Proper phase measurement

[RobertoG]

I would like to ask you to help me setup REW tool to do phase measurement. Would it be possible? I did it myself but results but are suspected. Topic is not too easy for me, but I would like to do it correctly. The issue concerns two-way speaker and I would like to re-design my crossover to achieve perfect time aligned speakers. I have problem with proper phase measurements. I use IR Winsows tool (left -1ms right - 3ms) to remove any reflection from walls - then phase wrapping dissapear on phase curve but I not sure wheather it is correct. For example low pass filter in wooffer path should give additonal negative phase shift above crossover frequency but it gave positve phase shift? I don't know what may be reason of that, so I stucked and I can not go further. May I asked you to take a look on my measurements and suggest me what was set inproperly.
On important thing - the crossover was done in analog domain - no additional DSP unit will be engaged in that process. Results presented in attachment was obtained in my first version od speaker crossover. Measurements have been done form 1m distance with using UMIK-1 (it was loacated in the middle between tweeter an woofer).
I'm aware that DSP allows me do to much more in filtering and delay signal, but I decided to do in in clasical way. The speakers are: 18W8531G00 and D3004/66200 in sealed box.

 

[John Mulcahy]

Your measurement system has an overall inversion, so you will need to select the option to invert the IR in REW's impulse graph controls.

You are not using a timing reference, so REW places t=0 at the peak of the impulse response. The position of t=0 has a large effect on measured phase, and not having a timing reference makes it difficult to make meaningful comparisons between different measurements. You can use REW's Estimate IR delay feature to get a good guess about where t=0 should be. If you use a timing reference you will need to find the delay offset to apply that corresponds to your mic distance and any system delays and apply it to all measurements. That is easier with the features of V5.20.

When making your measurements use the largest space you have available and place the speaker and mic as far from boundaries as possible - including the floor and ceiling.

 

[RobertoG]

Thank you John. My understanding is that I should us option marked by green arrow, am I right? Should I define Timng offset before measurements or can I do it later on, after measurements in "Impulse" menu as I marked below. Regarding distance to wall and floor I agree, but the most interesing for me i crossover cutoff region, so 1 - 4 kHz range plays the most impontant role. Later on I will try to measure it in big room to achieve good response form 300Hz, becaue I realized that I forgot about baffle step compensation and diffractions and their impact to my frequency response.

 

[jtalden]

RobertoG,
Maybe, but I don't think that is the whole story for setting up for acoustic timing measurements. There are several ways to do this, but here is one way:

• Your Output channel is set to L so the 'Ref Output' needs to set to R in the make a measurement dialog shown above.
• A R speaker with a TW needs to also be connected to the receiver, so if you have the R speaker already in the room and connected, you are good.
• Now measure the L TW driver. [The Ref sound will first chirp from the R TW and then the measurement sweep will then start for the L TW.]
• Using SPL&Phase/controls select 'Estimate IR delay'
• In the popup select 'shift and update timing offset'. [This will properly move the TW impulse to 0 ms and also update the needed value into the 'timing offset' field. That value will allow future measurements be properly timed to the L TW impulse.
• Measure the L woofer.
• Done - but you may want to now remeasure the TW and woofer again to assure the proper repeatability of the impulse locations.

 

[RobertoG]

I did simple experiment with new settings. I know that measurement space is limited, but I can't play in big room because my childrens are sleeping nearby. I will do proper measurments at the weekend, with longer distance and speaker disconnection. In this measurement disatance to microphone was~22cm. During first measurement i measured complete loudspeaker. In next measurement I covered tweeter. Will it be correct to set left and right windows to 0,6ms to gating unwanted reflections and whether phase is ready to further analysis? Or maybe other tools or functions would be better for that? I see that window type changes phase measurement ressult. Would the Tukey 0.1 be fine for that? My exemplary results are presented in the file. Thank you for all comments.

 

[RobertoG]

We were writing in the same time . Exemplary setup is at below picture. I will back to the topic next tomorrow and I hope to be ready do next step by myself. Thank you for your time.

