Any interest in an ASR community speaker project?

[dwkdnvr]

There's definitely something weird going on with those simulations. The DI varies far more within the passband of the mid than I'd expect - maybe it's just the scale, though. The design parameters aren't ideal, though - 3.5kHz is way too high to consider crossing a 125mm mid, and I'd hope to cross lower than 500 to the woofer as well. If you look at the Zaph ZA5.2, it uses a 14cm driver rather than a 12.5, but even crossing at 1.8kHz to a 25mm tweeter there is a wobble in the DI curve. When I was playing around with various options, I thought the 10cm Scan 10f might have a chance if crossed to the narrower-dispersion XT25 at ~3kHz, but even there it probably wasn't ideal. The SB10 might actually work with 500Hz and ~3kHz xovers to the Xt25 - would be a cheap pairing if it worked.

But yes - I think that if one buys into the Harman philosophy and the analysis produced by the Klippel system, the conclusion that a waveguide is the optimal strategy is pretty hard to avoid. It definitely complicates the discussion though, since DIY waveguide designs are still somewhat uncharted - there is some experimentation that will be needed.

I actually tried printing one of Augerpro's waveguides, but my printer apparently needs some TLC as all I got was a ball of goo - hopefully just stale PLA. There are potentially some other options - there is some good software over at diyaudio called 'ath4' that produces good looking waveguides and can simulate the response via Akabak for compression drivers at least - not sure how well it models dome tweeters, though. They need a fair bit of post-processing to get into printable form though since they're just a single surface - no thickness. CNC carving them is less work since you can just drop the surface onto the solid, but in either case you have to design the tweeter mounting arrangement. I had less luck trying to pull one of Augerpros models into my CNC software, but I may have to give it another try.

I'm actually planning on placing an order to PE and potentially Madisound in the next day or two, and so will potentially be able to grab a few things to experiment with to advance this discussion - some of the cheaper tweeter and mid options.

 

[617]

There's definitely something weird going on with those simulations. The DI varies far more within the passband of the mid than I'd expect - maybe it's just the scale, though. The design parameters aren't ideal, though - 3.5kHz is way too high to consider crossing a 125mm mid, and I'd hope to cross lower than 500 to the woofer as well. If you look at the Zaph ZA5.2, it uses a 14cm driver rather than a 12.5, but even crossing at 1.8kHz to a 25mm tweeter there is a wobble in the DI curve. When I was playing around with various options, I thought the 10cm Scan 10f might have a chance if crossed to the narrower-dispersion XT25 at ~3kHz, but even there it probably wasn't ideal. The SB10 might actually work with 500Hz and ~3kHz xovers to the Xt25 - would be a cheap pairing if it worked.

But yes - I think that if one buys into the Harman philosophy and the analysis produced by the Klippel system, the conclusion that a waveguide is the optimal strategy is pretty hard to avoid. It definitely complicates the discussion though, since DIY waveguide designs are still somewhat uncharted - there is some experimentation that will be needed.

I actually tried printing one of Augerpro's waveguides, but my printer apparently needs some TLC as all I got was a ball of goo - hopefully just stale PLA. There are potentially some other options - there is some good software over at diyaudio called 'ath4' that produces good looking waveguides and can simulate the response via Akabak for compression drivers at least - not sure how well it models dome tweeters, though. They need a fair bit of post-processing to get into printable form though since they're just a single surface - no thickness. CNC carving them is less work since you can just drop the surface onto the solid, but in either case you have to design the tweeter mounting arrangement. I had less luck trying to pull one of Augerpros models into my CNC software, but I may have to give it another try.

I'm actually planning on placing an order to PE and potentially Madisound in the next day or two, and so will potentially be able to grab a few things to experiment with to advance this discussion - some of the cheaper tweeter and mid options.

I got in touch with augerpro. He said the 4 inch wg he designed would work well with a 4 inch driver.

I traced his polar measurements and brought them into VituixCAD and the blend with the rs125 is perfect. Now, this doesn't include baffle effects, but I am pretty confident it will work well. The sb21 is very well behaved.

If you pick up an sb21sdc from madisound, I can get a 3d print of the waveguide and we can figure out how to test it. Right now my woodworking tools are not available, but I can get 3d printing done and make a very crappy box if needed. I also have an rs125 and a few other mids around that size to test.

 

[jurop]

Forgive me in advance since it is going to be a long one if you read it all. Perhaps a bit late on this but I like to share my view nonetheless before unsubscribing the thread for not being interested in it anymore if the project steers - like it is doig - towards a classic 2 or 3-ways.

Let me first pick some relevant quotes:

Well, this is to a degree the elephant in the room (or at least one of them). Are we aiming for a design that is a complete standalone system intended to be used without subs, or is it a design where we intend to enable integration with a multi-sub setup? IMHO advocating a DSP speaker design intended to be used with subs but not actually having anything in place to implement the integration is a bit of a problem. Enabling multi-sub integration isn't too hard if you use a DSP setup with enough channels (e.g. MiniDSP 4x10 or Nano-Digi), but is tougher to wrap your head around in the case of something like the Dayton DSP amps - how do you handle things like global delay alignment when you have N independent/unconnected DSP configurations? You can probably do it, but it's not going to be fun.

I think @617 was really hoping to have a design that was intended to stand alone, which is why he was going for a 10" driver. I have used the Dayton RSS210HF as the bass driver in a sealed system, and it actually did really well until I put it into our 10000+ cubic foot open space at which point it struggled. If we're targeting smaller spaces though, a sealed 8 with enough output can IMHO be a completely standalone system for music. If we're aiming for something that would handle a really big room though, IMHO we need more headroom.

And, going back to @Thomas savage original post, he had 'room correction' as part of the list of requirements. If we want to include at least of guidelines as to how to implement/integrate this design into a room correction system, we need to define what the expectations are. Certainly we aren't going to implement an actual automated RC system as a part of this, so it's a question of approach. IMHO this boils down to eiither
a) show how REW can be used to create correction filters that can be imported into this system
b) selection of a DSP platform that has Dirac or something equivalent already integrated.
c) (maybe) FIR filter integration so that DRC-FIR or Acourate filters can be used

Only a) really preserves the "DSP implementation independent" idea that I've been advocating, so that is my working assumption at this point. Since this is an open discussion about community direction though, alternative ideas/proposals are definitely welcome.

...and...

We're doing a lot of things with the active network which would be really expensive, difficult or impossible with passive. Having said that, if the drivers work well together from an acoustic perspective, a good passive design could probably be developed. A hyprid design could certainly be developed, where the mid/high is a passive network and the woofer is active.

It would be an interesting exercise for sure, to do a full passive design. If we use fairly benign driver such as the reference-paper woofers and a friendly midrange, it's possible the passive design could work pretty well; but if we're using dsp to smooth out little wiggles everywhere - good luck doing that with passive. And of course the bass alignment would have to be totally different.

Following the above quotes, let's focus on some points that had already been scavenged with some conclusions laid down:

a) 2 ways speaker: high frequencies crossover very difficult to directivity match, being it active or passive does not make any difference
b) 3 ways speaker: problem above solved but another one sports out: low frequency SPL difficult to obtain if a 30-40Hz bass extension on the lowers is desired

A lot of energy had already been spent in selecting drivers and making simulation without solving the balance above. Don't you think some hundreds of HiFi company already tried to do it in the past? Of course, being DYI, we are not concerned with market share, economy of scale and so on, but the issue still stand. The technology simply does not still allow to solve both a) and b) issues above at once.

A small addendum: my favourite speaker as today is a custom 4+1 way active 'full-horn' with distributed sub-woofers. It literaly TRASHES all the commercial speaker I heard till today, including JBL Everest 66000 and K2 and sporting a theorical output capability of 136dB SPL peak. And no, I can grant you that it doesn't sound 'nasal'. There's quite a bit of DSP power in front of it, including room treatment and correction. I am VERY spoiled towards very dynamic speakers, albeit my day-to-day companion is a trusty old LS3/5a.

Back to the core, my gut idea is:

sat: 3way for the 'top part' with passive x-over. Frequency response limited around 100Hz, actively high-passed
sub: 1way for the 'low part' with active lo-pass x-over.

Main issue being obliged to build 4x boxes instead of just two, let's have a look at such a combo:
1 - 6" mid with high order hipass around 80-100hz, efficient and cheap, crossed around 700Hz
2 - 3 or 4" mid with 700-3500/4000hz bandpass (700Hz is not casual choice, it's related to psycho-acoustic studies about human directivity perception)
3 - 6" midrange/midbass 80-100 to 700Hz
4 - 8/10/12" subwoofer

The above point means that:
- 1+2, we do not need a waveguide for the high frequency crossover
- 2+3, drivers are small enough to be fitted C-C very close; crossover point should not be an issue for directivity control
- 2+3, crossover point means that both breakups and Fs are far away to allow for 12dB/oct crossing with 'human' passive components
- 3, being a midrange the volume of the 'top box' would be quite small up to the point to be placed on a desk if built with a 5" instead of 6" sacrificing a bit of sensitivity
- 4 does speak for itself

ASR as a community could provide an active crossover EQ that works well when the satellite is placed on top of the sub. This would be the same of having a 'tower' three ways, just divided in two parts. One class-D amp per side, vertically amplified, and one hi.lo pass would do it. Let's call this step one.

Going from step one to step two would just be putting a stand below the satellite adding a DSP module for phase alignment, buying a measure system and to start placing the subs around the room for finding the best combination. Step three, just add two more subs. But the concept is that ASR community would provide a 'complete noob' starting system. The tradeoff is that we need 4x boxes instead of two but on the other hand the speakers could be dirt cheap and the only limit the SPL from the sub solution chosen (well, and the 4" inch mid since the market do offer very little for that component).

Not to underestimate the fact that such a solution DOES NOT EXIST (as far as I know) IN THE COMMERCIAL WORLD so it would be a 'prime'. Another +1 in favour of ASR.

Version A could be designed with a 10" sub + 6/3/1 inches combo. Just imagine what a version B using a 12-15 sub + 8/4/1inches combo could do. IMHO that is the peak of what could be done with today's technology without resourcing to horns/waveguide. The difference of course being price.

An example? All parts from partexpress:
- DaytonAudio PA255-8 --- $ 44
- Faital 6FE100 --- $ 45
- Faital 3FE22 --- $ 33
- Peerless XT25SC90 --- $18
Total speakers $ 140/channel
4x boxes
2x D-plates
1x hi/lo active bandpass - for the basic system, a splitter with an R-RC cell on the signal line could perhaps suffice?

Thanks to everybody who had enought patience to get here!

 

[TimW]

I have been using VituixCAD just for the SPL Trace tool. Now that I'm trying it out I realize I should have been using this for all of my DIY shenanigans. Not sure that I'm using it correctly though. What do you guys think of this? It's a 3-way that aims for wide directivity at the expense of output and low end extension.

@617 Have you been setting a baffle size in VituixCAD?

 

[HooStat]

it would be interesting of @Dennis Murphy had any input to offer on some of the driver size and cross-over questions.

 

[Lbstyling]

Some simulations. These use a 28mm tweeter, a 125mm mid and an 8" woofer. The mid and tweeter are offset, so there is a little assymetry in the horizontal directivity.

Here we have the response with LR4 crossovers. I equalized each driver so it was flat in both directions then applied textbook crossover slopes. This enabled me to move the crossover frequency around easily to see how the drivers blended.

I think I had a simulation with a better response with fourth order slopes, but I can't find it now. Around 1500k there is a bump in the early reflections and room response which could be equalized - I suspect it would be worth have a slight depression in the direct response to mitigate that a bit. There is also a big discontinuity around 3K, which is where the crossover is between the tweeter and mid.*

*Edit - the more I look at it, something is screwed up here. It's definitely possible to use fourth order slopes and getter a better result than this. I think the crossover between mid and tweeter wants to be really high or something.

View attachment 67062


What happens when you go to LR2? Evertything gets smoother. In reality I wouldn't use LR2 all the way from 0db to -infinity db, it's generally better to do a hybrid sort of approach where the first octave is gentle, and then it gets steeper and steeper. In this case, I'd pay particular attention to the lower bid bandwidth, which is pretty wide, I'd like to trim that back.

View attachment 67061

This is a lot better. . I wish I had a graph of the Revel Gem2, which uses the same driver sizes. I may try to overlay the Salon 2 with this speaker just to see how wavy the DI is compared to a very good design.

Now, it remains to be seen how this would work with a waveguide. I can't simulate those as easily as direct radiators, but I may speak with the WG author to see if he knows of any studies with the 4" WG and a 4" mid.

I suspect this is about as good as you can do with direct radiators, which isn't bad. Waveguides aren't going to help woofer/mid integration, so you'll always have a little waviness around 300hz, but you could probably get the DI very flat above that with care.

What software are you using to get this DI prediction? DI will be highly dependent on the baffle width and shape. Is there any data on how reliable this prediction is vs real world?

A waveguide for the tweeter is an excellent idea, and has the added bonus of allowing us to machine into it so the mid driver can get closer to the tweeter.

 

[Lbstyling]

Spinorama comparison. This is with the Salon 2, which isn't a perfect comparison because it is a 4 way speaker. Couldn't find a DI curve for the Gem2. Anyway, it's not worlds worse, but the issue at the tweeter crossover is not going to work. The lesson here? Even with LR2 crossovers and a big tweeter/small mid, getting a really smooth DI through the crossover region is difficult. I'm pretty certain a 4" waveguide is the way to go here.
View attachment 67065

I'm going to try to contact augerpro (is he here?) and get some input on these waveguides. The one designed for the sb21 is elliptical, but I can make the outer edge a rectangle for easier flush mounting.

One thing I do that has shown to be really useful is to make the crossover using shelf and peak filters. what this does is it allows me to shape the DI in the crossover region. If we use drivers that show smooth DI, and the shape uniformly contracts as freq raises, as long as we have significant overlap between drivers, we *should* be able to beat this. we may end up making a 4 way though!

 

[Lbstyling]

There's definitely something weird going on with those simulations. The DI varies far more within the passband of the mid than I'd expect - maybe it's just the scale, though. The design parameters aren't ideal, though - 3.5kHz is way too high to consider crossing a 125mm mid, and I'd hope to cross lower than 500 to the woofer as well. If you look at the Zaph ZA5.2, it uses a 14cm driver rather than a 12.5, but even crossing at 1.8kHz to a 25mm tweeter there is a wobble in the DI curve. When I was playing around with various options, I thought the 10cm Scan 10f might have a chance if crossed to the narrower-dispersion XT25 at ~3kHz, but even there it probably wasn't ideal. The SB10 might actually work with 500Hz and ~3kHz xovers to the Xt25 - would be a cheap pairing if it worked.

But yes - I think that if one buys into the Harman philosophy and the analysis produced by the Klippel system, the conclusion that a waveguide is the optimal strategy is pretty hard to avoid. It definitely complicates the discussion though, since DIY waveguide designs are still somewhat uncharted - there is some experimentation that will be needed.

I actually tried printing one of Augerpro's waveguides, but my printer apparently needs some TLC as all I got was a ball of goo - hopefully just stale PLA. There are potentially some other options - there is some good software over at diyaudio called 'ath4' that produces good looking waveguides and can simulate the response via Akabak for compression drivers at least - not sure how well it models dome tweeters, though. They need a fair bit of post-processing to get into printable form though since they're just a single surface - no thickness. CNC carving them is less work since you can just drop the surface onto the solid, but in either case you have to design the tweeter mounting arrangement. I had less luck trying to pull one of Augerpros models into my CNC software, but I may have to give it another try.

I'm actually planning on placing an order to PE and potentially Madisound in the next day or two, and so will potentially be able to grab a few things to experiment with to advance this discussion - some of the cheaper tweeter and mid options.

Agreed, the crossover is too high for a 5 inch. Given a choice, a mid driver covering 700-7000 would be awesome. Probably looking at a high output 3 inch for that though. crossover can be steep to the bass driver though, as the ear has low sensitivity to C to C spacing below this.

If the crossover is high enough, a really cheap, small tweeter should do great!

I am thinking the cash would be spent on the mid driver under this senario, the rest would be cheap.

 

[Lbstyling]

I got in touch with augerpro. He said the 4 inch wg he designed would work well with a 4 inch driver and that crossover would need to be above 2.5khz. This is no problem for many 4 inch mids.

I traced his polar measurements and brought them into VituixCAD and the blend with the rs125 is perfect. Now, this doesn't include baffle effects, but I am pretty confident it will work well. The sb21 is very well behaved.

If you pick up an sb21sdc from madisound, I can get a 3d print of the waveguide and we can figure out how to test it. Right now my woodworking tools are not available, but I can get 3d printing done and make a very crappy box if needed. I also have an rs125 and a few other mids around that size to test, a scan discovery 15cm, an rs150p, and an ne124.

Now THAT is exciting.

I still think a 3" with a WG would be the way to go as the WG will allow it to play lower, with less excusion, keeping the bass driver cheap. It will also give better measurement off axis in the vertical plane.

4" isn't bad though!

A mid WG....Eliptical?? would reduce the need/effect of the chamfering on the cabinet corners.

How wide is the waveguide? Would ideally need a cutout for the tweeter to reduce C to C spacing.

 

[Colonel7]

Now THAT is exciting.

I still think a 3" with a WG would be the way to go as the WG will allow it to play lower, with less excusion, keeping the bass driver cheap. It will also give better measurement off axis in the vertical plane.

4" isn't bad though!

A mid WG....Eliptical?? would reduce the need/effect of the chamfering on the cabinet corners.

How wide is the waveguide? Would ideally need a cutout for the tweeter to reduce C to C spacing.

Somewhat OT but the Peerless in your final 8 is today's deal of the day at PartsExpress - $34.25. You seemed to have moved beyond that one though

 

[KSTR]

Whatever crossover target you go for, make sure you have a constant phase offset tracking down to -40dB, it really does pay off sonically and is IHMO one important parameter often overlooked, compared to the magnitude-only centric design view we see so many times where most people only look at a proper -6dB intersect which should result in full cancellation when one way is inverted. The point is, adjust the magnitude responses until the phases track to within +-10deg even when this deviates from the "perfect" magnitude slope when looked at in isolation. In the XO region, phase is way more important than magnitude for proper integration.

The reason is that the phase is affected long before you can see a change in magnitude response. Simple cascade of two LR2 XO targets for a three way does NOT yield a LR2 3-way wrt to phase response. For mid/tweeter XO, anything else than true LR4 is typically compromised, we want the main lobe to be on-axis with only a small drop in energy curve, the latter being less important than the former. For a lower XO point a different phase offset can be used, anything from 0deg (LR) to 90deg(Bu) is useful, just check that it doesn't compromise directivity. I often went for a 60deg offset, a hybrid between LR and Bu, Bu target has the advantage of steeper final slopes while having the same total phase rotation (and hence group delay) than the next LR one order lower. Again, whatever phase offset, make sure it is maintained until the response is at least 30...40dB down.

The usual approach of straightening the driver's mag response first within some range beyoned the intended XO, then applying text-book XO function (and neglecting the impact of the adjancent XO or system roll-off) is systematically compromised, believe me. It does work to some extent when only a few bands with steep slopes are used.

My golden rule of succesful crossover design: clearly define the full set of target responses of the ways (including the final bass roll-off) and work from directly from there, that is determine the correction function directly by division (and make sure it can be realized in analog if DSP XO isn't planned, which means the correction must be minimum phase). Oftentimes, one can feed the correction target in an optimizer that adjusts the parameters of a set of transfer function building blocks until a fit is achieved. The choosen set of building blocks of course needs some educated guesses. From there, one can get the component values of actual circuits, again using the optmizer. This works really well for active XOs.

 

[B5234t]

I’d like to register my interest in this project and suggest a modular approach, with each driver housed individually. Start with a 4-inch two way aiming for smooth directivity and good predicted SQ down to ~70hz (low budget). Then aim for reference levels with a second 4-inch in an MTM arrangement (low-medium budget). Then add a 6~8 mid-woofer section underneath (and above) for improved mid-bass response (medium-high budget). Then add a subwoofer section/s at the bottom for improved bass response (high+ budget). Distortion, extension and power handling would be improved with each consecutive tier. This method could (potentially) please all members providing budget high performance to value SOTA.

Similar ideas:

My background:
I recently finished a degree in mechanical engineering. I live in NZ thus can’t contribute physically. However, I have plenty of CAD experience and some crossover design experience. I currently work part time as an audio repair technician.

 

[Biblob]

Whatever crossover target you go for, make sure you have a constant phase offset tracking down to -40dB, it really does pay off sonically and is IHMO one important parameter often overlooked, compared to the magnitude-only centric design view we see so many times where most people only look at a proper -6dB intersect which should result in full cancellation when one way is inverted. The point is, adjust the magnitude responses until the phases track to within +-10deg even when this deviates from the "perfect" magnitude slope when looked at in isolation. In the XO region, phase is way more important than magnitude for proper integration.

