Creating convolution files for Roon in the new Room EQ Wizard (REW) with Inversion

The soon to be launched REW includes new trace arithmetic features and I have uploaded a video in YouTube explaining a new room correction technique which does not use any EQ filters and creates Roon ready convolution files directly in REW with speaker response inversion over the Harman target curve.

There is also a very comprehensive supplementary text guide which replaces my 2.5 years old guide listed in this community. You can find the links to the new guide and the early access version of REW (v5.20.11) in the comments section of the video.

The correction results are quite promising and I would love to hear your feedback!

10 Likes

Looks like a promising way to create convolution filters. I tried quickly on my latest RTA measurement but could not get it to work (only silence when using the resulting convolution file). I will try later with new sweep measurements.

Hi Magnus,

It needs to have a phase response as well as FR to generate a convolution file. RTA measurements usually do not have phase response, maybe that’s the reason?

Btw, there are also lots of improvements in the new REW on the RTA measurement front which is your specialty and you should have a look:

I like RTA with moving-mic because it gives such an excellent average, sweeps (even when averaging several of them) seem less accurate. But I will do some new sweep measurements and try your method.

By the way, operations A/B and 1/A should work without phase, but I assume generating min phase won’t work. But these are new operations, so maybe they are a little buggy.

I know what you mean but it’s not really a bug. The regularisation percentage embedded causes some inaccuracies. However, with all the fast fourier transformation calculations behind the scenes, this was the only way. I know because I worked with John on the beta of this release for sometime, you can get absurd peaks and dips at extreme frequencies in vector calculation otherwise. Using 3.6 dB target level and 8% regularisation as explained in the guide, seem to yield very good results.

1 Like

I tried the whole process according to the video, and I am impressed with the result. Very natural and almost a hint of analog sounding compared to my old correction (I used your HK house curve, hard to know how much that affects).

If you have a well-treaded room you can use a lower regularization though, and even allow notches, this is the predicted response I get when using regularization 1% (0% don’t work btw) and not skipping notches (as you can see the notches are still handled well).

2 Likes

I am glad to hear that it worked for you Magnus, your approval of the process was important for me. Did you manage to set your RTA measurements as base? I will try different regularisation and notch option. Frankly, John Mulcahy advised me to use 8% and I didn’t try too many other values with the latest version (I had done some trials with previous betas). I am more focused now on target dB level of the target curve to amend the bass response. Btw we both have the same “impossible to get rid of 47Hz” dip!

1 Like

The trace arithmetic doesn’t seem to work with RTA, so this is with 2 sweeps (long ones, 40 seconds iirc) at the listening position.

Btw, you might want to remove the part about aligning SPL, since that’s done as part of the process anyway (you correct to the same target).

I had to add that part later because while I was convolving the correction impulses directly with Equalizer APO and re-measuring with REW, one time, I saw that L&R SPL changed by nearly 0.5db in the opposite direction after correction. It could well be EQ APO acting odd because It didn’t repeat afterwards but I decided to add that SPL check at the very end as equalizing SPL around 315Hz by 2 octaves each way ensures the phantom image is at dead centre.

Try generating minimum phase version of the RTA measurement and work with that may be.

Hi @Magnus
That’s a nice target curve. Can you please share the coordinates
Thanks

You can download it from the youtube guide if you look at the guide on youtube.

Thanks
I am just waiting to have a go using this method with REW

1 Like

Hi everyone, thanks for this detailed method of room correction, Serkan. I want to try it out.
Im a bloody Beginner in REW first I have some questions:

Is it possible to do this with a flat target courve or only with Harman?
I have a stereo setup with two aktiv subwoofers that are connected to my aktive studio monitors each side, cutoff-frequenz is set to 80Hz for both in the subs. How would I properly measure this setup? each loudspeaker and subwoofer for itself I guess? And then follow the instructions for the two bookshelf and subs separately?
How do I get it together and set the subs correctly in REW to get the filters in Roon after that?
I think using RePhase in the second step (the other Video) is not possible/usefull for aktiv speakers, right?

Thanks for your Help!

The inversion will work just the same around a flat target but you will hear less bass and more high frequencies at the end unless you listen to your system at exactly -83 db. The target curve intends to compensate for differences in our hearing of different frequencies at different volume levels.

As per sub integration with active speakers, I have neither in my stereo system so I can only speculate. I am guessing your active speakers take care of the crossing over digitally and they will not cause phase shifts. But overall phase still needs to be checked after integration with the subs imo. You could start by aligning your subs with your speakers using MSO (multiple sub optimizer - a free, very efficient tool) and then correct the overall response with this method in REW and that would work. In REW you can double check the time alignment of subs with your speakers in excess group delay graph. I am sure someone here with sub experience will be able to give more up to date information.

Ok, thanks for that advice Serkan! I will try MSO first.
But what about the measuring after that, ist it correct to measure subs and bookshelf separately after the subs are aligned and do the Inverted method twice (for speakers and subs separately) or should I take them as one full-range speaker for each side and measure them together?

I would treat them as one full range speaker with a quite low bass cut-off frequency and use that correction convolution in Roon. There are other ways of doing it, even using Roon DSP crossovers with two different convolution files for the subs and speakers, etc. but this will only complicate things.

to summarize: time and volume align the subs with speakers in MSO, then make two measurements for left and right and correct these with the method to create one common convolution.

OK, that’s the way I would try first, I will give you feedback when I have time to do it.
Last question for now: what do you mean by: “with a quite low bass cut-off frequency”? do I have to set the low bass cut-off inside REW although my Subs are set to 80Hz cut of already?
(sorry for my dumb questions but I try to understand and if I try it for the first time, I want to do it right :smiley: )
thank you very much for your tips!