Thread Rating:
  • 1 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Room correction
I have used Room Correction via Umik1 + REW + JRiverPEQ for the past couple of years, and having made dozens of measurements with speakers in various places, the one thing I have found is that applying the calculated filters to precisely match the 'house curve' NEVER works out for me in the long run.

I find that the filter amplitudes have to be at least halved, and often reduced to a third. It's as if the ear objects more to large corrections than it does to the natural undulations imposed by a room, so I have to find that magic spot where the adjustments are just enough to satisfy. But once I have found that point, it certainly is hard to accept those extreme Room Modes any more.

For some reason, I also prefer simple PEQ over Convolution - the latter sounds too sterile, even when I've gone to the trouble of tailoring REW's filters before exporting the Convolved WAV file to JRiver. Perhaps I'm not sensitive to non-linear phase shifts.

(p.s. I have done this with both White Phantoms and with D120+PMC speakers, although my listening is now entirely dominated by the latter).
JRiver v25 (Windows) >> 220Pro/CI >> PMC Twenty5.23 + twin KEF KC62 subs. One White Phantom.
Reply
It's definitely sad to see that HiFi experts claim for so many year and also competitors like LINN, Sonos or even lifestyle products like Apple HomePod are integrating room correction by default as an essential feature that has a massive impact on audio quality, Devialet is still not providing a solution to that crucial demand :-(
Reply
(08-Jun-2017, 13:15)JohnnySix Wrote: I have used Room Correction via Umik1 + REW + JRiverPEQ for the past couple of years, and having made dozens of measurements with speakers in various places, the one thing I have found is that applying the calculated filters to precisely match the 'house curve' NEVER works out for me in the long run.

I find that the filter amplitudes have to be at least halved, and often reduced to a third. It's as if the ear objects more to large corrections than it does to the natural undulations imposed by a room, so I have to find that magic spot where the adjustments are just enough to satisfy. But once I have found that point, it certainly is hard to accept those extreme Room Modes any more.

For some reason, I also prefer simple PEQ over Convolution - the latter sounds too sterile, even when I've gone to the trouble of tailoring REW's filters before exporting the Convolved WAV file to JRiver. Perhaps I'm not sensitive to non-linear phase shifts.

(p.s. I have done this with both White Phantoms and with D120+PMC speakers, although my listening is now entirely dominated by the latter).

That's interesting. I have also done many comparisons between PEQ and convolution and I have finally preferred convolution. I'm not sure is it fair to compare PEQ and convolution directly since those are not always doing the same thing. My convolution filters are also improving the pulse responses and that might be the main reason why I have preferred convolution. 

It takes some time to find a good target curve though. I definitely didn't like a totally flat target.
Bluesound Node > Matrix Audio X-SPDIF 2 > Genelec 8351B & 7360A
Devialet 1000 Pro
Bluesound Node 2i > Genelec 8330
Tampere, Finland
Reply
I now use the guide by Magnus on the Roon forum for using real time averaged measurements in REW. It does a surprisingly good good with the amplitude stuff (phase is ignored but can be added with a secondary process if you're that way inclined - amplitude is good enough for me).

I don't think any human would like a flat house curve - that would sound really bright. Simple downward sloping is good (like Dirac default), or use B&K curve, or even the REW default. I hover between REW default and B&K and sometimes add back in extra bass, although for overall 'it works on most things' setup nothing added works best for me. The bass can sound a bit lean after Correction so if you want the party it might need some added back in. I think one of the biggest benefits is both speakers balanced independently as best as possible, which really helps in a compromised space like mine. Plus you can tweak endlessly....

REWs convolution filters are essentially parametric EQ with the Magnus method - but it's easier to implement in Roon.

I never really find the corrected sterile - it's always better, but small changes to the house curve can have quite dramatic differences and it is hard to find the one you want. But I'm far happier with the result than without. I was going to buy Acourate but I'll just stick with the REW solution for now.

I'll use the Devialet system when the new streamerboard comes out. Oh no, wait.... my mistake.

