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

(Jeffrey Robbins) #61

Magnus, the reason I ask about separate measurements, particularly in a surround system, is because for my 2.2 system, I use the Audyssey Multi XT32 software built into my Marantz prepro. It runs a frequency sweep for each speaker from eight positions to calculate the average filters for each speaker. I can’t use that for my 5.2 surround because my multichannel DAC, a miniDSP, is analog out and the Marantz won’t apply Audyssey filtering to its multichannel analog input.

The other thing that REW alone can’t do, to my knowledge, is provide any bass management. As many of you know, SACD/DSF tracks are full range to all speakers in a surround format. My surround and center channels are smaller speakers and I don’t want the bass routed to them. In my setup, I am using my old fashioned Outlaw Audio ICBM-1 analog bass management device to route all bass from the front, center and surround speakers to the subs. It would be that bass managed output from the ICBM-1 that I would run REW on, although it seems strange to have all 7 speakers measured as one single speaker. But I guess you are suggesting that it can work pretty well that way.

Now, if anyone knows how I could export the Audyssey measurements in the Marantz and put them into Roon, that would be ideal. (Anybody?). Thanks. JCR

(Magnus) #62

This guide is made for 2.0 or 2.1 systems, where I measure left and right channel individually and create a stereo impulse wav file to use in Roon convolution. Maybe something similar can be done for a 5.1 or 7.1 system, except more channels and more measurements, but I am not the guy to ask about that I am afraid, since I have no experience with 5.1/7.1 systems.

Or maybe just this rather simple approach:

  1. Use the SPL Meter and test each speaker to make sure they play equally loud (make adjustments in Roon to correct if needed).
  2. Measure all speakers and bass at your listening position at the same time, using RTA and moving microphone around like I suggest in step 6, and apply the result to all speakers as a mono impulse wav in Roon.

I am sure this is not optimal, but its probably better than noting.

(Peter Sherman) #63

in step 14 you say, “save a .wav file for each frequency you use in Roon.” I’m not sure what this means. Can you please explain what you are referring to when you say each frequency. Thanks

(Magnus) #64

When you select that menu chose you get a small dialog window, where you can set the sample rate, you need one wav file for each sample rate you use for playback in Roon, then you will pack them together in a zip file and use that file from Roon convolution DSP. I added a picture to step 14, should make it more clear.

You can read about Roon covolution here: https://kb.roonlabs.com/DSP_Engine:_Convolution

(Morten) #65

hi all

For room correction - if it is based on measurements from only one microphone position, the correction applied will in theory be very (or extremely) specific to that single position in the room.

For measurements demonstrating the effect of moving the microphone (or your listening position) as little as 100 mm (4 inches) in any direction, see:

As i don’t regularly listen to music with my head stuck in a vice, I have been using a correction system based on multiple measurements in many random locations in the room and the corrective filters calculated from this process result in a sound which is better than the uncorrected signal, also if you move your head /seat, or walk around in the room.

This btw state of the ark, as my Lyngdorf amps with this system built-in are 15 years old.



I have just begun the process based on your great guide. Just wondering why you suggest that we convert to Flac since Roon supports WAV.


What adjustments did you make and how did you make them to arrive at the last two graphs?

(Ruud Verrijk) #68

Hi Magnus, thanks for the great tutorial. Using the RTA averaging is new to me. I will try it for sure.
I have never been able to equalize my room modes and dips so well as in your example. Could you share a picture and/or a list of filter settings that REW generated for this?


(Magnus) #69

Ahh, didn’t know Roon could play wav. Converting to flac with Audacity is till nice though, since you get to set the album, artists, etc. I named mine Measurement, so its easy to find from Roon.

The second to last measurement is done by playing Pink PN from Roon and measure with the room correction active. The last is after extra modifications done in Roon PEQ to make it even more flat, but even though the graph looks better I am not sure it sounded better (which is why that step is not in the guide).

(Magnus) #70

The pictures at the end of the guide, and in step 9 and 11, shows what it looked like for me. Generally speaking, in my room the lower frequencies need major adjustments, while the higher frequencies are pretty good as they are.

But remember that I use “var smoothing” which is what REW suggests, and that smoothing is more focused on higher frequencies. If I use for example 1/24 smoothing, the resulting measurement would look worse at higher frequencies.

(Ruud Verrijk) #71

What I am wondering is, how many EQ filters did REW generate for you, and are they extreme in terms of gain and Q to achieve such a nice result?


I have just completed my first REW filters, but I have already noted that I have a lot of room for improvements in my technique. Unbelievably, I can’t find a way to fully isolate L & R output – neither my amp or my DAC allow full L or R. I resorted to using the DAC’s balance controls, but some sound from the other speaker remained. Also, my technique o moving the mic around my listening area felt irregular and clunky, so I need to work on that. :wink:

The sound results were much more obvious than I expected. The sound is much less muddled and I’m really happy about that. But the sound level with convolution on is much lower than with it off, so comparisons are not very easy to make. I did find that these filters stripped out a lot of the music that should have remained. I’m listening to Tea for the Tillerman (Cat Stevens) and it sounds very thin – there is no body to the guitar playing. Same with Getz/Gilberto or anything else I have played.

Still, I see tons of potential. I noticed that the EQ filters were very aggressive with Q in a few places. I know you mentioned that I should try disabling filters with high Q slopes. Just wondering what you consider to be high? Most of the gains I ended up with were 5 or lower, but I have 6-22 as well.

