Background

I am currently using REW to measure, set filters and export the filters as a stereo convolution wav file.  I am exploring whether I can improve the room correction by using both REW & rePhase.  I am looking for guidance from rePhase users on whether I should take the next step.

 

Key question

Given I am new to rePhase and am struggling to understand phasing, what needs to be done etc., should I even attempt to use rePhase to correct the phase?

 

Details

Using Swissbear’s tutorial as the starting point, I can do the REW measurements, export measurements and filters to rePhase, generate the amplitude correction impulse, re-import back to REW and apply the correction impulse to the averaged measurement.  I can then generate the excess phase.  Basically, I got to page 6 of 8 of the tutorial and hit the wall!  Tutorial sparse on detail from here.

 

Do I export the excess phase from REW as a measurement? (Top of Page 7).  I assume that I will then use the “Filters Linearization” and “Phase EQ” to somehow  make the phase curve as flat as possible. But I think I have reflections in my data e.g. around 5kHz and possibly missed something in the tutorial ?

 

I can get rePhase to generate left and right amplitude correction impulses (wav file) and can combine them in REW to generate the stereo convolution wav file.  Maybe I should stop here and forget about correcting the phase?   However, if someone is game enough to help with rePhase, maybe I can get phase correction as well ????

Edited May 1, 20196 yr by Snoopy8
Typos

10 hours ago, Snoopy8 said:

Do I export the excess phase from REW as a measurement? (Top of Page 7).  I assume that I will then use the “Filters Linearization” and “Phase EQ” to somehow  make the phase curve as flat as possible. But I think I have reflections in my data e.g. around 5kHz and possibly missed something in the tutorial ?

Yes.   Load the excess phase version that you created in REW into RePhase.     Use phase EQ to create phase corrections for it above 100Hz.

 

Then load in to RePhase the averaged measurement.... the phase EQ you just created will still be loaded/applying .... now just re-add the PEQ .... and you should be done.

 

 

6 hours ago, davewantsmoore said:

Yes.   Load the excess phase version that you created in REW into RePhase.     Use phase EQ to create phase corrections for it above 100Hz.

Thank you @davewantsmoore for helping.  Before I export the excess phase,  I read that I should apply an IR Window with width of 7 cycles (or should be another number?).  I cannot claim I understand the  reasoning behind this.  Assuming that is correct, I can import the measurement into rePhase and the tutorial suggest avoid correcting the phase below 100 Hz to prevent pre-ringing.

 

Before

Not sure what the excess phase line shows, but I assume that is the area of focus to correct the phase?

 

After

I experimented with Filters Linearization and came up with above (which includes Pragraphic Phase EQ).  I am unable to adjust the phase at the 20K Hz range.  What have I done wrong?

 

Once I get the above steps correct, I should be able to complete the tutorial.

1 hour ago, Snoopy8 said:

What have I done wrong?

It's difficult to say.... but I wouldn't worry about it too much.

 

1 hour ago, Snoopy8 said:

Once I get the above steps correct, I should be able to complete the tutorial.

 

What I would do now, is export from rephase, apply it to your speaker, and take another sweep with REW.     Confirm you get a result similar to your AFTER pic above from rephase.

 

If you do - good.   If you don't - you have bigger issues.

 

If you do .... but you still want to try and get a better looking correction ..... then I would start the entire process again.

 

4 hours ago, davewantsmoore said:

What I would do now, is export from rephase, apply it to your speaker, and take another sweep with REW.     Confirm you get a result similar to your AFTER pic above from rephase.

 

If you do - good.   If you don't - you have bigger issues.

 

If you do .... but you still want to try and get a better looking correction ..... then I would start the entire process again.

Thank you for your help and advice.  

 

I re-did the whole process today, this time with a timing reference for the measurement. It was not in the instructions but given the first step is to time align the measurements before averaging, the revised measurements gave a more consistent average between the left and right channels.  I have now done a pass through the complete process and initial listening suggests that the results are good.  However,  I will do further testing and listening to confirm that I have done things properly and chosen the right options (of which there is much to explore!).

I now have @frednork trying this as well. Talk of the blind leading the blind down the rePhase alley...   ? 

 

Using Swissbear’s tutorial, we can do the following:

1. Measure 9 positions using REW
2. Time align and get vector average, export average measurement as text file
3. Apply EQ in REW and save filters as XML file
4. In rePhase, import average measurement, import REW filters and generate impulse
5. In REW, import and apply generated impulse to average measurement to get corrected measurement
6. With corrected measurement, generate the excess phase
7. Apply IR window, width in cycles =15 to excess phase and export excess phase as a measurement text file
8. in rePhase, import excess phase measurement text file

The tutorial gets vague at this point and need help from the rePhase experts including @davewantsmoore  to verify whether the following steps are correct and my description of the steps is close enough?  Please remember that I am new to rePhase and do not know enough of phase correction, terminology...