>>> 1st Place Award: Devialet, last decades most disappointing technology purchase.  <<<

Reply
To be clear, I don't go for a flat curve, but one that best follows the general trend of my room. My main point is that I don't think it's a good idea to iron out every peak & trough until it lies right on the target curve - some leeway must be allowed to allow them to "ride" the room's natural ebbs and flows, to keep it natural. (Ooooh, listen to me.)
JRiver v25 (Windows) >> 220Pro/CI >> PMC Twenty5.23 + twin KEF KC62 subs. One White Phantom.
Reply
(08-Jun-2017, 20:19)JohnnySix Wrote: I don't think it's a good idea to iron out every peak & trough until it lies right on the target curve - some leeway must be allowed to allow them to "ride" the room's natural ebbs and flows, to keep it natural. (Ooooh, listen to me.)

I don't agree with that personally. Leaving anything the room adds that can successfully be corrected defeats the point of doing it in my opinion. I don't feel there's a negative effect to removing everything that can be removed. Filling troughs is the hardest ones to judge and they should generally be left alone -especially in the low frequencies.

Obviously a lot of this is personal preference and Room/speaker dependent. The important thing is you end up with a sound you like and one that's better than you had uncorrected. I've certainly got that now, but there's always room to tweak/improve.

>>> 1st Place Award: Devialet, last decades most disappointing technology purchase.  <<<

Reply
(08-Jun-2017, 18:24)Hifi_swlon Wrote: I now use the guide by Magnus on the Roon forum for using real time averaged measurements in REW. It does a surprisingly good good with the amplitude stuff (phase is ignored but can be added with a secondary process if you're that way inclined - amplitude is good enough for me).

I don't think any human would like a flat house curve - that would sound really bright. Simple downward sloping is good (like Dirac default), or use B&K curve, or even the REW default. I hover between REW default and B&K and sometimes add back in extra bass, although for overall 'it works on most things' setup nothing added works best for me. The bass can sound a bit lean after Correction so if you want the party it might need some added back in. I think one of the biggest benefits is both speakers balanced independently as best as possible,  which really helps in a compromised space like mine. Plus you can tweak endlessly....

REWs convolution filters are essentially parametric EQ with the Magnus method - but it's easier to implement in Roon.

I never really find the corrected sterile - it's always better, but small changes to the house curve can have quite dramatic differences and it is hard to find the one you want. But I'm far happier with the result than without. I was going to buy Acourate but I'll just stick with the REW solution for now.

I'll use the Devialet system when the new streamerboard comes out. Oh no, wait.... my mistake.

Yeah, Magnus's guide is pretty awesome. I have tried it couple times and got nice results with it.

Finding an optimal house curve takes some time. I have tried dozens of different house curves. As you said, even small changes in the house curve have significant differences in sound. Some house curve might sound impressive at first but later on you might find it causing listening fatigue. I am currently using slightly tweaked Olive/Toole house curve and it sounds nice. Soon I'll compare it with B&K house curve and the one that was recommended by Mitch Barnett (flat to 1 khz and -6db at 20 khz). All these house curves are all similar in shape, so it's probably just a matter of personal preferences which one is the best.
Bluesound Node > Matrix Audio X-SPDIF 2 > Genelec 8351B & 7360A
Devialet 1000 Pro
Bluesound Node 2i > Genelec 8330
Tampere, Finland
Reply
try the Mathaudio Room EQ. It is free for foobar but also available as VST plugin for jRiver. It's the best solution I have ever used. Before I have used REW filters with jRiver, Acourate and Dirac. They all have a positive effect but also a negative impact. The Room EQ by mathaudio beats them all and is also very easy to use.
Reply
Mathaudio can not export to FIR filters so convolution in Roon's DSP or JRiver will not be possible. Players with support for VST plugins can use Mathaudio.
*
Devialetless!
Roon, ROCK/Audiolense XO/Music on NAS/EtherRegen/RoPieee/USPCB/ISORegen/USPCB/Sound Devices USBPre2/Tannoy GOLD 8
250 Pro CI, MicroRendu(1.4), Mutec MC-3+USB
Reply
I'm doing an offline folding of my audio files anyway. This does allow me to use any player, e.g. Bug Head. My favorite is jplay streamer with minimserver and Kazoo as remote which does also not support any convolution. Doing an online correction is not a good idea anyway as this adds a high processing load to the CPU which should be avoided during playback.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)