Dhar's Joseph Crowe Waveguide panel two way with TW29BN-B & PTT6.5W04

[JonMarsh]

One of our members, Dhar, shared with me a project he's being working on for a friend- it's a build using a Joseph Crowe designed front baffle for the SB and PuriFi drivers, with a waveguide for the tweeter milled into the front panel. Joseph Crowe sells a 3D CAD file for the baffle drawing, and then the buyer can either transform to a 3D printer or generate codes and get baffles machined using CNC.

My, how DIY has evolved...





An example for a completed baffle is shown, with suggested edge border round overs.



This cutaway in the vertical shows the vertical profile...




And here you can see the horizontal profile- of course, the polar patterns will differ. Those are not disclosed on the web site.

[JonMarsh]

An on axis graph of the tweeter + waveguide combination is provided:



One thing I noticed that was curious was the relatively pronounced roll off on axis between 10kHz and 20kHz. I was curious if that was consistent in the off axis plots- Dhar's measurements indicate that it is, and gets more pronounced.


The suggested crossover is a 4.7uF cap in series, plus an Load. This compensates for the waveguide rise in output, and then relies on the waveguide roll off and the tweeter fall off, with just 6 dB per octave attenuation of the electrical input. But this attention starts fairly high up for this nominal 4 ohm tweeter, which is actually closer to 3 ohms in working impedance.

[JonMarsh]

Here's a shot of the front panel that Dhar's friend had made up, on the cabinets which Dhar built for him.





Dhar chose a 2kHz crossover point, and given that the Z axis origins were nearly identical for the wooer and the tweeter with waveguide, he selected a 4th order LR target, as this would also provide more electrical power protection for the tweeter and increase the SPL capability compared with the "crossover" suggested by Joseph Crowe.

This is a VituixCAD schematic for the initial concept, based on modeling components already on hand.





Dhar measured data for the PuriFi woofer and the SB Be tweeter out to 60 degrees, and modeled this in VituixCAD V2.

This is the resulting calculated on axis SPL response...




Which looks like a fairly good first pass, and which exhibited a deep reverse null.


As is common for this crossover design approach (think Zaph waveguide two way; think Modula MT XE), without a LCR input impedance Zobel network, the impedance swings and current phase swings are substantial.

[JonMarsh]

Dhar updated his crossover with an input impedance LCR network, and this improved matters substantially.











Also from Dhar's initial measurements, the SPL with normal and inverted connection:






And polar measurements, showing the combined woofer with tweeter plus front panel waveguide:





Things are looking pretty good here, and certainly performing at a different level than what could be called the "minimalist" crossover solution at Joseph Crowe. But, before getting out the soldering iron, is there more that can be done? Is there a bit more of a "perfectionist" solution available without too much more effort?

[Evil Twin]

To quote an earlier episode...


"Once you were the teacher, but now, I am the master..."


Given some similarities between this project and one I have been working on when possible during the last two year's tumult, the "Darth Modula", it could well be that I have some useful suggestions to offer on Dhar's behalf. Do not doubt that I understand the meaning of a "perfectionist" crossover approach, even for a smaller project like this one...


So, in consideration of these issues, I offer this crossover proposal for consideration...



A few comments are in order...

• Given the mandate to produce a high accuracy "perfectionist" design, is it most desirable that the modeling be conducted as accurately as possible within the capabilities of VituixCAD; for this reason, the source output impedance has been adjusted from the default 1mOHm to a value representative of high performance solid state amplifiers such as the Cambridge Audio Edge W, or the NAD M33. Note that many so called cosmetic pretenders to that role may have much higher output impedance- it is not uncommon to see values such as 0.2 or 0.5 ohms output impedance for some products.
• Given the substantial impedance rise due to Fs of the tweeter, and the potential impact on roll off behaivor, it is suggested to employ an LCR zobel network to neutralize the impedance bump. Many might consider this unecessary, but this is the perfectionist approach...
• The "on hand" inductors for the trial woofer crossover are possessed of unacceptably high DCR, in consideration of the woofer load being a 4 ohm driver of low sensitivity. The inductor models have been updated to reflect high performance parts available from typical vendors such as Parts Express.
• As this is targeted as an acoustic LR4 crossover, it should be remembered that one weakness of that alignment is the narrow vertical lobe and the reduced room power response that results in the crossover region. Even the LR4's creators often recommend a slight bump in the crossover region in order to strike a balanced between direct sound and the room power response... this is implemented through the network adjustments.
• The tweeter crossover has been slightly rebalanced, as part of achieving the desired acoustic transfer function, and it was proven feasible to reduce the value of C6. Be thankful that there is at least one cost saving result in this perfectionist proposal...
• The input impedance LCR zobel is reworked, to bring it closer to true perfectionist performance levels.

