Directiva r1.2 design and build

[Alan J]

A hole more than 10 degrees below the listening axis is only a problem if you want lie on the floor, but a hole above can really change the sound if you stand up or move around the room.

In this design, in which I'm already cutting facets, this raises the question of angling the baffle or the tweeter. Though as you point out different crossovers (e.g., the active crossover in the r1 that Amir tested, and at least in simulation some of the passive crossovers that have been posted) put the hole below-axis. To be continued...

 

[fluid]

You can tilt the main lobe through phase difference between the drivers, LR has it straight on but it can be tilted up or down without making the baffle slanted which will have other consequences. Easy enough to see in Vituix as you can tilt it virtually.

 

[Alan J]

You can tilt the main lobe through phase difference between the drivers

...and though I was intrigued by the idea that angling the tweeter a bit might be useful, a quick look at what it would take suggests that it's not likely to be worthwhile - or at least out of the scope of what I want to do here.

I did want, before moving on, to convince myself that a passive crossover that, aside from tradeoffs between low-end response and sensitivity, could rival what was demonstrated with DSP. Below are the results of a couple more crossover tweaks:

HP and LP equivalent to LR4 at around 2300 Hz (150 mm) and 1700 Hz (202 mm), respectively. Purifi and DXT aren't in phase on-axis in the crossover range (thus the overlap between HP and LP). The central lobe is tilted above-axis, which is probably an advantage in the intended use.

And keeping LP and HP both at 2300 Hz LR4 - Slightly more ragged-looking perhaps, and I might want the upward tilt, but looks pretty good to my still-untrained eyes.

This is enough to make me feel good about cutting prototype wood. I'll be grateful for contrary more-informed opinions.

 

[fluid]

I would probably pick version 1 (top image) and look at raising the tweeter level by 1dB, it's normal to need to try a few variations of tweeter resistance to find the best subjective balance.

 

[Alan J]

I'm a bit over my estimate on DC resistance on the woofer side as well. If lower DCR is practical, the tweeter will need a corresponding boost. Even if there are no red flags re the widened crossover region (Jeff Bagby, for one, talked about drivers in quadrature in the crossover region as a normal thing), I do want to review distortion data before counting on the DXT to contribute too far down.

But onto a prototype baffle & box. Drivers and PRs are in hand.

 

[Alan J]

The first prototype enclosure is done. I made the baffle and rear panel removable to ease any experiments with different faceting or spacing, or with reinforcement/damping. The panels are sealed with hollow EDPM weather stripping. I don't know how much I need to worry about the exposed screw heads near the backs of the facets.

The volume uncorrected for drivers and the 3/4" denim batting that I laid around the perimeter is 12 L. The plan is to reduce that with solid blocks or bags of sand and to adjust PR mass. It needs some cleanup around the edges but meanwhile we get to listen during break-in.

Though the final crossover is to be passive, I'm using an AD1701 DSP (Tinysine) to make the prototype usable and perhaps to test approaches to the passive crossover. I loaded the crossover that Kimmo shared for r1. Per the data posted to the passive crossover thread, the mid-treble will be a few dB hot, and it sounds a bit like it is.

Next substantive tests might need to wait until after Thanksgiving.

 

[Alan J]

Small update: Loaded v20 from the Directiva active crossover thread, as well as @fluid's tweak of v20 which shouldn't be terribly far off for this enclosure. Listening/break-in continues and the more I listen, the more I like it.

 

[Alan J]

After maybe a dozen hours of mostly casual listening, Idecided to take an impedance trace to see what I can discern about the enclosure and to get an initial assessment of how the bass tuning compares with TSP-based predictions in VCAD.

The Purifi trace...

...looks pretty clean. There's a little bump around 55 Hz, and some wiggles (very hard to see at this scale) around 300-500 Hz. Zooming in...

