How to take reliable measurements from 100Hz to 500Hz to cross over a 3 way

[tktran]

I'm working on a 3 way WWMT, (2x10", 1x7", 1.25" sealed monopole) but encountering some challenges with the W-M crossover.
This is a dual woofer bin at the bottom of a sealed 2 way (think Wilson Watt-Puppy type arrangement)

Can anyone give me some guidance or refer me to some guidelines about to crossover in that critical 100-400Hz region properly?


So far, what i've been done is measure the woofers nearfield (mic distance- 1/2" from centre of cone). With the woofers in parallel I will assume +6dB gain in sensitivity. This give me reliable measurements up to approximately 500Hz. Then I measure the midrange/tweeter farfield, at approximately 2', which incorporates the effects of the baffle. In my room, the gate is long enough to allow measurements to be usable down to about 300Hz. What I'm stuck with now is how to use the nearfield woofer measurements with the farfield measurement of the midrange.

Am I going about this the right way, or have I gone off on a tangent?

[tcpip]

This is a challenge I too am facing. I too don't know what's the right way to do this. Here are some thoughts -- I'm hoping the experts will correct me.

The nice thing about these frequencies is that accurate phase coherence between drivers is less relevant here than between tweeter and midrange. So, splicing two nearfield measurements works here, even though this means you are not going to get the two graphs to be phase coherent with each other.

We can measure the drivers (midrange and woofer) nearfield, taking care to feed both with exactly the same amplitude of signal from our power amp, and keeping the mic the same distance from the cone in each case, say, 5mm. This will give us the nearfield SPL curves with relative amplitude. Then we can scale one of these curves as per the ratio of the surface areas of the two drivers. For instance, in your case we find that your two woofers have a combined surface area proportional to 50 sq-in, your mid has a surface area proportional to 12.25 sq-in, so your combined woofers have a surface area 4 times that of your midrange. In that case we realise that to get their farfield outputs to match, we must ensure that the nearfield woofer SPL must be 6dB lower than the nearfield midrange SPL for their farfield SPLs to match. We make a mental note to build this into the crossover.

Then we design the crossovers in the usual way, using the two nearfield SPL curves. We don't try to do any reverse-null checks to see if we get a deep notch, because we know that the two graphs are not phase coherent. And we ensure that the woofer output is 6dB lower than the midrange output when we model the combined crossover.

I don't think you should try to use the farfield measurement of the midrange.

Also, I think you should forget about factoring in the gain due to your paralleling the two woofers. Just feed in the same amplitude to both the midrange and the (woofers in parallel) and take nearfield measurements from both. Let the amp "figure out" how much more power it is feeding your woofers because of the lower impedance.

These are my current thoughts.

[Face]

I've had good luck with measuring at 1M and with a large bass trap(4"+) on the floor between the speaker and mic. That said, a lot still depends on the rest of the room.

Sent from my SM-G935V using Tapatalk

[TMM]

It seems a bit counter productive to try to eliminate baffle diffraction and floor bounce effects from your measurements unless you plan to listen to the finished speakers in the exact same way (nearfield or with a huge bass trap on the floor). I measure outside in an open space, with the mic at a fixed height and distance (approx 2.5m). Any edge diffraction or floor bounce effects become inherently designed into the crossover and therefore the optimal listening position from the finished speakers will probably be that of the measurement height/distance.

If you can't measure in a space that has minimal room effects (i.e. outside) then use the near-field measurement of the woofer and an edge diffraction sim to estimate how much baffle step compensation you'll need to build in to the design. You could also do a floor-bounce sim if you want.
Use only far field measurements at a fixed measurement position to design the crossover (W to M and M to T) that way you don't screw up your phase data. It doesn't matter if the low frequency looks messy on your far field data because you should have a fair idea of what needs to be done from your near field and sim data.

[kimmosto]

High reliability is difficult to reach at home. Short time window causes extra smoothing to mid-range which could hide possible acoustical/mechanical resonances. In addition, baffle diffraction simulation is not very reliable. Nevertheless, I'm using method described in VituixCAD Measurement Preparations.pdf document.
Exception: If short time window causes too high directivity index below 150 Hz, I will generate directivity (off-axis responses) with Diffraction tool and merge produced response set to measured directivity at about 150 Hz with blending of 3 octaves. The reason is that knowing of power response level at low frequencies helps adjusting frequency response (level & shape) below 200 Hz. This is not piece of cake operation but makes designing of conventional speakers more certain.

Few commercial speakers designed with this method: Taipuu Speakers

[Dave Bullet]

Speaker workshop is pretty handy for scaling nearfield / farfield. I presume you're familiar with Bagby and Laub's FRD blender tool?:
FRD Blender and Minimum Phase Extractor

My process - usually 1m distance is enough. I do outside measurements with tweeter at least 2.4m in the air to get 1,024 sample points with a >11msec gate at 96KHz. I use speaker workshop for measurements, but FRD tools to extract minimum phase since any "munging" of curves means recorded phase in SW is meaningless. Driver offsets must be factored then in the xover sim as Erik pointed out.

1. Gated tweeter (farfield)
2. Gated mid (farfield)
3. Gated woofer (farfield)
4. Nearfield mid
5. Nearfield mid port (if one)
6. Merge mid nearfield and port together (in SW)
7. Nearfield woofer
8. Nearfield woofer port (if one)
9. Merge woofer nearfield and port together (in SW)
10. Scale mid merged curve to mid farfield response
11. Scale woofer merged curve to woofer farfield response
12. Run FRD blender to splice merged nearfield into farfield for mid - check levels. Extract min phase
13. Run FRD blender to splice merged nearfield into farfield for woofer - check levels. Extract min phase
14. Extract min phase for tweeter
15. Re-import all into SW
16. have some xover fun.

[augerpro]

Don't forget to add the baffle diffraction signature to the near field responses before you merge them with the far field responses.

[JonMarsh]

well, this may be a topic that's a lot like politics- everyone has an opinion, and they seem to all be a bit different.

I'm with Kimmosto if you've got the time and wherewithal, but you can get pretty good results with a liberal application of the KISS principle, too, and a bit of flexibility in how you setup.


This thread briefly summarizes measurements and crossover behavior for the Isiris.

https://www.htguide.com/forum/showthread.php?43165-Isiris-evolution-to-Osiris-Jr-Summary-of-Isiris-build-and-new-update

Scroll down past the cabinet pictures. I'll supply some more comments later this weekend when I have time, but I do warn that I usually do measurement well out into the room, and initially positioning the LF cabinet so that there is reasonably minimal LF interaction with the room even with a longish measurement gate. No, this may not be where your wife/GF wants the speakers, but it might just be where they OUGHT to be for optimum listening results.

I'll usually use between 1M and 2M measurement distance, (depending on the vertical size of the system- farther for larger) and as Mike and I have explained for large systems like the Isiris/Isis/Osiris, measure at the listening height and a range of plausible distances. I also liberally use heavy synthetic comforters to quieten things in the midrange and highs. How you setup your actual listening setup is then up to you, of course, but trust me, as a few members fully understand, room treatment and speaker positioning is just about as important as the initial speaker design.

This is just an on axis result of one stage in the development design- the total power response does well in hitting my desired target slope for the complete system.




Yes, the 40Hz dip is a room mode, due to placement and the issues with that particular room; and for the intended design purpose, this was the placement that had the least issues in the areas of concern for the crossover and system development.

This is on axis and 30 degrees off axis.


Data is time aligned by also measuring all drivers at once to make a composite plot, and adjusting the delay parameters applied to minimum phase data in the design software until it matches the measurement.

My preferred measurement software is Fuzzmeaure, recently acquired by the Freedman group- have to see how that change works out.



ET takes issue with the above statement that the relative phase at those frequencies is not so important- it impacts transient definition and vertical off axis lobing and the total room power response. Just my 0.02. :W


Of course, it's a free forum and world- use what ever method gets you the results you want... and that you're comfortable with the effort and your understanding of it. I'm personally not a fan of the more complex techniques combining near and far field data. I think near field investigations are very useful at the driver level, but a wide axis power response is really what you need to understand the interaction of driver and enclosure, as Kimmotso states.

That's my story and I'm sticking to it... at least for now!

BR,

Jon

[Dave Bullet]

Don't forget to add the baffle diffraction signature to the near field responses before you merge them with the far field responses.

Thanks Brandon

I forget the BDS simulation bit. I use an old version of the Bagby BDS spreadsheet which allows odd shaped baffles which I needed for my 3 way design. Bagby and laubs FRD combiner imports neadfield (pre merged with port response) the BDS of the driver and gated farfield and merges them together (in steps 12 and 13 above)

[ergo]

Thanks tktran for starting this topic - it is interesting to read the various possible approaches and experiences.

Btw, bit OT but the Fuzzmeasure 4 is on sale (-20%) at http://www.rodetest.com That was enough for me to finally push the button to upgrade the 3.3 -> 4 as well. Hope it's worth it.

[Efalegalo]

Speaker workshop is pretty handy for scaling nearfield / farfield. I presume you're familiar with Bagby and Laub's FRD blender tool?:
FRD Blender and Minimum Phase Extractor

My process - usually 1m distance is enough. I do outside measurements with tweeter at least 2.4m in the air to get 1,024 sample points with a >11msec gate at 96KHz. I use speaker workshop for measurements, but FRD tools to extract minimum phase since any "munging" of curves means recorded phase in SW is meaningless. Driver offsets must be factored then in the xover sim as Erik pointed out.

1. Gated tweeter (farfield)
2. Gated mid (farfield)
3. Gated woofer (farfield)
4. Nearfield mid
5. Nearfield mid port (if one)
6. Merge mid nearfield and port together (in SW)
7. Nearfield woofer
8. Nearfield woofer port (if one)
9. Merge woofer nearfield and port together (in SW)
10. Scale mid merged curve to mid farfield response
11. Scale woofer merged curve to woofer farfield response
12. Run FRD blender to splice merged nearfield into farfield for mid - check levels. Extract min phase
13. Run FRD blender to splice merged nearfield into farfield for woofer - check levels. Extract min phase
14. Extract min phase for tweeter
15. Re-import all into SW
16. have some xover fun.

What an excellent post Dave. For someone who is slowing trying to get into this hobby (like myself), this post is immensely helpful. Thank you for sharing your process.

[Dave Bullet]

What an excellent post Dave. For someone who is slowing trying to get into this hobby (like myself), this post is immensely helpful. Thank you for sharing your process.

I've thought about writing an addendum to the excellent unofficial speaker workshop manual by Jay, Claudio and Nelson.

About adding a troubleshooting guide to measurements as it can drive you up the wall literally trying to find problems with your rig / setup.

Note: I am not an experienced builder / designer, but like most things DIY - if you want to do it free or cheap - you become the glue for multiple tools. My setup is all free software, a $25 PC and a bodged together "jig" that was meant to be a wallin 2 jig but due to variable contact resistance of the crappy switches became a "hard wired claudio negro jig in wallin 2 clothing" lol.

The most expensive part of the operation was the Behringer mic. I ended up doing a sort of self calibration using a known tweeter on known baffle - Seas 27TBFC/G from Zaph's measurements. This seems to have worked out well with other drivers measured.

I think outside farfield is underrated. Maybe due to my lack of knowledge, issues might show up with background noise fooling the system regarding non-linear distortion. But I haven't resorted to groundplane or other "Fancy" measurement approaches. Then again - this is my first 3 way, so I am no doubt going to have some "eat humble pie" moments :-) But isn't that half the fun (not humble pie, but learning).

[Matt M]

Thank you folks for writing this up!

+1 to the part of extracting min-phase after merging. As I understand it, VituixCAD isn't doing this in its merging tool. But it should IMO.

[Zvu]

The most expensive part of the operation was the Behringer mic. I ended up doing a sort of self calibration using a known tweeter on known baffle - Seas 27TBFC/G from Zaph's measurements. This seems to have worked out well with other drivers measured.

How did you generate .cal file ?

...I think outside farfield is underrated. Maybe due to my lack of knowledge, issues might show up with background noise fooling the system regarding non-linear distortion. But I haven't resorted to groundplane or other "Fancy" measurement approaches. Then again - this is my first 3 way, so I am no doubt going to have some "eat humble pie" moments :-) But isn't that half the fun (not humble pie, but learning).

Agreed. Closest to anechoic chamber for us mortals. >2m from the ground, 1-2m distance on a quiet day (or night - if you're like me).

[kimmosto]

+1 to the part of extracting min-phase after merging. As I understand it, VituixCAD isn't doing this in its merging tool. But it should IMO.

Definitely NOT, though you can convert all merged responses to minimum phase with Calculator tool if damaging of captured timing is essential. Minimum phase extraction is just a mistake with dual channel measurement data, especially if radiator is not minimum phase device acoustically or electrically or distance to acoustic center changes to off-axis. For example BMS coaxial 2" driver with integrated passive crossover installed to deep horn. Minimum phase extraction would normalize all responses (axial and off-axis) to equal distance by the shape of frequency response, and non-minimum phase features of both driver+horn -combination and integrated passive crossover are deleted. Timing and phase data would be totally ruined for off-axis, power and DI response simulation with common cone woofer, which maintains acoustic center while rotation. Also excess group delay calculation, step response etc. won't be fact even in axial response.
So, VituixCAD Merger is state of the art tool at the moment, but can be used as stupid and old-fashion way as the others.

[kimmosto]

^My horn example is not very relevant (though possible) when talking about merging. It just demonstrates obvious risks and faults in minimum phase extraction. Usually no harm is done if designer "knows" that DUT is very close to minimum phase device, and acoustic center is maintained while rotation of enclosure or speaker is designed by axial response only. Count of IF words is greater than zero and therefore minimum phase extraction should not be forced by merging tool. It is choice of designer, but not recommended (by me) if goal is to measure accurate timing with dual channel gear for acoustical 3D simulation where each driver is located in rotation center and timing/phase differences are included in measurement data.

[ergo]

Kimmo has a good measurement preparations doc walking through this in addition to VituixCAD manual.

In my process steps I also add
a) mid + tweeter in parallel far field response
b) for 3way also woofer + mid + tweeter in parallel far field response
like suggested in LspCAD tutorials and some other sources.

Then after all the steps above I'll input the measured paralleled response as a 'reference curve' and compare it to the simulated one pre-crossover. In mids and highs the response should match well. In low end if one uses the already merged curves for the woofer then it will deviate as the ref curve will have a time window induced LF drop in it and simulated one does not anymore. Still might be useful to check... One can also first use the pre-merged mid/woofer data to get the offset correct and then load the merged files and be more confident that sims match the reality.

[kimmosto]

^I've never measured drivers in parallel. Maybe because I've had dual channel or other latency stable gear for 15 years. No need to make any extra steps or tuning to investigate acoustic centers, and possible non-minimum phase features are captured with common timing reference (assuming that speed of sound stays constant while different measurements). Most severe problems which slow down total design time are FireWire connectors, static electricity and grounding. They tilt or confuse CLIO measurement system several times a day. Without these problems measurement and simulation time of for example Taipuu 4-way would be less than a workday. Crossover is expected to be final at once.

[ergo]

CLIO FW, when it works, is a more stable method indeed. I have that, but I also alter to using Fuzzmeasure, ARTA, HolmImpulse depending on a job and then it sort of helps me to 'put my mind at ease' if I have these extra files just to sanity check if I have a reason to doubt the results later.

I've been pondering if upgrading CLIO FW to USB would make sense or not... they do offer that, but according to the other forum it's 450EUR+VAT for QC version. It would be nice to get rid of the FW and related problems, but other than that they don't really offer much else to justify the asking price.

[Matt M]

I must apologize, Kimmo - I should have elaborated a bit more:

Bluntly replacing a dual channel measurement with computed minimum phase doesn't make much sense. I would have hoped that to be a bit more obvious. My experience is that the phase information of merged measurements is often divergent from a minimum phase behavior, but ONLY around the merging frequency (you will have to add a delay to one of the curves to be able to compare them, obviously ). That is hinting strongly that the way phase information is merged is insufficient. Not only the way VituixCad is doing it, but also the way I have been doing it before using other tools.

For people like me, who (most of the time) use single channel measurements, the approach I mentioned before would make sense - because the data is minimum phase to begin with. Actually what I would like to see is the following behavior - and that would also make sense with dual channel measurements:
- compute a minimum phase for the merged response.
- adjust the offset of the computed phase such that phase tracking is identical outside of the merging "zone", especially towards higher frequencies.
- replace the measured phase only inside the merging zone. Blend gently from low-freq-measurement to computed to high-frequency measurement.

That way the trustworthy part of the original phase information will be preserved. And the broken part repaired. I hope that clarifies what I meant.

As an example: attached is a merged measurement of a midrange driver (merged @ 400Hz). Red is the original phase. Green the computed one. You see, there is a deviation of up to 20° in the area you would be most interested in when crossing to a woofer.

- Matt

[Dave Bullet]

How did you generate .cal file ?

I used speaker workshop which has scaling, transformation and add/subtract functions to subtract the known measured tweeter (Zaph Seas 27tbfc/g) from my raw measurement of same tweeter on same bafffle. I ended up with a funny hump / dip / hump above ~ 13Khz as shown but in measuring other Vifa tweeters and comparing with those on-line, I think the low end and high end response seems to marry with other ECM8000 calibration files.

[kimmosto]

My experience is that the phase information of merged measurements is often divergent from a minimum phase behavior, but ONLY around the merging frequency (you will have to add a delay to one of the curves to be able to compare them, obviously ). That is hinting strongly that the way phase information is merged is insufficient.

I have seen that phenomenon too, but it has not been so serious that Merger would need changes. I just checked my last project (Taipuu 4-way) data and compared normal/measured phase to minimum phase of axial response. Greatest difference is around merging frequency, but it's only about two (2) degrees which is totally acceptable. This happens with lower mid-range driver (18WU) only. Phase difference with subwoofers (30W) and small mid-range driver (12MU) was zero around merging frequency. Difference comes from FFT calculation of time-windowed far field response. In that case from CLIO 11 with Hanning/2 time window of 4.50 ms.
User can adjust delay offset manually (after transition frequency is selected) to correct possible error in merged phase. Wide correction requires quite wide blending range, e.g. three octaves. Optional minimum phase plot in Merger tool would help that.
I would also investigate phase information of time windowed far field measurements, and select software and parameters which produce valid phase data with time window. Only this way designer can be sure that non-minimum phase features of DUT are captured properly, and forcing to minimum phase can be avoided.

[Matt M]

I agree that the better the measurements to begin with, the less the effect mentioned is even an issue. I suspect FFT to be the main culprit here, too. I usually use ARTA / UMIK-1 / ~3 ms / half-hanning. The maximum nearest-boundary is 1.2m. These are conditions a lot of DIYers have to deal with.

Regarding the topic of this thread, the application of the described technique often makes the difference for me between reliable/usable and questionable measurements when going below 500Hz.
VituixCAD is doing a lot of things quite comprehensively and automatically, so I figured it worthwhile to have it as an option in the merger. But it is no big issue since it can be done externally if required.

Btw. simulation data of Taipuu 4-way is looking great. Very promising!

[kimmosto]

...~3 ms / half-hanning. The maximum nearest-boundary is 1.2m.

Are you saying that you measure from longer than 2 meters?

[Matt M]

Typical "farfield" measuring distance is 60 to 70cm. With "Maximum nearest-boundary=1.2m" I meant I can never place the measured speaker further than 1.2m from the nearest boundary. Depending on circumstance the nearest boundary will be even closer. The "~3ms" is guess from memory, and may be wrong. The actual gate length will be determined by looking for the first reflection in the impulse window.

[kimmosto]

^Ok. Time window could be about 5.2 ms if the nearest boundary is at 1.2m and mic is at 70cm. Might be worth to verify maximum window length with calculator to avoid too short time window due to diffraction peak etc.

I just investigated my measurement archive containing time windowed measurements with justMLS, ARTA, SoundEasy, LspLAB and CLIO 10-11. Some systems are more sensitive to DC offset in sound card input or other L.F. disturbance than the others. Data in my archive is not systematically captured and comparable, but generally it looks that phase response of time windowed measurements is more reliable than magnitude response at low frequencies. My rapid and possibly biased conclusion is that phase matching style VituixCAD Merger is using could be more accurate than converting to minimum phase around transition frequency before merging. Merged response may not be exactly minimum phase in some cases where DUT is, but total phase shift and group delay could be closer to truth by matching measured phases.
Longer time window reduces possibility to magnitude problems at low frequencies -> measured far field response is closer to minimum phase in the beginning (if DUT is MP).

[Dave Bullet]

Typical "farfield" measuring distance is 60 to 70cm. With "Maximum nearest-boundary=1.2m" I meant I can never place the measured speaker further than 1.2m from the nearest boundary. Depending on circumstance the nearest boundary will be even closer. The "~3ms" is guess from memory, and may be wrong. The actual gate length will be determined by looking for the first reflection in the impulse window.

Why don't you just go outside? I've had great results hoisting the speaker to 2.4m+ into the air. Means I can get a 11msec+ gate

[kimmosto]

Temperature in my outside lab is -12 C at the moment, and 50...100 cm of snow. Other options are rainy, windy or nighttime while neighbors do not like measurement sweeps Few days a year would be okay but project schedule probably require some other time. 1.2 m to the nearest boundary is quite attractive in warm, silent and dry environment.

[Zvu]

^Use square noise - it is harder to detect. Turn your speakers away from the neighbours, directivity will do the job at higher frequencies (where neighbours are most likely to detect sounds), and at lower frequencies square noise resembles leafs in the wind or maybe sound of waterfall ? - maybe a bit louder but that kind of tone is not irritating as sweep and is familiar to our ears - much more than WHHOOOOOOOOOP !!!

[Matt M]

Why don't you just go outside? I've had great results hoisting the speaker to 2.4m+ into the air. Means I can get a 11msec+ gate

Right now...

1. the speakers weigh 35kg each.
2. they contain a 350 USD Accuton driver that I don't want to risk dropping from 2 meters height.
3. I am on the same page as Kimmo regarding environmental conditions.
4. I don't feel the need to do it outside, because the indoor measurements appear to do the job just fine.

If one has the opportunity to measure outside like you say, that's great.

I just investigated my measurement archive containing time windowed measurements with justMLS, ARTA, SoundEasy, LspLAB and CLIO 10-11. Some systems are more sensitive to DC offset in sound card input or other L.F. disturbance than the others. Data in my archive is not systematically captured and comparable, but generally it looks that phase response of time windowed measurements is more reliable than magnitude response at low frequencies. My rapid and possibly biased conclusion is that phase matching style VituixCAD Merger is using could be more accurate than converting to minimum phase around transition frequency before merging. Merged response may not be exactly minimum phase in some cases where DUT is, but total phase shift and group delay could be closer to truth by matching measured phases.
Longer time window reduces possibility to magnitude problems at low frequencies -> measured far field response is closer to minimum phase in the beginning (if DUT is MP).

What you describe sounds pretty reasonable to me. If your biggest deviation was 2°, your measurements must be way more accurate than mine in the relevant region. But I can't follow your conclusion. I have done two 3way-speakers now showing the same experience. There are several indicators that the computed phase behavior was better suited:

1. Drivers that are aligned by depth and filtered to ideal LR2 or LR4 slopes should display very good phase tracking in the crossover band. If they don't, something is seriously wrong. Computed phase showed the expected behavior.
2. Doing a vertical series of reverse polarity measurements at various distances (ungated) showed deeper and more symmetrical nulls for those crossovers that were based on computed phase.
3. This was confirmed by listening impressions from competent but uninformed listeners for one of the speakers. Crossover based on measurements were described as having a response tilted to the floor and drivers were considered individually locatable.

All this accounts only for the below 1kHz region. Beyond 1kHz I fully endorse measured phase data.

If you don't have the problem, there is no need to fix it. But if you experience it, it is a nice and easy fix without going outside.

Another footnote regarding the issue, likely controversial: VituixCAD does one thing, that I have wished simulation tools to do for a long time: compensating phase for driver-to-listener-distance individually for each driver. However, in practice it didn't quite have the benefits I had hoped for. In practice there are multiple listening positions in question. When adjusting things to the ideal stereo triangle 2.5m position, individual drivers became more easily discernible when walking around the room. Aligning acoustical centers physically and aiming for perfect slopes provided much better behavior, both at the sweet spot listening position as well as when moving around.

When VituixCAD compensates for positioning, my feeling of reference how things should look like gets diminished, or the intuition if there is even a problem with the measurement. Fortunately one can force VituixCAD to "dumb-down" by adjusting driver positions accordingly. I am so sorry, but that's what I did! ops: But I got better acoustical results that way. I think this is related to Troel's preference for stepped baffles as opposed to electrical solutions. He certainly knows how to do a ladder network, and is likely aware of the detrimental effect a stepped baffle has regarding diffraction. Yet he is doing it - food for thought!

This isn't a criticism on VituixCAD. IMO it is a great thing that the program is doing it the way it does.;x(

[kimmosto]

But I can't follow your conclusion. I have done two 3way-speakers now showing the same experience. There are several indicators that the computed phase behavior was better suited...

It's unfortunate that measured magnitude or phase has significant error, but fortunately minimum phase extraction can be done with Calculator tool after merging. Merger tool supports also merging of far field responses or already merged responses (including 0...720 off-axis responses at the same time). That feature enables merging of responses with computed phase and measured phase with selected blending bandwidth, as well as merging of simulated and measured far field responses. The latter is useful for eliminating excessive directivity index at low frequencies.
One important decision is transition frequency. Typical merging point in my projects has been 380...550 Hz. That range is quite good for minimizing excessive smoothing above transition and possible magnitude and phase errors below transition frequency due to time window function.

[Matt M]

Merging point has been around 400...450Hz, usually.

I will see to make use of the calculator tool next time and see how that goes.

That feature enables merging of responses with computed phase and measured phase with selected blending bandwidth, as well as merging of simulated and measured far field responses. The latter is useful for eliminating excessive directivity index at low frequencies.

I had the impression that this would be possible by now in VituixCAD. Also a very desirable feature. Keep up the good work!

[kimmosto]

I had the impression that this would be possible by now in VituixCAD

That is merging of far field responses - with or without off-axis responses. LF and HF responses can be measured or traced or simulated with Diffraction or Enclosure tool, or processed with Calculator or Merger tool or any external response tool. This has been supported since version 0.1.3.18 (2015-10-29).

[Matt M]

In that case, I have obviously missed that.

[tktran]

This has been a very useful thread and has helped me move forward with my crossover design.

My experiences have led to the following realisations-

1. Outdoor measurements- cons outweigh pros. Limited by setup/packup time (2 hours to complete a dozen measurements), noise and weather. It's not trivial to get a 100+ lbs 3 way WWMT speaker 2m above the ground, safely. My longest gate time was 8ms, which gives me measurements down to about 125Hz.

2. I need to learn how to merge far field and near field response properly, and incorporate baffle diffraction effects. Thank you kimmosto for your manual.

[kimmosto]

In that case, I have obviously missed that.

I just added Minimum phase option to VituixCAD Merger tool, rev 1.1.27.0. It's also possible to add group delay of HF response to minimum phase response by checking with GD of HF. That option saves delay adjustment of HF responses and possible offset in acoustic center if HF responses are measured with dual channel gear.

[Matt M]

Tried the new feature. Looks very useful to me so far. I will give it a closer look the next time the occasion arises. Thank you, Kimmo.

[tktran]

Weather conditions permitting, has anyone taken loudspeaker measurements outdoor, facing the sky?

I’m imagining the speaker elevated from the floor 1-2’, facing upwards, with the mic on a boom stand, facing down.

With the closest “walls” being the fence of the property, the first reflection should be about 20ms away...

[Zvu]

Floor bounce would be the one that would kill the response. There wouldn't be much of a gate there at 2 feet off the ground.

You could bury it in to the ground and factor in the baffle step in your design or, again, lift it up couple of meters.

[bvbellomo]

I am sure others will disagree, but I don't see a point in separating room/speaker issues to get something perfectly flat only to need to EQ it back to make it work in your room.

I'd design theoretically first, build the crossover, set up the speaker in the room and measure the woofer and be sure it isn't making any sound it shouldn't. For example, if your woofer has a harsh breakup problem at 1200Hz, throw a resister in place of the mid and play a test tone at 1200Hz at a voltage you'd hear 100dB (or a little above what you'd listen at) and see what comes out. I'd try several frequencies between the crossover frequency and whatever point you are sure no more sound comes out. This is the biggest problem I'd worry about.

Then I'd do a full sweep with both drivers at various points in the room and be sure I don't have a dip or spike near the crossover, adjust if needed, and retest the woofer above crossover if I changed anything.

Then I'd add resistors to pad down whichever driver is too loud. And I'd stop there. Anything further is too hard to measure and of little benefit. Will your crossover work well in an anechoic chamber? Probably not, but you aren't going to be listening in one. In any kind of real room, you either have to live with spikes and dips and either EQ or live with them.

[Carl V]

JA of Stereophile carried on about just this with the Designer of the Newly developed HARBETH 40.1
The back and forth was enlightening & the Correlation to what JA & Art Dudley heard/measured was
informative.

[kimmosto]

Anything further is too hard to measure

More advanced method is like a good investment, and not so difficult compared to multi-point room measurement "lottery". Well behaving and accurate product is more stable/universal to different locations, music genres, acoustics, personal tastes etc. But if you design only for yourself to a certain location, measurement and design method can be poor and narrow-minded. It's luxury of DIY which for example I can't afford.

[Efalegalo]

VituixCAD and Arta

Sorry to bump this thread, but I'm hoping to get a littler more insight on how measurements should be performed. I've used Xsim/WinPCD for crossover modeling before, but I'd like to learn a little more about VituixCAD.

My Measurement Equipment Includes: Focusrite Scarlette Solo, Behringer EMC8000 Mic, Dayton Audio DATs,
My Software:L ARTA - Demo Version - and Vituix CAD


Additional Facts:
1) I do not have the option to take outside measurements
2) Nearest boundaries for inside measurement are 4 feet (i.e., basically a room with 8 feet ceiling)
3) I do not (edited) have a turn table setup to take off-axis measurements (will build one in the near future.
4) Measurement gear is hooked-up using "Semi-Dual" diagram as provided in ARTA

Based on the above, what is the best way to use VituixCAD? My thinking below:


1) Take farfield measurements of the Tweeter, Woofer, and Tweeter + Woofer (paralleled) - all on axis with the tweeter (i.e. the design axis)
2) Use gating to remove reflections from the measurements in Step 1, and save all files with minimum phase
3) Using WinPCD - Acoustic Off-Set Tab, determine the Z-offset of the drivers
4) Now lets say all measurements in Step 1 were taken at 30" from baffle on axis with the tweeter ---> Now take another measurement of the woofer on the woofer axis at the same 30" baffle distance (if the speaker includes a stepped baffle to better align the woofer with the tweeter, make sure you are measuring distance from the stepped portion of the baffle). Again save woofer's response using minimum phase
5) Take near field measurements of the port and woofer.
6) Using VituixCAD's "Merger Tool", merge the nearfield Port Response and nearfield Woofer Response (adjusting for diffraction effect), with the Woofer's farfield response. Save file with "Minimum Phase".
7) Use DATs, measure tweeter and woofers (in-box) impedance
8) Because I did not take any off-axis measurements of the tweeter, using VituixCAD Diffraction Simulator, create the baffle, indicate the tweeter's radiating diameter, indicate the mic's listening position, load the tweeter's half space response, and check off - "Full Space", "Directivity", "Vertical Plane", "Negative Angles", and "Feed Speaker" ------> click okay

9) Because I also didn't take off-axis measurements of the woofer, perform the same steps above in Step 8.
10) Upload the tweeter and woofers impedance response as measured by DATs
11) Specify each driver's location relative to the design axis. For example, if the design axis is the tweeter axis, indicate 0, 0, 0 for X, Y, Z coordinates for the tweeter. If the woofer was 167mm below the tweeter, and 3mm in front of the speaker (because I used a stepped baffle on my current build), then indicate, 0 for X, -167mm for Y, and -3 for Z.

12) Begin crossover desing.

Based on the above steps, I have the following questions:

1) Do I check the checkbox next to ARTA's "Dual Channel Measurements Mode" when all my connection are in "Semi Dual" mode
2) Is the only time the response of the Woofer that measured on the Tweeter Axis utilized is when one is calculating Z-offset?
3) For VituixCAD, should all drivers measurements be on axis with each individual driver (while still maintaining same measurement distance from baffle)
4) Does VituixCAD calculates the design axis based on how one enters the location of the drivers on the baffle? For example, if I entered 0,0,0 for XYZ coordinators for the tweeter, will Vituix will simulate response based on that point in space?
5) Because no actual off axis measurements were performed, and instead the Diffraction tool was utilized to simulate off axis response relative to design (i.e. mic location) axis, is there any concern with indicating the design axis in two different locations (i.e., one of the main screen where each drivers' XYZ coordinates were indicated, and also within the Diffraction tool where the driver and mic locations need to be indicated).
6) In Step 8 above, when loading the woofer's "half-space" response, am I to use the woofer's response obtained on the tweeter axis or the woofer's response obtained on the woofer axis?

[kimmosto]

3) I do have a turn table setup to take off-axis measurements (will build one in the near future.

I suppose you do not yet have turning table because will build?

4) Measurement gear is hooked-up using "Semi-Dual" diagram as provided in ARTA

For this reason, you don't need item "3) Using WinPCD - Acoustic Off-Set Tab, determine the Z-offset of the drivers". Dual channel measurement with common timing reference (starting point of time window and PreDelay) eliminates need of software other than VituixCAD.

4) Again save woofer's response using minimum phase

Generally, minimum phase extraction is not needed with dual channel measurements and VituixCAD.

6) ... Save file with "Minimum Phase".

This is possible exception to previous answer. Minimum phase may be needed while merging near and far field because ARTA has some issues in phase angle at LF with short time windows. Check the difference between measured and minimum phase, but do not check Minimum phase as default.

8) Because I did not take any off-axis measurements of the tweeter, using VituixCAD Diffraction Simulator, create the baffle, indicate the tweeter's radiating diameter, indicate the mic's listening position, load the tweeter's half space response, and check off - "Full Space", "Directivity", "Vertical Plane", "Negative Angles", and "Feed Speaker" ------> click okay

You should build turning table and stop using Diffraction tool for directivity estimation for middle and tweeter. Diffraction tool is reliable enough until low middle but higher frequencies "deserve" decent polar response measurements with step of 10 deg.

1) Do I check the checkbox next to ARTA's "Dual Channel Measurements Mode" when all my connection are in "Semi Dual" mode

Yes. This is essential, in addition to loop cable in reference channel.

2) Is the only time the response of the Woofer that measured on the Tweeter Axis utilized is when one is calculating Z-offset?

If you refer to WinPCD, that is not needed. Z differences are included in phase response of dual channel measurements by keeping distance to baffle constant and starting point of time window and PreDelay constant. Z-location of stepped baffle is entered to driver list, relative to tweeter.

3) For VituixCAD, should all drivers measurements be on axis with each individual driver (while still maintaining same measurement distance from baffle)

Yes.

4) Does VituixCAD calculates the design axis based on how one enters the location of the drivers on the baffle? For example, if I entered 0,0,0 for XYZ coordinators for the tweeter, will Vituix will simulate response based on that point in space?

Yes. 0,0,0 mm is perpendicular end point design axis on baffle surface. Listening distance in Options window defines distance/radius to virtual mic, rotating in two planes around 0,0,0 point.

5) Because no actual off axis measurements were performed, and instead the Diffraction tool was utilized to simulate off axis response relative to design (i.e. mic location) axis, is there any concern with indicating the design axis in two different locations

Locate mic in Diffraction tool in the center of driver. This links diffraction data to (any) position you define for the driver in the main program.

6) In Step 8 above, when loading the woofer's "half-space" response, am I to use the woofer's response obtained on the tweeter axis or the woofer's response obtained on the woofer axis?

Do not load half space response to Diffraction tool if you are producing diffraction response to Merger tool.
For other cases: "the woofer's response obtained on the woofer axis" or half space response simulated with Enclosure tool.

---

Please rehearse this document for measurements, diffraction simulation and merging.

[draki]

My Measurement Equipment Includes: Focusrite Scarlette Solo, Behringer EMC8000 Mic, Dayton Audio DATs,
My Software:L ARTA - Demo Version - and Vituix CAD

Note that you can't save and export measurements with ARTA DEMO - you need to buy the licence.