• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Directiva r1.2 design and build

Alan J

Member
Joined
Jul 14, 2020
Messages
66
Likes
94
Having learned a ton from reading and re-reading the Directiva threads, and having my own slightly different goals, I'm posting here on my design/modification and (I hope) build process. Per the usual give-and-take, I hope this will be useful to some folks, and of course I'll be grateful for comments and guidance.

At the outset, thanks go to @Rick Sykora, @ctrl, and others for all of your work and for all that you've shared.

Primary departures from the r1 priorities are
  • Loosened (discarded?) the relatively easy to build constraint. I'm not relying on a flat pack. See below for more challenging-but-doable changes in the woodworking category.
  • Intent to push the volume down a bit.
  • Planning for a passive crossover.
All three of these derive from the overarching goal which is an heirloom gift to one of my children and his family. So the design has to fit their needs today while being something that remains "special" over the years.

The plan is for a faceted enclosure. In this rendering (please excuse random drivers where the Purifi and DXT go - does anyone have models?), baffle width at the bottom and driver placements are identical to r1. The height is a bit shorter than the Denovo flat pack: 13.75" instead of 16. Facets are cut back 30 degrees from the face as viewed from the top, along the tangent line that clears both drivers by 3 mm.

Directiva_minimal_faceted_2023-Aug-31_01-40-08PM-000_CustomizedView12321567161.png


Per the rendering, the intent is that the cabinet is solid wood (white oak, most likely), so special steps to deaden the enclosure might be necessary.

The PR goes on the back as in the r1. It'll probably be the SB Racetrack, but we shall see...
 
Enclosure alignment with reduced volume

The Directiva r1 baseline design in the Denovo box has an enclosure volume of around 14 liters. It uses the SB “racetrack” passive radiator with no added mass. With the intention of lowering the volume, we need to revisit the alignment.

Here’s the original, more or less :). For Qa (stuffing) and Ql (leakage), I’ve used what @ctrl found to match up with @Rick Sykora’s data (#323):

1695411390302.jpeg

Note that we’re driving @ 30W into 4 ohms (10.5V). That’s because we’ll be looking carefully at where distortion or (perish the thought) over-excursion damage could occur. At this baseline, the Purifi cone is hitting its linear excursion limit at 25.4 Hz and its mechanical limit at around 17 Hz (though of course the curve isn’t accurate in the nonlinear range).

There is a huge caveat here: The TSP-based model doesn’t apply at large (let alone very loud) signal levels. And it certainly doesn’t apply as drivers get outside of their linear ranges for excursion. The intent here, since we’re modifying a baseline design, is to discern trends and try to stay out of trouble with the changes we’re making.

Let’s save the baseline as an overlay, and start making changes. First, reduce the volume. Since we’re anticipating a passive crossover, I’ve also allowed (Rg) for some series resistance.

1695411447402.jpeg


The little rise below 100 Hz might be nice, considering that the driver sags a bit in that region. In any event, however, a bit of added mass on the PR brings us back closer to where we started with regard to bass extension:

1695411487479.jpeg

At this still-elevated power/SPL (and still remembering that we're looking at trends, not at numbers that we expect to hold up), the Purifi is hitting its linear excursion limit at about the same point, and that we've lost a tiny amount of headroom with regard to the mechanical limit.

We can keep going with added mass on the PR…

1695411824766.jpeg

As we add mass and push the box tuning down into the high 30's, the Purifi seems to be in safer territory with regard to music content (30 Hz and up) but the PR might be closer to getting into trouble.

This leads to the obvious question of whether a different PR would help, noting that aesthetically, we don't want PRs on the sides and we don't have room (or budget, really) for dual Purifi PRs on the back.
 

Attachments

  • 1695411447331.jpeg
    1695411447331.jpeg
    684.2 KB · Views: 100
Purifi -NAA (aluminum) driver

But before we look at PRs... Sticking with the SB Racetrack PR + 15 grams, let's look at the Purifi -NAA (aluminum cone) driver:

1695414735989.jpeg


A bit more mass gets us a flatter predicted response (less underdamped transient response? I'm not sure!):

1695415883434.jpeg


This looks promising, I think :), though I'm nervous about the higher PR excursions down near the bottom. Should I be, compared with where the baseline Directiva r1 landed?

