Home Theater Forum and Systems banner

IR response : multiple inpulse

5K views 22 replies 3 participants last post by  Stefc13 
#1 ·
Hello,

I would like to perform a phase measurement at the LP. But I have some difficulties to interpret the IR response:

1st:
There are 2 impulses quite similar within 1 ms. When I'm pressing "Estimate IR delay" sometimes the second is selected...

So I don't know which one I should select in order to fine tune the delay.

2nd
The phase seems to be very weird :yikes: I need to reduce a lot the cycle number in "IR windows" in order to get something acceptable but then I'm not sure this is realistic...

I attached the mdat file. Hope you can help me :smile:

The setup:
speakers are ~3 m from LP, LP is close from the back wall (10 cm between the wall and the sofa).

Thanks
Stef
 

Attachments

See less See more
#2 ·
The impulse response of both channels is very erratic for a 3 m LP measurement. This could be the result of reflections, but I am concerned there may be a problem with the REW settings. It would be good to review the settings. Be sure the same channel is selected as 'reference channel' for both measurement. Pick the appropriate 'measurement channel' for the left and the right measurements. Do not select 'both' for the measurement channel. That is one idea of a possible issue.

Assuming the objective is to adjust the delay such that the L and R sound to arrives at the mic position:
1. We could just measure the difference in distance from the mic to the 2 speaker baffles and adjust accordingly.
2. if these measurements are valid, we can look at the overlay of the ETC chart and note the offset in the 2 ETC is about 0.22 ms or 75mm with the left channel arriving first. A 0.22 ms delay increase on the left channel will work. A small movement of the mic (chair) position to the right will also work.

The following shows how the ETC overlay chart can be used.
Text Green Line Plot Font
 
#5 ·
The impulse response of both channels is very erratic for a 3 m LP measurement. This could be the result of reflections, but I am concerned there may be a problem with the REW settings. It would be good to review the settings. Be sure the same channel is selected as 'reference channel' for both measurement. Pick the appropriate 'measurement channel' for the left and the right measurements. Do not select 'both' for the measurement channel. That is one idea of a possible issue.
Thanks for your help !

I used the same channel as 'reference channel' and measured both speakers separately. I guess all this mess comes from the reflection.

Assuming the objective is to adjust the delay such that the L and R sound to arrives at the mic position:
1. We could just measure the difference in distance from the mic to the 2 speaker baffles and adjust accordingly.
2. if these measurements are valid, we can look at the overlay of the ETC chart and note the offset in the 2 ETC is about 0.22 ms or 75mm with the left channel arriving first. A 0.22 ms delay increase on the left channel will work. A small movement of the mic (chair) position to the right will also work.
My goal is to measure the phase and try to correct it with RePhase. Is it necessary to have such accuracy in the distance ? When I'm listening music I'm not exactly at the same place every time.

It looks that the second impulsion corresponds to the reflection, the mic was around 20 cm from the wall:


How can I get rid of this reflection ? Maybe by setting 1ms in the 'Right window' parameter of the 'IR windows' ? But the measurement will be still relevant ?

Thanks again for your help :smile:
 

Attachments

#3 · (Edited)
May I ask a stupid question here to see if I understand well the Rew index which says: "Measurements will have a time delay that corresponds to the difference in their distance from the microphone compared to the distance of the reference speaker - if the reference speaker is further away the delay would be negative."

Let's say I choose the right surround as reference and measure the front left and the front right. Suppose that rew tells me that there -100mm delay for the left and + 60mm for the right. Suppose I do not want to use the AVR distance setting and move the speakers to put them at equal distance from my MLP
Am I right to conclude that my left speaker is closer to the front wall ( further away of the mic) by 100mm?
What do I do? Do I calculate 100- 60mm= 40 mm and move forward the left speaker by 40mm and bacward the right speanker by 40mm? or
Do I move the speakers the same way but for the distances given by REW?:sweat:
 
#4 ·
FargateOne,
I don't know. I have never looked into the REW reported impulse positions and thus how to remember the directions needed for adjustments. I often get things backwards if I don't follow the procedure that I carefully worked out for myself. I confess that I sometimes get confused with direction even when using my own methods. I think REW reports time to the largest early impulse peak, but if there are several similar sized peaks then a small difference on one of the 2 channels may result in REW picking a different peak. This is partly passing observation, but mostly speculation as I really don't know how reliable it is use the REW reported offset numbers. I would need to investigate it.

