rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool

hi,
i have subwoofer measurement using REW, and the phase are not good.

attached please find my measurement, i measure the subwoofer in 7 position, and generated EQ using REW and applied to my 2x4HD.
the chart 10 is the measurement after EQ, and the phase has sharp drop.

so i use rephase to adjust it but i am not sure how to deal with a sharp drop in phase at around 72hz.

could any one please give me some hints?

thanks.
 

Attachments

  • 20181014 sub minidsp.zip
    965.6 KB · Views: 57
  • 123.png
    123.png
    54.7 KB · Views: 238
That phase shift at 72Hz corresponds to the magnitude sharp notch at the same frequency. If you could address that notch using minimum-phase EQ then you would take care of the associated phase shift in the same time.
Not sure you can and should address that notch though, and in that case the associated phase shift should also be left alone.
Is that notch still present with a close mic measurement ?
 
That phase shift at 72Hz corresponds to the magnitude sharp notch at the same frequency. If you could address that notch using minimum-phase EQ then you would take care of the associated phase shift in the same time.
Not sure you can and should address that notch though, and in that case the associated phase shift should also be left alone.
Is that notch still present with a close mic measurement ?

I am not very familiar with the minimum phase EQ, all I use is just the Auto EQ feature by REW. Would you mind tell me how to do a proper minimum phase EQ ?
I measured with mic on my MLP at ear level after EQ, and measured 7positions before EQ, also pos1 is the MLP.

Thanks.
 
...could any one please give me some hints...

Agree what pos hints and looking your mdat file shows some user stuff had to been understood in the long run, observed is IR (impulse response) which is ground for calculate phase is not alligned on timeline and actual about 11mS too early so for that use "Estimate IR delay" after each measurement or redo it for that mdat session you shared, also IR is inverted as far i can see so either invert it each time a new measurement has been done or go to "Preferences/Soundcard/Input Options" and tick on "Invert" so that any future measurements is inverted automatic. Info tells sweep is 2-200Hz lenght 256k and samplerate is 3000Hz, play around with these settings and mayby go a bit higher and think even a full bandwidth sweep will work, and that 3000Hz sample rate number sounds strange is that because source and output is listed as ASIO4ALL.

When pos talks minimum phase EQ also called IIR filter in digital domain he talks normal EQ as we know from analog gear and all EQ manupulation into REW is of minimum phase type, try understand normal our enviroment and musical instrument sound plus analog recording and reproducing gear is all of minimum phase type where phase pattern is predictable and follow amplitude domain, therefor if we EQ for example any non smooth amplitude error we also automatically repair phase, the confusing thing is probably in digital world we can also use FIR filters that can be either setup to behave as a minimum phase filter correction but it can also manipute phase on its own. In general use always IIR filters for any correction of a band pass and use only pure phase correction to repair if any real summing IIR XO point filters is used in system. In REW you can tell it to "Generate minimum Phase" and that trace show the phase that belongs to the measure amplitude and deviation relative to your measured phase is called excess phase and can for example be because measurments chain units is not calibrated or also in a acoustic domain measurment there can be tons of reflection that all sums or cancel into amplitude domain provided if they in perfect timed phase or some delayed cycles phase at arival time or out of phase. Try open your chart 10 again and "Estimate IR delay" as many times until dialog tells its at zero then "Generate minimum phase" and now see how phase for that measurment should have looked if it was perfect without any excess phase from non calibrated chain units and acoustic reflections. A help for acoustic reflections is adjust the time window especially at these low frq, try set "IR Windows" to FDW (frq dependant window) 6 widths in cycles before doing any EQ should help for filter out some reflections and not overdo EQ.

Other help can be generate textbook target files in Rephase as IR and import to REW so as to have some textbook graphs to look at and learn from, example is below where a band pass BW2 at 20Hz - LR8 at 100Hz was set to replicate something close to your curves.

