REW and Re-phase for dummies.

[racerxnet]

Guys,

I have created several filters with help from others using REW, and want to push this a bit further. Directions for Re-phase and creating the filters seem like a foreign language to me. Not sure if I need to record the room response with the filter and then export it as a .xml, use the .wav file generated in REW, use the .req file or what. Measurements have been taken with the Umik1 and saved. What I hear from the system sounds pretty good. Any advice is appreciated. Maybe I can help others in the future by learning for myself.

Thanks,

MAK

 

[Ron Texas]

Record your room response without a filter. The XML can be imported into Re-phase so you don't have to type it in. It's also possible to export the filters as text and use them as an include file in Equalizer APO. It's probably the way to go with foobar.

 

[racerxnet]

I use the convolver in foobar for the created files. Why don't I include the corrections made with REW in Re-phase? Do I import the phase corrections and add the correction files afterward?

 

[pkane]

Guys,

I have created several filters with help from others using REW, and want to push this a bit further. Directions for Re-phase and creating the filters seem like a foreign language to me. Not sure if I need to record the room response with the filter and then export it as a .xml, use the .wav file generated in REW, use the .req file or what. Measurements have been taken with the Umik1 and saved. What I hear from the system sounds pretty good. Any advice is appreciated. Maybe I can help others in the future by learning for myself.

Thanks,

MAK

Hey, Mark! Here are a few threads that might help you (sorry, forgot to include the links earlier):

https://audiophilestyle.com/forums/topic/30035-using-rew-and-rephase-to-generate-amplitude-and-time-domain-corrections
https://www.diyaudio.com/forums/mul...-linearization-eq-fir-filtering-tool-196.html

I still think there's more work that you may want to do to optimize speaker position and room treatment before you delve into excess phase correction. The CA/AS thread has a link with a detailed tutorial by swissbear on how to do the REW/Rephase correction in his post.

 

[noobie1]

I use the convolver in foobar for the created files. Why don't I include the corrections made with REW in Re-phase? Do I import the phase corrections and add the correction files afterward?

You import filters created in REW to RePhase and make phase corrections to the imported filters. You then export the frequency and phase corrections as an impulse response which you can use with you convolver. You can check that your final impulse response file is correct by importing it into REW and performing a convolution with your room measurement.

 

[racerxnet]

Paul,

I will address the room once again and measure the response. As I get more accustomed to using REW and its functions, I will get a better handle on the problems. I will measure the before and after results later this week.

Happy New Year.

 

[JoachimStrobel]

Paul,

I will address the room once again and measure the response. As I get more accustomed to using REW and its functions, I will get a better handle on the problems. I will measure the before and after results later this week.

Happy New Year.

You would want to be very careful about measuring and possibly averaging the phase response above 200—600Hz. Measuring the correct phase response around your LP is the most tricky part, the manipulations in RePhase can be learnt with the links given.
You could start with just a single LP measurement and then play with that. Then you would take several measurements and vector average them. But review the phase of each to be averaged measurement and kick out the off response. A frequency dependent filter will give much nicer phase measurements. But I am not sure if that is not fake. A 1/48 filter might more honest but will leave a complicated phase response. My evolving understanding is, that Phase should be best measured with a microphone array.

 

[Le Concombre]

Quick tips:

Hand pick corrections below Schroeder or slightly above (until measurements become less messy and L & R channels yield similar responses (because the speaker, not the potentially different interactions of the speakers with the room, is in control) while looking at the Steady state measurements average (no gating/FDW in REW)

Hand pick corrections above Schroeder when measurements become less messy and L & R channels yield similar responses while looking at a gated Average (ie FDW 3) so you only correct Direct Sound/Early reflections (FDW 3 means that above 300 Hz only things happening before 10 ms is taken into account, above 3K it's 1 ms etc). Basically you keep the profile and just allow yourself tiny corrections to smooth it

I ended up adding REW automated corrections below Schroeder for more precise corrections. It's debatable. Try.

