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

I stand by the fact that you should not be in the room…unless perhaps you have a mic placed in your ear (like the Smyth Research guys do) … see the video here https://ksr-video.imgix.net/projects/2525427/video-690844-h264_high.mp4 buts this an entirely different means to an end but the principals are still there.

I have one of these coming…I’ve done the demo and and demo’d the unit to others…its insanely mind blowing. If there is a CANJAM or other show on nearby go check it out if they are participating.

But I digress … I guess you could sit in the listening position and have the mic on a stand and be very quiet…but Ideally most procedures suggest being out of the room or at least far away from the mic.

Also turn of other sources of noise like fans and A/C etc. Close windows and try to do it when there as little external noise interference as possible like passing traffic or construction work :smiley:

There is typically a noise floor of 40+ dB, and in that case whatever sound you do (breathing etc) will not make a lot of difference. Also, you are present when listening normally, and your body influence how the sound reacts in the room, so from that perspective you should sit in your listening position when doing measurements.

I would guess its different when doing very precise measurement in quite rooms like a studio, maybe for phase adjusting speakers etc. But for normal room correction in a home environment, I would say its better if you are at the listening position than outside the room.

Using the moving-mic with Pink PN sound makes the reading more reliable, since it automatically averages the measurement over an area. But if you rather do 8+ sweeps for each channel and average them out in REW that also works (but will take longer time). My suggestion then would be to use the positions Dirac uses (see picture):

Measure for each channel for spot 2-9, and one measurement for both channels on position 1. Then average them for each channel (+ spot 1) in REW and perform equalization.

1 Like

I agree that having your body in the listening position is probably going to be the most accurate, but most programs, Dirac included don’t suggest as the fist measurement for Dirac would npmean the mic would be in your body/head.

Most guides will also say you should be out of the way, if not out of the room.

Out of the room is pretty impractical.

When I last measured, I put myself well back behind the mic with mic on an arm, and took measurements with me in several different positions to see what the effect was - it was negligible - they were all essentially the same.

I know some systems say to remove key furniture that’s in the listening area. Personally I ignore this since they’re there when I listen. Ok they may move slightly, but why correct for theoretical Room when you have the exact one at your disoosal. A dummy (or volunteer) at the listening seat could be a good idea, but I’ve never gone that far, and I remove myself because I measure over a wide area - I like listening generally in the room, including standing up.

I think it also depends on what kind of room correction you want to perform. For example, if you measure a big couch in a living room, and want room correction to cover a wide area, then its probably best to stay out of the way (or at least behind the mic). But for a specific listening position, I think you should be at the listening position.

But it also depends on how you measure. I use the moving mic, 2 spirals in opposite directions outside each ear, and one spiral in front of my face, which is much easier to perform with some accuracy if you sit at your listening position.

I tried the Dirac measurements earlier, and by using the 90 degree microphone correction it was easy (and convenient) to do that when sitting at your listening position. The central one I did by leaning the head back a little and measure at the nose tip. The result was to close to tell apart from the moving mic method, but took much longer to perform in REW (a total of 17 sweep measurements vs 2 60 seconds measurements).

1 Like

Just discovered this thread, fantastic tutorial, an easy and efficient way to perform room/speakers correction in the frequency domain. If there is a Roon Award contest, @magnus you have my vote!

There is no doubt that the spatial averaging technique of the moving mike technique across your preferred listening area allows for a very efficient correction of the speakers and room in amplitude, at least as good as Dirac & Co.

Amplitude correction is the indisputable STEP 1 that any “dematerialised” audiophile has to implement. The cost is ridiculous against the benefits. Consider it before buying any cable of any kind above 100$ (price of the microphone). STEP 0 is of course to ensure proper placement of the speakers and listener, as well as room treatment is feasible.

My experience is that the audio message and soudstage can be further improved with additional corrections/convolutions, that can come on top of the amplitude correction (Roon allows to chain convolutions so you can pile up as many as your processing power allows).

  • STEP 2 : correction of the phase curve and time alignement of speakers : the impact goes from “no change” to “wow” depending on your speakers and personal sensitivity. I recommend to try it if you listen a lot of acoustic recordings, percussions, guitars… The dynamic and clarity should improve. Dirac/Acourate/HAF/Audiolense do it pretty well with a cost (evaluation is free though). You can implement it for free using REW in combination with Rephase but it requires some time and is a bit more technical than the RTA/MMM amplitude correction presented by Magnus.

