Directiva r1.2 design and build

[Alan J]

Progress! On carefully re-taking on-axis measurements, measured response with the current crossover is very close to predictions. We're up about a dB at the top.

And here's the six pack from a full data collection - every 10 degrees all the way around, plus every 5 degrees between -20 and +20 vertical. For the Preference Rating fans out there, it's 7.09. On-axis and LW slopes are a bit high (slightly positive instead of slightly negative), and sound power is declining at just under 1 dB/octave instead of the textbook 1.2.

Overall, aside from being a dB or so hot at the high end, everything adheres closely to the V.CAD predictions that are reproduced below:

Note that the crossover region lobe isn't tilted upward in the as-built speaker as much as in the model. This suggests some phase error between Purifi and DXT data. Around 15-20 us would explain the difference. That's a lot - around 5-8 mm error in microphone placement - and I thought I'd done better than that! If I was that far off, V.CAD prediction, other than the more forward-facing lobe, is a slight improvement to that 1-3 KHz sound power dip.

While I wanted to wrap this up, it makes sense to bump up the treble slope resistor and re-test.

 

[Alan J]

[EDITS for completeness in this post - added predicted - measured comparison and updated comments on slopes]

Here's what we get by bumping R3 (tweeter slope) up an ohm. On axis, the measured response now matches the modeled response:

Comparing slopes (6 pack below) with Full Space targets...

Curve	Textbook dB/oct	Actual dB/oct
ON	-0.04	-0.1
LW	-0.12	-0.18
PIR	-0.69	-0.22
SP	-1.2	-1.22

In the design, I focused a lot on LW and SP, both of which are pretty close. I could fuss with this more, though I think we're down to rapidly diminishing returns on further effort. Experience with attempts to correct the small sag in SP indicates that this would result in an unacceptable rise in direct sound on axis and in the listening window in that region. This is a matter, I think, of the dispersion characteristics of the drivers and - even more, perhaps - the driver separation and phase relationship in the crossover region: where there is notable lobing but not an elevated response on axis, there is, overall, less total power.

FWIW, Preference score for this limited frequency range is 7.86.

I'll check the impedance to be sure seals are tight (the PR sheet metal rim requires mounting screws to be carefully torqued), re-take near field, and merge so that we can have a full range response.

 

[Alan J]

And here's what we get with near field merged. Preference score with VituixCAD full space defaults (I know it's not the end-all but it's so hard to not look) is 8.1 (OK, 8.096), which is surprising considering that smoothness doesn't look spectacular. Reference axis linearity is 2.7 dB.

 

[Alan J]

I'm wrapping up this project at last, just in time to deliver finished speakers on Thanksgiving day.

Here's the final whole-speaker 6 pack - same as above with just a bit more care in aligning near and far field.

 

[Alan J]

As I close this project out, just a few lessons, observations, and words of thanks...

Lessons and observations...

• Be diligent and stay organized. Double-check setup and settings before every measurement. Use the VituixCAD project file structure. Label files carefully. Don't rush. One of my major misadventures here involved somehow choosing the wrong cal file for a microphone.
• Don't get ahead of yourself. I spent a lot of time playing with crossover designs prior to having driver data with my actual baffle. This was instructive, I suppose, and it's a bit addicting, but not great use of time.
• Assume that components of your test system are imperfect, and eliminate sources of error whenever possible. My other major misadventure involved assuming that an amp that reviewed well wouldn't introduce significant errors - even as Amir's test showed load dependence at high frequencies. Moving my loopback to the amp output eliminated the amp as a source of error.
• Templates and tools make a difference. Since I have a 3D printer, in addition to making jigs for cutting facets, routing openings, drilling terminal and mounting holes, etc., etc., I was also able to make tools like end caps for a meter stick to mate with the baffle on one end and my measurement mic on the other.
• I did the crossover design using just far field responses. It would, in this case,have been better to merge nearfield responses with the woofer far field response so that the whole spectrum was visible.
• Seek simplicity and modest parts count in crossover design.

And some words of thanks...

This is a really remarkable online community. Everyone has been helpful, constructive, and kind. Thanks to @RickS and the Directiva team for what was the starting point for this design. This was my first almost start-to-finish design (the Directiva team selected the drivers) and it would not have been possible without the repository of knowledge and communal support that I found here. I'm grateful as well to the people at Purifi and Seas, as well as Madisound and PartsExpress who provide components, distribution, and support for the DIY market.

 

[kemmler3D]

The FR, directivity, and speaker itself look fantastic - congrats on a successful project. A DIY build that needs make no excuses.

