Directiva r1.2 design and build

[terryforsythe]

It was suggested before that this is attributable to a known resonance in the Purifi basket in that general area.

Perhaps. If you feel inclined, you could get an accelerometer and measure the enclosure vibrations to confirm it is not the enclosure. Also, you could try experimenting with internal bracing to see if it mitigates the resonance - maybe an H-shaped brace extending front to back internally with the back of the woofer up against the brace with some rubber between the two. If the woofer has a vented pole piece, you would need to have a cutout in the brace so as not to block the air flow.

Another question for those with more experience: My focus in steps 1-3 is on the far field data (>400 Hz or so). Is there any need to fuss with sealing the wire passage through the binding post holes?

Probably not. I doubt that is responsible for the 410Hz resonance. But, you could try sealing the opening around the wire with putty or something to see if that changes that resonance.

 

[Alan J]

Thanks, @terryforsythe. With the PR in the rear, I don't have room to brace front to back, but I might try a horizontal brace. I have some neoprene and I might try threaded steel rod like Alexander Heissmann does since that would be easy to insert and remove.

Regarding the sealing, I don't suspect it as a cause of the resonance (it was there in the well-sealed prototype). The question is whether it matters for data collection > 400 Hz for crossover design and I'm thinking that the answer is no.

 

[Lars Risbo]

It was suggested before that this is attributable to a known resonance in the Purifi basket in that general area.

there is no known resonances in the basket. However, if the basket is clamped hard and the motor is freely suspended then basket and motor forms a mass-spring oscillator. this goes for all drivers on the planet. This issue was eg describe by the late Linkwitz (see his web page). By supporting the motor and or making a softer clamp, this resonance can be avoided. Good system design practice use such techniques.

however, the blip might also be the spider resonance. typically around 400Hz for our 6.5” family.

 

[Alan J]

Thanks, @Lars Risbo ! I apologize for the mis-statement. I guess that formally it's a resonance - mass and spring - but neither a property of the basket itself nor avoidable without extreme measures to stiffen the basket (or imagined very low-mass magnets).

I've installed a simple horizontal brace. It's a tight pressure fit, with a dense 3 mm EDPM pad on one end and 2 mm neoprene (40 durometer) on the other. The top trace (purple) is unbraced and the lower trace (blue, separated in REW) is with the brace. There's little difference, so whatever the source this single brace has no effect.

I'm thinking that this is too small to worry about. The disturbance, at its max, is around 2.5% of the impedance at that frequency. Distortion traces from the prototype are at the noise floor (-60dB) at 80 dB SPL and run parallel the noise floor at SPL 85 and 90. There's no suggestion of any issues around 410 or harmonics/subharmonics of 410. So I am thinking that I can move on.

Opinions to the contrary are welcomed!

The blunted impedance peaks, by the way, were due more to poor seating of the PR than to the (still not sealed) test wire passages.

 

[terryforsythe]

With the PR in the rear, I don't have room to brace front to back, but I might try a horizontal brace.

Here is a very rough sketch to what I was referring. The brace is yellow.

 

[Alan J]

Got it. Thanks. I could fit something close to that (would need to be tapered at the Purifi end). I'm feeling doubtful regarding this little 410 Hz impedance blip.

Knock test gives a wide side panel peak at 622 Hz, top panel at around 1.1 KHz. To the degree that's a valid indicator of panel resonance, it's not in the neighborhood of 410.

Here's the distortion scan from the prototype at 85 dB, taken unfortunately in a pretty noisy environment...

 

[terryforsythe]

If the basket is ringing, I think the H-brace and rubber isolation pads as in my drawing will help dampen the resonance. I would look at the spectral decay (Waterfall in REW) and look to see if the resonance is ringing. If not, it probably is not a problem.

 

[Alan J]

Thanks. I think I can get set up in my test space tomorrow.

 

[Alan J]

I'm not sure whether I'm producing the waterfall correctly, or whether my test environment will support it. My first reflection is at around 5.1 ms. FWIW, here's the burst decay with 5.0 ms Hamming windows at the left and right (per the REW help file, the burst decay is computed using whatever window was applied to the measurement).

Also FWIW, here's the Fourier plot with the default 500 ms left and right windows, so it's for the speaker in the room with its 5.1 ms floor/wall reflection.

 

[terryforsythe]

I would try with the smoothing turned off. If the resonance at 419Hz does not show up, I think you are good.

 

[Alan J]

I would try with the smoothing turned off. If the resonance at 419Hz does not show up, I think you are good.

Driver and room with 1/48 octave smoothing (the least REW will do)...

Despite waking at 3 AM thinking about ways to construct a magnet-supporting brace, I think I will move on!

 

[Okankav]

Hi Alan, you can try this product inside the box vertically one 10*20 cm piece enough

KNAUF INSULATION MP Acousticboard KDB032​

İt worked in my case

 

[Okankav]

For bump around 400-500 hz region

 

[Alan J]

Hi Alan, you can try this product inside the box vertically one 10*20 cm piece enough

KNAUF INSULATION MP Acousticboard KDB032​

İt worked in my case

Thanks, @Okankav

 

[RickS]

Not likely worth your time. If no major distortion corresponds, it is unlikely audible. If you think it is a related to the mounting to the front baffle, can try different torque levels on the screws to see if it alters.

