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

have you tried the waw export formats in rephase?

Yes - the .wav format works with Roon - but when I export in this format from rePhase I get a message saying that the level has been lowered to avoid clipping - and the amount of level reduction seems to vary between different sample rates. The message suggested that this could be avoided by using '.dbl' format which is why I tried this approach.
 
Looks like the problem was with the way that I created the .cfg file which Roon uses to parse the filters (Roon requires all filters and .cfg files to be contained in a single .zip file).

My first attempt was to create one .cfg file with links to all 12 .dbl files (44.1, 48, 88.2, 96, 176.4 and 192 for L and R channels). This didn't work.

My second attempt was to create a separate .cfg file for each pair of LR filters - so the .zip file for use in Roon now had 18 files - 6 .cfg files and 12 .dbl files). This works.

Interesting to note that, when looking at the signal path in Roon, filters created using rePhase show as using 16K taps - this is consistent with what I exported - I used the default setting in rePhase of 16,384. If I use .wav files from rePhase they show the same - 16K taps - as expected.

If, however, I export .wav impulse files from REW then the size of the .wav files is bigger - 584Kb for 44.1KHz - and varies by sample rate - 2.1Mb for 192KHz. When the filters exported from REW are used Roon signal path shows 131K taps. The .wav files exported from rePhase are 16Kb.

Should I be using more taps when exporting from rePhase?
 
You should probably use more taps for the higher sampling rates.
Think in term of FIR duration, not pure taps, because duration is the parameter that defines the filtering capability (resolution) of a given FIR.

For example an arbitrary 1Hz resolution requires a 1sec FIR, ie 48000 taps at 48kHz and 192000 taps at 192kHz.
 
Last edited:
Delay Video in Windows

Hello, I'm new here, but not new to rePhase.

I have a Windows based system, running rePhase (+REW & eq APO). It drives my 3 way active open baffle speakers very well. Been running great for ~ 2 years now. Thank you very much Pos.

rePhase does my FIR filters, driver EQ, and some room EQ, all (approx) linear phase. But this means I need lots of taps, and so have 103mS delay.

It works absolutely brilliant for music. Not so much for video with the delay.

Does anyone know a way to delay video in Windows to align with the audio?

Sorry if this has been covered before, or would be better on another forum. I did search, but didn't see anything relevant - mostly suggestions on reducing audio delay, which I obviously can't do here.
 
I run JRiver as a media player that allows me to sync audio and video. Works quite well and has a convolver as standard tool to use the FIR files. I bet there are more media players out there that allow you to sync.

Some streaming services allow audio/video sync, even Netflix has it available behind a certain key combination (I forgot wich, I don't use it anymore). You'd have to re-sync for every new file though.

Just google for it and you come across solutions like this:
YouTube Audio Sync Tool - Chrome Web Store
 
Hi,

I am rather new to Rephase and REW and have been trying to linearise my passive speakers. I am using vs 1.4.3 of Rephase and 5,19 of REW. I have read some tutorials but cannot export the automated EQ from REW to Rephase as .txt files, it isn´t generated in the expected format. What can I do apart from copying all the settings by hannd and inserting them on the paragraphic eq ?

As an aside, and probably the wrong place to ask, but as I try to time-align in REW all the measurements without windowing using the " estimate IR delay" and then accepting the shift , I see that not only some don´t land on the first big step response peak / zero ref. time, the ones that don´t are at diferent, and sometimes largish time distances from the relevant peak...Then if I manually shift them to properly line 0 ref time precisely with the peak often the zero, dotted red line doesn´t line with zero time and the peaks curve sort of "wiggle" and slightly change shape as I try to precisely get the zero on top of the peak...Is this normal, am I doing the alignement right or should I leave it at where the estimated time shift lands?

Thanks, newbie posting here, if infringing normal forum behaviour, let me know, and thanks in advance for any input. :)
 
Right, thanks PMental.

Anyway, I am afraid to get the phase all screwed up on REW with the impulse response delay estimation and the problems described, it doesn't even seem to settle in the same peaks, sometimes tbe first, others on the second, so I am not sure I am getting proper summation of responses when I vector averge them before applying smoothing and equing the result to export to Rephase..
 
:eek:
Is it you Thomas ?

191109020547278424.png
 
Right, thanks PMental.

Anyway, I am afraid to get the phase all screwed up on REW with the impulse response delay estimation and the problems described, it doesn't even seem to settle in the same peaks, sometimes tbe first, others on the second, so I am not sure I am getting proper summation of responses when I vector averge them before applying smoothing and equing the result to export to Rephase..