The reason is that the phase is affected long before you can see a change in magnitude response. Simple cascade of two LR2 XO targets for a three way does NOT yield a LR2 3-way wrt to phase response. For mid/tweeter XO, anything else than true LR4 is typically compromised, we want the main lobe to be on-axis with only a small drop in energy curve, the latter being less important than the former. For a lower XO point a different phase offset can be used, anything from 0deg (LR) to 90deg(Bu) is useful, just check that it doesn't compromise directivity. I often went for a 60deg offset, a hybrid between LR and Bu, Bu target has the advantage of steeper final slopes while having the same total phase rotation (and hence group delay) than the next LR one order lower. Again, whatever phase offset, make sure it is maintained until the response is at least 30...40dB down.

The usual approach of straightening the driver's mag response first within some range beyoned the intended XO, then applying text-book XO function (and neglecting the impact of the adjancent XO or system roll-off) is systematically compromised, believe me. It does work to some extent when only a few bands with steep slopes are used.

My golden rule of succesful crossover design: clearly define the full set of target responses of the ways (including the final bass roll-off) and work from directly from there, that is determine the correction function directly by division (and make sure it can be realized in analog if DSP XO isn't planned, which means the correction must be minimum phase). Oftentimes, one can feed the correction target in an optimizer that adjusts the parameters of a set of transfer function building blocks until a fit is achieved. The choosen set of building blocks of course needs some educated guesses. From there, one can get the component values of actual circuits, again using the optmizer. This works really well for active XOs.

Wow, this is very technical and goes above my understanding. Is it possible you can explain this using some visuals, like FR and phase slopes? Right now, I don't really understand what to do and what to aim for.

 

[617]

I have been using VituixCAD just for the SPL Trace tool. Now that I'm trying it out I realize I should have been using this for all of my DIY shenanigans. Not sure that I'm using it correctly though. What do you guys think of this? It's a 3-way that aims for wide directivity at the expense of output and low end extension.
View attachment 67087
@617 Have you been setting a baffle size in VituixCAD?

Yes, to simulate directivity I bring driver responses into the baffle tool, which then spits out 360 degrees of FR files which account for baffle diffraction and baffle step (same thing) and also a simple model of driver directivity based on driver dimensions. Works well for simple things like pistons, but can't simulate drivers which break the rules, like waveguides and bmrs (and drivers whose directivity becomes wider and chaotic at breakup)

 

[617]

Whatever crossover target you go for, make sure you have a constant phase offset tracking down to -40dB, it really does pay off sonically and is IHMO one important parameter often overlooked, compared to the magnitude-only centric design view we see so many times where most people only look at a proper -6dB intersect which should result in full cancellation when one way is inverted. The point is, adjust the magnitude responses until the phases track to within +-10deg even when this deviates from the "perfect" magnitude slope when looked at in isolation. In the XO region, phase is way more important than magnitude for proper integration.

The reason is that the phase is affected long before you can see a change in magnitude response. Simple cascade of two LR2 XO targets for a three way does NOT yield a LR2 3-way wrt to phase response. For mid/tweeter XO, anything else than true LR4 is typically compromised, we want the main lobe to be on-axis with only a small drop in energy curve, the latter being less important than the former. For a lower XO point a different phase offset can be used, anything from 0deg (LR) to 90deg(Bu) is useful, just check that it doesn't compromise directivity. I often went for a 60deg offset, a hybrid between LR and Bu, Bu target has the advantage of steeper final slopes while having the same total phase rotation (and hence group delay) than the next LR one order lower. Again, whatever phase offset, make sure it is maintained until the response is at least 30...40dB down.

The usual approach of straightening the driver's mag response first within some range beyoned the intended XO, then applying text-book XO function (and neglecting the impact of the adjancent XO or system roll-off) is systematically compromised, believe me. It does work to some extent when only a few bands with steep slopes are used.

My golden rule of succesful crossover design: clearly define the full set of target responses of the ways (including the final bass roll-off) and work from directly from there, that is determine the correction function directly by division (and make sure it can be realized in analog if DSP XO isn't planned, which means the correction must be minimum phase). Oftentimes, one can feed the correction target in an optimizer that adjusts the parameters of a set of transfer function building blocks until a fit is achieved. The choosen set of building blocks of course needs some educated guesses. From there, one can get the component values of actual circuits, again using the optmizer. This works really well for active XOs.

