Home Theater Forum and Systems banner

21 - 40 of 45 Posts

·
Premium Member
Joined
·
1,562 Posts
This info helps me a lot. Thanks.
> The phase difference we see over 10 kHz is due to the Right Ear (red trace) being wrapped at ±180 and the Left Ear (green trace) being unwrapped. They both measure the same. Below, in the second chart, I show the impact of unwrapping both the traces using a cursor position at 200 Hz. Any cursor position below the ~13 kHz phase divergence will achieve the same result.
> This is a case where the SPL response rolls off steeply at ~13 kHz and thus the HF Phase response rolls up steeply as a result. In this case you have placed the 2 IR's properly for export into rePhase. You will just not try to correct the phase above 13 kHz.
> This is a single driver with no XO or room impact so the measured phase response is the expected minimum phase response.

It all makes sense to me now.

Phase1.png

Phase2.png
 

·
Registered
Joined
·
43 Posts
Discussion Starter #22
Thank you for your analysis.
I attached one chart which original data is the same one as I sent. The display on the curves are SPL, Phase(IR Delayed and minimum phase).
The difference between [IR Delay] and [mini.phase] curves is caused by your previous explanation, right?
Then, I can export the TF(.txt) file after [IR Delay] process to rePhase, and I can just ignore over 13KHz.
I got you. Thanks a lot.

Yuichi
 

Attachments

·
Premium Member
Joined
·
1,562 Posts
The SPL/phase measurement (Blue traces) are the correct traces to export. The IR is located well for export to rePhase. The phase falls to the 0° axis, i, e., to where the SPL rolls of at ~13 kHz. So yes this is a good measurement to export, and yes, it is correct to avoid any SPL/phase correction in rePhase above 13 kHz. Just now I am looking more closely at the SPL trace, the rapid SPL falloff is actually nearer 11 kHz so that would be a little better upper limit for any SPL/phase correction.

The minimum phase trace is of no practical interest for this purpose.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #25
Hi
I have one other question regarding IR and phase display.

Attached is a 2 way speaker measurement. The tweeter location is miss aligned.
In this case, the phase curve at 2.74KHz and 3.17KHz slipped down sharply. I guess on these frequencies, SPL is very small, so it is difficult to compare the phase. If so, why so large phase difference exists before and after the 2.74 and 3.17KHz??
Minimum phase display shows reasonable curve.

In case of original SPL curve is reasonably flat (no sharp drop:the tweeter is reasonably located), IR phase curve looks reasonable and close to the minimum phase curve.

Yuichi Arai
 

Attachments

·
Premium Member
Joined
·
1,562 Posts
Assuming this is a typical speaker, this is normal for LP measurements in a room.
> The phase is 'correct', but includes the impact of the strong reflections in the room.
> We normally have little use for the plot in this format. We normally want to see the phase of the direct (first arriving) sound.

I loaded your file in REW 3 times and then changed the settings on the second 2 copies to illustrate the changes needed to put this measurement in a more usable form.

1) Below, we show what happens with the phase unwrapped (red) Vs what would be displayed with the phase wrapped (blue). REW unwrapped the phase based on your current cursor position, so the cursor must have been in the range of the green arrow around 15Hz. We know that because the 2 traces are aligned in that range. It is standard practice to set the highest frequencies measured near 0°. REW normally places the IR location near 0 ms so that will happen. Had your cursor been maybe 10 kHz when it was unwrapped there would be the expected change in the relative offset. This is just a side comment to explain why the 2 traces are located where they are.

