Anarchists Build Log

[wogg]

I had just a few minutes last night to do some DATS sweeps and run some high power sine waves.

Tuning
The full length (11") Precision Port kit landed the cabinet at 32Hz. That's just a little low, dropping below -3dB and making the modeled F3 higher. I will want to trim the length a bit and raise the tuning just a few Hz. I compared the response in BB6 for 32, 34 and 37Hz tuning. They all converge at 34Hz with the obvious higher tuning = a couple dB gain above tuning and lower tuning = a little more extension. Since I'm going for the "holy crap that bass is coming from there?" reaction, the couple dB of gain between 35 and 40Hz is what I want rather than moving in the extended bass shelf direction. For whatever reason, most of my builds end up tuned low and EBS aligned, so this will be a good thing to get right this time.



In order to determine how much to cut off the 2" Precision Port kit, I used their calculator and set it to land where the actual tuning landed with the 11" kit by adjusting the effective volume a bit to get the numbers to line up.

Full length shows the effective volume at 0.72 cu ft.


Tweaking tuning up to 37Hz for the higher output shows the length dropping by 2.9 inches.


Perhaps I'll take just an inch off first to validate how this calculation worked before ending up tuned too high accidentally and having to find another 2" tube

Bass OUTPUT
Jeebus these Anarchies are sweet. This is the first time I heard one with power in a cabinet and it does not disappoint. With the 32Hz tuning I was able to pump a good amount of output deep to that tuning frequency. Very impressive!

Impedance Sweep
There are some anomalies in both the woofer and mid at around 500Hz on the sweeps that are not in the free air sweeps. I also felt the slightest movement of the RS100 mid when pumping bass in the woofer so I do think I have a little bit of an internal leak. Then I remembered in the middle of the night sometime that I didn't seal the little drill hole to get the wiring through. I'm thinking that leakage is causing the ripples, so I'll hot glue seal that sucker and try again. I'm also a bit concerned about the gasketing, the straight cuts have a little potential to leak as well. I have an idea to help improve the seal by cutting an additional square of gasket into the seams to make the air path more complex. I'll do that to the back side, but I'm not taking that damn baffle off again until I'm ready to apply finish.

[wogg]

I did some more test work on this last night:
Added denim lining around the woofer cab, ~1" thick lining the sides and top covering the area opposite the internal port terminus.
Stuffed the mid chamber with pink insulation + a couple denim chunks at the back
Hot glue sealed the wiring hole from mid to woofer chamber
Re-run impedance sweeps
Close mic measurements of woofer and port

First, that wire hole was definitely the leak causing the 500Hz impedance blips. After sealing that, the curves are smooth!

Here's the woofer with a 9.5" PP length. Funny, after lining the cab and taking 1.5" off the port length the tuning still lands at 32Hz. The lining dropped the frequency more than I expected.


Here's the mid, nothing of note it's nice and smooth with a low Q resonance.


For the woofer, I cut the port down more to raise the tuning. Took a couple rounds to get there and ended up with about 8 1/8" effective length, landing the tuning near 36Hz.


In my close mic measurements merged in Virtuix, I was using the 9.5" long port which did confirm the slow down slope of the deep bass so I'll have to re-do those with the higher tuning to really see how the deep bass response worked out. I also saw a hefty port resonance at 600Hz that hopefully may have calmed down with the shorter tube and slightly different terminus location in the cabinet. I'll redo those later and see where we are. The resonance was under 10dB down from the cone output so I'll want to try to tame that if it's still that big. Even though I'm crossing 500Hz or lower, that kind of output contribution will affect THD when excited with harmonics from 300, 150, 75Hz etc. I'll put those up when I do the whoop whoops later.

This attention to detail on the woofer tuning is a little new to me, I'm more of a calculate the port and shove it in there type guy. Going the extra mile on this build!