Thanks for your input. The reason I straightened the responses with a hundred peaking filters is just to make changing the crossover points easier. In an actual design, my experience is with passive networks, where I always strive for good phase integration across as wide a band as possible, using no more filters than needed.

For me the only reason to do this is to ensure the loudest point is the reference axis; if the crossover point is perfectly in phase on axis, it guarantees that off axis there is no point where it is more in phase, which would cause a peak.

Another thing I really dislike in 3 ways is all 3 drivers playing at once. Sometimes when you remove the mid you can still see output at the mid region, due to the use of very shallow crossover slopes. I much prefer to steepen the roll off after an octave or two so you get better out of band rejection.

This is very technical stuff though, and the beauty of dsp is that we can easily change it and listen for ourselves.

 

[617]

I’d like to register my interest in this project and suggest a modular approach, with each driver housed individually. Start with a 4-inch two way aiming for smooth directivity and good predicted SQ down to ~70hz (low budget). Then aim for reference levels with a second 4-inch in an MTM arrangement (low-medium budget). Then add a 6~8 mid-woofer section underneath (and above) for improved mid-bass response (medium-high budget). Then add a subwoofer section/s at the bottom for improved bass response (high+ budget). Distortion, extension and power handling would be improved with each consecutive tier. This method could (potentially) please all members providing budget high performance to value SOTA.

Similar ideas:

My background:
I recently finished a degree in mechanical engineering. I live in NZ thus can’t contribute physically. However, I have plenty of CAD experience and some crossover design experience. I currently work part time as an audio repair technician.

Thanks for your interest, as a repair person your perspective will be valuable.

The idea of modularity is an interesting one, @jurop also proposed the idea of a 6+2+1 or some other small 3 way as a satellite. A 6+2-3+1 format can give you very smooth off axis response without using waveguides, and if you assume the use of subwoofers, can go satisfyingly loud above say, 80hz.

I do however value simplicity as a design parameter in this project, and making a speaker which doesn't have good bass extension on its own (a true satellite) or bass extension at the expense of SPL (a 6+1) also limits appeal somewhat.

Having said that, the lessons we learn here, and the resources we gather, would make the design of such a speaker very easy, especially considering that dsp is used. If we provide useful data about what sizes of driver can meld with each-other, making changes and substituting drivers becomes a bit easier. For example, I thought that using a 10" with a 3" mid would work well, since there are a million speakers which do this. Not the case! It suffers from the exact same problems that using a 8" woofer and a 1" tweeter suffers from - poor off axis response. To get things really smooth, you need 8-4, or maybe 10-5.

(The conclusion of this thinking is the Revel Salon 2, or any number of other 4-way designs. A 3 way is perfectly adequate for getting great on axis response and headroom, but if you want full range, good off axis without horns, you need 4 speakers. Examine the Salon 2 and notice how close the driver sizes are between woofer and lower mid, lower mid and upper mid, mid and tweeter.)

One of the initial ideas for this speaker was the 'ultimate satellite' in some form or another - very small, but very dynamic, giving bass down to 80hz. That is still an interest of mine, but it turns out that 'ultimate satellite' can't be small, even if you restrict bass output to 80hz. The speaker we're currently looking at, with an 8" driver, can't do 'reference levels' even at 80hz I don't think. But it can go loud enough at 40hz to satisfy most people, and can go loud as hell at 80hz, all while maintaining excellent off axis behavior.

Ironically, I'm almost done building the 'ultimate satellite' speaker, which is a Gedlee Abbey kit. It's gigantic and hurts to move around, it's a two way with a simple crossover and a 12" woofer, and it's only goes down to 60hz or so. Part of the reason I'm so focussed on dynamics with this project is experience with that and other speakers with high output - even at low volume it sounds exciting and real in a way which 6" midwoofers don't.

 

[617]

Since everyone likes pictures, this is what the sealed 8" looks like. This is the baffle design I used for the latest simulations. Due to the fact that the woofer is big and the cabinet volume is low, it is quite shallow. A vented version would be probably twice as deep or taller, depending on builder preference and vent location. If you made it a floor stander you might be able to put two PRs on the back. The drivers are offset; this helps smooth the on axis response of the mid but with the waveguide I'm not sure if it's strictly necessary. This really isn't a very large speaker.