As of the bass boost target below Schroeder I use 4 "targets" : EBU, BK, Harman's "Trained Listeners" and JBL Arcos. There's no such thing as a sound respectable Full Range "target". Good speakers just tend to fit BK and Toole's Steady State predictions.

So called target RR1 that come from misreading Harman's research should be avoided if you listen to anything acoustical because of its fat 100/200 region. Maybe it's no hazard that recommendation for advanced Time domain correction come from people who claim extended bass response based on RR1. In my system (active speakers recently adjusted by Cabasse, the manufacturer) adding a High Pass filter at 11 Hz and strictly avoiding to beef the 100/200 region yield the kind of faster mean and lean bass response you get from advanced Time correction less the headache and debatable side effects of either promoting multiple phase wraps in the bass region or adding preringing

 

[racerxnet]

Hey, Mark! Here are a few threads that might help you (sorry, forgot to include the links earlier):

https://audiophilestyle.com/forums/topic/30035-using-rew-and-rephase-to-generate-amplitude-and-time-domain-corrections
https://www.diyaudio.com/forums/mul...-linearization-eq-fir-filtering-tool-196.html

I still think there's more work that you may want to do to optimize speaker position and room treatment before you delve into excess phase correction. The CA/AS thread has a link with a detailed tutorial by swissbear on how to do the REW/Rephase correction in his post.

Here are some sweeps after moving the Bass columns around for the afternoon. A much improved response to me based on the sweeps taken. Next is how to correct the bottom end so I have some results that sound good.

MAK

 

[ernestcarl]

Hand pick corrections above Schroeder when measurements become less messy and L & R channels yield similar responses while looking at a gated Average (ie FDW 3) so you only correct Direct Sound/Early reflections (FDW 3 means that above 300 Hz only things happening before 10 ms is taken into account, above 3K it's 1 ms etc). Basically you keep the profile and just allow yourself tiny corrections to smooth it

Not related to rephase exactly... but, the results via MMM and FDW 10 for EQing each channel (at the listening position) for me yields similar level of improvement in phase matching -- results differ mainly in the magnitude correction. Nevertheless, both result in a flatter response.

 

[racerxnet]

Guys,

I tried several times to correct the hump at 32hz and the loaded filter sounds bad. The bass columns are sealed, 4 drivers per column x 2. Not sure what the target settings should be. I am wanting to correct up to 300 hz. Crossover it at 85 hz. Phase at 5. Your input is appreciated.

MAK

 

[pkane]

Here are some sweeps after moving the Bass columns around for the afternoon. A much improved response to me based on the sweeps taken. Next is how to correct the bottom end so I have some results that sound good.

MAKView attachment 105142View attachment 105143View attachment 105144

Can you share the REW file with the capture of left and right that includes phase?

Try a straight tilted line target first, and then bump up mid-bass/bass if the result sounds deficient in that area. Don’t correct sharp peaks or dips.

If you measure the corrected speaker response, you’ll be able to see if the filters you picked are working.

 

[racerxnet]

Paul,

I hope that this is what you are asking for. I cannot attach the .mdat files to look at here. In the "target settings pane" are there specific settings I should be using for the low end correction.

MAK

 

[pkane]

View attachment 105220View attachment 105221

Paul,

I hope that this is what you are asking for. I cannot attach the .mdat files to look at here. In the "target settings pane" are there specific settings I should be using for the low end correction.

MAK

Well, an .mdat file would make it easier, but you have a room mode at 33Hz or so. Since you are not using subs, I think you'll just need to attenuate that peak. What happens when you add, say a -10dB gain, Q=7 or 8 parametric filter at 33Hz?

Here's what your original file looked like, before room changes, look around 33Hz to see the extra decay time for that frequency:

The rest of the response looks OK to me, except for the 120Hz peak which you may also want to try to correct.