And what about Purifi's recommended alignments for the -NFA and -NAA drivers? For the -NFA Purifi lists a typical Vb of 10.8 liters - check. But for the -NAA, with Purifi's PRs, it's 14.8. Are those "typicals" relevant here? I don't think so, but I'm not certain, so input welcomed!
 
Alternative PR’s

The SB Racetrack PR has a Vd of 196 cm^2, about 1.5 times that of the Purifi driver and less than the typical guideline that the PR Vd should be about twice that of the driver. Our goal here is to determine whether the PR presents a meaningful constraint, and, if so, whether an alternative exists that would fit our design and present less of a constraint. The answer so far from Directiva r1 experience seems to be that no - the PR is basically fine. But we've changed the volume and provisionally moved to the aluminum cone.

There are a few ways of looking at the question of “meaningful constraint.” And however we look at it, it depends upon the content. When we’re playing music, we might count upon there being little or no content below around 30 Hz (assuming a rumble filter if playing vinyl, as my customer in this case likes to do). With cinema content, we need to at least accommodate lower frequencies, without distortion or, at least, safely. Indeed, my customer has HDMI from the TV running to the same Sonos amp that will drive the speakers, and since we will not be using the Sonos subwoofer settings I don't know if there's a way to prevent LFE content from passing through.

That leads to three ways of evaluating a PR (or driver, alignment, and PR combination):
  • SPL at which neither the driver cone nor the PR exceed their linear Xmax, at any frequency > 5Hz. This is the “clean cinema” level.
  • SPL at which neither the driver cone nor the PR exceed their mechanical limits, at any frequency > 5 Hz, AND the driver cone remains within its linear range for all frequencies above 25 Hz. This is the “clean music, cinema-safe” level.
  • SPL at which neither the driver cone nor the PR exceed their linear Xmax, at any frequency > 25 Hz. This is the musical level. For simplicity, we’ll use a HP filter in VCAD here (20 Hz, 2nd order).
Here’s our system at 11 L volume with the Purifi aluminum cone and SB racetrack PR with 80 grams added (a bit flatter than before). I’ve left the half ohm allowance for a passive crossover. Ql is bumped down to 7 which perhaps is more realistic.

The level is set to the "clean cinema" max: The cone maxes out its linear range with cinema content, i.e., low-frequency effects all the way down to 5 Hz. SPL is 94 dB (at 200 Hz). The Purifi cone is the limiting factor: At its maximum extension, the PR has 3.9 mm left to go (35% of its full travel).

1695585654651.jpeg


Bump down to the "clean music, cinema-safe" max. Driver cone movement is still the limiting factor, though the PR is coming close to its Xmax as well:

1695585777629.jpeg


And finally, "clean cinema" max, where both driver and PR are within linear limits. Again, the two hit their limits at almost the same level:

1695585909904.jpeg


Wash, rinse, repeat, for different PR possibilities, and we get the following. In each case, we're "cone-limited," though at the higher levels the SB Racetrack PR is getting close - it's headroom is less than 10% of available mechanical range.

PRAdded massVbf3Clean cinema SPLlimited byheadroomClean music, cinema- safe SPLlimited byheadroomClean music SPLlimited byheadroom
SB SB15SFCR-00801146.994 dBCone35%97 dBCone7%100Cone2%
Purifi PTT6.5PR-NA-02701146.2 Hz94 dBCone63%97 dBCone46%99Cone46%
Dayton RSS210-PR801145.6 Hz94 dBCone67%97 dBCone53%99Cone53%
Dayton DSA-215PR1201145.694 dBCone41%98 dBCone10%99Cone17%

The results aren't surprising: The box is tuned about the same in each case, so the driver cone and PR share the burden similarly. The driver cone runs out of room first, as intended. It's then a question of whether we have enough headroom in the PR movement to be sure it's not limiting.

If we would measure the "clean music" level by trusting that the source material has nothing below 25 Hz rather than by applying the HF filter, we could eek out a couple of dB for music with either of the Dayton or the Purifi PR's, but not with the SB.

