VituixCAD v2

[Reet]

This was discussed in October on diyaudio. Few quotes from there:

Preamp. gain is not included in PIR file so it must be 1. Gain of soundcard input should be included in mic sensitivity in order to deliver correct sensitivity value to PIR file. This happens automatically when sensitivity is measured with microphone calibrator and 'Estimate Mic. Sensitivity' button, and accepted with Accept button.

Main purpose of frequency response measurement is not to show dB/2.83V/1m of the driver or speaker. It should show sound pressure level at the mic - no matter voltage in terminals and measurement distance. And this value should be delivered to frequency response graph or post processing for example with VCAD.

Main problem with ARTA is the dB re 20uPa/2.83V units selection in my opinion. The way I look at it, dB SPL has nothing to do with voltage, it's just sound pressure represented as dB relative to 20uPa. Mic sensitivity is recorded as mV/Pa, so common sense would be to simply multiply the input voltage from the mic by the sensitivity value to get sound pressure in Pa, then convert to dB re 20uPa, in which case, dB re 20uPa/1V is the correct units.

I believe the "problem" is solved in ARTA by simply taking the mic sensitivity value, divide by 2.83, then change units to dB re 20uPa/1V. By doing so, same SPL/V will still be presented without complete recalibration, and VituixCAD SPL will match ARTA without needing a 9dB adjustment (multiply by 2.83).

[Nil L]

Happy New Year everyone!

[kimmosto]

Just a couple notes on REW as it relates to use with VituixCAD, the latest early access 5.20.14ea12 build of REW includes a a few more windowing functions to align with the IR processing features in VCAD: Tukey 0.5, 0.75, and cosine is now available, among other UI improvements and features.

Measurements with REW is now updated (2023-01-01) but it's still quite short and coarse. Probably not good enough for beginners. Most of these instructions are already dated due to changes in VituixCAD so user manual and testing is probably needed to get the latest information about Convert IR to FR, Diffraction, Merger and Options. My energy level seem to be too low for maintaining four different versions and user manual with partly overlapping information. One option is to split chapters related to each measurement program, far field + near field measurements and post processing with VCAD to separate documents to avoid double work and unnecessary differences in the same task.

[Reet]

Happy New Year!

One option to reduce duplicate instructions would be to change REW process to simply measure and export impulse, so workflow is the same as ARTA. Unfortunately exporting impulse from REW with timing in tact requires a couple extra steps beyond just the save button. I have put together these simple steps here:
https://www.htguide.com/forum/articl...e-in-vituixcad

This allows REW process to be same as ARTA instruction from page 7, utilizing the IR - FR tool in VituixCAD. Unlike the measurement guide, this instruction assumes that timing offset was not used when measuring, so t=0 is the loopback reference. I have attempted to get John to implement an impulse export function that captures the timing information without requiring extra windowing steps, so there's a small possibility of seeing that in the near future before final release of 5.20.14

[arivel]

among the various types of cabinet that can be inserted is there also the karlson?

[kimmosto]

Few small fixes and additions done to Measurements with REW.pdf. For example comment about High Pass checkbox in Preferences.
Merge loopback... looks correct option in dual channel mode to compensate measured response with active or passive high pass for protection.

[kimmosto]

2.0.98.3 (2023-01-03)

Convert IR to FR

• Sample rate, data length and start time detected from REW's impulse response export as text. Skip samples set to start time - 2 ms (spare time for Left window). Scaling set to 100 dB.
• Added Cosine 0.50 and 0.75 time window functions.

Impulse response

• Added Cosine 0.50 and 0.75 time window functions.

[kimmosto]

among the various types of cabinet that can be inserted is there also the karlson?

No. Enclosure tool does not include ultra-fidelity enclosures by Karlson. Crossover designing is possible for any cabinet type, but designer must load actual measurements for it.

[kimmosto]

Sample rate, data length and start time detected from REW's impulse response export as text. Skip samples set to start time - 2 ms (spare time for Left window). Scaling set to 100 dB.