Unwrapped Vs Wrapped.png

2) Below, we show how to eliminate the impact of the room so we can better see the phase rotation of the direct sound. We can use the REW Frequency Dependent Window (FDW) feature to mitigate the room impact. This feature allows us to look at the phase of the first few cycles at each frequency. In this case I used a 5 cycle setting as shown below. Larger settings still had at least one of the shifts. I often find that 5 cycles is a good setting for this purpose as it often removes the room influences and still does not distort the phase at the extreme frequencies. Some shift of the phase will occur at the extreme high an low frequencies with lower FDW settings. I zoomed in on the wrapped phase trace (blue) and made the needed setting changes on the FDW (green) trace. If we look closely we see that at about 2700, and 3100 Hz strong room reflections resulted in 360° sudden phase changes in the trace. These sudden 360° shifts are more easily seen the first chart where the unwrapped (red) trace drops 360° suddenly at these 3 frequencies.

No-FDW Vs FDW.PNG

3) Below is just to show what the FDW trace looks like when it is properly unwrapped.

FWD Unrapped.PNG

Now for the big disclaimer. This looks like more than one driver was active in this measurement. If there is an XO near 3 kHz and these 2 driver are not on the same baffle then the large phase shift near 3k may be due to distance offsets of the drives rather than room influences. Also, If timing delays are being applied to align the drivers phase then the delays may be set wrong. It is impossible to be sure of the situation by just looking at a measurement. There is not enough information about the details of the setup.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #29
Hi

Is there any way to calcurate two measurement data such as "add" or "subtract" or "Invert".... I like to use this function to subtract the free space data from the headphone data.

Yuichi Arai
 

·
Premium Member
Joined
·
1,562 Posts
Yes, REW refers to these IR functions as 'Trace Arithmetic'. They are located in 'Graph Controls' of the 'All SPL' graph panel. REW 'Help' has more info and screen shots of that panel if needed.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #32
I like to ask about the impulse response waveform.
Attached is the IR data taken by REW. DUT is a 12cm full range speaker mounted on a enclosure.
The purpose is to adjust accurate dual speakers positioning. So, as the first step, I try to capture accurate impulse waveform.
Attached F120AbyREW is taken by REW. I wonder why I have the first negative spike in front of main positive wave on time zero. Is this negative spike true?
As a reference, attached F120AbyDirac taken by Dirac Live measurement facility. This wave is before EQ. The DUT is identical including miniDSP UMIK-1 mic and other signal chain. for both measurements. This has no negative spike in this case.
Any one knows, why? or any specific settings on REW for this purpose?

Yuichi Arai
 

Attachments

·
Premium Member
Joined
·
1,562 Posts
any specific settings on REW for this purpose?
Yes, see REW 'preference/soundcard'. Under 'input option' there is an 'invert' check box. This box in normally 'clear'. If it 'checked' the IR of all measurements will be inverted. This is there to correct for some low cost soundcards that invert the signal as a result of its design.

Also REW provides the option to invert individual measurements (handy for some types of analysis). This control is only seen when the 'Impulse' graph panel is selected. It is found when 'show graph controls' is open. The box is labeled 'Invert Impulse'. The box inverts the IR of the active measurement if it is 'checked'.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #34
Thank you for your comments.

I did not set the 'Invert Impluse' check both in the 'Sound Card' tab and the 'Analysis tab' in the Preferences settings. Those are both clear.
Is there any effects generated by digital calculation? Or speaker diaphragm behavior?
It looks to me that the large positive pulse looks the first arrival sound like the Dirac one?
How do you think?

Yuichi
 

·
Registered
Joined
·
43 Posts
Discussion Starter #35
On top of Impulse waveform question, I have following one.

First, settings are as pic 'Analysis', SoundCard and Measurement for all through following measurements.
I have some times the 'NotGoodSPL' curves. In this case, impulse is like 'NotGood Impulse'.
When it works fine, the curves are like 'GoodSPL' and 'GoodImpulse'
The provability is ..let's say 50% to 50%.
It looks like that the Sweep does not reach to the high or the sweep does not synchronize??
If you have any suggestion, please let me have it.