• The dips in the excess phase indicate where phase correction can be applied.  In the "Filters Linearization" tab, apply a filter at the first dip (above 100 Hz) with a starting crossover of LR 24 dB/oct and continue adding a few filters until the phase (dotted lines) have no more swings.You will need to experiment with the crossover and frequency.  Usually, 3 or 4 filters is enough, with maybe the last filter as rotate.  
  
  
  Am I doing this rePhase step correctly?  Above phase correction is as flat as I can get it..
• Then in "Paragraphic Phase EQ", make the final adjustments to flatten the phase as much as possible.
  
• After the phase filters are completed, go to  "Paragraphic Gains EQ" tab and import the REW filters.  
   
• Have now applied both phase and impulse correction.  The convolution correction impulse can now be generated.

Are the steps correct?  How can the writeup be improved?  What steps have been missed?

20 hours ago, Snoopy8 said:

Are the steps correct?  How can the writeup be improved?  What steps have been missed?

It seems like you are on the correct path ... but I cannot be sure.

 

It might help, if we simplify that what first .... and then once that is understood, we can worry about the how.

 

1.   Generate and export the excess phase from the correct amplitude response in REW.

2.   Import the excess phase into rePhase

3.   Use the phase editing tools (filters linearise / paragraphic phase EQ) ... or whatever they're called .... to flatten the phase

4.   Now you have the phase corrections in rePhase, import the averaged measurement... and import the the amplitude correction filters....  and then export the complete correction impulse response.

 

 

You appear to have corrected your phase below 100Hz  (eg. 80Hz).     That's not a good idea

 

... probably the biggest question I would have at this stage, is "did you generate the excess phase curve correctly?".     What does the excess phase chart look like (in REW)?

 

Quote

At this stage, you have generated the amplitude correction impulse and applied it to the averaged measurement. The result is an amplitude corrected measurement, which will be used to generate the phase correction.

Show us this chart from REW  (SPL and phase).     The "amplitude corrected measurement".    ie. the averaged measurement, with the amplitude correction impulse (that you created with rePhase) applied.

 

Then show us (or show it on the same chart, if you can do so clearly) .... what you get when you generate excess phase.

 

Then show us what rephase looks like once you import the excess phase data.

Unfortunately (and I'm likely stating the obvious here) these sort of tools are very difficult to drive, when you don't understand the result to expect.

 

When you're following instructions and you do "step 5A" .... it helps to be able to know what the result of 5A should look like and why.    Because then if you see something you don't expect, you know something went wrong.

2 hours ago, davewantsmoore said:

It seems like you are on the correct path ... but I cannot be sure.

 

It might help, if we simplify that what first .... and then once that is understood, we can worry about the how.

 

1.   Generate and export the excess phase from the correct amplitude response in REW.

2.   Import the excess phase into rePhase

3.   Use the phase editing tools (filters linearise / paragraphic phase EQ) ... or whatever they're called .... to flatten the phase

4.   Now you have the phase corrections in rePhase, import the averaged measurement... and import the the amplitude correction filters....  and then export the complete correction impulse response.

 

 

You appear to have corrected your phase below 100Hz  (eg. 80Hz).     That's not a good idea

 

... probably the biggest question I would have at this stage, is "did you generate the excess phase curve correctly?".     What does the excess phase chart look like (in REW)?

 

Show us this chart from REW  (SPL and phase).     The "amplitude corrected measurement".    ie. the averaged measurement, with the amplitude correction impulse (that you created with rePhase) applied.

 

Then show us (or show it on the same chart, if you can do so clearly) .... what you get when you generate excess phase.

 

Then show us what rephase looks like once you import the excess phase data.

I have been trying this with @Snoopy8's help

 

 

below is excess phase imported into rephase with no further manipulation

Edited May 26, 20196 yr by frednork

and to finish the story below is the amplitude and phase corrected measurement which was then exported as an impulse for each speaker and joined together in REW for various sampling frequencies 

 

then re measured (quickly today) only single measurement, will need to do multiple again and do average to really know how accurate it is.

 

 

 

How does it sound? I also did a non phase corrected version which sounds really good to my ears,  but the phase corrected version seems smoother with a more cohesive soundstage. Maybe I can describe it as where the room which the recording was done in feels always present whereas the non phase corrected has individual instruments popping up out of (less connected sounding) blackness when their level is loud enough.

 

Also for your info most graphs are 1/12 smoothed

Edited May 26, 20196 yr by frednork

12 hours ago, davewantsmoore said:

Unfortunately (and I'm likely stating the obvious here) these sort of tools are very difficult to drive, when you don't understand the result to expect.

 

When you're following instructions and you do "step 5A" .... it helps to be able to know what the result of 5A should look like and why.    Because then if you see something you don't expect, you know something went wrong.