This makes reading of REW's impulse response exports easier as text files, but I don't like to recommend using Convert IR to FR with REW which already has very good FFT and FRD export functions. We may hope that JohnPM will change timing concept in the future:

* In dual channel mode with electrical or acoustical timing reference: IR time scale is adjusted automatically so that t=0 when reference input starts to receive feedback. Time scale should have short negative period e.g. 50 ms (or extra samples in the beginning such as in ARTA) to ensure that everything including possible pre-ringing and whole rising of reference signal/impulse is captured. Measurement window does not need and should not have any time scale adjustments in dual channel mode with timing reference.

* In single channel mode without timing reference: IR time scale can be adjusted automatically so that t=0 when impulse starts to rise (for example 10 % from max). t=0 at the peak is not the best option with LF radiators and other low-passes because reference time producing minimum phase response could leave quite/too much on the negative side. Also single channel mode needs some extra samples in the beginning.

* Reference time setting in impulse response graph defines phase response in frequency response graphs and frequency response exports. Not t=0 to safe t=0 fixed at reference time. This is valid method for both single and dual channel mode.

[kimmosto]

Sample rate, data length and start time detected from REW's impulse response export as text. Skip samples set to start time - 2 ms (spare time for Left window). Scaling set to 100 dB.

This makes reading of REW's impulse response exports easier as text files, but I don't like to recommend using Convert IR to FR with REW which already has very good FFT and FRD export functions. We may hope that JohnPM will change timing concept in the future:

* In dual channel mode with electrical or acoustical timing reference: IR time scale is adjusted automatically so that t=0 when reference input starts to receive feedback. Time scale should have short negative period e.g. 50 ms (or extra samples in the beginning such as in ARTA) to ensure that everything including possible pre-ringing and whole rising of reference signal/impulse is captured. Measurement window does not need and should not have any time scale adjustments in dual channel mode with timing reference.

* In single channel mode without timing reference: IR time scale can be adjusted automatically so that t=0 when impulse starts to rise (for example 10 % from max). t=0 at the peak is not the best option with LF radiators and other low-passes because reference time producing minimum phase response could leave quite/too much on the negative side. Also single channel mode needs some extra samples in the beginning.

* Reference time setting in impulse response graph defines phase response in frequency response graphs and frequency response exports. Not t=0 to safe t=0 fixed at reference time. This is valid method for both single and dual channel mode.

[Reet]

2.0.98.3 (2023-01-03)

Convert IR to FR

• Sample rate, data length and start time detected from REW's impulse response export as text. Skip samples set to start time - 2 ms (spare time for Left window). Scaling set to 100 dB.

Excellent! Though, I would prefer that skip samples is simply set to start time without 2ms offset to place t=0 at same t=0 represented in REW. One could remove the timing offset when measuring to include full time of flight, and exported IR will be represented in the same manner that skipping first 300 samples with ARTA PIR files does.

In any case, I'll update my how to for REW impulse response export to apply to only VituixCAD versions prior to 2.0.98.3.

[Reet]

Excellent! Though, I would prefer that skip samples is simply set to start time without 2ms offset to place t=0 at same t=0 represented in REW. One could remove the timing offset when measuring to include full time of flight, and exported IR will be represented in the same manner that skipping first 300 samples with ARTA PIR files does.

In any case, I'll update my how to for REW impulse response export to apply to only VituixCAD versions prior to 2.0.98.3.

On a related note, the latest release of REW 5.20.14 early access includes some additional changes to impulse response export, that may be a bit redundant given the recent update to VituixCAD, but I'll list the highlights here for good measure. In any case, it's now possible to bring in impulse response data from REW using either text or wav export with preserved timing.
Original post with full details:
https://www.avnirvana.com/threads/v5...154/post-87167

• Added: Export impulse as wav has an option to place t=0 at a specified sample index, default is 256
• Changed: When exporting an impulse response as WAV without windowing or setting t=0 at a specified sample index the peak is no longer placed at 1 second but left wherever it appears in the raw data

[kimmosto]

I would prefer that skip samples is simply set to start time without 2ms offset to place t=0 at same t=0 represented in REW

Problem or challenge is that impulse response text reader should support also REW files measured with no timing reference option. Impulse starts before t=0 in those files almost without exceptions so everything before Start time can't be automatically skipped. Also "cal and timing" has tiny (~1 sample) difference to timing ref only. 2 ms time offset supports all three time reference modes without comment text detection.

