My mic was horizontal, pointing towards the speakers, then I did the 2 spirals as Magnus described in his posts. 2 spirals in the region of each ear, with pink noise generator in REW, one measurement with right speaker only (chosen in generator) and the other with left speaker only.
UMIK1 with calibration 0°
Did you do a test with a sweep on a loop back or maybe that doesn’t work for a umik…not used mine yet…a loop back should give you a flat line essentially…
Edit/update seems you can’t do this with the umik
No, pink noise, no sweep.
The flat line is what I expected from the B&W diamond high tweeters in their special housing. But the fall down from 75db to 60db starting at 10kHz?
What distance are you measuring at?
I measure at round about 3m. Anyway: the speakers sound brilliant and the example measurements in the link above look similar.
What you think about my SPL graph?
I’m generating filters 20-600hz to 72dB, no Q above 7. Gives a lot more control in the bass area, no distortion anymore.
The dip in higher frequency is probably ok, it might be because of DAC filter or tubes (if you have any), and it also looks like those speakers are designed a little like that. The result should be a little less airy but more smooth.
When doing partial correction, I think its best if the end of the correction matches the continuation. In this case, looking at the picture, I would go for 75dB for 20-600Hz (unless you use a house curve that don’t have 0 at 600Hz).
Then try once with correction all the way (it won’t correct the dip at higher frequency) and compare and select what you like best. In either case, 75dB looks like a good target if you use default slopes in REW.
New house, new system! The measurement is about what I was expecting, given speaker placement (photo in this post)…
I auto-generated 20-800hZ, and then manually added 2 more points at 2.5kHz and 15.8kHz. First listen just now was pretty great. In my old system I did a lot of fiddling around before I got to something I really enjoyed, so any tips would be appreciated.
No house curve?
No - not in REW. I’ve always done the correction flat, and then had a house curve as another EQ in Roon. I’m not sure if that gives the same result or not?
It also woks, and can be good while experimenting with different house curves. But once you found a house curve you like, you should do it all in REW to get best sound quality.
The correction looks good though, except you have a high Q filter at 15khz which is not optimal. Q up to about 5 is ok at lower frequency, but the higher frequency you correct the lower the Q should be.
I’m going to learn about phase corrections by reading this thread - but this is really, really hard stuff to understand. 
Oh I didn’t realise doing the house curve as another EQ in Roon would be lower quality! I’ll have to settle on a house curve now!
Thanks for the advice on the high frequencies. I’m going to play around with things to minimise the number of filters, and probably keep them all under 800hZ I think. I’ll post back with the results.
My advice would be: don’t bother. I have experimented with phase corrections, but most of the time a normal minimum-phase EQ -> convolution file is best (which is what you get from for example REW). The reason minimum-phase EQ is often best is that the impact from the room is often minimum-phase, so a correction often fixes both amplitude and phase at the same time.
Having said this, phase/timing fixes often will produce better sound, its just that neither REW nor RePhase is made for this. You need software like Dirac or Accurate for that, or use the service provided by Thierry at https://www.homeaudiofidelity.com/
You can try to improve the crossover for your speakers with REW/RePhase though, but that’s not related to digital room correction.
Btw, I got most of this information by chatting with the developer of RePhase, Thomas Drugeon.
In my opinion, PEQ in Roon is not very high quality. But even if it was, you still want to keep the total number of DSP filters and changes to a minimum to get best sound quality.
Here is how my correction looks like, it uses 8 filters and is not very flat but it sounds better than auto-corrected 16 filters. Notice how I have left 2 dips untouched, one is a null-node and one is from speaker-boundary interface.
Audiolense too (XO version with the True Time Domain feature). It’s a bit easier to use than Acourate.
Audiolense is cool - but only Windows. They should build a MacOS version. I think Mac users are not so rare in audio/HiFi circles…
Made some adjustments: XTZ subwoofer/KEF LS 50W
Lower cross-over settings, ports closed on XTZ and XTZ EQ flat.
Also new convolver filter: measurements before and after. Keep in mind this is a smal subwoofer with 8” woofer (dual setup) flat response all the way up to 25 Hz. Even before correction it sounds really good.
Correctiion is set up to max. 500 Hz
Impressive flat and linear curve with just a 8" woofer. Is this a real measurement or an estimation by REW?