From over here software the below attached pdf white paper should be good reading for frq as low as these also because frq is so low it should be possible using white papers nearfield method to set a perfect band pass at enclosure location and then via the shared spreadsheet over there called "Baffle Diffraction and Boundary Simulator v1.2" calculate and export a frd file to Rephase to be reverse EQed so that room nasty gain and location problems is corrected for.
 

Attachments

  • 1000b.png
    1000b.png
    297.9 KB · Views: 225
  • White Paper - Accurate In-Room Frequency Response to 10Hz.pdf
    754.4 KB · Views: 97
Last edited:
  • Like
Reactions: 1 user
Hi BYRTT,
Thanks a lot for your reply.
I am a newbie in using REW and EQ tools, reading a lot of related articles and trying to understand what has happened to my system and trying to fix them.

Yes I am using ASIO4All to do measurement, and using lefft speaker as reference timing to measure the subwoofer.

I will try the raw measure again with your suggestion and correct the inverted IR and align the time before posting it again.

Thanks for sharing the readings, it is very useful.

Thanks.
 
Different results using REW on different computers

Hello:

First I would like to thank POS for RePhase... what a great program which opens many fascinating options for the DIY community.

Here is my question. I have a 3 way active system (tri-amp) using a DIY op amp based analog crossover (LR4 @ 300 Hz and 3000 Hz). I have been using RePhase for fine tuning the EQ and to phase align they system.
I have REW loaded on the computer that also has my music and music player (Foobar 2000) on it. In addition I have REW loaded on a lap top. I have a UMIK-1 USB microphone I use for measurements.

After a lot of trial and errors using both tools (REW and RePhase) I have been able to generate impulse wave corrections and using " foobar convolver" with Foobar they seem to be working. The problem is if I take measurements using REW from the computer foobar is on to generate the corrections and then "test" the correction with REW on the lap top the phase is not aligned. If I use the lap top to measure and create the correction and then "test" that correction using the lap top the phase correction "works". If I "test" the correction generating using the REW on the playback computer it also works. I have gone through the REW settings on the two computers and they both appear to be the same to me.

So how can I tell which or if either of the REW measurements are telling me the truth? The laptop measurements and the corrections they generate "seem" more sensible to me and are much easier to correct in RePhase i.e. once I set the crossover correction filters it takes care of 90% of the phase alignment where as the REW measurements taken with the "Playback computer" require more corrections with the phase equalizer. Any help in reliably being able to confirm my RePhase correction is working would be appreciated.

Thanks again.
 
Hi levimax,

Agree it can sometimes be a bit longhaired process get same computer do all loops of playback/correction/recording flawless at the same time, i'm on JRiver myself so know nothing about feature settings into Foobar so will try hint the best i can. In REW/JRiver chain a good test is a hardwire loop from soundcard output to input and if that doesn't look clean with JRiver DSP turned on and off when looking into REW impulse or distortion tabs then forget run the hole acoustic chain, what helps get it clean is adjust buffer sizes into JRiver which also helps on sound quality should there be a flaw, also if sound is running via Windows sound system then ensure windows is setup to same bitdepth and samplerate as player expect because converted samplerate even it works give some smaller distortion especially for precise phase extraction. In REW "Make a measurement" window theres a setting named "Lenght" and it can mean alot when chain is long with a lot of delay so try set it to 512k and 1M and see what happens there.

Should above not help your situation there is a kind of failsafe way called "Offline measurements" and i often use it myself when using soundcards that is split into two seperate physical units (USB) where one is for input and the other is for output, not 100% shure but think that setups culprit is they each on their own non syncronized clock plus the seperated USB buss interface. In REW go to help/Show help/Offline measurements to be guided, in short you will end play a wav file into Foobar that REW has generated, then on same computer or another computer you record that sweep for example using free Audacity program and end drag the sweep file plus recording file into a window in REW that calculate its difference. Tip if you still get wrong phase or frq range is somehow not extended as expected then add some silence in start and end of that REW genereated sweepfile, Audacity is again our friend doing that trick.
 