The latest revision 2.0.98.4 has comment text detection for keywords "loopback", "timing reference" and "calibration". All three options set different initial value to Skip samples. Impulse peak is located to ca. 2.907 ms ~ 1000 mm with "no timing reference" option. "Loopback as timing ref" skips -startTime/sampleInterval - 1 samples. "Loopback as cal and timing ref" skips -startTime/sampleInterval samples. This solution is very sensitive to possible format changes in REW's text exports. Another approach would be reading directly REW's mdat files, but I have not found or asked spec for it.

Test measurements with loopback cable in both channels (not acoustical measurement):

[kimmosto]

2.0.98.4 (2023-01-04)

Convert IR to FR

• "loopback", "timing reference" and "calibration" comment texts detected from REW's impulse response export as text. Skip samples set to:
  startTime-2.907 ms with 'no timing reference',
  startTime/sampleInterval-1 with 'loopback as timing reference' and
  startTime/sampleInterval with 'loopback as cal and timing reference'.

[Reet]

Problem or challenge is that impulse response text reader should support also REW files measured with no timing reference option. Impulse starts before t=0 in those files almost without exceptions so everything before Start time can't be automatically skipped. Also "cal and timing" has tiny (~1 sample) difference to timing ref only. 2 ms time offset supports all three time reference modes without comment text detection.

Ok, I get that. I should test for myself, but I would expect the same problem to exist for ARTA with single channel measurements as well. Maybe ARTA is clever and bumps the impulse a few samples ahead of 300 for single channel measurements.

[kimmosto]

I would expect the same problem to exist for ARTA with single channel measurements as well. Maybe ARTA is clever and bumps the impulse a few samples ahead of 300 for single channel measurements.

ARTA aligns single channel measurement (as pir) by moving IR so that impulse starts at sample 300. Peak is further, but I don't remember how smart that logic is with low passed signal where impulse could rise very slowly.


Get Reference time button in VCAD sets cursor to sample where level is 20% of impulse maximum. That usually gives quite close to minimum phase, but not exactly. Minimum phase is too subjective to be goal in this game.

[kimmosto]

Forgot to say that skipping fixed 300 samples with ARTA pir is based on assumption that no one measures in single channel mode, though it is accepted for near field.

This also reminds me that REW could require dual channel and time reference also for near field measurements because impulse maximum could be at significantly different location for cone and port/passive. This has not been a big problem with ARTA in single channel mode due to IR movement logic, but copy-pasting instructions from ARTA to REW version and my lazy testing could cause small error or at least risk.

[Reet]

This also reminds me that REW could require dual channel and time reference also for near field measurements because impulse maximum could be at significantly different location for cone and port/passive. This has not been a big problem with ARTA in single channel mode due to IR movement logic, but copy-pasting instructions from ARTA to REW version and my lazy testing could cause small error or at least risk.

I should mention that REW includes option to set t=0 at impulse start as well, which alleviates this issue as it operates similar to ARTA, and for nearfield and port it does capture the impulse without any data required prior to t=0. Below is a quick test of various single channel methods of determining t=0 in REW, green traces show both nearfield and port output with t=0 at IR start.



Here's the same in ARTA, single channel mode with cursor at sample 300:

[Reet]

One thing I noticed with REW, is that when saving multiple measurements under same conditions, start time in the exported impulse is recorded differently. When loading into VituixCAD IR - FR tool, it seems to only look at start time for the first file, and "skip samples" is set the same for all files, which isn't right. Start time and skip samples will need to be set for each file individually to maintain correct timing information.