  • STEP 3: further help your brain to feel you’re with the musicians
    Hifi is about creating an illusion of you being with the musicians playing in a studio, a concert hall, a jazz club… Frequency balancing of the reverberant sound corrects the unnatural tone of the reverberation caused by the directivity pattern of your speakers and the specifics of your room (for example if you have an irregular reverberation time over the spectrum, check the RT60 screen in REW under “Overlays”). The brain doesn’t like it because it’s not natural of a studio/concert hall/jazz club. I have implemented this correction in my (difficult) room and got a terrific improvement in the soundstage despite the absence of any room treatment. The walls have disappeared, the soundstage is super-wide (goes well beyond the walls !).
    X-talk reduction is in my view is a must if you listen a lot of acoustic “live” recordings made with dummy heads or standard stereo recordings (two microphones, one for “left ear”, one for “right ear”). These recordings are fine when listened to via headphones. But when listened with speakers, what comes from the left speakers goes to your left ear AND right ear, same for right speaker ! This is the X-talk. Quite strangely it is not proposed in standard by DSPs even though it is not that difficult to implement.

In my case going through STEP 3 provided a lot of benefits on most of my records, with no negatives. I went through HAFservice and they took care of STEP 1/2/3. (NB: I am not affiliated to HAF in any way, and paid for the service, actually 3 times cheaper than Dirac). It would be good if bigger players like Dirac could provide such advanced features ! We discuss it here and also here.

EDIT: tor those interested here is my view on the current market offering for room/correction/convolution. One could add a column for the REW/MMM/RTA approach documented in this thread : Amplitude correction only (the most important thing to do anyway), quite easy to apprehend and implement, and completely free, except for the cost the microphone of course.

2 Likes

I love that statement. Time and time again I see people discussing spending sometimes thousands on speaker cables, interconnects, digital cables, and power cables, yet have no idea what/how these cables intend to improve. Placebo aside, it could well just be a case that any particular cable has a frequency reduction/increase right where you need it, and is why they sound ‘better’ rather than due to any expensive magic. Of course the trial and error to find the right combination is probably worse than needle in a haystack.

Hey, I’ve been there (haven’t we all?) - maybe not thousands of pounds but I’ve spent money chasing cable and ‘decrapifier’ improvements that would be completely insignificant to Room Correction - I was literally blown away when I discovered Dirac.

I know room design/treatment is the best and also simplifies things (it just works with whatever you put in there) but for so many of us it’s just not an option. Those who are in this position but ignore DRC because it ‘takes the life out of the sound’ or whatever criticisms you hear - really should try a modern implementation. I can only assume all the bad press is from poor implementations, maybe old ones using early technology/maths.

I ignored many times when more knowledgeable people said ‘measure your room’. I paid a hefty price with kit changes that were never going to fix the problem. Now I say it to everyone! Measure!!

And after that the biggest difference has been speaker isolation (isoacoustic gaia footers to be precise), which has been similarly transformational for my suspended floor setup and the second best money I’ve ever spent (or will spend, they’re on loan). :slight_smile:

2 Likes

Most likely a silly question but did you have to change your DSP settings because the gaia footers raised the speaker somewhat?

.sjb

Yea, speaker isolation is nice, mine costed 10$ and made a noticeable difference. I also believe that better cables and USB purifiers makes a difference, but those changes are very small and should only be considered once you have a good room with DRC, good speakers, amp and dac.

I think there are several reasons why some people don’t like digital room correction:

  1. They did it wrong (as did I first time), usually by only doing one measurement/channel, which will make the music at higher frequencies sound false and mess up stereo imaging. The reason for this is the short waves at higher frequencies, so moving your head/ear just an inch from the measurement spot will give a different sound.
  2. They did it right, but is so used to the old sound that the improvements sounds worse. For example, if your room gives a peak at around 100 Hz, many instruments like acoustic guitars will sound thinner when corrected.
  3. They believe that bit perfect is the way to go, and that whatever they hear when using bit perfect and expensive audio equipment equals what the artist intended (that reasoning is only valid in a perfectly acoustic treated room like a mastering studio).
  4. It feels expensive to spend money on a software like Dirac (better to spend them on cables right? :slight_smile: )

I listened to some expensive stuff in an HiFi stora a while back, but in a mildly treated room with no digital room correction, and despite that the speakers, amps and dac used was about 10 times as expensive compared to what I have, it sounded much worse (some parts in higher frequencies did sound better though).