Last edited:
Thank you very much for your helpful reply.

Thinking about my playback computer even though I use WASPI I do have a USB DAC (not a sound card) and then a USB microphone and I am also running DSP's in Foobar (Convolver and Reclock) so it does seem that trying to get that to all work perfectly enough to measure phase is a long shot.

I am going to try the "Fail Safe/ Offline measurement" method you mention using a separate computer to make sure I am getting accurate phase measurements to start with and to verify the correction is working.
 
...Anyone implemented crossfeed for headphones with FIR and/or rePhase?

Not on a permanent basis but in its build into JRiver player its easy to try out and get feet wet from a subjective standpoint, if you wonna try it and haven't any JRiver license can say a trial version will work unlimited and full featured for 30 days so plenty of time to evaluate effect, in below we look into DSP settings where if out at left you tick on "Headphones" you get those comments and settings over in right side.

attachment.php


Having subjective fun using effects, that headphone effect can also be turned on for speaker use but in speaker regard think its better use the "Effects" container over at left where one can set various strength of surround field and also a reverbation enviroment.

That said lately got a EARS unit from miniDSP and hope it will help understand headphone domain a bit better and set some more or less perfect DSP corection schemes down the road, what i discovered so far is if we model and set up headphone enviroment into one of Jeff Bagby's spreadsheets that is normal for room enviroment use we get some curves myself never thought of before and if they kind of true to reality reminds me why bass area into headphones is often subjective sensed not have the same realism a real musical instrument performs, below graphs is on axis models of room gain plus baffle into headphone domain verse a normal 25 sq meter room, looks in headphone domain enviroment we have huge gain all from DC point up into kilo hertz.

attachment.php
 

Attachments

  • 1000.png
    1000.png
    31 KB · Views: 533
  • 1001.png
    1001.png
    16.9 KB · Views: 528
Last edited:
Hey Pos, or anyone else in the know :)

I've been trying to really learn how minimum phase crossovers time align and sum.
So I've been generating impulses in rePhase, and studying them and their summations in REW.
Many aspects are making sense!

But here's something simple I don't get...

I made a LR4 high-pass at 6300Hz in rePhase.
The REW impulse response looks inverted to me, but I know it's correct by the way it sums textbook-like with a 6300Hz low-pass when properly time aligned.

Why does this 6300Hz HPF impulse appear inverted?
Many thx, Mark
 

Attachments

  • rephase 6300 HP.JPG
    rephase 6300 HP.JPG
    172.4 KB · Views: 347
  • rew 6300 HP impulse.jpg
    rew 6300 HP impulse.jpg
    44.6 KB · Views: 119
Last edited:
  • Like
Reactions: 1 user
Is it roof of nyquist or what mark100, i was in the middle of test myself and got same result at lower rates where 96kHz and up got better, then fired up new software to me in free VituixCAD and it confirmed with below rates 44/48/96/192kHz.
 

Attachments

  • 1000.png
    1000.png
    142.8 KB · Views: 135
Last edited:
Is it roof of nyquist or what mark100, i was in the middle of test myself and got same result at lower rates where 96kHz and up got better, then fired up new software to me in free VituixCAD and it confirmed with below rates 44/48/96/192kHz.

BYRTT, thanks for looking.

Maybe I didn't get it...you've given me new doubt ..lol
But that said, i didn't get a change in polarity by going from 48 to 96kHz ...???

After seeing my rePhase/REW sim results, I'm now looking at real transfer functions under Smaart...and see the same inverted impulse on the HP side.

I rationalized it was due to a 4th order LR making 360 degrees rotation...which has to be split on both sides of xover...so +180 low, -180 high. With -180 being an inversion. Maybe that is full of sh&&t haha

But to test the theory, I tried using LR8's where it I'm thinking 720 degrees gets split +360, -360. Both these should be positive polarity..(i hope :)
They do measure that way....so maybe....

Please do check the results at low sample vs high...dunno what to think there...rate shouldn't matter should it ???