[kimmosto]

I should mention that REW includes option to set t=0 at impulse start as well, which alleviates this issue as it operates similar to ARTA

It's IR manipulation in Control window, and should be done separately for every NF measurement. Already 3rd place where we have to make settings to get this simple task done. Moving IR in time scale is the most confusing and unnecessary especially when time reference is activated. Maybe I should remove Measurements with REW from my web site until whole timing concept is simplified and working...

[kimmosto]

One thing I noticed with REW, is that when saving multiple measurements under same conditions, start time in the exported impulse is recorded differently. When loading into VituixCAD IR - FR tool, it seems to only look at start time for the first file, and "skip samples" is set the same for all files, which isn't right. Start time and skip samples will need to be set for each file individually to maintain correct timing information.

All measurements loaded to the file list must be measured and exported with the same time scale settings, and timing must be locked to reference channel with dual channel and IR start or delay estimation with single channel which become t=0. All text exports should have the same Start time e.g. -1.0 s. If not, export or measurement does not work properly. All responses will be exported with the same settings to the same directory so original data cannot have any deviations in settings, and Skip samples can/should be constant; typical for measurement program such as 300 for ARTA. Not for individual IR file.

[fluid]

Is it possible to create subtractive delay derived crossovers directly in Vituix? I have tried to input something like in the attached document but I can't get a derived highpass to work.

Matched Delay 4th Order Crossover 12252022.pdf

[Nil L]

REW measures multiple drivers normally while maintaining their timing. Exporting such measurements to VituixCad does not have any annoyances or irregularities. Everything works as it should. After the preliminary measurement, 'estimate IR delay' is applied, then all drivers are measured. They have the correct timing.

[Nil L]

Below is a quick test of various single channel methods of determining t=0 in REW, green traces show both nearfield and port output with t=0 at IR start.

In your screenshots from REW and from ARTA, there is a difference in the response behavior of the port. Between 3.5ms and 4.5ms. In the screenshot from REW, the port response starts when the driver response has reached its maximum. This is right. In the screenshot from ARTA, the port response starts at the same time as the driver response. Which cannot be in reality.

[Reet]

All measurements loaded to the file list must be measured and exported with the same time scale settings, and timing must be locked to reference channel with dual channel and IR start or delay estimation with single channel which become t=0. All text exports should have the same Start time e.g. -1.0 s. If not, export or measurement does not work properly. All responses will be exported with the same settings to the same directory so original data cannot have any deviations in settings, and Skip samples can/should be constant; typical for measurement program such as 300 for ARTA. Not for individual IR file.

I agree with that sentiment 100%, however its not the reality of the simple testing I've done with near field driver measurement and port output in REW, whether 2ch loopback signal is used or not. I tested again this morning to make sure I didn't do something silly. All measurements are under same conditions, all I've done between nearfield and port measurement is spin the speaker around and relocate the mic at the port output.

For an example using single channel measurement, t=0 at IR start.
nearfield driver start time: -0.99978125 // Start time (seconds)
port output start time: -0.9986875 // Start time (seconds)

When loading into VituixCAD, nearfield measurement is top of the list, and port output is shifted back in time:



Expected result, from REW overlay:


For 2channel measurement with loopback, the behaviour is the same, port output starts before 0ms point in VituixCAD which is worst result possible. I'm not sure if there is some obvious setting in REW I am missing, but using windowing for IR export to force t=0 location is the only way I am able to get reliable results with IR export for this task.

Testing completed with REW 5.20.14ea18, and VituixCAD 2.0.98.4.

[Reet]

REW measures multiple drivers normally while maintaining their timing. Exporting such measurements to VituixCad does not have any annoyances or irregularities. Everything works as it should. After the preliminary measurement, 'estimate IR delay' is applied, then all drivers are measured. They have the correct timing.

I'm not exporting frequency response, I'm exporting impulse response.

[Reet]

In your screenshots from REW and from ARTA, there is a difference in the response behavior of the port. Between 3.5ms and 4.5ms. In the screenshot from REW, the port response starts when the driver response has reached its maximum. This is right. In the screenshot from ARTA, the port response starts at the same time as the driver response. Which cannot be in reality.