...the disturbance around 55 Hz remains really subtle. The wiggles in the 300-600 Hz region are more visible but pretty subtle. Those were more prominent when the box was sitting on a somewhat wobbly stand, got still smaller when moved to a stable but lightweight desk, and still smaller when the box was placed on a thick pad. This suggests to me that movement of the enclosure on the stand is a factor here, though of course movement on the stand could be interacting with something in (for example) the mounting of the Purifi. I did not experiment with tightness of the mounting screws.

If I zoom in more closely, there's some raggedness at the minimum between the two impedance peaks.

As noted in a prior post, the enclosure is inexpensive big box store 3/4" plywood, with a double-thickness front baffle. The interior perimeter (sides, top, bottom) are lined with 3/4" recycled denim batting. The sides and top/bottom were glued with biscuit joints and polyurethane. Front and rear baffles are screwed on (epoxied brass inserts and machine screws) and sealed with foam tape (replacing the hollow EDPM weatherstripping that I used initially).

There is, at this point, no interior reinforcement. I'm a bit surprised by the absence of more obvious resonances. I'll be grateful for opinions from those with more experience.

 

[Alan J]

A quick look at enclosure tuning relative to predictions:

I haven't attempted a measurement of actual volume with the drivers taken into account. By eye, the volumes taken up by the drivers might roughly equal the volume contributed by the baffle openings, especially considering the flared opening for the Purifi in the double-thickness front baffle. Nor have I tried to account for the denim batting.

At the nominal 12 L, measured impedance (green) is fairly close to predictions (black). The box Q values need to be set pretty high to get the widths of the peaks to approximately match. The frequencies of the peaks match almost perfectly if setting the volume to 14 L.

 

[fluid]

The box is small, and the double thickness baffle makes quite a difference. Some reasonable damping in the interior helps a lot.
Comparing free air impedance of the woofer to the in box measurement can help to show you what comes from where. The Purifi driver does have a basket resonance at around the frequency of the higher wiggle.

It can be tricky to get simulated impedance to match real life, there is quite a few parameters that can have an influence. For example the qms of the passive radiator changes the lower peak quite considerably. Changing the Qa value to try and model the damping that is in the cabinet can help too, but ultimately it is really not something to worry about too much.

Measure the woofer and passive radiator nearfield and combine them with some baffle diffraction to see what you got and if you are happy with it. Try it in the type of osition it is likely to be used in and see if the bass sounds OK or too boomy etc.

 

[Alan J]

Continuing the initial look at tuning, these are quicky obtained nearfield measurements. The SPL scale is uncalibrated. The Xmax of the Purifi made me nervous, so both Purifi and PR were measured at 15 mm from the dust cap. This seems to me to be a pretty good correspondence between predicted and measured. VCAD's merger tool export button doesn't show in the absence of a farfield measurement, so I don't have the curves on a single plot.

I assume the signal from the PR above 150 Hz to be leakage from the back of the woofer. 20 dB down at the worst point seems OK. [EDIT} At this enclosure depth, it gives rise to shows as 1 dB loss in summed SPL, unfortunately at a spot where there's already a dip in the Purifi response. (it gives rise to a 1 dB contribution to total SPL at that point). The drop-out around 5 KHz is around the point where the microphone distance is 1/4 wavelength.

[EDIT - Nearfield FR shown below with no baffle step correction.]

 

[Alan J]

Thanksgiving guests are long gone and I'm back to evaluation of the prototype enclosure.

First, some near field bass tuning data...

Red is the baseline nominal 12L enclosure. Compensated for cutouts, drivers, and PR, estimated actual volume before installation of the denim batting is 11.7L.
Purple is estimated 10.2 L, obtained by placing wooden blocks inside.
Green is also 10.2L, but with 22 grams added to the PR.

Caveat: SPLs were uncalibrated and not the same for the three runs (my setup is susceptible to moving a level control; I failed to leave it in one place and didn't repeat cal). These are aligned at 200 Hz, which seems sensible but isn't the same as repeat measurements with calibrated SPL.