 

[jtalden]

Yes, that setup and those windows settings on the TW+W measurement do show the general phase response above 1.7 kHz pretty clearly.
The chart below shows how the window settings fit compared to the impulse response position. The left window must be located prior to the start of the impulse. It is, and it could even be a little closer, or much further away without causing any issues in this case. The right window is just close enough to filter out the first reflection. The frequencies below 1.7 kHz are outside the window and thus not included. FYI, I moved the 'Window Ref Time' to -0.081 ms as it should properly be placed just before the rise of the first significant impulse peak. This minor adjustment was not large enough to make any noticeable difference in the chart, but the Ref time should not stray too far from that position or the charts can be misleading.
This view also shows how the window setting can be displayed and dragged into position.
The default Tukey 0.25 window works well for this type of work.

 

[RobertoG]

Thank you. I understood your latest post, but for the same settings I achieve Frequency resolution 833,33Hz and I don't know why. Regarding previous post, I suspect that you reffered to acoustic timing reference measurements but I selected loopback timing reference. Am I right? I'm carefully reading your and John posts and if I understand all points, then I will confirm that all is clear. You are right I'm not bright in this topic, but I'm impressive about what REW can do and I do my best to grab new knowlegde.

 

[jtalden]

Good Question. I am confused - again.

In my REW session the frequency resolution is dependent only upon the setting of the right window. When set to 0.6 ms it results in the 1666.67 Hz value (1 s / 0.6 ms = 1666.67 Hz). that value is not impacted by the position of 0 ms, the position of the reference time, or the left window. I am not able to get any other number with that 0.6 ms entered for right window, so I don't know how your dialog shows 833.33 Hz. We are both running REW V5.20 I think. The 833.33 value is for the total window length. I think REW used that total window width in an earlier version.

The correct value in my opinion is the distance of the right window from the point just at the start of the rise of the impulse. It would be up to us to position the impulse and the reference time at t=0 if we want frequency resolution value to be correct for our situation.

 

[John Mulcahy]

There is an Analysis preference (selected by default) to use only the right hand IR window width when deciding the minimum valid frequency in the response.

 

[RobertoG]

Thank you. With new setting I achieved the same resolution frequency. I have one doubt regarding the second and third suggestion:

• A R speaker with a TW needs to also be connected to the receiver, so if you have the R speaker already in the room and connected, you are good.
• Now measure the L TW driver. [The Ref sound will first chirp from the R TW and then the measurement sweep will then start for the L TW.]

What is receiver in this particular case? Should I connect Right OUT to an amplifier to generates an auxilialy sound from another speaker in the new location?

When I selected "use loopback as timing reference" one microphone and one speaker connected to amplifier output was enough to do measurement.

When I select "use acoustic timing reference" should I use two amplifier outputs and two speakers? The first in final enclosure, which will be characterized, and the second is only the auxiliary tweeter? If yes, what distance and location should be chosen? Thank you.

 

[jtalden]

Thank you. With new setting I achieved the same resolution frequency. I have one doubt regarding the second and third suggestion:

• A R speaker with a TW needs to also be connected to the receiver, so if you have the R speaker already in the room and connected, you are good.
• Now measure the L TW driver. [The Ref sound will first chirp from the R TW and then the measurement sweep will then start for the L TW.]

What is receiver in this particular case? Should I connect Right OUT to an amplifier to generates an auxilialy sound from another speaker in the new location?

Yes.
By 'receiver' (that term shows my age), I just meant whatever box or computer controls routing the REW signal to L and R channels.

When I selected "use loopback as timing reference" one microphone and one speaker connected to amplifier output was enough to do measurement.

My understanding is that loopback timing does not work with the UMIK-1 or any other USB connected mic. The computer may have a different delay each time a new measurement is made.

When I select "use acoustic timing reference" should I use two amplifier outputs and two speakers? The first in final enclosure, which will be characterized, and the second is only the auxiliary tweeter? If yes, what distance and location should be chosen? Thank you.

Yes, that is correct. The other speaker is just for the timing reference sound. The sound is >3 kHz I think. Any normal loudspeaker at any location in the room will work fine as the reference so long as the mic hears the sound easily. Even a small desk speaker will probably work fine. Just test the repeatability of the measurements as suggested above to assure it's working properly.

Also be aware that using the UMIK-1 mic we need to use either ASIO4ALL or another ASIO driver instead of using the Java driver. Others members in this forum can point you to setup instructions or possibly a video if you have trouble getting it to work correctly. There is some info in the REW help also. I don't use a USB mic and keep no reference links at hand.