 

[Le Concombre]

Maybe you should consider correcting L+R as a whole in the Low end.
Generate filters working for that summation (should be 3 or 4, very few) up to 500 Hz
Generate maniac automated filters in REW for each channel with 1/24 resolution for correcting up to 500 Hz
Delete as many as you can in RePhase (generate FIR and then play with cursors, corrections appear in a different colour) ; mainly keep eQ points downing peaks.
As in attachement BK IX I it might look far from ideal but it's important not to generate blotted bass response by trying to fill narrow dips
Even at 1/24 resolution the summation is very satisfying IMO, as in L+R BK vs TOOLE AND BK PREDICTIONS.
The kind of Q and amplitude I allowed can be seen in the "capture d'écran". Please note that above a -1.5 4.36 eQ point @369 there are only 8 eQ points beyond, with amplitudes less than 1 dB and Qs no higher than 2.87.
BTW they reveal there's sort of terra incognita around 400 Hz, in my case where correcting the Steady Sate is not perfectly kosher since we're above Schroeder and where just keeping the FDW 3 smooth might be misleading. I strongly suggest you consider that above Schroeder the evil is in overcorrection. Hence I'm keeping Aligned FIRs regardless of how corrected Steady States or FDW3 look. The map is not the territory.
Last but not least, while we are looking at the corrections we generate and not to the resulting convolved Steady States or FDW responses, I suggest you smooth them as much as you can while keeping the same enveloppe. I identify and lower peaks at 1/24 resolution but good looking psychoacoustically smoothed bass response should be favoured IMO over perfect looking correction at high resolution (see smoothing FIR)
First thing smooth down the peaks at 33 90 180 (or so it seems) as seen in the both view and you should be good

 

[pkane]

Here are @racerxnet 's latest measurements (left, right, and combined) in REW .mdat format.

I'm not used to making correction filters remotely. For me it's an interactive process with corrections, measurements, and listening followed by many iterations. As I see it, there are some room modes, early reflections and cancellations that should all be addressed. Asking for your help

 

[Le Concombre]

Here are @racerxnet 's latest measurements (left, right, and combined) in REW .mdat format.

I'm not used to making correction filters remotely. For me it's an interactive process with corrections, measurements, and listening followed by many iterations. As I see it, there are some room modes, early reflections and cancellations that should all be addressed. Asking for your help

I once thought of doing it for money ; I could do it for free to show my know how but would require at least 9 measurements taken "à la Dirac" of each channel
RePhase could certainly help with time domain in this situation. I stopped messing with phase with my system but Cabasse did a wonderful job to start with' For reference I attach pictures of my impulses steps and GD WITHOUT any time domain correction after amplitude correction

 

[racerxnet]

Here are some corrections I made from 15Hz to 300hz. I did not try to boost the dip in the Left channel at 74Hz. With the combined corrections the graph looks much better to me. Paul, I added the Left spectrogram with my corrections applied. Comments welcome.

 

[pkane]

Here are some corrections I made from 15Hz to 300hz. I did not try to boost the dip in the Left channel at 74Hz. With the combined corrections the graph looks much better to me. Paul, I added the Left spectrogram with my corrections applied. Comments welcome.

View attachment 105648

Looks better, Mark. Try smoothing the SPL plot, say 1/12 octave or use Var Smoothing. Then, also zoom in on the vertical scale on your graph, so that 1-2dB variations become easier to see. Phase may need to be smoothed too, but this will require a larger number of sweep captures from different positions, combined using vector average.

 

[ernestcarl]

Here are some corrections I made from 15Hz to 300hz. I did not try to boost the dip in the Left channel at 74Hz. With the combined corrections the graph looks much better to me. Paul, I added the Left spectrogram with my corrections applied. Comments welcome.

View attachment 105648

I think in many cases, one is just better off unchecking the phase trace altogether to focus in on the magnitude response. You can "linearize" the phase later with little to no noticeable change in the SPL trace. Phase becomes really important to look at when you are summing multiple speakers e.g. sub(s)+mains