Consistent with the TSP-based model, low-end response with the baseline 12L enclosure rises slightly at the low end. With reduced volume, this is a bit more pronounced. Adding mass eliminates the rise and extends the low end. fB for both the 11.7L enclosure and the 10.2 with added PR mass is in the high 30's.

I'll be making a decision as to whether the reduce the enclosure volume a bit, or leaving plenty of volume to spare for the crossover, for any need to add volume due to crossover series resistance, or any needed reinforcement to deal with any resonances in the final enclosure.

 

[Alan J]

And some FR measurements...

In short, my measurements with my baseline (approx 11.7 L) enclosure with facets on either side of the DXT are very similar to those obtained for Directiva r1 (I'm using the responses from the v20 crossover VCAD project file).

Compared with r1 (dashed trace), axial response (solid trace) shows a bit less bass extension, and some upward tilt as we approach the low-end rolloff, both consistent with the somewhat smaller enclosure. As indicated in the last post, I do expect to get some bass extension back by tuning of the PR.

I'm not sure what to make of the high-Q dips at 4500 and 7400 Hz, the latter of which is also present in the r1 response.

Directivity of the unfiltered Purifi is similar as well. With the notable exception of those high-Q dips, vertical directivity in particular might be a bit improved.

For the DXT, on-axis response is, again, similar:

Directivity of the unfiltered DXT is pretty similar to what was obtained for r1. The only notable - but probably not very significant - difference I'm seeing is the undulation in the horizontal ERDI in the 2.5-4 KHz range going in the opposite direction.

Overall, I'm not seeing the faceted baffle providing an improvement over the rectangular face of the Denovo enclosure r1. But we (the eventual owners and I) like the aesthetics. The plan now is to cut facets below the Purifi. We expect this to be aesthetically preferable and to result in relatively subtle changes in response and directivity.

 

[Alan J]

Meanwhile, a quick look at crossovers... As mentioned above, we'll need to know how much volume to reserve for the passive crossover (if it's to be inside the enclosure). And the DC resistance on the Purifi side (which is higher if using inductors that take up less volume) affects the bass alignment.

I'm prototyping with DSP, both to be able to listen and also to sanity-check approaches. I'm starting by sticking to DSP functions that are nominally available with passive components, which means no delays and no boost (notch filters OK, peak filters not OK).

Here's an example that looks promising. Note the two blocks shorted out that didn't seem worth the component count - especially the notch at 725 Hz that would require a very big inductor. The plan here is, using DSP, to

1. measure (validate the model)
2. listen (be sure we're happy!)
3. design the passive crossover starting with an approximation of the DSP
4. decide on crossover placement and finalize the enclosure volume

 

[Rick Sykora]

And some FR measurements...

In short, my measurements with my baseline (approx 11.7 L) enclosure with facets on either side of the DXT are very similar to those obtained for Directiva r1 (I'm using the responses from the v20 crossover VCAD project file).

Compared with r1 (dashed trace), axial response (solid trace) shows a bit less bass extension, and some upward tilt as we approach the low-end rolloff, both consistent with the somewhat smaller enclosure. As indicated in the last post, I do expect to get some bass extension back by tuning of the PR.
View attachment 331718
I'm not sure what to make of the high-Q dips at 4500 and 7400 Hz, the latter of which is also present in the r1 response.

Not sure I recall these dips either. In any case, they are likely harmless - either measurement artifacts or inaudible in any case.

Please forgive my late comments. For some reason, the forum alerts are not working for me on this thread.

Directivity of the unfiltered Purifi is similar as well. With the notable exception of those high-Q dips, vertical directivity in particular might be a bit improved.

