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

Hi Markus, this is also what i have somehow ā€œsensedā€ by trying this method in the past. s there a way to restrict the Qā€™s or restrict the gains in REW? I understand ā€œflatness curveā€ acts as a minimum deviation threshold but I have not seen ā€œmax deviation allowedā€ type of settings? Is that the reason you revert to manual (and yes, I have seen the relevant post of yours).

I have also noticed that REW cannot correct the major dips (especially the R channel ones). I understand this should not be attempted for null points but I cannot tell if my dips are due to null points or not (have not done the excess phase analysis to detect them)

Gains can be restricted but thats not the biggest problem, steepness (high Q) is. But try auto-correct with only a few filters available and then you can see the details in the EQ filters window. Its typically the narrow corrections that requires high Q so with to few filters available REW will only correct the bigger problems.

Or experiment and do it manually, Q=5 filters should work pretty good in your case except for a less narrow filter (q=2.5 maybe) to lower 150Hz and surrounding in first channel, and similar Q in second channel to raise 200Hz with surrounding. No need to touch anything above 300Hz. But I think you can get good results by lowering available filters and let REW do it.

Hi Markus, would something like that make sense ?

I tried to keep Q at 5 with one exception but i have used ā€œmodalā€ for the dips (in which case Q is not an input)

I never experimented with ā€œmodalā€ filtertype in REW so not sure how that works, and with var-smoothing which does very little at lower frequencies its hard to see because of all the small dips and peaks. Maybe you should use 1/12 or even psycho acoustics.

Try this:

  • Set smoothing to 1/12
  • Disable all but 5 filters in EQ filter window
  • Let REW do auto-correct betweem 20 and 700Hz
    I think that should give a conservative but good result

And yes, I know I suggested var-smoothing earlier but in your case its probably not optimal :slight_smile:

OK :slight_smile: will try

So having had the acoustics guy in this is where heā€™s managed to get the room to without any panels or other treatment - just placing the speakers and setting the sub by ear. And honestly, it now sounds so good I am not sure whether to mess around with convolution files etc.

Turns out the guy is a bit of a genius when it comes to sorting room acoustics out having worked on a host of recording studios and most super-clubs worldwide. Was nice to have him round to sort out my modest little room :slight_smile:

I listen to a good chunk of vinyl as well so having the room well set up means it sounds good regardless of source. May give one a go though just because I can :).

3 Likes

@dannybgoode
Got a good bit smoother, but the coarse scaling is deceptive here.
Thereā€™s still almost a 20dB fall from the low bass room mode to the upper bass to lower midrange average level, and then almost 10 dB of a rise thereafter.
I strongly encourage you doing the REW shuffle and whipping up some convolution files! And if your room layout is not symmetrical, do separate filters for left and right channels.
The improvement, again, will blow your mind, I guarantee you!

The room was tuned with me in it and listening to an hour or so of music and honestly it works for me just fine. Itā€™s been set up for my sonic taste as it is.

I may mess around with convolution because I can but I have no pressing desire to

1 Like

And sorry should have said that is the left and right traced just the room as it is

Hey, everythingā€™s fine that makes you happy!
No pressing from my sideā€¦enjoy!

Iā€™ll have a play for sure because I can however the room was tuned for how I like music to sound. The whole idea in the first place was to mess with such things after all. It was only when asking for help on how to interpret the results did I get this guy recommended to me and he has done a solid job just working with what I already have.

I have already applied a ā€˜perfectā€™ convolution file (flat, no house curve etc) and found it too bright at the top end which is exactly what we tuned out. I suspect that if I play with a bit of EQ Iā€™ll just end up back where I am with no convolution :rofl:

That dip in red is gigantic, luckily its at 70Hz and you have no dip in other channel so it wonā€™t matter that much (you donā€™t hear direction of 70Hz, at least not very good). The main thing I would correct is to raise the 70-250 area, it is a few dB to low which will lead to music sounding a little less warm.

A flat correction will sound bright and thin, thatā€™s why you have a house curve. It has to do with how the ear perceive sound and frequencies (we hear higher frequency as louder than lower frequency with the same volume).

