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

(Magnus) #161

Sweeps also include reflections, but you can use various techniques to get rid of them. However, my thinking is that whatever reaches your ears should be room corrected, and that includes reflections. To get rid of reflections, you need to treat your room with acoustic treatment (Auralex etc), but whatever reflections you have should be included.

An interesting thing about reflection: if you have a specific listening position, its better to have speakers with a narrow sound dispersion, since they focus the sound more towards your ear and will result in less reflections (usually the reasoning is that wide dispersion for speakers is good).

Phase matters less than most thinks in normal room correction, I used a trial of Dirac and compared and could not hear any difference between my own REW room correction. I also talked to the developer or software rePhase and he said that room correction should be minimum-phase EQ correction and nothing else.

(Steve) #162

Built another set based on my interpretation of the B&K curves, which is what I ended up preferring with Dirac.

I’ll then add some extra bass hump in PEQ until I settle on something I like - too much LF addition and my speakers go a bit ‘boxy’, too little and you just don’t get the bass ‘thump’. Well, nothing really thumps in my lounge and with these speakers - especially at the volumes I listen at these days - but you know what I mean.

Beyond that I’m going to just enjoy some music and wait until I can measure again.

If anyone’s interested I aded a house curve following Magnus’ example up above, and used these numbers which I think reasonably reflect the B&K curve.

20 0.5
65 1
200 0.5
650 -1
2000 -2
20000 -5.5

(see B&K diagram here)


As I was driving to work yesterday, I began thinking about what I would have to do to bring Roon with room correction into my car. I really can’t wait for Roon to provide support for remote/mobile playback and control!

In the meantime, I might apply this guide to in-car measurements and apply the EQ filters to a PEQ app on my iPhone – if I can find one that works with HF Player.

@Magnus, I blame you for this! :wink:

(Steve) #164

I’m no expert on DRC but when I spent a fair bit of time setting up Dirac in the past, I had similar experiences to yours when toggling back to the uncorrected sound.

The thing is, looking at both our measurements we’ve probably been hearing too much atmosphere - due to room, or speaker manufacturers choice, or both. So at first listen the corrected sound seems too dull, but for me anyway after taking time to get used to the corrected sound, reverting to uncorrected seemed horribly bright.

I’ve learnt that I take quite a bit of time to adjust - like weeks or even months - so flicking between A/B type comparisons is less helpful in this scenario.

As an aside, I do often hear people saying they’ve heard DRC but didn’t like it because it ‘sucked the life’ out of the music. Maybe they’ve just become accustomed to ‘too much life’, or they’ve heard bad (I guess earlier generations using very simplified algorithms) DRC. Maybe they had near perfect environments already. So far even using this free approach has bought much more life back to my music. Even if I’m still dialling in the settings and getting used to it, it’s much more real than I had before. Stereo image is much better, and I can go louder without fatigue. It’s much more fun to listen to - pretty much like when you do a big hifi upgrade…

Plus it’s a tinkerers delight. :slight_smile:


i completely agree. my updated settings 20-1k with the PEQ adjustment are much more natural than the uncorrected sound. but i am still experimenting. and that’s not a bad thing.

(Steve) #166

I have to confess, having used the B&K curve, I’ve ended up adding a little steeper HF roll off in PEQ - so probably more like the REW default that end. A slightly darker sound maybe, but I prefer it - after a year with a bright system it’s nice to go a bit dark…

(Steve) #167

There hasn’t been much traffic here this week, so I just wanted to say I’m really happy with the result here. I’ve been really enjoying music the last week, and the ability to make subtle but effective tweaks to the sound is very satisfying. Low-level listening (which sadly I’m doing a lot of) is particularly enjoyable.

It’s got me thinking how much better full convolution filters from the likes of Acourate might be, but at the same time I’m happy enough that I’m going to save the money and rest here for a while. Sometimes it’s good just to enjoy!

How can we encourage more people to do it?

(Magnus) #168

REW produce full convolution filters as far as I know, but without any time/phase modifications. But in a normal room, phase modifications aren’t that useful anyway. I compared REW with Dirac (trial) using the same house curve and I could not tell any difference (Dirac claims to do much more than frequency/amplitude correction). Having said that, its possible there is a difference and I could not hear it.

I recently upgraded my speakers from $300 ones to $1600 ones (Dynaudio X14A) and they sound awesome, but the old speakers sounds much better with room correction than the new ones does without room correction. Just a hint of how much room correction can do.

I feel confident that, in my room, there is no audio equipment that sound better than what I have (no matter the price) without room correction, compared to my relatively cheap stuff with room correction.

(Mark Allen) #169

100% agree with you and @hifi_swlon

I think people can really maximize their speakers value with some room correction! No need for me to upgrade speakers at this point in large part to Roon DSP/REW room correction. I added a house curve as well and the results are outstanding.