My primary method is a graphical one as shown above. I prefer charts to tables of data. I usually use the impulse chart rather than the ETC chart, but sometimes the ETC is easier to interpret as it was in this example. Things that arrive earlier appear more to the left of the chart, so to move the red/left sound to the right to match the green/right ETC position we need to add delay to the left channel. If we instead want to change a speaker distance to make the adjustment then we need to move the left speaker further away from the mic (by 75 mm in this case). We could instead move the right speaker closer by 75 mm so the green curve falls at the red curve position.

Where I sometimes get confused is when I am shifting the impulse position manually within REW rather than changing the delay in a DSP box. To simulate adding delay to the left channel in REW we can manually shift the impulse position 0.22 ms to the right. To do that we need to shift it -0.22 ms. This seems backward to me as I am always thinking about adding delay time. I does make sense recognizing that this is a time scale not a delay scale so negative time is a positive delay? It still is a little confusing to me. I have finally made peace with it however. :)

I hope that helps. It works pretty well for me. I suspect you will need to do some of you own experimentation to become comfortable with a methodology that makes sense to you.
 
#6 ·
FargateOne,
(...)
My primary method is a graphical one as shown above. I prefer charts to tables of data. I usually use the impulse chart rather than the ETC chart, but sometimes the ETC is easier to interpret as it was in this example. Things that arrive earlier appear more to the left of the chart, so to move the red/left sound to the right to match the green/right ETC position we need to add delay to the left channel. If we instead want to change a speaker distance to make the adjustment then we need to move the left speaker further away from the mic (by 75 mm in this case). We could instead move the right speaker closer by 75 mm so the green curve falls at the red curve position.
I will try to interpret graphs and impulse chart to see if I better understand.

I would need to investigate it.
If you do, please give us some thoughts it would be greatly appreciated.

I hope that helps. It works pretty well for me. I suspect you will need to do some of you own experimentation to become comfortable with a methodology that makes sense to you.
I am glad to know that I am not the only one to be confused! And, after I read the MiniDSP information about acoustic reference with REW getting me more confused!. I will do test (moving forward or backward one speaker at a time and remeasure) this week-end and report here.

Thank you!
 
#11 · (Edited)
Hello,

So I gave a try with the mic at 1m from the back wall. I think the results are more consistent :smile:.

By the way, it seems that the previous measurements were done with the 2 speakers playing at the same time... :duh:

I'm a bit confuse about the way to set the reference in the IR windows. I read different ways of doing it.
- From a previous post from you: the phase should go smoothly to 0° at 20KHz
- I read also that the step response can be used, the first positive raise indicates the pic where the 0 reference should be
- the step reference should be positive, if it goes to negative then the impulse is inverted.

Should I invert the impulse ?


Thanks again for your support :T
 

Attachments

#12 ·
By the way, it seems that the previous measurements were done with the 2 speakers playing at the same time.
I'm glad you got that sorted out. The IR looks more normal now - well done. Note that, I did warn you in post 2 not to select 'both' as the 'measurement channel'.

I'm a bit confuse about the way to set the reference in the IR windows. I read different ways of doing it.
- From a previous post from you: the phase should go smoothly to 0° at 20KHz.
- I read also that the step response can be used, the first positive raise indicates the pic where the 0 reference should be
- the step reference should be positive, if it goes to negative then the impulse is inverted.

Should I invert the impulse ?
First, I note that acoustic timing was still active for these measurements. In post-7 I mentioned that acoustic timing should be off for rePhase phase correction measurements. This is no problem as in REW we can still adjust the IR position wherever we want it.

Regarding Inversion:
After aligning the 2 IR's on top of each other and placing them near 0 ms we can set the reference time to 0.00 ms and also apply an FDW of 5 cycles as shown below.

Text Green Blue Line Font


From this IR shape (and the step response not shown) it is impossible to tell for sure the correct polarity. There are still several early peaks that confuse the issue on what IR position to choose. It may have been better to measure at 1 m from the baffle as recommended above, but that also may not have helped. Regardless, we can look at the phase and see how that looks.

Text Green Line Blue Plot
---

Note that the phase rises a little before reaching 20 kHz. This suggest we are probably better shifting the IR's so the next earlier peak is at 0 ms to arrive at this:

Text Green Blue Line Font


Text Green Line Blue Plot


This looks as good as any setting to get phase top fall smoothly to 0°. I tried inverting the impulses and tried other IR locations, but that did not help either. Remember, the IR location can be shifted and inverted in rePhase during the analyze so we can deal with the options there where we can see what works best for the phase correction.