View attachment 331721

For the DXT, on-axis response is, again, similar:
View attachment 331724
Directivity of the unfiltered DXT is pretty similar to what was obtained for r1. The only notable - but probably not very significant - difference I'm seeing is the undulation in the horizontal ERDI in the 2.5-4 KHz range going in the opposite direction.
View attachment 331728

Overall, I'm not seeing the faceted baffle providing an improvement over the rectangular face of the Denovo enclosure r1. But we (the eventual owners and I) like the aesthetics. The plan now is to cut facets below the Purifi. We expect this to be aesthetically preferable and to result in relatively subtle changes in response and directivity.

The faceted baffle was at best marginal for r1. As one of the goals was build simplicity, decided to forgo.

Please post the VCAD files for your speaker (when you are ready). Notably sooner if you want others to suggest potential optimizations. As you likely know, the directivity could be more consistent. Was wondering if your voicing played a role in this or not?

 

[Colonel7]

Meanwhile, a quick look at crossovers... As mentioned above, we'll need to know how much volume to reserve for the passive crossover (if it's to be inside the enclosure). And the DC resistance on the Purifi side (which is higher if using inductors that take up less volume) affects the bass alignment.

I'm prototyping with DSP, both to be able to listen and also to sanity-check approaches. I'm starting by sticking to DSP functions that are nominally available with passive components, which means no delays and no boost (notch filters OK, peak filters not OK).

Here's an example that looks promising. Note the two blocks shorted out that didn't seem worth the component count - especially the notch at 725 Hz that would require a very big inductor. The plan here is, using DSP, to

1. measure (validate the model)
2. listen (be sure we're happy!)
3. design the passive crossover starting with an approximation of the DSP
4. decide on crossover placement and finalize the enclosure volume

View attachment 331732

What DSP software are you using? Here’s a way to do it with Equalizer APO

 

[Alan J]

What DSP software are you using? Here’s a way to do it with Equalizer APO

I'm using the little AD1701 board from Tinysine. After a few hours with SigmaStudio it became easy enough to work with. You can do things like have one of the on-board pots function as a selector for A/B (or A/B/C) testing. As noted, the goal in this case is just to listen enough to sanity-check what I'm doing in VCAD. Then, I hope, I'll have decent confidence in the in silico passive crossover design.

 

[Alan J]

Please post the VCAD files for your speaker (when you are ready). Notably sooner if you want others to suggest potential optimizations. As you likely know, the directivity could be more consistent.

Here it is....

Was wondering if your voicing played a role in this or not?

Nope. In fact, I'm suspicious of the subtle sag below 300 Hz and of it being a teeny bit hot in and around the crossover region, especially in the in-room response. TBH, this version was the result of optimizing for preference score, which is perhaps a silly thing to do .

 

[Rick Sykora]

After maybe a dozen hours of mostly casual listening, Idecided to take an impedance trace to see what I can discern about the enclosure and to get an initial assessment of how the bass tuning compares with TSP-based predictions in VCAD.

The Purifi trace...


...looks pretty clean. There's a little bump around 55 Hz, and some wiggles (very hard to see at this scale) around 300-500 Hz. Zooming in...

View attachment 323480
...the disturbance around 55 Hz remains really subtle. The wiggles in the 300-600 Hz region are more visible but pretty subtle. Those were more prominent when the box was sitting on a somewhat wobbly stand, got still smaller when moved to a stable but lightweight desk, and still smaller when the box was placed on a thick pad. This suggests to me that movement of the enclosure on the stand is a factor here, though of course movement on the stand could be interacting with something in (for example) the mounting of the Purifi. I did not experiment with tightness of the mounting screws…

… and the 380 Hz is likely a known resonance in the Purifi driver frame. It showed in Amir’s Directiva measures and we thought it was his test setup. Purifi has supposedly improved but has stated it is common problem in strong motor drivers. If you want to reduce, you need bolts and nuts to mount the woofer. I did improve with some very long screws, but only worked as the left and right holes extended into a brace. Otherwise, you will likely strip the holes trying to tighten enough.

 

[fluid]

Here it is....

I think the rabbit must have escaped from the hat