Of course, there's no way that this simulation, using the small signal TSP model with large signals, is that precise! If we remain uncertain, either of the Dayton PR's are possibilities. But an 8" PR would require a cabinet width of at least 10".
 
Last edited:
Having learned a ton from reading and re-reading the Directiva threads, and having my own slightly different goals, I'm posting here on my design/modification and (I hope) build process. Per the usual give-and-take, I hope this will be useful to some folks, and of course I'll be grateful for comments and guidance.

At the outset, thanks go to @Rick Sykora, @ctrl, and others for all of your work and for all that you've shared.

Primary departures from the r1 priorities are
  • Loosened (discarded?) the relatively easy to build constraint. I'm not relying on a flat pack. See below for more challenging-but-doable changes in the woodworking category.
  • Intent to push the volume down a bit.
  • Planning for a passive crossover.
All three of these derive from the overarching goal which is an heirloom gift to one of my children and his family. So the design has to fit their needs today while being something that remains "special" over the years.

The plan is for a faceted enclosure. In this rendering (please excuse random drivers where the Purifi and DXT go - does anyone have models?), baffle width at the bottom and driver placements are identical to r1. The height is a bit shorter than the Denovo flat pack: 13.75" instead of 16. Facets are cut back 30 degrees from the face as viewed from the top, along the tangent line that clears both drivers by 3 mm.

View attachment 313895

Per the rendering, the intent is that the cabinet is solid wood (white oak, most likely), so special steps to deaden the enclosure might be necessary.

The PR goes on the back as in the r1. It'll probably be the SB Racetrack, but we shall see...
Cool project. I’ve been waiting for the R2 to move along. Are you a woodworking enthusiast? How thick is the baffle going to be with solid wood and 30 degree chamfers? Solid wood can be a polarizing subject due to the chance of splitting. Never tried it but I’d probably just go for veneer to be safe bc I’d be afraid for the risk. White oak should look great. Looking forward to seeing your choices and progress.
 
Except for the newer woofer, recall this is closer to the design that @ctrl preferred.

Look forward to seeing your results!:)
 
Are you a woodworking enthusiast?
I'm enthusiastic about that piece of the project. If you're building your own cabinets, and doing electroacoustic design, and maybe some electronics, speaker building is truly mixed media.

How thick is the baffle going to be with solid wood and 30 degree chamfers? Solid wood can be a polarizing subject due to the chance of splitting.
In the rendering in Post #1, the baffle needs to be at least 38 mm thick. That's with the chamfers cut at 30 degrees.

I'm not too worried about integrity of solid wood stock for boxes this small, especially if I pay attention to grain direction. I do think that joinery will require some thought, and I'll need to consider interior treatment or CLD. In describing a desired aesthetic, my son used the term "honest," which is really interesting. What it looks like should reflect what it is. So if it's BB plywood, expose the edges (which doesn't mean that BB with big facets is going to be the right choice here). And if it presents as solid wood, the grain on the facets should be consistent with that.

All of this is relevant to the early stages. My first thought was to prototype in MDF or void-free plywood. But I might consider prototyping with oak from the big boz stores so that I can evaluate for resonance at the same time.
 
Last edited:
I want to settle the PR question and move on. Because reasonable alternatives are 8" PRs, the change would require a change in the box dimensions. Before we even get to the diffraction questions, and possibly some additional technical questions, let's look at aesthetics.

On the left: With room for an 8" PR on the back the baffle is now 10.25" x 15" and the box is 8.4" deep. With the larger width, the facets are now cut back at 20 degrees.

On the right: The original starting point has a 8.5"x13.7" baffle and is 10.8" deep.
Comparison.png
 
In this rendering (please excuse random drivers where the Purifi and DXT go - does anyone have models?)
Purifi has step files of all their models on their website. Scanspeak does too.
 

Attachments

  • Baffle Surface Ortho.png
    Baffle Surface Ortho.png
    2.2 MB · Views: 227
I want to settle the PR question and move on. Because reasonable alternatives are 8" PRs, the change would require a change in the box dimensions. Before we even get to the diffraction questions, and possibly some additional technical questions, let's look at aesthetics.