If you can take a screenshot of the differences in impulse timing it will be easier to help.

The time align function will usually get them within one sample delay, it is possible to get them closer but the difference only shows above 10Khz due to the wavelengths being so short. This shows the difference blue is REW alignment red is manual. 5dB per grey line. Not really anything to worry about.

attachment.php
 

Attachments

  • Screen Shot 2019-11-09 at 11.58.31 am.png
    Screen Shot 2019-11-09 at 11.58.31 am.png
    114.9 KB · Views: 775
Question about @SwissBear's Tutorial PDF. In the last step, it appears there are 3 files for each channel.

1) Averaged measurement file exported from REW (as .txt).
2) AutoEQ file exported from REW (as .xml).
3) Excess Phase file exported from REW (as .txt).

As I read it, it then instructs importing all 3 of these files into rePhase to generate the final convolver file (Averaged Measurement, AutoEQ into "Parametric Gain EQ" and Excess Phase into minimum phase corrections).

How can you load the Excess Phase (.txt) file into rePhase's minimum phase corrections? I only see a way to load the REW AutoEQ and averaged measurement file.

There is an "import REW filter settings (.xml)" pulldown menu option under "Parametric Gain EQ", but I don't see an equivalent for the excess phase import.

I might be missing something.

TIA

Using:
  • rePhase 1.4.3
  • REW V5.20 Beta 27
 
Last edited:
As I read it, it then instructs importing all 3 of these files into rePhase to generate the final convolver file (Averaged Measurement, AutoEQ into "Parametric Gain EQ" and Excess Phase into minimum phase corrections).

How can you load the Excess Phase (.txt) file into rePhase's minimum phase corrections? I only see a way to load the REW AutoEQ and averaged measurement file.

There is an "import REW filter settings (.xml)" pulldown menu option under "Parametric Gain EQ", but I don't see an equivalent for the excess phase import.

I might be missing something.

TIA

The excess phase measurement should be loaded separately, then by using the paragraphic phase EQ and filters linearization the idea is to make the measurement closer to a flat line to remove the excess phase and leave you with an approximate minimum phase result.

The workflow to follow the tutorial should be:
1) Create vector averaged measurement
2) Use FDW and smoothing to avoid overcorrecting later
3) Use REW auto EQ or manual to correct the windowed response to your target
4) Save the filters to file and export the averaged measurement as text
5) Open the measurement in Rephase and import the REW filters from file
6) Generate the correction filter
7) Apply that filter in REW by using A*B function in All SPL tab then reduce level back to original
8) Generate excess phase version in REW
9) Export excess phase measurement as text, import to rephase
10) Correct excess phase back to flat or as much phase EQ as you are willing to use
11) Generate an overall filter in rephase by combining the REW and phase correction

Make more sense?
 
The excess phase measurement should be loaded separately, then by using the paragraphic phase EQ and filters linearization the idea is to make the measurement closer to a flat line to remove the excess phase and leave you with an approximate minimum phase result.

The workflow to follow the tutorial should be:
1) Create vector averaged measurement
2) Use FDW and smoothing to avoid overcorrecting later
3) Use REW auto EQ or manual to correct the windowed response to your target
4) Save the filters to file and export the averaged measurement as text
5) Open the measurement in Rephase and import the REW filters from file
6) Generate the correction filter
7) Apply that filter in REW by using A*B function in All SPL tab then reduce level back to original
8) Generate excess phase version in REW
9) Export excess phase measurement as text, import to rephase
10) Correct excess phase back to flat or as much phase EQ as you are willing to use
11) Generate an overall filter in rephase by combining the REW and phase correction

Make more sense?

Thanks Fluid.

I thought the EXCESS PHASE file was the correction (analogous to the AutoEQ file), not a template guide for manual correction.

The PDF in step 4 indicates loading the averaged file again which threw me off.

"4) ... Once you have finalized the phase correction, you can import the initial averaged measurement in rePhase and apply the REW filters. ..."​
 
Question on linearization

So I have a question on linearization

I have a sub that has a LR4@80hz

My sub is QTC.83 and a f3 of 45hz (sealed obviously)

(18” sundown SA-18)


The box linearization looks a lot like the filter linearization, they are very close as far as shape in rephase


Can I just do the box linearization and get the filter linearization simultaneously with one go at it OR do I need to do both linearizations

I listen to both and both sound good.

What is proper in this scenario? As one is a box linearization and one is a filter , does that change anything or do they both need to be accounted for separately,

Thanks in advance:)
 
Last edited: