In @perrymarshall's thread about his open baffle/bass reflex hybrid I accidentally hijacked it a bit, with Perry suggesting I consider his Cottonwood Live Edge Beryllium design with the drivers raised.
Funds are a little tight right now so I decided to attempt the build but with less expensive tweeters. I don't really know what I'm doing here yet, so thoughts are welcome!
Here's what I've cobbled together thus far:
Now that I have the raw materials on hand I'm a bit lost on how to start getting this set up in the Flex 8.
Anything you could suggest to get me in the ballpark would be deeply appreciated!
Funds are a little tight right now so I decided to attempt the build but with less expensive tweeters. I don't really know what I'm doing here yet, so thoughts are welcome!
Here's what I've cobbled together thus far:
- a used miniDSP Flex 8 and Umik-1 microphone
- a pair of SB 15OB350 woofers
- a pair of SB26CDC tweeters, and working on getting a pair of 8" oval somasonus waveguides. I couldn't swing the TW29BWGN, but that or the TX could be future upgrades.
- a pair of the PRV WG175PH horn tweeters Perry's used for multiple designs, for the back side of the baffle
- passives to equal to the Live Edge Beryllium Cottonwood design (47 uf film caps for the front tweeter, exact cap and resistor values for the rear tweeter)
- a 5-channel amp stolen from my HT system, to be replaced with Eric A's 2x150W amp modules in the near future
- free 1.75" thick used doors to chop into test baffles. First thought is 15" woofer at 24" off the floor, slightly tapered toward the tweeter at the top, triangular wings for the woofer zone.
Now that I have the raw materials on hand I'm a bit lost on how to start getting this set up in the Flex 8.
Anything you could suggest to get me in the ballpark would be deeply appreciated!
This is a very solid project.
Go with large waveguides. Even 8” diameter round is not too big. It will assist the radiation pattern match between the 15OB350 and the tweeters.
Xover should be around 1500 Hz.
In the cottonwood design I take a very particular, unusual approach to crossover slopes, which look like this:
The purpose is to minimize phase distortion, by using shelf filters instead of low or high pass filters as much as possible. This is because they only have half as much phase shift.
The immediate 1500Hz transition on both sides is accomplished with a shelf, and the woofer only needs a shelf filter; while the tweeter which needs more protection also has a 12dB high pass that kicks in below 1000.
This achieves minimal overlap to reduce lobing at the crossover frequency.
In the cottonwood dipoles I did this with FIR filters which are somewhat advanced, but you can also implement with standard IIR DSP filters.
You can use an L-pad instead of the 10 ohm resistor and 2.7uF capacitor if you want to adjust the level of the rear tweeter.
You’ll need to add other parametric and/or shelf filters for each driver to achieve flat response. I also did this with the Cottonwoods, even though I don’t discuss the details in the PDF. You can download the 2X4HD XML files and examine what I did.
If your wings are comparable in depth to the ones in the Cottonwood Dipole design, then with 10-15dB of bass boost from the DSP at 45 Hz you can push them down to about 40Hz. You should roll them off below that point.
This will make a very high resolution system with quite respectable bass and great imaging anywhere in the room.
Go with large waveguides. Even 8” diameter round is not too big. It will assist the radiation pattern match between the 15OB350 and the tweeters.
Xover should be around 1500 Hz.
In the cottonwood design I take a very particular, unusual approach to crossover slopes, which look like this:
The purpose is to minimize phase distortion, by using shelf filters instead of low or high pass filters as much as possible. This is because they only have half as much phase shift.
The immediate 1500Hz transition on both sides is accomplished with a shelf, and the woofer only needs a shelf filter; while the tweeter which needs more protection also has a 12dB high pass that kicks in below 1000.
This achieves minimal overlap to reduce lobing at the crossover frequency.
In the cottonwood dipoles I did this with FIR filters which are somewhat advanced, but you can also implement with standard IIR DSP filters.
You can use an L-pad instead of the 10 ohm resistor and 2.7uF capacitor if you want to adjust the level of the rear tweeter.
You’ll need to add other parametric and/or shelf filters for each driver to achieve flat response. I also did this with the Cottonwoods, even though I don’t discuss the details in the PDF. You can download the 2X4HD XML files and examine what I did.
If your wings are comparable in depth to the ones in the Cottonwood Dipole design, then with 10-15dB of bass boost from the DSP at 45 Hz you can push them down to about 40Hz. You should roll them off below that point.
This will make a very high resolution system with quite respectable bass and great imaging anywhere in the room.
Last edited:
Thanks for all of the helpful information!
I'm not sure how FIR files are generated. And I believe I read that the HD and Flex handle them differently... outputs vs inputs. That might be part of a phase 2.
Unfortunately the first round of waveguides got messed up, so I'll have to start with the flat tweeter bezel. Will be ordering again soon.
I'm getting a 404 page not found from the PDF link. Does the file have a new home?
I'm not sure how FIR files are generated. And I believe I read that the HD and Flex handle them differently... outputs vs inputs. That might be part of a phase 2.
Unfortunately the first round of waveguides got messed up, so I'll have to start with the flat tweeter bezel. Will be ordering again soon.
The URL of the PDF is https://psma-website-assets.s3.amazonaws.com/live_edge_beryllium_dipoles.pdf
I generate the FIRs with a program called Eclipse FIR designer, which is a subscription program. I understand there are free programs that can do it too. Another approach is to use DIRAC.
HD and Flex DSPs handle them differently, that is correct.
It's not the end of the world if you don't add the FIR correction at the end. They'll sound fantastic without it.
I generate the FIRs with a program called Eclipse FIR designer, which is a subscription program. I understand there are free programs that can do it too. Another approach is to use DIRAC.
HD and Flex DSPs handle them differently, that is correct.
It's not the end of the world if you don't add the FIR correction at the end. They'll sound fantastic without it.
I must not have expressed that well... inside the PDF there's a link to your miniDSP HD file. That's the link that gives the 404 not found error.
https://tinyurl.com/berylliumdipoledsp When you click on this link it should prompt you to download a zip file
Thanks, that did trigger a download. Haven't looked in detail yet... was busy making sawdust.
So, my first sketch of the layout ended up having a small error that put me off by an inch for my prototype baffle raw material = old door. Unfortunately, it was an inch in the wrong spot.
Guess who spent the next hour trying to figure a different version that would still let me fit 2 of them on 1/2 of a free door because... why? 🤔
So, my first sketch of the layout ended up having a small error that put me off by an inch for my prototype baffle raw material = old door. Unfortunately, it was an inch in the wrong spot.
Guess who spent the next hour trying to figure a different version that would still let me fit 2 of them on 1/2 of a free door because... why? 🤔
So I bravely tried a new form and cut everything and then my wife walked out and, with barely a glance to the side, asked me why I was making coffins. 🤣
I had no rebuttal.
Later, I discovered that it also wasn't wise to assume that the tweeters were 104 mm in diameter. These are proto baffles and will get routed over with waveguides soon, but still kind of annoying!
Not yet mounted, but dropped into place... my first OB coffin prototype:
Hopefully it's not a dead end.
I had no rebuttal.
Later, I discovered that it also wasn't wise to assume that the tweeters were 104 mm in diameter. These are proto baffles and will get routed over with waveguides soon, but still kind of annoying!
Not yet mounted, but dropped into place... my first OB coffin prototype:
Hopefully it's not a dead end.
Last edited:
For the final version I recommend you assume the waveguide is 8” diameter. The larger waveguide ensures good integration and consistent radiation pattern at the crossover point.
You also want to make sure the wings have at least 8” depth at the centerline of the woofer (at least if you want to use them down to 40Hz).
You also want to make sure the wings have at least 8” depth at the centerline of the woofer (at least if you want to use them down to 40Hz).
Not sure why this particular step took so long for just a prototype, but anyway, the baffle stabilizers are roughed out.
These will hold the baffle angle as well as provide mounting for the rear horn tweeter, the binding posts, and hide the simple crossover (albeit pretty big capacitor).
Kind of just for fun, though... probably won't keep this design over time. But who knows?
Separately, got the software download approved by miniDSP and started my learning curve of their software. Feels really odd not to be able to see the cumulative effect of the PEQ and crossover settings. Am I doing something wrong?
I also tried to dig through the Cottonwoods file... hard to understand, although I did eventually pick out the woofer bass boost section.
Since I don't yet have a waveguide for the tweeter, would it be reasonable to cross at approximately 1900 Hz for now, with the tweeters recessed into a flat baffle?
These will hold the baffle angle as well as provide mounting for the rear horn tweeter, the binding posts, and hide the simple crossover (albeit pretty big capacitor).
Kind of just for fun, though... probably won't keep this design over time. But who knows?
Separately, got the software download approved by miniDSP and started my learning curve of their software. Feels really odd not to be able to see the cumulative effect of the PEQ and crossover settings. Am I doing something wrong?
I also tried to dig through the Cottonwoods file... hard to understand, although I did eventually pick out the woofer bass boost section.
Since I don't yet have a waveguide for the tweeter, would it be reasonable to cross at approximately 1900 Hz for now, with the tweeters recessed into a flat baffle?
Yeah that's fine. PEQ and crossover and FIR are all on separate screens, they don't combine them. In Cottonwoods, look at the FIR tab too.
That feels a bit strange that you can't fully see what you're doing in the miniDSP software! Thanks for confirming, though.
For your Cottonwood file, I'm not able to see it inside Device Console since I don't have a 2x4 HD. Or at least, it gave me an error when I tried to import it. I opened it in a web browser, which is quite a mess to behold. At least it has some structure to it.
So far I tried to visually match your curves based off of your descriptions. It was a surprise to see that the steep shelf filters gave huge spikes on either side, which I then tried to flatten out with a couple of peak filters.
Make that past tense--after going to preset 2 and trying to import your file, I returned to preset 1 and the Device Console had lost most of the tweeter PEQ filters. What the ? 😕
For your Cottonwood file, I'm not able to see it inside Device Console since I don't have a 2x4 HD. Or at least, it gave me an error when I tried to import it. I opened it in a web browser, which is quite a mess to behold. At least it has some structure to it.
So far I tried to visually match your curves based off of your descriptions. It was a surprise to see that the steep shelf filters gave huge spikes on either side, which I then tried to flatten out with a couple of peak filters.
Make that past tense--after going to preset 2 and trying to import your file, I returned to preset 1 and the Device Console had lost most of the tweeter PEQ filters. What the ? 😕
Did you save the settings as a file in your computer? If you did, you can reload it.
A great way to make those shelf curves in post #2 above is to add 2 shelf filters together, one with a Q of about 1.5 or 2, and the other with a Q of 0.5. Play with the frequencies, shelf sizes and Qs of each and you can easily find a combination where it drops quickly by about 10dB then levels off some.
What I do during development is save often, then once you get a config that seems viable, load it to a 2nd preset, and use that to keep developing, and save it as a different file. Then you can go back to #1 as your baseline to check/compare progress... ad nauseum. 😅
Gets harder to work this way if you have multiple configs for different speakers. Then, just save at different points of development under different file names, reload files to the same preset as needed.
Gets harder to work this way if you have multiple configs for different speakers. Then, just save at different points of development under different file names, reload files to the same preset as needed.
This is how I build steep crossovers with low phase shift, using Shelf Filters.
This is a combo of 2 low shelf filters. Each is -8dB. Filter 1 is 1500Hz, Q=0.5. Filter 2 is 1200Hz, Q=1.6.
This isn't quite enough to complete a tweeter crossover. You still need to add a 6dB/octave high pass with -3dB at 1KHz, and the result will be roughly equivalent to an 18 or 24dB/octave filter in terms of protection for the tweeter, but will have half the phase shift - roughly the same phase shift as a 12dB filter.
This graph in the MiniDSP doesn't include the 6dB high pass, that's in a separate menu.
You can build a woofer crossover that is symmetrical by transposing the frequencies around 1.5KHz.
This is exactly how I built the FIR filter targets in Eclipse FIR software, which I used in the Cottonwood Dipoles.
I don't like steep crossovers, so I use only enough to do the job at hand. My philosophy is, don't use nuclear weapons when a bow and arrow will do.
This is a combo of 2 low shelf filters. Each is -8dB. Filter 1 is 1500Hz, Q=0.5. Filter 2 is 1200Hz, Q=1.6.
This isn't quite enough to complete a tweeter crossover. You still need to add a 6dB/octave high pass with -3dB at 1KHz, and the result will be roughly equivalent to an 18 or 24dB/octave filter in terms of protection for the tweeter, but will have half the phase shift - roughly the same phase shift as a 12dB filter.
This graph in the MiniDSP doesn't include the 6dB high pass, that's in a separate menu.
You can build a woofer crossover that is symmetrical by transposing the frequencies around 1.5KHz.
This is exactly how I built the FIR filter targets in Eclipse FIR software, which I used in the Cottonwood Dipoles.
I don't like steep crossovers, so I use only enough to do the job at hand. My philosophy is, don't use nuclear weapons when a bow and arrow will do.
Last edited:
@mikessi and @perrymarshall, thanks for sharing your methodology suggestions and approaches for working with the miniDSP! Very, very helpful to see.
I haven't had much time to try those ideas out yet, but I did manage to get 2 prototype speakers wired up. I'm not sure the wings will be enough, but they're foam core elegantly held on with electrical tape, so easy to swap out if needed.
I'd debated about rounding off the corners near the drivers and decided to leave them for the first version here. Seeing it now, I'm not that convinced. When I router for the waveguides I think I'll round them off at the same time.
I had some small-fingered assistance wiring up the speaker binding posts.
I invested (or wasted 🤔) some time seeing if I could hide most of the wiring. In parallel I discovered that with my 1.75" thick baffle I didn't actually need to punch through the back for the tweeter! But, that's going to change with the waveguide anyway. For the next version I'll figure out a way to have the rest of the tweeter wire disappear even better.
Not yet in the listening room, but here is the pair for prototype 1.0!
A rear quarter view. A little quirky for sure, but also kind of fun to be around. When I drop in the waveguides I might even paint them all white... though they'd probably lose some of their grungy appeal. 😉
I haven't had much time to try those ideas out yet, but I did manage to get 2 prototype speakers wired up. I'm not sure the wings will be enough, but they're foam core elegantly held on with electrical tape, so easy to swap out if needed.
I'd debated about rounding off the corners near the drivers and decided to leave them for the first version here. Seeing it now, I'm not that convinced. When I router for the waveguides I think I'll round them off at the same time.
I had some small-fingered assistance wiring up the speaker binding posts.
I invested (or wasted 🤔) some time seeing if I could hide most of the wiring. In parallel I discovered that with my 1.75" thick baffle I didn't actually need to punch through the back for the tweeter! But, that's going to change with the waveguide anyway. For the next version I'll figure out a way to have the rest of the tweeter wire disappear even better.
Not yet in the listening room, but here is the pair for prototype 1.0!
A rear quarter view. A little quirky for sure, but also kind of fun to be around. When I drop in the waveguides I might even paint them all white... though they'd probably lose some of their grungy appeal. 😉
In contrast to dipole bass configs I've used (H & Linkwitz-W), yours looks like the dipole cancellation frequency will vary quite a bit more. (The dipole peak might be less pronounced, too, if there is one.) IE, the distance for waves travelling around the baffle varies depending on path. Over the top is shortest, around the sides varies, with horizontal being longest due the the shape of the wings. Not sure how that affects the bass response. Probably need lots of measurements with different mic positions to tell.
Happy to see your project has started !
A 8 inch waveguide for tw29txn-b-8 is almost ready for me. (Need to glue 2 parts because of power failure in the village during the 3d printing )
A 8 inch waveguide for tw29txn-b-8 is almost ready for me. (Need to glue 2 parts because of power failure in the village during the 3d printing )
Actually when traveling around the baffle varies depending on path, have very little effect on dipole peak dispersion, pressure always find shortest way, specially when wave length is longer than frame/baffle.