High frequency roll-off with DIRAC Live Processor

[PersechiniA]

I have tried a variety of settings, i.e., ASIO vs wasapi, and have been unable to eliminate high frequency losses through Dirac Live Processor (lower curves), compare with straight through to a topping E30. Both sets of curves were collected with REW using a USB connection to a topping E30, with umik-1 and the manufacturer's correction file. Please note that I applied at 300 Hz cut-off to the filter. Psychoacoustic filtering. This difference in FR is audible to me. While I realize that these results suggest I can do without Dirac, which is what I am doing, I did pay a pretty penny for the software to correct my KEF Q350 speakers (I am now using LS50 meta speakers) and this behavior appears to be a significant problem. I overlooked this until now because I did not make measurements outside Dirac.

My specific questions are:
(1) Has anyone else seen this kind of thing?
(2) What might I be doing wrong that could account for this?

Tony

 

[ppataki]

Did you perform a 9-point (or more) measurement?
If yes, try to perform a single-point measurement
Just place your umik at the MLP (Main Listening Position) and run the first 3 measurements at that single point and then check the results
Also this will enforce a more aggressive correction of your dips between 50 and 90Hz (which you might actually find beneficial, if no excessive ringing will occur)

 

[PersechiniA]

Yes, I did do several measurements. But I applied a 300 Hz cut-off on the filter, so the high frequencies should have passed unaffected. This information is in the image caption, but I forgot to add it to the post body. Sorry about that!
I just edited the post to make this clear.

 

[theyellowspecial]

Is the sampling rate consistently 48,000Hz or higher across the audio chain?

 

[PersechiniA]

That is a good question. It is set to 48000 in REW, and the display on the E30 DAC reads 4x that value for the Dirac Processor output (because I selected that rate to take advantage of Qubuz hifi output). What I do not know is whether there is a lower sampling rate somewhere in the chain. It should be said that there is a similar roll-off in the Dirac Live measurements, I just wasn’t paying close attention until now. The sampling rates being applied definitely warrant a closer look—perhaps there is a problem with a windows setting somewhere. Or some sort of sampling rate mismatch.

 

[jonfitch]

I would imagine there would be no audible difference since they is barely any content in most music above 8k in most music, but have you tried gating corrections above 10k? Does that still cause an FR drop-off?

 

[Χ Ξ Σ]

I have observed the roll-off as well when I were using Dirac.

Mostly likely this is because the sample rate of REW didn't match the sample rate of your computer and/or dac. I have done measurements showing a similar roll-off in my older setup. Have you checked if the sample rates displayed on REW, E30, and computer are the same?

Another possibility could be the VST/AU plugin host software you use for Dirac Live Processor plugin. When I used Audio Hijack to host the DLP AU plugin, it always showed a high frequency roll-off starting around 14-16khz. Switching to SoundSource the roll-off was gone, although I can still see a bit of zigzag above that region. I am not sure if you are using the VST version or the newer standalone software which I have never tried.

 

[PersechiniA]

Turns out to be a sample rate mismatch. Clearly an newbie mistake, which is what I am. As far as the "audible" difference I thought I heard, well, the curves made me hear it.

Thank you all for the input. This is a great forum.