC interpolation. That can be disabled using the IR oversampler setting in the Analysis preferences.