(Magnus) #73

I have a 2.1 system, and I increased the volume and crossover on the sub first. This I did to make sure I didn’t try to put +dB filters on destructive bass nodes. I don’t have a crossover either (both sub and monitors on full signal), so by giving the sub higher volume it will do most of the work at its frequencies (< 80) after correction.

So yes, I had many -15 dB filters, but not a lot of high +dB filters. I also did a final measurement, playing the Pink PN from Roon with the room correction in place, and did some fine tuning in the Roon EQ. I will add this step as an optional step to the guide. The second to last picture is without this fine tuning, the last is after this fine tuning.

(Magnus) #74

To play L or R, can’t you just remove a signal cable, or a speaker cable?

About Q and high frequencies, if you do the moving-microphone measurement I describe in 5 and 6, you don’t need to worry about that. The only thing you need to make sure is to don’t boost steep holes at lower frequencies (but you don’t seem to have any of those). The moving-mic technique seems to work very well with high-frequency modifications.

I think you have the target a little low though, I always try to set the target so as much of the measurement as possible is already correct (but same targets on both channels). It’s a little harder for you do to that, but try setting the target at around 75 dB or so.

REW does not correct frequencies that dip below the target and never comes above target again, so if you know your speakers can handle it, add 1 manual target to boost the bass, in your case, you could add a filter at maybe 55 Hz with a +10dB and Q or around 5 (experiment and see what produce the best result). To do this, disable one filter in the “EQ filters” window, let REW generate 19 filters and then enable and set that filter manually.

But make sure you listen critically afterwards, in case this turns out to be a boost of a destructive node in which case its better to remove it (it will sound like distorted bass or something similar).

(Magnus) #75

Updated the guide with 2 optional steps as part of step 16, based on feedback above.


I can unplug the cables from the output from the DAC – I’ll try that with my next go (with the speakers, I’d have to move them to reach the cables, so I want to avoid that for obvious reasons).

I was using the moving mic technique – or attempting to. I’ll try to vary it more next time.

I had set the target at 75 but I noticed that when I clicked the Set Target it was reset back to 68.3. Next time I will manually set it at 75 and not click on the Set Target link. Should the Target entered match the SPL level I started with (74.6 in my case)?

These speakers are actually my L/R (fully sealed) speakers in a 5.1 setup that I happen to be running as L/R for music duty temporarily. Normally I would run them in 2.1 for music, but my sub is down for repairs. The L/R can handle down to about 40hz if I remember correctly, but due to their compromised location in a cabinet, I normally run them with the cutoff at 80hz. So REW has been an intriguing excericise of how much it can compensate for my severely compromised arrangement and room. I’ll try the 55hz boost – nothing to lose and easy to revert.

Next zone to calibrate will be a very large loft-like space with no specific listening spot. Results could be great or gruesome.

Now if I could only extend my REW rorrections to the rest of my 5.1 system for other AV sources. (Do not respond to this :blush:)

Thanks again for your extremely helpful guide and all of your assistance and feedback! You have been amazing.

(Magnus) #77

Just realized: you set the output in the Generator window in REW when playing the Pink PN noise. You can see it in the first picture. And the target will usually be around the SPL level, but I try to set it to help REW by placing it on a good spot. In the picture in step 11 you can see I set it on the higher frequencies.

(Anders Vinberg) #78

This seemed reasonable to me (no expert), but then I used Acourate which does use only one and that made me wonder. @Uli_Brueggeman, the author of Acourate, has written at length about this, arguing for the precision of single-point measurement. As I understand it, Acourate can do a better job by doing more careful measurements from a single point, taking time into account.

It also occurred to me that if the various listening locations, like the three seats in the sofa, are truly different due to room interactions, DSP cannot change that. You apply one correction, you get one sound, and the location differences would just overlay on top of that. Making many measurements and averaging them doesn’t change this fact. It is possible that sophisticated analysis of multiple measurements can improve on this, but averaging cannot be the answer.

As I said, I’m no expert. I used Acourate, created a convolution from the center position in the sofa. Measuring afterwards, that location became very good. Other locations differed +/- 4 dB around the center position, which is a reasonable result. The corrected curve was a reasonable average. I don’t know if this result comes from a reasonable room.

(Magnus) #79

The problem with one measurement is that you get very localized reading, moving the microphone as little as an inch will probably give big differences on some frequencies.

When I started with REW, I did the like that: one measurement where you head is, generate filters from 20 to 20 khz and then thinking I was done. It sounded horrible though, female voices, electric guitars, and other things in that frequency range was all wrong. Turned out my measurement had given me a +8 dB at around 1 khz, but when doing a more proper measurement it turned out to be a -dB at that frequency. So instead of boosting that area, I lowered it.

Multiple sweeps, like Dirac does, seems to work good, but from my own testing the moving-mic method works best.

(Anders Vinberg) #80

Yes, I saw in some of the discussion that they talked about multi-point as a way to avoid increases. But that seems unrelated: you should not make increase changes in any case. Even with many measurements you might see a big dip because of a room resonance, and if you blindly compensate forest you get that problem. Acourate uses a different approsch: you set the target curve below the measurement, which means the correctiononly pulls down, no Increases. If you very deep notches you don’t want to go below all that, you can set the target at dome reasonable point and then explicitly limit or block any increases. I had read about this technique, and I tried to replicate it in REW. Was not successful, and bit the bullet and went with Acourate, learning cure or no.