I agree, hence the title of this thread on whether I should be using rePhase in the first place!

 

@frednork appears to be making more progress than I am, so I will follow in his footsteps.  Any comments on what he has done so far?

"should?"

 

Depends on what you do with it (hammer vs nail)   

3 hours ago, Snoopy8 said:

Any comments on what he has done so far?

Looks ok.... have some (not simple) questions.   eg  "why does that look like that?"

1 hour ago, davewantsmoore said:

Looks ok.... have some (not simple) questions.   eg  "why does that look like that?"

happy to try

21 hours ago, frednork said:

happy to try

What's the solid black line which has a 2dB bump in it around 8k?

 

You amplitude correction introduces some high Q errors?  (eg. 130, 400, 16000Hz)  Why?   (although it's hard to see as you're not using a consistent dB scale in each chart)

 

Your "convolution filter applied" doesn't look like what you had in rephase?   Why?     You seem to indicate that you remeasured (with mic) to get this data?.......  Did you take the measurement under identical conditions?  (eg. there is a massive lump as 2khz, etc.)

 

You appear to have done the process correctly, etc.   Excess phase, etc. looks good.....  but it seems like either your original measurements, or your "remeasure" have an issue in one or both.

 

... what that all means, is I can't really tell whether you've got good result or not.    The various lumps and bumps I'm talking about are not good - but I can tell if they really exist, or are just an artifact of your process.

 

 

What happens if in REW you take the original averaged measurement, and apply the total consolution/impulse  (ie. the one with both the amplitude and phase corrections built into it) ?

First just wanna say thanks @davewantsmoore for tackling this. I do realise its a bit like diagnosing throat cancer by looking up someones a*s*h*le.

 

5 hours ago, davewantsmoore said:

What's the solid black line which has a 2dB bump in it around 8k?

Thats the mic calibration, should have turned off that line

 

5 hours ago, davewantsmoore said:

You amplitude correction introduces some high Q errors?  (eg. 130, 400, 16000Hz)  Why?   (although it's hard to see as you're not using a consistent dB scale in each chart)

This is an artifact of how REW does corrections, it doesnt boost low frequencies much , I suppose due to the possibility it is a null. I can try boosting manually. the one at 16k is the fall off of the tweeter which it also doesnt touch

 

Ignore above and see below

Edited May 28, 20196 yr by frednork

3 hours ago, davewantsmoore said:

Your "convolution filter applied" doesn't look like what you had in rephase?   Why?     You seem to indicate that you remeasured (with mic) to get this data?.......  Did you take the measurement under identical conditions?  (eg. there is a massive lump as 2khz, etc.)

Yes, as this is a single measurement I still need to do an averaged measurement in a similar way to how the averaged original was done which will hopefully fix that.

3 hours ago, davewantsmoore said:

What happens if in REW you take the original averaged measurement, and apply the total consolution/impulse  (ie. the one with both the amplitude and phase corrections built into it) ?

Ok I think I stuffed up the graphs of the amplitude correction by using the wrong impulse. My versioning has failed me (again)

this is what I get when I redo it the 2 corrected traces have been moved apart slightly for readbility

 

 

One thing that is becoming apparent is I am not 100% sure of when I should be using windowing. The turorial recommends that no windowing should be used for the amplitude correction, but to be used on the av measurement transformed by the amplitude  impulse for the determination of excess phase. But there are plenty of other opportunities to use it ie the impulse x av measurement trace. the excess phase trace,  

 

The other question is how to determine the optimal window length?

18 hours ago, frednork said:

?

On 28/05/2019 at 11:05 AM, davewantsmoore said:

Your "convolution filter applied" doesn't look like what you had in rephase?   Why?     You seem to indicate that you remeasured (with mic) to get this data?.......  Did you take the measurement under identical conditions?  (eg. there is a massive lump as 2khz, etc.)

Below is averaged measurement with amplitude and phase convolution used 1/12 smoothed

 

Above shows the 9 individual measurements made for the corrected average trace using the suggested positions in swissbears tutorial

Edited May 29, 20196 yr by frednork

The initial measurements are critical (garbage in garbage out), so it is an interesting exercise to repeat the process with different inputs.

 

How does it sound?

1 hour ago, davewantsmoore said:

The initial measurements are critical (garbage in garbage out), so it is an interesting exercise to repeat the process with different inputs.

 

How does it sound?

It seems to have more body and life and realness than the no phase correction version. Its also smoother sounding.and as I mentioned before individual instruments seem to take up more of a 3d space rather than flat images in a 3d space likethe difference between well and poorly made 3d movies. Overall pretty happy. I did run a DEQX for a while and wasnt able to achieve as good a result but I suspect my understanding and skills were not up to it and my equipment has gone up  a level or 2 since then so maybe more noticeable also.

 

I will do another correction for a different seating position so should be interesting.

Any thoughts on when to or not to use windowing