I may have missed it (didn't follow the whole thread) but how does the distortion look from the final product? I know that's one of the major selling points of these Purifi drivers...

 

[Alan J]

I didn't do stepped sine, unfortunately. I'm not sure how to properly show distortion from regular measurement sweeps at 1m, but here's what I get from REW, on axis at around 85 dB SPL. The frequency response is valid down to 200 Hz. Distortion is said to be fairly accurate across the whole spectrum, though not as good as with stepped sine.

 

[a4eaudio]

A comment on showing distortion measurement:

• Change the presentation to dBr with a scale something like 0 to -90dB or -100dB.
• Add HD3, HD4 and HD5 to the chart. The odd-order distortions (HD3 and HD5) are likely to be the most audible.
• IF you want to get the noise floor down, measure at 31.5cm. If you record the SPL at 31.5cm just deduct 10db for the 1 meter comparison (i.e., 96dB at 31.5cm is equivalent to 86dB at 1 meter, but being much closer to the speaker keeps the noise floor lower). If you don't want to take new measurements just for distortion purposes, I'd just hide the noise floor and show at HD3 using your data above (HD5 is probably going to be very low anyways).

 

[Alan J]

Thanks to @a4eaudio for the guidance. Here's a plot from the same ~85 dB sweep, as dBr and 2nd through 5th harmonics.

To the degree it's useful, here's the same plot with traces masked below the noise floor.

 

[a4eaudio]

...To the degree it's useful, here's the same plot with traces masked below the noise floor...

That is quite impressively low distortion!

By the way, I don't know how REW does the distortion below the noise floor, but I have compared measurements like yours at 1 meter (where the distortion was below the noise floor and I assumed it may not be very accurate) to the same measurement at 31.5cm where the noise floor is lower. The "masked" portion turned out to be very accurate when I compared it to the closer measurement and the same distortion when it did not need to be masked..

 

[kemmler3D]

Thanks to @a4eaudio for the guidance. Here's a plot from the same ~85 dB sweep, as dBr and 2nd through 5th harmonics.

View attachment 492419
To the degree it's useful, here's the same plot with traces masked below the noise floor.View attachment 492420

Distortion 50dB down for a small 2-way is pretty damn good in my opinion, especially with odd harmonics being even lower. I bet these sound like a million bucks. Nice work all around.

 

[Alan J]

Delivered on Thanksgiving day. They do sound wonderful. Tears were shed. We didn't have time to take any in-room measurements or to experiment with placement.

 

[kemmler3D]

Delivered on Thanksgiving day. They do sound wonderful. Tears were shed. We didn't have time to take any in-room measurements or to experiment with placement.View attachment 493816

They look great in situ. Looks like mission accomplished. A great example of how and why to do DIY.

 

[Paweł L]

Delivered on Thanksgiving day. They do sound wonderful. Tears were shed. We didn't have time to take any in-room measurements or to experiment with placement.View attachment 493816

Alan any updates on the project: listening impressions (I assume this is not permanent placement for the set ) , measurements, etc. And did you try them with different amplification. These are indeed sexy pair.

 

[Alan J]

This is the placement until my son and family rearrange their apartment or move! Placement on the countertop was intended from the start, which is why I wasn't concerned about the couple of dB drop in the full-space bass response below 350 Hz. The room is about 15' wide and 12' deep, with double French door openings along one side and the back that are normally kept open. I think the placement could be a bit wider. They're being driven by a Sonos amp (per Amir's review, unimpressive but OK).

They sound stellar. They fill the room beautifully, as we hoped (attention in the crossover design on linear Sound Power). For me it's common to walk through the room when music is playing and feel drawn to sit and listen. I think that the lobing in the crossover region is detectable when going from sitting to standing within 6' or so, but it's not obvious and no objectionable, and I think that the vertical nulls and horizontal dispersion came out well for the room. I suspect that the low distortion of the Purifi mid-woofer has a lot to do with the listening experience.

 

[Paweł L]

Yeah, the 6.5" cone and 1" will always have a suck out in x-o area, can't beat the physics here, but good distortions numbers in low freq resp. I thought you will have them sitting on nice metal stands, meter away from the front wall in acoustically treated room - I get it, life is tough, but visually this are very pretty spots, do most likely no complaints from your significant other. I was looking at some amps modules on AliE, NC252 or Danish Pascal modules look tempting (I'm getting tired from passive X-overs, just not enough flexibility, expensive, if you like perfection).