1 Like

Bit of a cut and paste from another forum this as Iā€™ve been discussing the same thing but to give you an idea of what the process was with the acoustics guys. This was also my first attempt in measuring with REW so my technique may be a bit off as well :slight_smile:. Doubt it would make a huge amount of difference but I didnā€™t use your spiral method hear rather just pointed the mic at the speakers.

We spent a good two hours listening to my music in the room though and discussing what I liked and did not like about the sound with Mike & Nico fettling until it sounded right to me. Not what measured right or what they thought was the ideal; it was about how I liked things. I have absolutely no doubt that had I instructed them to make the room measure flat they would have done so, albeit perhaps that would have necessitated either some form of DSP and/or panels etc.

To me my system now sounds pretty much perfect - I donā€™t find it lacking warmth, itā€™s just beautifully balance to my ear. The dip is interesting and is something to do with the big wardrobe but donā€™t ask me to explain what!

For completeness I have done a combined L&R measurement 1/6 smoothed and overlaid over a ā€˜perfectā€™ house curve and also against a completely flat response as well as a 1/1 smoothed one. Note a perfectly flat response sounds pretty horrid to me - I have tried it using a convolution file and HQP - and it is pretty common for people to dislike a perfectly flat response hence the development of the so called ā€˜house curveā€™ which. I felt an L&R combined measurement pertinent as the reflections from both speakers together interact with each other.

The pale blue line is a flat response and the dark blue/turquoise line is the theoretical house curve and the red my measurements

The first graph is with my REW graph averaged to 1/1 octave which is the overall mean output from the speakers, the second is 1/6 smoothing to compare directly with the separate L&R graphs previously. The curve I used is the Harman curve and is available here for anyone that is interested - Harman Curve

The 1/1 smoothed one pretty much follows the ā€˜idealā€™ perfectly. Sure I could tweak a bit with a convolution file but given I use other sources than digital it is nice having the room largely sorted.


Great guide! Tack @Magnus!

Would be wonderful with a REW integration into Roon. Any thoughts about this @danny? Maybe something to put on the roadmap? Like automatic transfer of convolution files to core and Roon as playback for test sounds. :blush:

2 Likes

Since Roon supports convolution files directly from REW, I think the next step is to add full support for room correction in Roon (measurement, create filters etc). Even if it only handled frequency adjustment it would be a great help if the GUI was easy to use, like for example Dirac step-by-step guide (REW is not exactly easy to get into).

Or maybe a Dirac integration with a one-time cost for ā€œDirac Roon editionā€ which would enable Dirac and Roon to work together seamlessly (as opposite to Roon and Dirac today).

3 Likes

Hi dannybgoode, could you describe the type of optimizations the acoustics expert did for you that worked?

Was it for example speakers placement ā€œradicalā€ type of changes (moving them more than 50cm) or ā€œfine tuningā€ (couple of cm which several people claim have significant effect)?

Thx

I had got it pretty good already so it was minor changes and also tuning the sub to integrate much better.

Thereā€™s a lot of furniture in the room including a rather large wardrobe which was causing issues so it was a case of manipulating the modes and dips to counteract each other through repositioning.

It was an interesting process; a couple of hours of listening to my music as adjustments were made until I was happy. I could have ending up at a similar point through trial and error by having someone there who does this for a living made the whole process quicker.

Thatā€™d be a paradigm shift for Dirac: their entire model is based on a per-location license. Thereā€™s also the question of what their deal is with automakers, and user experience consistency: with the (hopefully, eventually) upcoming Roon-on-the-go, how do you deal with users who wouldnā€™t understand why they canā€™t use their nice and expensive Dirac license in their car ? Finally, thereā€™s the, letā€™s say, aloofness : while @flak offered a way around Diracā€™s limitations with the VST suggestion, which is great and exactly what oneā€™d expect from a coms guy, he not only (and very admittedly) didnā€™t take the time to actually look at ROCK in-depth before answering, but also didnā€™t seem to care to address the specificities of the different products. Itā€™s great for them that their business is so vibrant they can ignore a platform with tens of thousands of users, and a disappointing conclusion for Roon users, but given what weā€™ve seen, itā€™d seem like the logical conclusion is that Dirac consider that theyā€™re on top of the world and that Roon are the ones who should come kiss the ring, and thus that everyoneā€™s best interest would be better suited by NOT pushing for Dirac-Roon-Edition.