Let us now examine the basic results of these proposed updates...




I believe you may find this result most satisfactory, considering the limitations of the tweeter and waveguide interface.




Examination of reverse phase connection confirms the expected deep phase null with consistent crossover phase tracking.

[Evil Twin]

Additionally, we will review the other VituixCAD plots of simulated performance:


First, the room Power and Directivity index:





Next, review the group delay and phase response:





The actual filter functions as implemented:





And the calculated input impedance:





I believe we can all agree that at the least, the flatness of impedance magnitude and current phase constitutes perfectionist levels of performance. This must await verification through construction of a test crossover, of course.

[Efalegalo]

ET,

Thanks for posting this. In regards to the tweeter's response drop between 10Khz - 20Khz, I don't see the same droop in HiFiCompass's Measurements. I also don't see it in Troel's measurements ( we see about 3db there).

For those that are interested, my measurements setup consists of Focusrite 2i2 (Gen 1) + Dayton Audio APA150 Amp + Cross Spectrum Labs Behringer ECM8000 mic.

Some have questioned the accuracy of the CSL mics over the years. My mic was calibrated in April of 2018. I had compared it to MiniDSP Umik (Gen 1) - and they were pretty much bang on until 15Khz/16khz.

For those considering this design, my 2 cents would be to pass on purchasing the full baffle or at least waive the request to have the woofer cut out machined. Crowe has the woofer rear-mounted, and the effective thickness of the baffle where the woofer sits is likely no more than 7-8mm (useless IMHO).

If I was going to have Joseph make the baffle, I would insist it the woofer be front-flush-mounted, with at least 12-15 mm for screws to bite in, and then chamfer.

Doing so would result in the "Z" value specification in VituixCAD to change, and correspondingly will require tweaks to the network. The good thing is that ET already has the measurement files.

Lastly, the polars were only taken out to 60 degrees and only in the horizontal direction. I believe the tweeter waveguide has a slightly smaller vertical angle than the horizontal angle. In absence of providing VituixCAD vertical measurements, it (by default) assumes horizontal polars approximate vertical polars (I think). I will try and take virtical measurements as well.

[Evil Twin]

ET,

Thanks for posting this. In regards to the tweeter's response drop between 10Khz - 20Khz, I don't see the same droop in HiFiCompass's Measurements. I also don't see it in Troel's measurements ( we see about 3db there).

For those that are interested, my measurements setup consists of Focusrite 2i2 (Gen 1) + Dayton Audio APA150 Amp + Cross Spectrum Labs Behringer ECM8000 mic.

Some have questioned the accuracy of the CSL mics over the years. My mic was calibrated in April of 2018. I had compared it to MiniDSP Umik (Gen 1) - and they were pretty much bang on until 15Khz/16khz.

For those considering this design, my 2 cents would be to pass on purchasing the full baffle or at least waive the request to have the woofer cut out machined. Crowe has the woofer rear-mounted, and the effective thickness of the baffle where the woofer sits is likely no more than 7-8mm (useless IMHO).

If I was going to have Joseph make the baffle, I would insist it the woofer be front-flush-mounted, with at least 12-15 mm for screws to bite in, and then chamfer.

Doing so would result in the "Z" value specification in VituixCAD to change, and correspondingly will require tweaks to the network. The good thing is that ET already has the measurement files.

Lastly, the polars were only taken out to 60 degrees and only in the horizontal direction. I believe the tweeter waveguide has a slightly smaller vertical angle than the horizontal angle. In absence of providing VituixCAD vertical measurements, it (by default) assumes horizontal polars approximate vertical polars (I think). I will try and take virtical measurements as well.

I do not believe this is a function of YOUR measurements, but a function of the interaction between the Joseph Crowe waveguide design and the SB tweeter.

Note that this drop off is quite evident in the data he provides...


Both for the raw measurement,





And for his minimalist crossover response plot.

[Efalegalo]

Here are the horizontal polar measurements (0 - 90) with the revised xover. I didn't have exact parts, but these are very close (still high DCRs on woofer inductors).

These measurements were taken at low volume (late night)- about 1M distance. Gate time of 3.5ms. (click to enlarge)



Also - some vertical polars +/-25 degrees (increments of 5 degrees). There might be some be artifacts included as a result of my turntable (with the tweeter now closer to rotational surface)

[Evil Twin]

It would be interesting to see an impedance plot, if that is possible.

[Efalegalo]

It would be interesting to see an impedance plot, if that is possible.

