A guide how to do room correction and use it in Roon

Hello! I apologize if my question has already been asked by someone. When using Convolution files, I have problems playing DSD. The sound quality is greatly degraded. Is it possible to use Covolution for DSD files?
Thank you in advance for your answer and help.

It’s not possible to do DSP on a DSD stream, so what Roon does (when needed) is to convert DSD to PCM, perform DSP on PCM, and then back to DSD.

Up-sampling to DSD works better though since then DSP is performed on PCM before the conversion to DSD.

Respectfully, some setups support convolution DSP to native DSD with no conversion to PCM. There’s some explanation here in the community about how this works, but I don’t have the reference handy.

Requires a lot of CPU, but it works and sounds great in my system.

Hmm, I don’t know the tech stuff deep enough to say for certain, but I think some sort of internal conversion is needed, and they seem to add the modulator again later on which confirms this.

Wow, @Magnus this is great. Thanks so much!

1 Like

Hi @Magnus - I finally found a post from @brian that explains this:

Have a look and let me know what you think.

1 Like

I actually read that before, forgot it until now. I would guess that 64-bit wide DSD and PCM are rather similar, but the conversion becomes a little faster this way.

…a year later

Just rereading these guid posts, thank you @Magnus for your time and attention to helping skill up the community.

When generating and saving the IR filters, why not “Normalize to peak values”? Would this not produce a more detailed filter, without compromising on the selected measurement SPL?

Thanks again for this great guide!
I just went through the process and am happy with the result, sound-wise. However, I had to push headroom to -7dB to completely prevent clipping… the clipping indicator turned on just occasionally and for short moments on some tracks, but still. So I’m just wondering: Is there any problem caused by using that much headroom?

Also wanted to share my results with you. EQ limited from 20 to 800Hz. Open for any further improvement advice or comments from the pros!

It depends on if you use integers or floats, someone else who has more technical knowledge might be able to answer this better, but when using float there is no use with normalizing to peak value so best to leave it unchecked.

Personally, I always use 32-bit float and no normalizing.

1 Like

Good day! Just started to deal with REW.

This is how my setup looks like:

Trifonik, which consists of an acoustic speaker Airpulse A100 and a Bowers & Wilkins PV1D subwoofer.

Measured but with different microphone positions.
The first measurement is the 90 ° calibration file and the microphone in an upright position. The second is a regular calibration file and a microphone in a horizontal position. True, in this measurement, the subwoofer gain was set to + 2dB. Usually set to 0, so anything from 20 to 100Hz can be skipped.

The question is - which of the above measurement options will be more correct?

Here are the links to the measurements:

For clarity, here is the equalization according to the @Magnus instruction for the left and right channels.

When measuring the microphone horizontally:

When measuring a microphone in a vertical position:

And here you can see that there is a difference in the measurement results depending on the position of the microphone … This is especially noticeable from 2000Hz, just at the cutoff frequency of the tweeter. Therefore, it is interesting to know in which position the measurement will be more correct?

I take measurements using a UMIK-1 USB microphone.

Finally, 2 questions:
1) How do I get REW to remove the lower hump at 85Hz?
2) Is it possible with REW’s capabilities to raise the middle from 2000Hz?

And how can I best put them together in one big picture? As for example, Dirak does it when it equalizes two channels at once so that there is no difference in perception.

Thanks in advance for your answers!


Cool system. PM sent.

1 Like

@Magnus have something to say?)

The only filter where REW automatically corrected from 2000Hz is 1/48 smoothing.

But again, the hump at 85Hz does not want to get out of it.

And in general, the frequency response looks ragged ((

How can I tweak the frequency response neatly and smoothly?

use psychoacoustic smoothing instead

it’s a dip, and you shouldn’t correct it, because that’s very likely a standing wave null at your measurement/listening position.

did you measure with the moving mic procedure or a static mic position?

1 Like

Thanks! How can I get REW to do the correction from 2000Hz?

No, I used a static microphone position. To be honest, I didn’t understand how to measure with a moving microphone.

In Dirac, I used 9 statistical zones to measure:

And can you please in more detail about the standing wave? How did it come about?

Seems, you did the measurements with the subwoofer contributing, which could cause phase cancellation with the desktop speakers
And since the sub is offset to one side, the effect is different for both channels.
Standing waves are a characteristic of all listening spaces.
Move your mic around more and do measurements to see the wild differences in bass response caused by standing waves in your room.

This way, you’ll better even out position dependent dips and peaks which results in a better match between measurement and listening impression.

1 Like

Does that make sense to do in a desktop setup where the listener will always be in the same seat/position?

I think very much so, since there will be a lot of early reflections from nearby boundaries contributing to combing effects.
Imagine that the microphone is a small capsule at a singular point in space vs. ones head and pinnae, and you’re likely not sitting with your head in a vice.
Also, due to the small distances involved, minimal movements equate to large relative changes.

I’m coming from having read relevant books and my practical experience with measuring speakers and room acoustics in a DIY setting since about '96.
Let’s see what @Evgeny finds out, if he chooses to try and get back to us with real world experience!

1 Like