(Steve) #170

Well, what it produces is just the combined result of all the active parametric filters you see in REW - its a very clean’ impulse response curve. So saving each out as different sample rates and using them as convolution filters is unnecessary - you might as well just load them into Roon as parametric EQs. Of course the problem there is its not easy to import them.

From my understanding, a true convolution filter would use the parametric EQ curves to get the resulting curve you see in REW, but then it would build a final filter by negating the resultant waveform and combining this with the EQ curves, so that you’d get a ‘perfect’ and detailed filter with all the nuances the PEQ couldn’t deal with. I could of course be wrong here and it not being my field have probably explained it badly.

In any event it was more curiosity since it was a while ago I used Dirac so need a more up to date comparison. I do think the phase adjustments are worthwhile, but I don’t really want to pay £300 and far around with another complex package just to find out there’s not much in it!

So I’m going to stay happy and let it remain unknown… its more mysterious that way anyway.

(Magnus) #171

Yes, the difference is very big, but it not always you get the perfect result even with a fairly detailed guide like this. But what scares some people is the notion that any modification is bad, since it makes it less bit-perfect and clean. It might also be hard to switch from the old sound to the room corrected sound, since its so different and people get used to things sounding wrong.

About convolution filters: it does only use the EQ settings that’s corect, but it has 131k taps so the resolution is quite high. But yes, the same setting in Roon EQ should produce the same (or very similar) results. And I suggested that Roon should be able to import filter settings from REW which was put on the “TODO” list, so maybe a future version or Roon will be able to import EQ settings from REW (which will be a benefit for high sample rates, since REW has a rather low limit on its export).

Not sure how DSD and convolution/EQ works, have to get a DSD DAC first to try :slight_smile:

(Steve) #172

That’s kind of what I mean. 131k taps but aren’t really needed since PEQ curves are defined mathematically and can be stored in their perfect precision for all sample rates with minimal data (Q, gain, frequency).

I saw your suggestion and it’s a good one. I’d love to compare convolution to PEQ, but there’s no way I’m typing them in!

I’ve got so few DSD albums and Devialet’s AIR is broken with DSD on a Mac so I’m out of luck anyway.

(Steve) #173

Got a window to get some new (better) measurements today - much better conditions, and done at a decent volume. It was a lot easier second time round.

I also finally measured the result of the previous filters, but at first glance the results weren’t as good as I’d expected from listening. Then I remembered I tweaked the speaker positions very slightly last night in prep for doing the new measurements. I wouldn’t have thought it would make the difference I’m seeing, but it may have done, so am just going to ignore that for now and concentrate on the new set.

Sadly, not sure when I’ll have time to make the new filters, let alone measure the results, but at least I’ve got the measurements done!

Will post some more graphs etc when I’m a bit further along.

(Gustav Persson) #174

I have been reading this thread closely and tried to do some measurements today. I set the SPL to around 76 db before I did the measurements in the RTA. Why do I get this kind of measurements with lower levels? Do these seem like valid numbers? My kid was a sleep so I didn’t wanna play the for too long, but will try again tomorrow. I also used a too short USB-cable so my mic movement might not have been as systematic as I intended.

Any feedback for my next attempt is greatly appreciated. I don’t wanna spend too much time working with filters before I’m confident i’m doing proper measurements!

(Jeff Young) #175

@Gustav_Persson, that looks about like mine did. If you zoom the X axis into 20/20K (there’s a button in the lower right to do that), and then zoom the Y axis in till you’re just looking at (say) 20dB to 80dB, then it should look more like what you expect.

As to why the mean output is about 50dB when you set the SPL to 76, I don’t know. It was similar when I did it.

(Magnus) #176

Is that graph from corrected sound (Roon playing Pink PN), or directly from REW? If it is from Roon, the lower level is probably due to automatic volume correction. If the measurement is from REW, I don’t know why the levels are so low, but the graph itself looks good.

(Gustav Persson) #177

It is directly from REW. I ran the PINK-PN signal from my macbook to my Xeo 2 speakers through bluetooth. I didn’t have time to try the filters I created through Roon.

In my main window it shows I have an input of around 75 db (I think?). I will see if can get a longer USB-cable today and redo it tonight. But it seems reasonable that there is a spike around 50 hz for the left speaker as it is almost placed in a corner.

(Gustav Persson) #178

Did you find that it was fine that the mean output was lower than expected when doing filters?

(Jeff Young) #179

I didn’t set up a convolution filter. I’m using PEQ instead, so that I can see the corrections:

A couple of notes:

  • Using PEQ does mean that you have to play the PN or sweep through Roon to see the effects.
  • You’ll note that I do have a 2 dB gain set, but that’s to correct for volume levelling.

(Gustav Persson) #180

I tried again with the same result of low levels in the measurements. HoweverI also made a sweep from 10-22khz (?) for each speaker, through the measurement ikon and got the following measurements. Also by adjusting levels beforehand to 75 db. They look more reasonable. No idea why this is?