Here's the W15 bandpass. It was a lot trickier to trace the MCA12 curve, and maybe it can be done more simply.
Observations:
-mid is inverted as per Troels
-don't recognize the filtered slopes as particularly close to any textbook shapes; just have to trust the differences come from baffle step and diffraction, which I don't know how to verify using this approach
-attenuation resistor was removed for W15 to match filtered amplitude of MCA12
-W15 needs 4th order electrical HP and 3rd order electrical LP to match MCA12 filtered slopes
-there's a small peak on the W15 around 5k - might need a notch filter
-so I agree with the concern that by not doing a native design bespoke for the specific driver, you're perhaps precluding a simpler crossover
-then again, the overall transfer function does seem to be able to be replicated
Observations:
-mid is inverted as per Troels
-don't recognize the filtered slopes as particularly close to any textbook shapes; just have to trust the differences come from baffle step and diffraction, which I don't know how to verify using this approach
-attenuation resistor was removed for W15 to match filtered amplitude of MCA12
-W15 needs 4th order electrical HP and 3rd order electrical LP to match MCA12 filtered slopes
-there's a small peak on the W15 around 5k - might need a notch filter
-so I agree with the concern that by not doing a native design bespoke for the specific driver, you're perhaps precluding a simpler crossover
-then again, the overall transfer function does seem to be able to be replicated
Last edited:
Moment of truth:
Original XO and transfer function, summed response:
Refitted XO with W22 + W15 Nextel + T25 Millenium:
Quite close! Probably can be optimized. But hmm, looks like this would work as a starting point, and overall it should sound similar to the original.
The Millenium can also be swapped for the T25CF001 -- might address the recess near the XO, but I need to take another break!
Original XO and transfer function, summed response:
Refitted XO with W22 + W15 Nextel + T25 Millenium:
Quite close! Probably can be optimized. But hmm, looks like this would work as a starting point, and overall it should sound similar to the original.
The Millenium can also be swapped for the T25CF001 -- might address the recess near the XO, but I need to take another break!
Last edited:
Swapped in the T25CF001. It tracks the T27 curve easily, but there's still a final response wrinkle around 6.5khz where one may need to just optimize it separately, ditching the original curve from here. Impedance also looks smoother with the CF001 instead of the Millenium. I think it'd be a good choice.
I also learned how to use a target curve by creating SPL files from the original XO per driver; attached below with the T25CF001 FRD and ZMA.
I also learned how to use a target curve by creating SPL files from the original XO per driver; attached below with the T25CF001 FRD and ZMA.
Last edited:
Hi TroutBear,
see, I'm not trying to convince you of an opinion, but serve you some facts from reality. So you may get some new ideas that might fit your situation better than the ones you have right now.
Combinations with passive x-over networks can sound absolutely perfect, even better if you find a matching amp. Not any active speaker is sounding good, many are awful.
I use both, active and passive, even mixed, like active subwoofer with passive 2-way.
Funny you mention the Hafler, as I used the DIYS version of it for two decades and considered it my reverence point. Anyway, objective testing showed me that a LM3886 chip amp or some very low priced D-amps beat it in any A-B test. Not like some minor "maybe better if you listen closely", but clear and repeatable on different speakers with any music. This result was pretty shocking for me, as the Hafler was one of the best all round amps I ever had. The competing A/B and D amps cost was less than its cabinet or heat sink. Time to face reality... some A/B chip amps are not cheap substitutes and a lot of D-amps don't sound digital, cold or synthetic, but simply perfect. On the other hand, there are some D-amp brands mentioned here, that I admit to dislike and consider overpriced.
Some here seem to have some very romantic views about passive loudspeaker design.
If there is anyone who doubts that developing a 3-way speaker is a job of many weeks concentrated work, he has no clue. Maybe an x-over that prevents instant destruction of the chassis is done in a day, but really making some combination of world class speakers sing, is a different pair of shoes. By the way, even the cheapest SEAS speakers are world class and used in extremely expensive commercial products. Often optically modified.
On the other hand, active speaker systems are set up in a few hours. If anyone does not believe, go to a concert of your favorite super group. Such speaker systems, made to please thousands of paying guests, are set up in a few hours. In a different location every evening, with different gear. 100% active. Of course, these guys who can do such jobs don't come cheap and use all kinds of measuring equipment and DSP's. Today bad sound on a concert has become rare.
I understand that software, DSP's and D-amps are considered devils toys by some fundamentalists. May Allah be with you.
I may give you an example of what you have to do for your specific speaker combination. If the speaker tells you "please do a 18dB high pass with a shelf filter to compensate for the baffle step and a 12dB low pass," what is simpler for you to do: Finding the point " 400Hz 18dB high pass", "shelf filter" and "3000Hz 12dB low pass" in a simple software or constructing one from capacitors, coils and resistors. This is for a resonating resistance that changes over frequency!
In a DSP software you can do this in less than two minutes, maybe 10, if you do it for the first time. Such software is not complicated and only people who do not measure what they do have problems with it.
Using passive parts, a basic result will take you days of trying and passive parts worth a few hundred $ at least, just for something that goes in the direction of the correct filter for a such a mid chassis. The reality is even more complicated and frustrating. Very often professional developers change chassis, as they simply don't fit in a combination.
Then, forget about changing a driver and modifying the existing x-over. This is something that may work for the tweeter in a very simple two way system or, but a bit more complicated, if you add a second identical woofer to a two way. Changing the mid in a 3-way is a no go and you end up developing a new x-over. If you do not trust me, ask Troels or any competent developer with a reputation.
