I spent a lot of time to find the best convolution in my setup, testing the various solutions out there, including a deep dive into REW/Rephase which gave a performance comparable to Dirac Live, at the expense of 10-15 hours of my time. The best results I got so far was with HomeAudioFidelity (HAF) as their proprietary “reverberation toning” and “cross-talk cancellation” techniques lead to spectacular results in my untreated reverberant living room. Here is a table summarising the current convolution offering, let me know if there are any mistakes.
I do not see any mistakes in your comparisons, but Audiolense definitely belongs here. Quite similar in performance to Acourate, somewhat easier to use.
1 Like
Rephase is not really made for room correction though, I talked to the implementer of Rephase and he said that fixing phase in a typical listening position is pretty worthless due to all reflections. Rephase is more useful for fixing phase between drivers, for example when building your own speakers (and then you measure a few inches from the driver).
As long as you get frequency amplitude corrected you will have come a long way, and to get that good is mostly about correct measurements together with software to generate filters and a convolution engine (or PEQ) in your listening app.
I agree that the prime intention of POS when he created Rephase was to correct speakers/filters, not the room. But coupled with REW it can be very useful to generate FIR filters correcting the amplitude as well (room+speakers). REW can now natively generate filters that can be exported to Rephase for amplitude correction. Performing multiple measurements around the listening position (5 to 9) allows to almost eliminate reflections and actually measure the phase behaviour of the speakers (in REW / ALL SPL / Time Align measurements / Vector average).
EDIT: here is a link to the fantastic tutorial of pda0 on the French Hifi Forum on how to use Rephase for room and speaker correction. In French unfortunately. http://forum-hifi.fr/attachment.php?aid=354
I investigated this when I wrote the REW guide, hence my email conversation with Thomas Drudgen (creator of Rephase), and just importing the REW filters into Rephase and creating FIRs does not do any phase adjustments. You need to fiddle with the phase-EQ to do that, and the phase (at least in my room) is all over the place.
Here is a quote fro Thomas:
I believe room-specific phase correction should never be done.
For room correction minimum-phase EQ is all that is needed, and that kind of correction should only be done once the (anechoic response of the) loudspeaker itself has been properly EQed, possibly including phase linearization.
Reading that. and comparing Dirac to room correction done in REW (I could not tell them apart), I am somewhat suspicious about all claims from room correction software how much more you gain with phase corrections compared to frequency amplitude corrections.
Thanks for sharing that link. I used google translate to translate it to English and saved as an english PDF. Will give it a read this evening. (64 pages!). Will also thank the authors on that forum.
You’re welcome. See also here in English https://www.computeraudiophile.com/forums/topic/30035-using-rew-and-rephase-to-generate-amplitude-and-time-domain-corrections/
Yes I have seen that before but have not tackled rePhase yet. I have only used REW to generate frequency amplitude convolution filters for Roon - with pretty good success. I was unsure if using an average of multiple measurements (which I think is crucial) in REW would lose the phase information and thus make rePhase moot but apparently not. (Unfortunately, using a moving pink noise signal as per @Magnus’s guide does not capture phase data).
@Magnus to be clear: Thomas is right in saying that “room-specific phase correction should never be done”. This is not what Dirac, Acourate or the new-kid-on-the-block-from France Home Audio Fidelity do : they do not correct the “phase” of the the room but the “phase” or “time domain alignment” of the speakers themselves. That’s why they require to do 5 to 9 measurements around the sweet spot. Averaging them allows to remove reflexions and have access to the intrinsic behaviour of the speakers.
The amplitude correction although is taking into consideration both the room and the speakers. Additionally, HAF filters correct the “amplitude” of the reflexions at certain frequencies to make the reverberant field more “natural” for the brain, correcting defects from the speakers (directivity) and the room. Nothing to do with phase here.
1 Like
And yet doesn’t Dirac claim to do both frequency and time domain corrections??
Frequency correction : Room and Speakers
Time domain correction : Speakers only
[quote=“alec_eiffel, post:29, topic:30383, full:true”]
@Magnus to be clear: Thomas is right in saying that “room-specific phase correction should never be done”. This is not what Dirac, Acourate or the new-kid-on-the-block-from France Home Audio Fidelity do : they do not correct the “phase” of the the room but the “phase” or “time domain alignment” of the speakers themselves.[/quote]Despite the name, nothing corrects the room except real acoustic treatment, digital room correction is all about correcting the signal into the speakers. And since crossover between elements is done inside the speakers, nothing can be done about those either (unless you have some very specialized multi-channel bi-wiring setup).