On the left: With room for an 8" PR on the back the baffle is now 10.25" x 15" and the box is 8.4" deep. With the larger width, the facets are now cut back at 20 degrees.

On the right: The original starting point has a 8.5"x13.7" baffle and is 10.8" deep.
View attachment 314722

The larger one has a more massive look to it imo. My wife would veto but she does not like bevels at all...

For a bookshelf speaker, my approach has been to keep the size more in line with a large book. For the active r1, I pushed the bass design. Aside from wanting solid bass from it, the practical side is that a second speaker will add more output as will the room. This is even more true if the design actually sits on a bookshelf as I recall you intend. For most other use cases, adding a subwoofer is the best approach for addressing lower and more bass output.

For r1 out in stands in my room, the bass is very solid. I would only use a subwoofer to improve the lowest octave. From my reading across ASR, adding a subwoofer is often the most utilized approach for many members.
 
Last edited:
  • Like
Reactions: pma
Purifi has step files of all their models on their website.
Thanks. Don't know how I missed it :facepalm:. 3DWarehouse (Sketchup only?) has the DXT but login/download doesn't seem to work...
Directiva_minimal_faceted_2023-Sep-28_05-37-34PM-000_CustomizedView47852796462.png
 
OK - some decisions...
  1. The wide/shallow enclosure to accommodate an 8" PR on the back is rejected for aesthetic reasons.
  2. After a dive into bass alignments, we're sticking with the paper/fiber driver that was used in r1. Using the approach of starting with a textbook vented alignment and substituting a PR for a vent of equal area, the -NAA calls for a larger box volume, especially when factoring in the of effect of series resistance on Qt. I'm comfortable with reducing the volume relative to r1, but became uncomfortable with doing so when also changing to the aluminum cone.
 
An update on the enclosure design, likely to be followed by a period of silence as I build the first prototype enclosure.

The approach I'm taking here, with little effort at simulation, is to
  • minimize the baffle around the DXT tweeter
  • maintain minimal woofer-tweeter spacing
  • avoid complicated cuts and angles between the woofer and tweeter - so no faceting between woofer and tweeter, and no attempts at on-axis alignment of effective acoustic centers
The result is a front baffle surface reminiscent of the angled side enclosure that was tested during Directiva r1 development. That enclosure was abandoned due to a resonance issue. The hope here is that the 30 degree faceting provides the intended benefits without introducing other issues.

In the rendering below, I've moved the woofer and tweeter down and added a top facet. In its intended location, the speaker will sit on a low console and if pushed back away from the edge we'll get some cancellation from the reflected wave. Moving the woofer down won't fix it, but it pushes the frequency further out into the crossover range where the tweeter is contributing.

This is my most likely first MDF or plywood prototype.

Directiva_minimal_faceted_2023-Oct-04_03-26-18PM-000_CustomizedView21584320366.png
 
maintain minimal woofer-tweeter spacing
It is always a good idea to check on the centre to centre spacing with the intended crossover frequency. Sometimes pushing the drivers as close together as possible can be counterproductive. Ratios that have the spacing at the crossover frequency in the 0.5 to 0.7 wavelength range produce the least desirable directivity in general. Those spacings can be used to create specific directivities that might be useful in some designs but they are not 2 way bookshelves.

Following the accepted wisdom of pushing the drivers close together and lowering the crossover frequency might not give the desired response.

Perhaps you prefer the look and are willing to accept the compromise, something to think about before cutting the wood.
 
Thanks so much for raising this! The current as-drawn spacing is 145 mm, same as r1 if I've correctly copied the dimensions from the r1 build thread. Right now, I'm figuring on crossing between 2000 and 2500 Hz - so wavelengths in the 137-171 range. If at the high end (see below), that puts the spacing at 0.85 wavelengths.