Sure - but again - I didn't have exact values (only close). Impedance is mostly 5.2/5.3 ohms (lowest is 4ohm at 150hz, -18 phase angle). I think the bump in the impedance at 200hz is likely due to a lack of sufficient stuffing, or possibly an air leak. These speakers aren't currently all the way put together (e.g., woofer is held down currently with only two screws, real removable panel is currently secured with only 4 screws, etc.)

[Scottg]

I do not believe this is a function of YOUR measurements, but a function of the interaction between the Joseph Crowe waveguide design and the SB tweeter.

:yesnod:

..it's a result just about all dome waveguides have.

Really the only way to *"improve" that is a diffraction slot in front of the dome itself (..which will also have freq. problems). Compression drivers "get around" this by effectively having a diffraction slot via the exit of the driver itself (..and the phasing-path helps remove the freq. problems that would occur with the diffraction slot in front of the dome tweeter.

*and it's NOT an improvement overall.

[JonMarsh]

Mmmm, I'm not sure I understand exactly what you mean, but that steep fall off between 10kHz and 20kHz is NOT a result of the more effective dome tweeter waveguide builds I've worked with.

One example, the Module MT XE with the SS Discovery tweeter:






Another example, a waveguide (CNC milled from layered MDF) that Steve Manning and I developed, with the Dayton RSe28a tweeter on axis:




And a polar set:




I'm not saying that this is the be all and end all of dome tweeter waveguides, but the difference between that and the Joseph Crowe part is substantial.

[Scottg]

It's largely about how close the walls of the waveguide are to the center of the tweeter - and this includes the profile of the dome tweeter (..and notably the driver's own pressure loss off-axis at higher freq.s).

AND

It's in relation to the total/combined response off-axis: particularly how *uniform the loss is off-axis throughout the passband.

In this case the Satori tweeter has a fairly shallow dome and a lot of pressure loss off-axis at its higher freq.s - which means there isn't as much pressure to (literally) squeeze to compensate the roll-off behavior and still achieve a uniform pressure loss. When I first looked at the design I sort of mentally went "oh no, why didn't he use the smaller diameter Bliesma driver?" ops: Of course he might not have know about it at the time. :unsure:

*uniform doesn't have to mean "constant", but rather a smooth progressive loss from on-axis to off (..accepting concentric/conical designs that tend to the constant/straight-wall variety that have the typical "dip" or "hole" in response on-axis only somewhere above 4kHz.)

Joe's overall design is actually quite good - at least within a 10 degree (+/-5) vertical window on-axis. He might have been able to "pinch" a bit more vertically at the waveguide entrance (next to the dome) to get something a little better horizontally at that top octave, but vertically it would result in much greater loss off-axis at those highest freq.s.. It's all "give and take". ops:

[Scottg]

It's interesting looking over Brandon's Waveguide's for 3D printers thread, reminds me of Dave Pellegrene's waveguide progress.

One of the more interesting qualities is that close-coupling of the driver to the wavguide at the "throat"/entrance seen here:

https://www.htguide.com/forum/showthread.php?39523-Dave-P-Wavguide-testing-5-tweeters

Note that Blogs & Groups have that mm+ "lip" or "tunnel" that bends-out to the overall angle of the waveguide (as it progresses to the "mouth"/exit of the waveguide). To me this is literally the "starting point" to a better waveguide - it couples better to the dome itself and reduces that diffraction and resonant condition. Notably it's not dissimilar from what the manufacturer's face-plate does, and it's quite likely that the manufacturer's face-plate depth is a decent starting point for determining that "tunnel's" depth before bending-out to the proposed waveguide (..and then slightly extending that depth with various iterations to see the result).

Also, with respect to Dave P's progress it's nice to see one of his last and perhaps greatest waveguide achievement, and that it (ironically) utilizes a tweeter that's moderately similar to the SB Be tweeter (though with less pressure loss off-axis in the top octave):

https://techtalk.parts-express.com/forum/tech-talk-forum/61272-scan-speak-d2604-833000-waveguides?241965-Scan-Speak-D2604-833000-Waveguides=

The easiest graph to view (to see the uniformity) is with the eq *"ON" (8" round waveguide with the SS tweeter). The result is almost constant - better than the Josephs but also larger and because of the concentric/round nature is more difficult to get good verticals because of the extended distance between it and the midrange/midbass, though as a practical matter you can "cut-into" a round waveguide to move the midrange/midbass closer to the tweeter AND you can also lower the crossover freq.. Of course Dave's was also tested on a larger baffle, so it's not really a fair comparison.

*8" round guide 2" deep 1" throat
0-90 degrees EQ'd flat XO 600Hz