Room correction with Roon

I’ve started using Roon again recently have once again looked into room correction. I am using REW to take measurements with a Umik-1 microphone. I think the measurements I have taken are OK.

I’m interested to see who is using DSP for room correction and what settings they use? Specifically,

  1. In REW, what values do you use for EQ > Filter Tasks > Match Range? I’ve seen a tutorial with 20Hz to 20kHz being used whereas others say only go up to 300Hz.
  2. Do you define a house curve in REW or in Roon’s DSP? I suppose doing it in Roon makes it easier to tweak later? In that way REW can try to achieve the flat curve.
  3. Do you use the Convolution engine or manually apply EQ using PEQ (Parametric EQ)?
  4. In REW, what values do you use for EQ > Filter Tasks > Individual/Overall Max Boost? I understand that boosting nulls is a no-no, but do you do any boosting at all?
  5. What value do you have for EQ > Target Level? I chose the auto “Set Target Level” and it came to 83.3. This seems much higher than the 75 others get.

I think thats all! Looking forward to your feedback on room correction as its been out in the wild for ~2 years now.

1 Like

I assume you know this link which describe a simple method for room correction with Roon using RTA.

Well, it depends if you want to correct the full range or only bass. If mids/highs look ok when it’s better not to correct than to correct. If possible in the mid/highs try to apply the same correction to both speakers to preserve imaging.

Important: REW’s house curve is only used to set the target correction, and is not an actual correction. My personal preference is to implement the house curve in REW. This post can help to calculate filter parameters.

I am using the convolution engine as it’s exhausting to copy all parameters in Roon DSP, especially since Roon 1.6 (nice UI for DSP overall, but PEQ s*cks). I have ran some test and the outcome is strictly equivalent between REW generated convolution file and Roon PEQ implementation. Convolution is more costly in terms of processing but usually not an issue.

Moderation is de rigueur. I never go above 4dB.

Well it all depends on the level of your recorded sweep! Don’t bother. 83dB is a bit loud for an acoustic measure, maybe try to stay between 75 and 80dB.

1 Like

Here are the measurements I get with 3 per speaker and an averaged:

Do the mids and highs look as though they need to be corrected? On some recordings I do feel I could do with less treble.

And here are the EQ settings:

I guess I need to re-run the measurements if 83 is too loud - I may take more than 3 while I am at it.

I’m going to keep a close eye on this thread. I’m really interested in using REW with Roon DSP but knowledge is fairly limited right now.

83dB is OK don’t bother. The sweeps were probably a bit loud :slight_smile:

  • For amplitude correction, make sure you use “average the responses” and not “vector average” which is more for phase correction.
  • Do an average of the 3 measures for each speaker, and use this “Average_L” and “Average_R” curves for EQ (it seems that the average you posted is the average of both L and R measures).
  • Smoothing: Var smoothing is better for EQ, keeps the bass accurate for room mode identification, and smooths the treble.
  • EQ settings: avoid 1/6 smoothing, too much, Var smoothing is much better here. Keep in mind the correction is applied on the smoothed curve! House curve looks OK, but if you limit the EQ to 500Hz then the mid/treble won’t be corrected and will be too high level. So either you do EQ full range or raise the target level to match the mid/highs (around 83-84dB), and correct the highs in REW, slowly decreasing as per your house curve. You can apply full range EQ in REW but make sure the filters for L and R channels are very close to each other, and do not generate group delay variances about 35microsecondes after 1kHz.
  • EQ settings: you can allow for max boosts of 3dB, and a flatness target of 2dB, 1dB can lead to lots of filters, potentially harming the phase response.
  • make sure you save your filter as a stereo filter as corrections for L and R will be different.
1 Like

Hi Alec

Thanks so much for your help. Can you explain in more detail what you mean here:

When I first tried a full range EQ it cut the treble A LOT, but I will try again.

I re-ran my measurements this morning using 9 positions for each speaker all around my listening position.

Here is the left:

And here is the right:

There seem to be quite a lot of nulls! I’m using EQ across the full range and am getting good results. The Left speaker has 13 filters applied to it while the Right speaker has 9. I bounced them down as Stereo 32 bit.

For deeper analysis, it would be good to see the spectrograms. Maybe post your .mdat file on a shared drive, Dropbox/Google Drive?

Hi @alec_eiffel, thank you for your kind offer. Here is my mdat file: https://www.dropbox.com/s/637u99rdzuoaohj/Untitled.mdat.zip?dl=0

I really fancy having a play with REW and room correction, but do I really need to spend £90+ for a Umik-1 microphone? I may well abandon the process part way through (it looks pretty complex taking the DIY route), and even if I do perform a real good job of creating those convolution files, I’m unlikely to use the mic for anything else any time soon - just seems a bit excessive if one is not going into the room correction business freelance? :blush:

PS. I’m aware of the HAF service thanks and if I make a half decent job could well use them, but for now I just want to play…

No, you’d need a UMIK-1 with a proper calibration file like this: http://cross-spectrum.com/measurement/calibrated_umik.html
I use an EMM-6 from the same guys (and a Sound Devices USB Pre2). Calibration is very important. I started with the downloadable calibration from Dayton, but got much better sound when I created filters from measurements with the Cross Spectrum calibration. $100 is cheap. A pro grade mic like this: https://earthworksaudio.com/products/microphones/qtc-series/qtc40/ is around $1000 or this https://www.bhphotovideo.com/c/product/828117-REG/DPA_Microphones_4006_A_4006A_Omnidirectional_Microphone.html:open_mouth:

1 Like

Hello,
I saw this thread, and another one like that.
I use Dirac for my main system and like to comment here, that the most important part of the room correction is the room or house curve. You can ruin everything, without knowing it, by selecting a wrong house curve. The topic fills terabytes, a good start is this:

In short, having the liberty to eq to any target cmes with the burden of how? Limiting efforts below 500 hz might be wise. For a full frequency DRC, you are looking at a convolute of the loudspeaker maker’s effort to build this already in for most living room (with a smooth azimuthal roll off) and the special situation in your place. But you would need to know how your speakers behaved in a perfect room and then only correct that bit. Which makes DRC an interesting exercise…

Now, what thrills me with Roon is, that I could apply DRC for my other listening places, not just the main system. I will give that a try…

I’ll have to take a look at that ASR forum link. So much to read on this topic, so many opinions. I bought a UMIK-1, utilized a spare laptop and a long powered USB extension cable so I could sit out of the way while measuring. Anyway, I may have been a bit excessive but I used the moving-mic-measurement process and ended up with a 175 averages per channel. After creating the filters and importing into Roon, I get a really nice detailed presentation. I was just playing around to get to know the process and ended up with a nice result. This will all be re-done after I add some room treatments. I don’t have any because my Magnepans did not need it and I knew I’d be replacing them soon in my new home, so I figured I’d wait until the new dynamic speakers (Tekton Moab) arrived and settled those in. In room response isn’t bad but there a few mid bass/ humps to be tackled and a little suck out around 60hz from one speaker. Using DSP now has paid dividends and I’m almost very satisfied. I want to see if I can get a little more bass energy and roll the highs off a little. It’s not fatiguing, but it’s on the lively side, a tad forward. This is for educational purposes right now so I get the hang of it and can spend less time tooling around with the software after I get he treatments up which I think will probably go a long way towards taming the liveliness of the response. Imaging is great, but I wonder if it gets any better after treatment.