Looking at some r1 results, here's what was posted as v12c of the active crossover:
Screenshot 2023-10-05 at 9.34.37 AM.png


...and the polar plot at the crossover frequency:

1696514270493.png


For the passive crossover anticipated here, unless I use unequal slopes to try to bring the Purifi and DXT phases together in the crossover region, the LP cutoff will be higher than the HP. For example, here's a crossover posted by @TimVG for r1, with some further optimization in VCAD (flaws are, therefore, owned by me):

Screenshot 2023-10-05 at 9.52.11 AM.png

As noted, the Purifi filter is close to LR4 at 2600 Hz and the DXT close to LR4 at 1600 (the original posed crossover was less extreme - around 2300 and 1700). Here's the polar plot at the point where we find the deepest above-axis hole:

1696514689300.png


I'd be willing to put in some simulation effort. Can you suggest the best tools? I think that in VCAD, I'd need to (1) find appropriate bare driver responses, (2) add in an estimate of diffraction for each driver at a test position, and (3) apply a candidate crossover. Do I have that right?
 

Attachments

  • 1696514623150.png
    1696514623150.png
    51.9 KB · Views: 63
  • Screenshot 2023-10-05 at 9.34.37 AM.png
    Screenshot 2023-10-05 at 9.34.37 AM.png
    567.3 KB · Views: 59
  • 1696512964838.png
    1696512964838.png
    51.9 KB · Views: 78
Last edited:
I'd be willing to put in some simulation effort. Can you suggest the best tools? I think that in VCAD, I'd need to (1) find appropriate bare driver responses, (2) add in an estimate of diffraction for each driver at a test position, and (3) apply a candidate crossover. Do I have that right?
Vituix is the right tool to use. You can use the Directiva data to see what the effects are, it won't be exactly the same but still in the ball park. On a small baffle the diffraction signature won't change enough to be too worried.

The Y value is what you need to change. 0 represents the listening axis, in the example above the tweeter is positive Y above the axis and woofer -Y below. The difference is the CTC.

This moves the null from the vertical spacing as you see in the polar charts. 1.2 wavelengths is a good number but it can look a bit odd depending on the crossover frequency. Anything from 1 to 1.4 is usually the best compromise In a two way it is difficult to make the mid woofer get high enough with directivity or break up issues so it is a balancing act.

You can smooth a directivity mismatch by putting a phase offset in the components, like everything it is a compromise.
 
Here are vertical directivity plots for different woofer-tweeter spacing. As @fluid suggested, these were done using r1 data, changing only the driver spacing. The listening position was kept midway between - though simply moving the woofer down is essentially the same when the listening position is 3 m away. As before, driver responses are what was posted by @ctrl to map Amir's NFS data to @Rick Sykora's representative sample of the Purifi woofer. The crossover in this case is the one that Rick built and posted in the r1 passive crossover thread. The Purifi response is very close to 4th order at 2300 Hz, and the DXT response very close to 5th order at the same frequency.

Coaxial (make believe) spacing

1696605931891.png


Approximately minimal - 145 mm spacing (for which the crossover was designed)

1696605995516.png


Expanded - 195 mm spacing

1696606065088.png


Keeping a fixed tweeter position at the original +72.5 mm and indexing the woofer position downward for greater spacing indicates that
  • as spacing increases, the below-axis "hole"forms
  • very narrow-band dips, like the one near the center of the below-axis "hole" in the last image, form at very specific spacings.
Using VCAD's optimization function, keeping the tweeter in its fixed position 72.5mm above-axis (original center), whether choosing 50/50 axial and in-room response, or preference score, the optimum spacing settles in on 151 mm. (Interesting side note: if allowing the tweeter to move, the optimization function will move the tweeter up as far as I allow it, which I think comes down to moving off of the tweeter axis.

Conclusion (?): The r1 spacing of 145 mm seems fine. Since we've gone through this exercise, may as well go with 151 (a bit more robust woodworking-wise?), even as the difference is indistinguishable.

Thanks to @fluid for recommending that I do this.
 
Last edited:
You're welcome. It's important to consider these things in their entire context even thought the most obvious way to see it is in the vertical polar plot. Have a look at the PIR and early reflection responses in combination with the DI, certain spacings will improve these or make them worse. Consider too if having the listening axis be between the drivers will work out in their intended position or on stands. 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. I am more convinced now that a design that has a significant difference between sitting and standing makes certain types of music sound wrong to me.
 
Back
Top Bottom