 

[dwkdnvr]

Since everyone likes pictures, this is what the sealed 8" looks like.

Ok, starting to catch up. The dimensions you have here are actually really close to what ended up with for my RSS265/PR test box. After spending a lot of time trying to get the best performance out of the smallest volume, I am really surprised at just how small it is. Maybe it'll seem bigger when I bring it in from the garage, but I do agree - it's not a large speaker. Of course, if you use it as a stand mount it takes up just as much space as adding another 12-16" to the bottom to make it a floor stander, but that's another discussion.

For aesthetics, I was thinking Avalon style facets rather than a flat baffle. The DXT-Mon uses this style even with the DXT waveguide tweeter, although I don't believe there are any measurements showing how much impact it has. https://heissmann-acoustics.de/wp-content/uploads/dxt_mon_grau_800.jpg . Of course, that's only a 6" driver, so scaling it up would involve making it wider too. A better way to think of it might be taking this http://www.htguide.com/forum/showth...nt-Reference-Thread-How-we-realized-the-Dream and dropping one woofer and shrinking it. The facets do elevate the construction requirements, though. In a perfect world, measuring a flat baffle and a faceted baffle would be an interesting exercise to see how much if any difference it makes. The TODO list on this is already pretty long though, so we'll have to see whether that makes the cut.

 

[dwkdnvr]

Agreed, the crossover is too high for a 5 inch. Given a choice, a mid driver covering 700-7000 would be awesome. Probably looking at a high output 3 inch for that though. crossover can be steep to the bass driver though, as the ear has low sensitivity to C to C spacing below this.

If the crossover is high enough, a really cheap, small tweeter should do great!

I am thinking the cash would be spent on the mid driver under this senario, the rest would be cheap.

Nah - there's no way to push even a 3" driver up that high without compromising directivity. Check the Scan 10f off-axis response at hificompass. At 60 degrees out, it starts crashing above 4k. https://hificompass.com/en/speakers/measurements/scan-speak/scanspeak-10f4424g00 . Even the Dayton RS52 which is a 2" dome doesn't really hold up at 7k - it's over 10dB down at 60 degrees. (and that specific driver isn't usable that high anyway due to the massive break-up) https://hificompass.com/en/speakers/measurements/dayton-audio/dayton-audio-rs52an-8

In other words, the idea of a very-wide-band mid is a different design philosophy from the nominal ASR approved Harman/Toole approach that we are trying to base this project on.

Having said that, if you're OK with narrowing directivity, the RS225/Scan10F 'FAST' design by xrk971 on diyaudio is an interesting option.

 

[dwkdnvr]

My golden rule of succesful crossover design: clearly define the full set of target responses of the ways (including the final bass roll-off) and work from directly from there, that is determine the correction function directly by division (and make sure it can be realized in analog if DSP XO isn't planned, which means the correction must be minimum phase). Oftentimes, one can feed the correction target in an optimizer that adjusts the parameters of a set of transfer function building blocks until a fit is achieved. The choosen set of building blocks of course needs some educated guesses. From there, one can get the component values of actual circuits, again using the optmizer. This works really well for active XOs.

I trimmed your post to avoid over-quoting, but there is a lot to think about here. I think the ultimate example of what you're describing is the Duelund xover design that JonMarsh/ET over at HTGuide are using in a couple places. This strategy defines all 3 target response functions in terms of 2 parameters - a midrange center frequency, and an aleph value describing the steepness of the midrange rolloff curves. The entire system phase response ends up being equivalent to a 2-way LR2 response - only 180 degrees of phase rotation through the entire spectrum. I find this approach very intriguing, but it puts a lot of demand on the drivers as their pass-bands are very broad, and it's VERY difficult to realize passively - as Jon Marsh said "Dueland crossovers are not for the faint of heart or light of wallet". However, if we decide to undertake a full 3-way active DSP design, it might well be worth considering. It would absolutely influence driver choice, though.

Here is an overview of the concept http://www.htguide.com/forum/showthread.php?41692-Proposal-for-a-compact-three-way&highlight=duelund