1 Like

Not a silly question.

I didn’t, because I haven’t had time and I’ve been experimenting (they’re on loan). But yes I will have to. I don’t think the 2-3 cm height increase is the predominant factor to be honest. The isolation works so well I’ve got completely new depths to the bass, depth I’ve never heard before in this house, and based on that I assume that other peaks and troughs caused by speaker/floor interaction have changed too.

I plan to measure with and without and see what the deal is. Another Gaia/Devialet user has done this and shown some bass peaks have been reduced. But it’s obviously a bit of a PITA to swap the feet over.

Frankly, I kind of like what I hear but I want to re-do it anyway as I want to maximise their potential. I also plan to try the homedidelityaudio service so I get full convolution with phase and reverb, rather than the basic REW one.

But I need to buy the footers first! Then I need time. So scarce to have the house to myself with time to mess around with microphones. Sadly.

Anyway the footers by themselves make a big difference. Consider me a sceptic that’s been won over - and isoacoustics are pro audio and there was no trace of foo in our email exchanges.

2 Likes

@magnus thanks for this brilliant guide! I realise this is probably a really silly question from a newbie here but when you do the measurements do you pause recording as you transition to each location (left ear, right ear, face) or just record continuously when moving to these positions? Thanks!

Position mike at starting position, hit “record”, then start moving slowly following the suggested patterns, then stop record, save. Then do it again for the other speaker.

1 Like

Great, thanks for clarifying - will give it a try!

Record continuously, I move the microphone in 2 spirals outside each ear, one spiral is on the left-right plane and one is on the front-back plane, and with increasing radius. About 10 seconds/spiral, and 5 spirals in total so 50 seconds of continuously measurement for each speaker (in some cases its enough to to do one measurements with both speakers, so experiment and compare whats best in your case).

You can try other patterns as well, the important thing is to get a good average outside each ear, and to do it the same for both channel and for left/right side of head. Getting a good average is the main improvement over spot-sweep measurements, and can make a fairly big difference.

1 Like

Thanks @magnus, all clear now and rest of the guide was very easy to follow too.

Sigh. This gives me a headache.

3 Likes

One tricky part of doing room correction is to know where its safe or not to apply a boost. Applying a boost to a dead node just leads to increased distortion. For example, my 48 Hz is stone dead, and if I try to boost it I don’t get any more 48 Hz, but I get very audible distortion, and increased ear pressure (very easily noticed on some tunes).

Here is how you can use REW to find out where its safe to boost. Its called excess phase, and where the excess phase is spiky and far from zero, you should not try to boost. To get the excess phase, do this:

  1. Do sweep measurements, one on each side of the head, right outside the ear
  2. Generate minimum phase in REW (its in the controls section, button to the right).
  3. Select the GD tab (Group Delay), and check the excess phase

More information here: https://www.roomeqwizard.com/help/help_en-GB/html/minimumphase.html

Don’t bother about excess phase on higher frequencies though, they will be so very localized and besides if you use variable averaging you will average them out. But below 500 and especially below 100 is very important to not try and boost excess phase areas.

Btw, no amount of phase/time alignment trickery in software is going to fix this, you will have the exact same type of problems in Dirac, HomeAudioFidelity, Acourate etc. The only way to solve this is to do room related stuff like moving the subvoofer, moving the listening position or acoustic treatment in the room (bass traps), and even then it might be very hard to fix.

3 Likes

The notch filters (Roon calls them band stop) turned out to work very well, not even Lorde - Royals present any problem (the part at 25 secs is especially difficult). Just pure bass, despite that I sit in a concrete bunker :slight_smile:

Here is my EQ, applied to a flat response (if you follow the guide, then set LF Raise slope and HF fall slope to 0 to get a flat target):


Of course, the placement of the notch filters depends on the room, so don’t copy this. Follow my previous post to find out where you need them.

2 Likes

why do you have a 1dB Peak added at 75Hz?

To get a little extra warmth, not that it matters much. It has nothing to do with notch filters or dead nodes.