[wogg]

Acquired some data yesterday! This is the cool part IMO. I took the following measurements to build the crossover, all executed on tweeter axis at ~1M:
Tweeter
Tweeter + Mid
Mid
Mid + Woofer
Woofer
Woofer close mic
Port close mic

The basement measurement setup is tight but I can get to about a 3ms gate window before the first reflection. The close mic and port will be blended with the far field woofer in order to capture below ~300Hz where the gate invalidates the measurement. I found that important with my last design to get the baffle step and final bass response right, and in a 3 way with the crossover point not too far above the gate limit it's even more important. The combo files will be used to calculate the Z offset for the acoustic centers. Here's the updated setup using my studio pre and the Dayton EMM6 mic. I'm using ARTA with the calibration file loaded.


Eyeballed on axis and 1M out. Bonus peep at the music production end of the basement. I freaking hate that desk and old metal rack down there, my spring build is planned as a new studio desk instead of another set of speakers. The speakers on stands down there are Dynaudio BM15 passives I've had for a really long time.



View from the speaker, this is the reminder of my little basement cave, that little couch is my listening spot these will be pointing at when done. The far end is the work from home USB dock setup.

Polars were done improvised and eyeballed on a rotating stool. The center of rotation was not perfect on the y axis, a little behind the drivers and probably a little off center to keep the stability on the chair. These will be used to guestimate the power response in Virtuix so I figure these are still better than the theoretical dispersion calculations based on diameter like WinPCD does without polar measurements. I did +- measurements at 0, 15, 30 and 45 degrees, all eyeballed against my little Starrett angle gauge.

[wogg]

I've been playing with crossover design this week as I have time. Here's the process so far:
Initial on axis design in WinPCD, I'm just faster at getting started in that software and can get close to targets there before using other software for final tweaks and verification. Then I used VituixCAD2 to do the final stuff such as pulling in my polar measurements.

To min phase or not...
On axis design is being done with the close mic merged woofer / port data. That output was done in min phase using the Merger tool in Vituix. I found that the I needed to apply min phase to the other files to get them to align, and in fact since all measurements were done single channel without time of flight data and I found the X, Y, Z offsets with good confidence, running min phase files seems to make sense across the board. In Vituix this is just a checkbox but for WinPCD I'd have to process the FRD files externally. The old Bagby files don't work since I updated to 64bit Office so I've dropped the continued development in WinPCD at this point in favor of Vituix.

Schematic
Turning out pretty simple, 2nd order electrical across the board with L pad on tweeter and mid. The woofer likes a Zobel and a notch tank to drop the aggressive breakup way down into the grass. Without the notch, the 5kHz breakup was about 35dB down, with the tank it's way down >50dB.

Note: pre-filter series resistors in tweeter and mid circuit are shorted, removed from circuit.

Here's the on axis response 6 pack image from Vituix.

Crossover points landing just above 500Hz and 4kHz. One thing bugging me here, that slope down under 80Hz. I made some valiant attempts at making the woofer low pass more of a Bessel function, but try as I might I always seemed to increase the 100Hz lump or get a nasty peak at the knee. I tried 3rd order and some ridiculous component value notch filtering... not much luck. I'm also not super confident in the port mic sum on that, the little peak seems odd and I believe the issue is really the mic placement for the port measurement. I found just a few mm of movement into the port caused a large difference in the sum and I'm not sure exactly where to place it for the best accuracy. If this is to be believed, my F3 is way up at 50Hz with F6 way down at 30Hz. Otherwise, I'm quite pleased with this potential, the reverse nulls are plenty deep and summation looks great. Impedance is a tiny bit low, I'd call this a 6 ohm nominal speaker but that shouldn't cause too much trouble for your average AVR.

Here's the same crossover, using the polar measurements I took and showing only 350Hz + where the measurement data is valid.


Turns out I could have used more polar measurements out to 90 degrees to get a better picture. I wasn't sure what Vituix would do, it's odd that I get a polar waterfall from a single on axis measurement but this is cut off at 45. I wonder how it's estimating that for the on axis only measurement? With this limited set to 45 degrees though, it looks pretty good to me. No significant tweeter bloom or odd power off axis that I can tell.

[a4eaudio]

I didn't comment earlier when you just used it once but thought I'd point it out after this post... VituixCAD has no "r".

[wogg]

Whoops! Thanks Dave :

[a4eaudio]

Did you get the EMM6 (like you say in post 33) or the UMM6 (USB mic)? Just wondering because of your other thread comparing WinPCD and VituixCAD.

[wogg]

Did you get the EMM6 (like you say in post 33) or the UMM6 (USB mic)? Just wondering because of your other thread comparing WinPCD and VituixCAD.

I think I'm having a eureka moment,,,, exactly what you're alluding to.

I did get the EMM-6, and am running it through a Focusrite Scarlett with 12 available mic pre inputs for my basement studio. Why am I not doing 2 channel measurements? Duh....

To be continued. I'm going to do some learning from the other thread, starting here down and come back with 2 channel measurements and a re-do of the crossover. I'll have a pile of parts ordered that will likely change dramatically, but this will be a much better learning exercise on stepping up design to another level.

[a4eaudio]

I think I'm having a eureka moment,,,, Why am I not doing 2 channel measurements?

Haha, that was going to be my next question if it was the EMM6.

Here is a recommendation...get/build a turntable. Since your drivers are not all vertically aligned, take 0 to 360 degree measurements in 10 degree increments horizontally and vertically. It is very tedious and if you make mistakes like I do, you end up having to do them all over again (a time or two or three sigh). The VituixCAD measuring guides for REW, ARTA, etc. are good. And then you have such powerful data to take advantage of the tools available.

Below is a DIY turntable I built, I plan on posting a thread at some point, But the key was the 15" Heavy Duty swivel stand that can handle 250lbs. I posted a pic from Amazon but no link to respect forum rules.

[wogg]

Christmas activities have definitely derailed the continuation of this project for now. I did confirm that it's super easy to just pop a 1/4" TRS jumper in my audio interface, check a box in ARTA, and take 2 channel measurements. Next step is really digesting the measurement instructions and getting better data to use for crossover design part 2. Hopefully at least a few of the values I bought end up useful

Merry Christmas everyone! Stay warm out there, pretty sure there are a whole lot of us in the large swath of middle America under a super deep freeze right now.

[wogg]

Attempting some 2 channel measurements today to get a better set of data for crossover design tweaking. I have limitations keeping me from the whole spin-o-rama method exactly as described, so I'm doing a little improvisation based on my own understanding, which is potentially bad

Working on-axis at 1M to do design first. With 2 channel measurement, the offset calculation isn't needed any more since I have time of flight data included. Here's what I've done / all the issues. Adding ? to the bullets I'm not super confident in.

• Using ARTA (unlicensed demo mode)
  • Can't save PIR files due to the licensing, so ARTA has to save out the FRD data instead and I can't use Vituix to convert
  • Perhaps could use REW for this instead
• In ARTA to preserve timing
  • Did sine sweeps of all 3 drivers with the mic in the same position
  • 2.8V was too hot for the tweeter, clipped the mic so I ended up around 1.7VRMS. Went back T - M - W after determining this to get all at the same level
  • ? Set left cursor at 9ms, before the start of the impulse but chosen to be an easy, consistent number. I figured a little extra phase wouldn't be a huge deal as long as all were in line.
  • Set gate right cursor based on the tweeter first reflection, which was the earliest significantly
  • Exported FRD files for all 3 with phase from ARTA
• Did a new close mic woofer and port measurement
  • Level was too hot, had to reduce further
  • Left cursor ended up at 6.5ms instead of 9, no right side gate
  • Placed the mic in the precision port a bit to approximately where the flare stops
  • ? Merged using Vituix tool, applied 3.5ms delay to the close mic files to align the left cursors
  • Adjusted close mic levels to join at the 400Hz merge spot since I had to adjust level
  • ? Output merged FRD without any minimum phase processing, hoping that would preserve the same timing as originally measured

Now I have a new set of FRD files for all drivers. I've pulled them into Vituix and PCD, but am finding myself making some other assumptions that I'm not confident in:

• With 2 channel data, the Z offset isn't needed and should be cleared to 0?
• The X and Y values are still useful for simulation in either / both?

If I can clear up those, I should have enough to make pass Forum at crossover design.

[wogg]

I have some design progress now. Ultimately the solution for Vituix is to use the PIR output from ARTA and process all files at one time using the PIR to FRD tool. That ensures all the files are processed the same and saves time too.

I made a new set of 2 channel polar measurements, this time going with 30 / 60 / 90 degrees. Not as much as recommended, but my measurement setup and basement dimensions are the limit. This still gets me a decent polar map, and is much better than a single point on axis design.



After a while of messing with the data and figuring out how the optimizer works I have a better crossover design. The goal here is to mix on axis flat response with power response, to achieve the best smoothness in the room I can.

Vituix notes:
Follow the measurement guide exactly.
Use the mic calibration file when processing the PIR files, that is applied when doing the time domain to frequency domain calculations
The optimizer is not intuitive and requires careful setting of target and frequency range to work
The simulation is done at 2.5M to work best, and reverse null phase alignment is not the main objective (don't worry about that, focus on power and on axis response)


I spent a good amount of time adjusting manually, adding variants, using the optimizer, and working against values readily available to get in real life. As I went more complicated, the return wasn't any better so by opening parts in Vituix I ended up simplifying significantly with no Zobel or tank on the woofer at the expense of a little mid and tweeter extra parts. Still, with only 2 parts count addition to a 2nd order electric filter on all drivers + L pads for mid and tweeter this is about as simple as it gets.



There is still some directivity bloom at the tweeter crossover point, but on axis I'm near +-2dB. There's an interesting lean on the directivity toward + degree movement caused by the asymmetrical baffle arrangement. The opposite speaker is complementary, and I don't suspect that will cause a problem, may actually be a bonus depending on the room. We'll see.

Today I've disassembled the test speaker and am now testing some more dye and finishing options for the baffles. I've got RIT red dye on oak rolling, and will have an ebony stain on top of that for a layered deep red. We'll see what that looks like on my extra board before hitting the baffles.

[Wolf]

Water based ebony will be black, oil based will be brown.
Another option is black shoe ink.
Wolf

[davidroberts]

Might also try ebony ink from a local craft store. IIRC it is water based so you can easily dilute it.

[wogg]

I've been abusing the extra oak board I have from the baffle with a variety of Minwax colors, conditioner, and now RIT dye. I think this one is looking the best, a good soak with RIT followed by a pair of 10 minutes soaks of Minwax ebony, then another RIT soak. This is the powder version of the dye, one full box mixed with about 8 ounces of water. I think I'll try the poly on it after it's fully dry this week and see what it looks like with a satin seal on it.

[wogg]

I think I have a decent process now. Here's the test board with just about the final look. You can see other stain test patches all over.



It'll be dye, stain, dye, stain... repeat until happy. Then satin poly multiple coats. This does take extensive dry time between cycles, but the results are pretty cool. Bench reset to start the first stain soak on the actual baffles today.

[Billet]

Looks great.

[a4eaudio]

I have to admit I wasn't really grasping the RIT dye, stain process. But that looks nice!

[a4eaudio]

Double post

[4thtry]

I noticed that you did a partial (0/30/60/90) instead of the full spinorama due to space limitations. This got me thinking as to how big a difference this would make in the various modeling curves. Which curves would be affected the most and why? So I loaded my Diffractorama speaker model into VituixCAD to generate a comparison. The first graph below shows several VituixCAD curves with my full 0/15/30/45/60/75/90/105/120/135/150/165/180 data set loaded for all drivers. The 2nd graph below shows the same curves, but this time with only the 0/15/30/45/60/75/90 partial spin data set loaded for all drivers.

As you can see, the biggest change is with the slope of the dark blue power response curve. The In-Room Response (orange) slope drops a little bit too, but no where near as much as the power response curve. The Listening Window (green) curves seem to show little change. This makes sense, because there is much less high frequency energy in the 105 to 180 degree data curves. Also, the Directivity Index (red) shifts down significantly while the ERDI total curve (Salmon/orange) seems to show little change.