How do you get Roon to see the test files? I added them to a watched folder but i can’t find them in Roon.
Carl
Please disregard my question. I got it by forcing a rescan.
Carl
1 Like
Great guide which i’ve been following thanks with great results.
quick question, what do you mean by “vector multiply”, there’s no obvious setting I can see in REW. is this the trace arithmetic function A * B? or did you mean vector average?
Thanks
Thank you. It is the trace arithmetic A * B which actually is the vector multiplication of the speaker response with the impulse.
1 Like
Is there a reason to manually adjust phase with minimum-phase filters in rePhase instead of just using linear-phase filters?
For example, put all corrections at or below 100Hz in bank 1 in rePhase, and the corrections for higher frequencies in bank 2, then set bank 2 to linear-phase filters (se pic)
Hi Magnus,
I have never thought of that but sounds like a great idea. In one combined convolution, you could get rid of the so called transient smearing effect of min. phase filters and pre-ringing effect of linear phase filters. Speaker crossover induced phase shifts (if any) would still need adjusting in paragraphic eq. though. Let me know how it unfolds please.
I just tested it on a moving mic RTA measurement, so I can’t do more than adjust the phase so filters above 100Hz gets a flat phase. So I put the 3 filters I had below 100Hz in bank 1 with minimum-phase setting, and the rest in bank 2 with linear phase setting and generated convolution file.
The sound signature is pretty much the same, but some background details seems clearer and more detailed (less smeared).
Imported the convolution file into REW and it looks like it works as it should:
Sounds good. I will try the same soon as I can. One thing just popped to my mind, linear phase filters will intorduce a group delay to high frequencies (they are all delayed the same amount so wouldn’t matter if applied to all frequencies) compared to min phase of bass frequencies. Did you hear any such oddity?
Serkan and Magnus,
very happy to have come across the room correction threads on ROON and both of you. I need lots of guidance to ensure I am not messing up.
My rig consists of a pair of newly rebuilt Magnepan 3.6s actively crossed over at 80 Hz using a Bryston crossover. My subs are triple 12 inch open baffle design from Gr research. I am using my corei7 lap top as core. RPI4 with Roopie is feeding my R2R DAC via USB.
Below please find the frequency response for both channels at LP.
Here is the pahse for both…
impulse for both…12 ms spike is front wall reflection (dipole speakers)
and here is the spectrograms for one of the channels (they are pretty much identical)…
Here is what I do.
1- I match phase of subs and speakers at 80 hz
2- I purposely increase gain of subs to allow me to flatten bass curve by cuts only using REW
3- I only equalize in REW up to 200 Hz. Please note that my Maggies internal crossover for bass to mids is 240 hz. I stay away from this frequency.
4- I import filters and frequncy sweep into Rephase.
5- I adjust phase starting at 300 Hz, staying away from the crossover
6- I generate wav files from 44 to 192khz all at 65536 taps using 32bit LPCM mono.
7- I import the zip file into roon dsp
Please critique at will. Should I be doing something with the phase shift at the 245 crossover? Should I adjust phase below the crossover down to 100? Anything else, plase let me know…I am just a newbie trying to make my system sound as good as it can.
Sergio,
Congrats on picking up a carefully selected and quite different system. I have no experience with panel speakers or open baffle subs and they could have pretty different phase characteristics. But if you insist 
1) I would do sub phase alignment at the very end after everything else is fixed-80Hz makes sense 2) makes sense but I would still not touch frequencies illustrated with excess group delay graph (#2 in my guide above) 3-5) I am pretty certain you should fix that phase rotation at 240 Hz in rePhase with a simple min-phase linearization filter like this:
(you might need LR 24 db/oct just check and see on graph). Then you can adjust equalization induced phase shifts freely above 100 Hz. Your room’s schroder frequency must be somewhere between 200-250 Hz anyways so up to this frequency is all you can eq without worrying about phase shifts 6) you need a lot more taps for smooth scaling of 192 Hz-try 1048576, 65536 is only enough for 44.1 & 48
Don’t forget to look into room acoustic treatments as well, some things, like to much reverberation, cannot be fixed in DSP (looking at your spectrogram indicates you have to much, but a waterfall graph will show it better).
The difference in phase at higher frequency is also a little troublesome since its so easily creates comb filtering and cancellation. Treating first reflexes might help.
In the first graph left speaker seems to be 1.6 db louder than the right which will cause center image to shift to left. You can lower the left channel volume exactly that much while creating the impulse response in rePhase but as Magnus says you should check and treat the room first. This could simply be left speaker being closer to LP than the right but more likely left speaker is in a denser part of the room , boosting its volume.