Comments:
You can export either of the these channels for the rePhase work as they are very similar. I would not phase correct phase below 100 Hz in this case as there is apparently an abrupt phase shift around 80 Hz. This is possibly due to a SW handoff or possibly a room mode. Either of these will shift with the position of the mic so this is not representative of the situation at the LP. It's safest to leave that low bass unchanged unless you do a lot of work to investigate the situation in detail.

The Adjusted File FYI:
View attachment jaPost11.mdat
 
#14 ·
If by, "more measurements" you mean a measurement at 1 m mic distance then that may be a useful check. No need to measure both channels. The current data shows the 2 speakers are identical. I am not expecting any practical difference. The commnet was primarily just an observation the IR is still a little unusual. There should be no problem using either one the current measurements for the rePhase work.
 
#15 ·
Hello,

Yes, I tried to follow all your recommandations. I also put blankets on the floor to avoid reflections.

I think it looks now better :smile:

But I still have some doubt about setting the 0 reference in the impulse windows. When I'm using the 'estimate' function, REW estimation seems to me a bit weird :dontknow:

Here are my results, hope this time I did it correctly. The mdat file is the original one without any tuning.
 

Attachments

#16 ·
Yes, this looks identical to the last measurements for phase rotation. This is thus the accurate phase behavior of the direct sound. This response (or the last one) can be exported from REW for import into rePhase. I provided an mdat in Post-12 that shifted the IR to a reasonable position to start the work in rePhase. It may well be necessary to shift it again in rephase as it is not really clear where the XO's are actually located and how to best locate the IR. It will take some research on the XO points in the speaker and experimentation in rePhase as to how to best compensate for them. I do not have the time to explain the detail on how to do this. This is just more work than I have available time for.

If you look up and provide the XO points of the speakers. I can do an analysis in rePhase to provide one possible solution for a phase correction filter FYI. It would work just fine in your setup. There will probably be other solutions that will also work fine also. It also can be used as an example that may be helpful in your learning the process.
 
#17 ·
Here are the XO values: 350Hz, 4kHz

Thanks a lot for your support.

if you have some links available, I will be happy to try to understand by myself. If we are shifting the phase, is it to determine the "real" measurement ? Or the phase is relative, and we are trying to find the best position for an easy correction ?
 
#18 ·
The rePhase thread on diyAudio.com contains several links to various guides. It's a big thread though and I didn't review or save any links. I don't even know if any of the them step through creating a simple phase correction filter for a commercial passive XO'ed speaker such as yours. They at least briefly describe how to do it if you can find the right posts.

I will see what I can do with your provided data and XO info. If I have some success I will try to provide a brief summary.

I also need to know:
> What sample rate do you use for your music?
> What music server do you use?
> Is there a limit on the size of the FIR filter that can be applied in that music server? [32k is plenty for this, I would expect, and we can do well with much less if necessary.]
 
#20 ·
I also need to know:
> What sample rate do you use for your music?
> What music server do you use?
> Is there a limit on the size of the FIR filter that can be applied in that music server? [32k is plenty for this, I would expect, and we can do well with much less if necessary.]
> Mainly 44.1 (CD ripped in FLAC)
> Foobar
> Not sure to understand but it's running on a PC, I guess to issue to expect with the CPU load.

Thanks again :T
 
#22 ·
Here is the result of my effort.
I started with the Left channel of the REW measurement from Post-12 and exported it as text with the settings as shown below:
Text Line Plot Slope Font


I imported it into rePhase and using trial and error adjusted the 'filters linearization' 'Crossover' and 'time offset' controls to make the phase a flat as possible. I hoped the XO info you provided might assist the placement and selection of the filters shown below, but it didn't. The rePhase controls are not really designed for use with commercial passive XO speakers in mind. The chosen filters worked pretty well to remove the phase rotation. I did shift the 'time offset' by -0.01 ms as shown to best flatten the resulting phase response. This filter could be used 'as is' in Foobar.
Text Line Font Design Parallel


There is still some minor ripple in the phase response. Some of this is because the actual physical XO filters do not correspond with the choices I needed to make to get a reasonable result. Some of it is due to box diffraction and room reflections. If we assume most of this is related to the direct sound rather than room effect and want to further smooth the response, we can do that. Below the 'parametric phase' filter tab was used in addition to the above settings. We can spend as much or as little effort as we want to smooth the response further using these controls:
Text Line Design Font Parallel


Text Line Font Design Parallel


Below is the resulting filter as shown in Audacity. The second chart zooms in on the amplitude scale to show that the tails of the filter are not truncated using this 2048 tap filter.
Text Blue Line Parallel Screenshot


Text Blue Line Font Plot


Finally, below is a zip of the rePhase and the created FIR phase correction filters:
View attachment FIRs and rePhase.zip
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top