The single channel measurements are all "wrong" as far as perfect timing goes, however using t=0 at IR start provides close enough result to not have any significant error <200Hz, similar behaviour of "time align" button in all SPL. 2ch measurement with mic located well provides best result, below is what I would consider to be the correct impulse behaviour and phase alignment is correct, 100-200Hz region overlap above tuning is what I would expect:

[Nil L]

Is the measurement with an acoustic timing reference equal to a two-channel measurement?
You have moved the port response 0.5 ms late. It's 22 cm. How do you physically implement it?

[Reet]

No acoustic reference, loopback reference. I really don't have any interest in acoustic reference for loudspeaker design, it's just making more work for yourself, but if you really want I can create that comparison, I don't expect it to be any different. \

I didn't "move" any response, just measured 2mm from driver dust cap, and 2mm from port exit. Phase response overlap in 100-200Hz region is what I would call "correct", perhaps is incorrect thinking but merged results match the enclosure model with great accuracy.

[Nil L]

I didn't "move" any response

In the screenshot (from REW) of post 700, the port response start time is about 0.25 ms. In the screenshot of post 702, the port response start time is about 0.75 ms. So I figured that you shifted the port response to merge the phases.

[Reet]

In the screenshot (from REW) of post 700, the port response start time is about 0.25 ms. In the screenshot of post 702, the port response start time is about 0.75 ms. So I figured that you shifted the port response to merge the phases.

Post 700 is single channel measurement with "set t=0 at IR start". Post 702 is 2ch measurement with loopback. Nothing is moved, only post 702 includes correct timing as mentioned.

[Reet]

No acoustic reference, loopback reference. I really don't have any interest in acoustic reference for loudspeaker design, it's just making more work for yourself,

Main problem with acoustic reference is that mic needs to move to measure port output, so keeping acoustic reference constant distance from mic becomes a challenge. I would avoid this measurement method if at all possible for measuring loudspeakers. Good to use for HT room setup only.

[Reet]

...
For an example using single channel measurement, t=0 at IR start.
nearfield driver start time: -0.99978125 // Start time (seconds)
port output start time: -0.9986875 // Start time (seconds)
...

Testing completed with REW 5.20.14ea18, and VituixCAD 2.0.98.4.

FWIW, exporting impulse response from REW as WAV file does maintain correct timing. I've questioned the "start time" in the text file on the REW forum, since it does not appear to actually represent t=0.

[Reet]

FWIW, exporting impulse response from REW as WAV file does maintain correct timing. I've questioned the "start time" in the text file on the REW forum, since it does not appear to actually represent t=0.

Sorry for bombardment of posts today.

Response from John Mulcahy: "Start time is the time at which the entire data record starts, generally 1 second before the IR peak. The time that corresponds to depends on where the peak is relative to t=0."

With this in mind, for use with VituixCAD, I think either export from latest REW version as WAV, with checkbox for "place t=0 at sample index: 256" which does work well for keeping timing in tact, or export as text file with IR window to force start time of impulse response export.

[kimmosto]

Sorry for bombardment of posts today.

Response from John Mulcahy: "Start time is the time at which the entire data record starts, generally 1 second before the IR peak. The time that corresponds to depends on where the peak is relative to t=0."

With this in mind, for use with VituixCAD, I think either export from latest REW version as WAV, with checkbox for "place t=0 at sample index: 256" which does work well for keeping timing in tact, or export as text file with IR window to force start time of impulse response export.

I might understand wrong, but IR text export looks total mess where start time or time with estimated delay are not known at least in single channel mode. Just time from the beginning of sample buffer to IR peak is known and commented with "//Start time". Then also name is misleading because that is peak time from the beginning of sample buffer. We can just hope that t=0 equals to constant 1s with timing reference, but that could be variable too because user can move IR in time scale with Control window.
I still don't know what to do with this import. Probably the best is to set initial Skip samples to 1s/sampleInterval - 1 and that's it. And fix the problem with generic time series informed by FrancoB on diyaudio.