So what remains is to make sure that the sound coming from the speakers and reaching your ears as closely resembles an ideal room as possible, and that includes reflections. And the main reasons for taking multiple measurements (or using the moving mic technique) is not to correct phase, it’s to make sure short wave frequencies aren’t measured wrong.
Dirac does this very good, probably by partially averaging speakers but giving extra weight to the center measurement, and it does it in an easy to use guide.
Think about it: you can’t change the phase of a reflection without either changing or treating the room, or changing the signal to the speakers, and if you change the signal than the original non-reflected sound will also change phase, and you will have gained nothing. In other words, you can’t get rid of destructive nodes or cancelling sound waves with any type of digital room correction (if you could, then real acoustic treatment would never be needed).
Having said this, Dirac and some others claims do to phase/time corrections, and I don’t doubt that they do. But they can’t break the laws of physics. Whatever extra they do beside frequency amplitude correction is very minor. I bet the main reason they try to emphasis those adjustments is to try and distance themselves from REW.
But try yourself: download a trial from Dirac, and compare to a god measurement done in REW, and see what difference there is. My current room correction (done with moving mic and some trials and error) sound better than the one I have for Dirac, but then I spent a lot of time to experiment.
I read that guide several times when experimenting with room correction (and later writing the guide about it). Its only the last step that deals with time/phase alignment, and whatever measurement is shown in that screenshot is definitely not from a normal room (more likely from an anechoic chamber). If I do one measurement and check the phase (which can be done in REW), the phase information is all over the place, wrapping itself around 360 degrees hundreds of times. Try to manually correct that if you like 
I struggled with that as well initially. The room is a normal room.
- measure only one speaker
- align impulse response at t=0 (REW can do it, see “impulse” view)
- apply Frequency Dependant Window, to eliminate reflexions, start with 6 cycles / octave, you can go even lower
The phase curve should now look much better.
By the way @Magnus I’ve just seen your tutorial regarding Room Correction with REW in Roon. Fantastic work ! The RTA/moving mike technique is actually very powerful and simple to implement. In most cases the result are probably very good and comparable to Dirac, HAF, REW/Rephase and Co.
The alternative and more complex techniques of measurements I referred to earlier allow in addition to perform time domain analysis and, if needed correction. The impulse/step/ETC/delay/spectrometers views of REW are a very good tool to “see” the problems you have like early reflexions on walls / tables at given frequencies, and implement corrections like bass-traps, absorbers etc… Secondly, phase correction of the speakers is possible. Depending on your setup, the effect of this correction can be invisible or spectacular (mostly on the soundstage and dynamic). Third, the group delay analysis helps you to see the frequencies where your system behaves as a “minimum phase” system, and where it’s not. This is helpful in fine tuning the amplitude correction, and leave non-minimum phase areas out of the correction (especially boosts).
1 Like
Danny - I think this functionality would significantly enhance Roon. Appreciate the business case needs to stack up against all the other stuff you are doing; if it helps I’d be prepared to pay extra for something akin to Dirac that was within Roon.
4 Likes
I agree, there are some very nice tools in REW, and I have used them to test the room acoustic and to determine the effect of acoustic treatments. For example, sitting in a small room with concrete walls, my waterfall graph on 50 Hz is not very good despite bass traps.
I haven’t checked speaker/driver phase, I have just assumed that Dynaudio knows how to do that on their speakers.
+1
I’d also be prepared to pay extra for something akin to Dirac that was within Roon.
3 Likes
His tutorial (assuming @Magnus is male) has been wonderful and has made a huge difference in my listening experience.
1 Like
Hi @Martin_Maisey
I use a pair of MA-201 from around 1990. I’ve modified them quite a bit, (better filter components) but the basic design is the same. With room correction I get incredible performance from the speakers. RC is a (Cross Spectrum Labs) calibrated Dayton EMM6 mic, Sound Devices USBPre2 and Audiolense XO. Amp is a Devialet 250 Pro.
I think the MA-201 where £200 at the time…
My RC chain would be around USD1400 today. I can not imagine living without RC. I am sure your Bronze 5s would keep up!
2 Likes