Thank you in advance. Yuichi
 

Attachments

·
Premium Member
Joined
·
1,562 Posts
The negative going initial IR peak is just due to the extra phase rotation when the sweep extends well past the point where the SPL starts to roll off significantly. This is not surprising. We normally set the sweep so we are testing within the bandpass range of the speaker. There is no problem sweeping beyond that range, but then we may see the initial IR peak be negative as you found above. Again - no problem. If the purpose is to confirm the polarity using the IR graph we should chose a sweep that stops within the bandpass range of the speaker. The IR chart has some uses, but the other charts show polarity and almost everything else much more clearly. The charts just above show the phase to be at -720° from 300 Hz to 15 kHz. That is the same as 0° so the speaker is definitely wired with positive polarity. Ignore the IR chart in this case. The likely reason that the Dirac Live IR did not show the same initial peak is that its sweep did not extend past the bandpass range of the speaker.

Regarding the high frequency truncation of some of the sweeps:
The only idea I have is to try reducing the buffers from 32k to maybe 16k. It's possible that there is too much delay and the high frequency is not being captured. This is just a guess based on some recent comments in other threads. I think it was only happening on some Mac OS X computers. This may not help. If reducing the buffer sizes doesn't work, you could post an mdat containing the 2 sweeps and we can take a closer look. You can also look at the scope chart to see if the current measurement looks to be properly captured. Only the last sweep is available to view in the scope chart and it is not saved with the file.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #37
Thank you for your quick reply.

I use Mac OS X platform in this case. So, reduction of the buffer size to 16K gave me a good result. I will try it on Windows platform. Also will try the negative spike question, too.

Many thanks,

Yuichi
 

·
Registered
Joined
·
43 Posts
Discussion Starter #38
On Windows platform I obtained desired result. Attached are what I want. Also, I have not experienced the sweep problem on this platform.

Thanks for your great support. Yuichi
 

Attachments

·
Registered
Joined
·
43 Posts
Discussion Starter #39
Subj:Travel time measurement for miniDSP delay setting

I am using REW V5.19-Beta4.
I have tried to measure the sound fly time from the diaphragm to the mic. Purpose is to find out the acoustical difference between two diaphragms, so the accuracy of absolute value is not needed.

Impulse response Calculation setting...I did was..
(1): [No timing reference] and [no check] on [Set t=0 at IR peak] -> Always Peak is set to 0.
(2): [Use loopback] the value does not depend on the distance
(3): [Use acoustic timing...] ->Measurement value is not reliable such as 1 meter shows 0.33 msec and 10 cm shows 0.31 msec something like these.

In the case(3), I measured 1 meter and approx. 10cm distance, the following numbers are shown.
1m :number on the measurement Panel is 0.33msec delay estimation on control is 0.361 msec
10cm : 0.31msec and 0.350 msec respectively.

Please advice correct settings for this purpose.

Yuichi Arai
 

·
Premium Member
Joined
·
1,562 Posts
You want to measuring the difference in delay between 2 channels? Why would it be different unless the mic is slightly off center?

I am not understanding your issue with #3. It is the correct method. I do not look at the REW reported delay number so I don't know how accurate and repeatable it is. You can evaluate that yourself.

The basic method I suggest is:
> Set your music player/AVR/DAC to stereo mode.
> confirm stereo connections from the PC soundcard through your system to the L & R speakers
> Set REW acoustic timing on.
> Set Left channel as the REW reference channel.
> Set Left channel as the REW measurement channel.
> Measure [a chirp will occur in the left channel before the full sweep occurs in the left channel. The right channel will be silent.]
> Set Right channel as the REW measurement channel.
> Measure [a chirp will in occur the left channel before the full sweep occurs in the right channel.]
> Open the REW overlay impulse chart and find the 2 impulses. [You can measure any offset accurately by zooming in then using 'ctrl-right mouse button' top drag between 2 identical locations on the 2 IRs. The distance between these points should agree with REW reported difference in the impulse locations.]

Note:
Take a few repeated measurements with this process to assure good repeatability. If the result changes significantly there is a problem that needs to be corrected.
 
21 - 40 of 45 Posts
Top