[Reet]

I might understand wrong, but IR text export looks total mess where start time or time with estimated delay are not known at least in single channel mode. Just time from the beginning of sample buffer to IR peak is known and commented with "//Start time". Then also name is misleading because that is peak time from the beginning of sample buffer. We can just hope that t=0 equals to constant 1s with timing reference, but that could be variable too because user can move IR in time scale with Control window.

Yes, this is what I've found, impulse response export from REW as text file is frustrating, but somehow REW will re-import these files with timing in tact.

I still don't know what to do with this import. Probably the best is to set initial Skip samples to 1s/sampleInterval - 1 and that's it. And fix the problem with generic time series informed by FrancoB on diyaudio.

For REW I have 2 solutions that work well. First is to use IR window for IR export, other is to export impulse response as WAV file with "place t=0 at sample index" option selected. Both of these options work well, exporting text file as raw data without windowing is frustrating me, so I think I'll just avoid it for the time being.

[kimmosto]

2.0.98.5 (2023-01-08)

Convert IR to FR

• Import of REW's impulse response as text modified: Skip samples set to: 0.997093s/sampleInterval with 'no timing reference', 1s/sampleInterval-1 with 'loopback as timing reference' and 1s/sampleInterval with 'loopback as cal and timing reference'.
• Enabled import of generic impulse response text file without any header or with illegal MLSSA header. Was disabled for few revisions.

Lets try with assumption that t=0 equals to 1.000000000000 s.
Also generic import is now fixed back how it was 2-3 revisions ago, before this REW lottery.

[Nil L]

Kimmo, when working with the Impulse Response Viewer, the graphs are very different in amplitude, so viewing a new graph requires a lot of level adjustment. Wouldn't it be useful to create a normalize button that, when used, will change the amplitudes to a form that does not require increasing or decreasing the amplitude?

[kimmosto]

Kimmo, when working with the Impulse Response Viewer, the graphs are very different in amplitude, so viewing a new graph requires a lot of level adjustment. Wouldn't it be useful to create a normalize button that, when used, will change the amplitudes to a form that does not require increasing or decreasing the amplitude?

I will not add any normalization because Impulse response window already has automatic scaling of Y axis. Program scans maximum from both Impulse and Step (before 50 ms). Usually Step hits higher than Impulse, especially with low-passed signal.

Is that the problem i.e. Step is hidden and you want auto scaling by Impulse response only?

[Nil L]

Yes. I was talking about displaying the impulse graph. Autoscaling is always performed as if impulse and step were always on. Even if the step is disabled, scaling occurs as if it were enabled. Because of this, the impulse may not be visible at all.

[kimmosto]

Yes. I was talking about displaying the impulse graph. Autoscaling is always performed as if impulse and step were always on. Even if the step is disabled, scaling occurs as if it were enabled. Because of this, the impulse may not be visible at all.

Ok. Download and install the latest build 2023-01-09 where Y is scaled by IR only if Step is not checked.

[Nil L]

It worked. The peak of each graph is clearly visible. If you don't save overlay. If you follow the ratio of the graphs using overlay, it's also good, until the tweeter graph appears. As soon as it appears, all others become almost unreadable. Peaks are difficult to determine. I have no idea how to solve both. Therefore, if only this improvement remains, which is now, it will be good. Thank you.

[kimmosto]

I don't usually "work with" or look impulse response only so automatic scaling was not separated for all three options (IR, Step or both). Program has always calculated individual maximum of both IR and Step so this was one of the easiest mods.

[Nil L]

It is interesting to know your opinion. What is more important to coordinate between drivers - impulses or step?

[kimmosto]

It is interesting to know your opinion. What is more important to coordinate between drivers - impulses or step?

Timing is designed with main window by looking phase, normal and excess GD, frequency responses and directivity. IR window is for exporting impulse responses for FIR XO and analyzing timing of total SPL as ETC and Step. Impulse responses are quite invaluable for analyzing.