Are you using any internal damping material (notably between the woofer and pr)?

 

[Alan J]

Not likely worth your time. If no major distortion corresponds, it is unlikely audible. If you think it is a related to the mounting to the front baffle, can try different torque levels on the screws to see if it alters.

Are you using any internal damping material (notably between the woofer and pr)?

Nothing right now - just the 1" denim around the periphery. There are definitely various bumps in the impedance trace without it, larger as I recall than the little on at 410.

 

[Alan J]

EDIT - what's below was a misadventure. The idea was basically this: Is there a way to estimate the on-axis driver offsets without moving the microphone? What I tried, which was basically a geometric approach in which the mythical acoustic center is modeled as a fixed point. Consistent with many discussion of acoustic center, and whether the concept is even useful, that model fails. Calculated from the difference between Purifi and DXT IR delays, as seen from a fixed microphone at the Purifi axis - even with a geometric analysis a bit better than what I have below - the DXT is only around 4 mm in front of the Purifi. Measured from IR delay at the respective axes, with careful positioning of the microphone at a replicable distance from the baffle surface, the DXT is 16.7 mm in front of the Purifi. Per the VCAD files I have from Directiva, this is in agreement with what was found by the folks who collected the Directiva dersign data.

So - I'm set up for measurements and wondering what others' strategies have been for preserving phase relationships of the two drivers. I've failed to find a clear discussion of the topic in the forums.

I'm using a Dayton EMM-6 with a loopback timing reference. So as long as I don't move anything I have an independent reference. But I need to remove and remount the speaker to do "vertical" scans, and I think it's way better to capture each driver relative to its own axis, lest the data be valid only at one distance.

I'm wary of trying to move the microphone and get a sufficiently accurate distance with a tape measure, as a few mm can make a noticeable difference. My planned approach - but maybe there are better ones...

1. Set up on the Purifi axis and collect data. Get an estimated IR delay, relative to the loopback reference, for the on-axis (Hor 0) scan (but don't apply any changes yet).
2. Without moving anything, grab a sweep from the DXT and get an estimated IR delay. Because the driver is delay corresponds to the hypotenuse of the triangle formed by the on-axis distance to the DXT and the driver separation. Calculate what the on-axis delay would be.
3. Move to the DXT axis, making a best effort to replicate the distance. Take a scan and get an estimated IR delay, which should match the prediction from step 2. Adjust and repeat as necessary.
4. Set up on the Purifi axis for the vertical data. Adjust as needed to replicate the IR delay from step 1.
5. Set up on the DXT axis for the vertical data. Adjust as needed to replicate the IR delay from step 3.
6. Adjust all data using the DXT delay. Purifi data will have some remnant delay due to the depth of the driver.

This is a lot like what I did with the prototype. In that case, I separately corrected Purify and DXT data for microphone distance, then applied an adjustment to just the Purifi (or just the DXT) corresponding to the IR delay difference between the two.

Any better ways??? I've read kimmosoto's guide over and over and there is no mention of maintaining timing when re-aligning the microphone with driver axes. 10 degrees in the upper part of the crossover region is just a couple of mm.

 

[Alan J]

Ugh! Just blew out the DXT. I have no idea what happened but while checking the position on the stand there was a very loud pure tone then nothing. Regrettably, I had the DAC I use for testing (TASCAM US-1x2HR) set as the default output on the computer so I suppose it could have come from anywhere - though the volume was set near zero.

The DXT prices looks to be unchanged at Madisound, so that's good. Meanwhile, I'm proceeding with the other driver. And I calibrated the output voltage in REW with a true RMS meter, turned the amp gain way down, adjusted the DAC output to restore the original test level, and verified that a new scan falls right on top of the prior one. 0 dB is now 2.5 V and I'm testing at -10 dB which is around 0.79 V. So, I hope, no more accidents.

[EDIT] I removed the waveguide - diaphragm - voice coil assembly from the DXT. The voice coil wire had failed between the voice coil and the terminal and there was a corresponding melted area in the foam pad that sits between the front plate and the magnet assembly. I was able to repair the connection (pics available if anyone's interested) but the DC resistance is now lower than normal and the impedance trace is bumpy and has a blunted peak. So even as the voice coil looked OK, there's probably damage among the turns.

 

[Alan J]

The drivers on the finished baffle are about 15 mm lower than they were on the prototype. The DXT is no longer so close to the top edge of the enclosure, while the Purifi is closer to the bottom with a steeper facet below it. In the new measurements, the on-axis response of the DXT is significantly different.

[NOTES:
1. These are far field measurements with a resolution of about 225 Hz - not yet merged with nearfield.
2. I shifted traces in REW for the purpose of comparison. In doing so, I didn't retain the Purifi and DXT levels at a fixed signal level.]

 

[Alan J]

Here's a candidate crossover from my first day of fiddling, with the aid of XMachina which bettered my initial blank-sheet-of-paper attempts. The Purifi needs very little, and there are lower component count solutions, but I'm biased toward eliminating its involvement at higher frequencies. The DXT is trickier and this is a pretty low parts count solution. Nothing here is dissipating much power.

Optimization here was for 50/50 listening window and sound power. There's that pesky but perhaps acceptably small dip in sound power, with corresponding non-monotonic DI - but without the big crossover region rise in vertical DI that was present in some other solutions.

Inevitably, I will play with this a lot more...