Excellent feedback you guys…I have extensive room treatment, the front wall is diffusive, the back is totally absorbant as recomended by Linkwitz (http://www.linkwitzlab.com/rooms.htm) for open baffle design speakers. I will take pictures of my set up and upload them today. I think the difference in volume/phase between the left and right is likely due to wall construction. The left speaker has a side wall of bricks, the right speaker has side wall of plaster. Both of them treated, but perhaps not enough. I tried to adjust the phase a rotation at 240 with a simple min phase as recommended, but if did not do much at all. After literally hours of tweaking both the minimum phase filters and the paragraphic phase bands, here is what I was able to accomplish, but I have no idea as of yet how it will sound…A reminder that I have done this a million times but this is the first time I venture into the crossover and below. My previous phase adjustments were very simple and minimum as long as I addressed 300 Hz and above.
before
after
Believe it or not, with all those phase changes, the sound is pretty decent. I am very surprised. I thought I had just made a holy mess!
1 Like
I think you would get better result by more conservative changes. Absolute phase is not important (i.e it wont matter if you are at 0 or 90 or something else), what matter is to avoid fast phase changes.
I am pretty sure a slightly declining or increasing line is much better than a straight line with lots of small bumps on them.
Your Qs seem to be too small (large) and probably cancelling out each other hence the need for that many corrections. But it wouldn’t really matter as long as the overall result is good (rePhase comes with 16 banks for each channel for a reason). It would just create more taps and delay overall but nothing your i7 would have trouble chewing through.
Treatment to absorb lower bass frequencies more than 3 db is very hard and looks very ugly as well. I had similar left/right reflection material difference and eventually could only fix it in rePhase.
If you send me your initial left and right channel REW measurements (all of them - not the vector averages) I can have a close look and give you better information.
Much appeciated…I will send them tomorrow morning. Very eager to have an expert look at my measurements and perhaps suggest a better way to optimize the sound quality.
Guys,
here is the mdat file for my room. I have used a 6 inch grid aroud the LP. I am also including pictures of my system/room. Most people make fun of it, comparing it to a CIA interrogation room. As you can see, it is a dedicated basement room which I can use however i how want. At this time I am still experimenting with set up. I have the Maggies in their raw look as I am monitoring any Mylar delamination or wrinkling. My goal is to get it to a point where I can get the most of it and then spruce it up to make a bit comfortable and warm. The room is 19’ wide, 35’ long, and 8.5’ high. Of interest is that the factory actually voices these speakers in a similar room (side walls are both cinder blocks though). I would appreciate any insigth you may have on where you would start optimizing. The bass region is purposely high in order to get it ready for eq. using REW. The system sounds awsome to me with just REW equalization from 20 to 200 and no rephase tweaks. I have been using rephase lately and i did notice a bit more clearity in the high frequencies. The subs are crossedover at 90 Hz. The Maggies are used with their external crossovers which have a 240 Hz bass to mids xover.
A reminder that since these are dipole speakers, they produce sound equally front and back. The 12ms reflection is the reflection from the front wall.
One of the misteries I am trying to understand is the phase shift for the left speaker off axis. ooks good for center positions but once the mike is offset, I see the shift in phase. Is it due to time alignment? If I do time align all, it goes away. I am using the right speaker as the timing reference.
https://drive.google.com/file/d/1yUQ-c0H418o6wijU8v-CZienLcdEmwrR/view?usp=sharing
1 Like
Sergio, can comfortably say from the looks of this room, you’re a true audiophile I will try and create a quick and reasonable convolution for you when I find the time. At the moment I am still working as I am based in Germany (in a bank btw). I guess from your distance units you’re in the US. Some quick points:
-
your right speaker is 1.09 in closer to LP than your left
-
Those two major dips at around 190 Hz and 232 Hz cannot be equalized (confirmed by the exceess phase graph). These correspond to 1.6m and 1.3m respectively,. I am assuming one is the front wall and the other is either the floor or the ceiling. Btw I don’t understand much from bass traps but I had read an expert article once that treatment to corners, FW, RW, Ceiling and floor all improve the sound but side walls are not a good idea. Just for you to consider
-
There is a strange 5db drop in the right speaker treble from around 1500 Hz to 12.5 Khz which could be a defect on the right speaker:
-
Lastly boosting the sub volume so as not to use EQ boosts might be a good idea but you’re leveling the bass dips only on the smoothed graph, those dips are usually infinitely deep in the unsmoothed graph and rising them up wouldn’t really make a difference in reality. I am not sure though, just an idea.
Thank you very much Serkan…I will try to find the reason for the left vs right difference in the treble region. Perhaps the fuse. How did you determine that the right speker was closer than the left? Impulse response? what are the units for the 1.09? Distance or time? Fortunately I do not equalize 232 nor 190 because I set all my boosts to 0.