Oh, one more thing: Most DSP software can be downloaded and used for free, even without a DSP connected. Just like measuring software. Give it a try, all you can lose is a wrong opinion. Try the cheap Dayton DSP, a good example and well working.
Last: If you are afraid of DSP chips as well as A/D and D/A conversions: Any music you listen to has gone through dozens of them until the recording is finished. Studio gear uses quite cheap ones, that may surprise you. If you have a few hundreds of them in an average professional studio, they have to be cheap!
Sorry, it's not 1999, but 2022! Digital stuff is here to stay...
see, I'm not trying to convince you of an opinion, but serve you some facts from reality. So you may get some new ideas that might fit your situation better than the ones you have right now.
Combinations with passive x-over networks can sound absolutely perfect, even better if you find a matching amp. Not any active speaker is sounding good, many are awful.
I use both, active and passive, even mixed, like active subwoofer with passive 2-way.
Funny you mention the Hafler, as I used the DIYS version of it for two decades and considered it my reverence point. Anyway, objective testing showed me that a LM3886 chip amp or some very low priced D-amps beat it in any A-B test. Not like some minor "maybe better if you listen closely", but clear and repeatable on different speakers with any music. This result was pretty shocking for me, as the Hafler was one of the best all round amps I ever had. The competing A/B and D amps cost was less than its cabinet or heat sink. Time to face reality... some A/B chip amps are not cheap substitutes and a lot of D-amps don't sound digital, cold or synthetic, but simply perfect. On the other hand, there are some D-amp brands mentioned here, that I admit to dislike and consider overpriced.
Some here seem to have some very romantic views about passive loudspeaker design.
If there is anyone who doubts that developing a 3-way speaker is a job of many weeks concentrated work, he has no clue. Maybe an x-over that prevents instant destruction of the chassis is done in a day, but really making some combination of world class speakers sing, is a different pair of shoes. By the way, even the cheapest SEAS speakers are world class and used in extremely expensive commercial products. Often optically modified.
On the other hand, active speaker systems are set up in a few hours. If anyone does not believe, go to a concert of your favorite super group. Such speaker systems, made to please thousands of paying guests, are set up in a few hours. In a different location every evening, with different gear. 100% active. Of course, these guys who can do such jobs don't come cheap and use all kinds of measuring equipment and DSP's. Today bad sound on a concert has become rare.
I understand that software, DSP's and D-amps are considered devils toys by some fundamentalists. May Allah be with you.
I may give you an example of what you have to do for your specific speaker combination. If the speaker tells you "please do a 18dB high pass with a shelf filter to compensate for the baffle step and a 12dB low pass," what is simpler for you to do: Finding the point " 400Hz 18dB high pass", "shelf filter" and "3000Hz 12dB low pass" in a simple software or constructing one from capacitors, coils and resistors. This is for a resonating resistance that changes over frequency!
In a DSP software you can do this in less than two minutes, maybe 10, if you do it for the first time. Such software is not complicated and only people who do not measure what they do have problems with it.
Using passive parts, a basic result will take you days of trying and passive parts worth a few hundred $ at least, just for something that goes in the direction of the correct filter for a such a mid chassis. The reality is even more complicated and frustrating. Very often professional developers change chassis, as they simply don't fit in a combination.
Then, forget about changing a driver and modifying the existing x-over. This is something that may work for the tweeter in a very simple two way system or, but a bit more complicated, if you add a second identical woofer to a two way. Changing the mid in a 3-way is a no go and you end up developing a new x-over. If you do not trust me, ask Troels or any competent developer with a reputation.
Oh, one more thing: Most DSP software can be downloaded and used for free, even without a DSP connected. Just like measuring software. Give it a try, all you can lose is a wrong opinion. Try the cheap Dayton DSP, a good example and well working.
Last: If you are afraid of DSP chips as well as A/D and D/A conversions: Any music you listen to has gone through dozens of them until the recording is finished. Studio gear uses quite cheap ones, that may surprise you. If you have a few hundreds of them in an average professional studio, they have to be cheap!
Sorry, it's not 1999, but 2022! Digital stuff is here to stay...
Some good work shown here motokok.
Either you force the woofer down, or you bump up the target curve by the needed dB. In that case duplicate and increase the tweeter level as well, and alter the tweeter crossover to that target.
The breakup region can have axial differences which don't show on a response plot. Fortunately the aim is to cut the breakup region out anyway, and the response is down. The drivers are the same size so it happens around the same frequencies and what you see shouldn't be unexpected. Looking at phase, the response peak hasn't taken it for too much of a ride.
No problem.
Thanks, @AllenB! I hate manual SPL tracing but I learned a lot from this exercise, and I think the 3WC is a very cool, public DIY design.
I think I'm done with modelling for now. Would you mind giving your perspective on some of the objections to the OP's design intentions here? Specifically, the worthiness of replicating transfer functions in simulation?
@Turbowatch2 - you make some fair points I agree with on but I also don't care for your unilateral and didactic view on this fun exploration. I also don't like the use of words like fundamentalism and referring to Allah in an audio forum discussion thread. You make some contradictory points, citing Troels as the genius-level expert (yeah, with LSPCAD optimizers) + decrying boutique passive parts and shiz -- who even brought that up? And what does Troels himself like to use in his kits? Dayton and Bennic caps? Don 't tihnk so.
I personally don't like the Alliexpress type / "ASR recommended" digital amps, DACs and associated codswallop. They're not for me, and the OP already said they wanted to use their Hafler/Mac + passive. I couldn't give a fig if the ASR crowd thinks Revel speakers are the dogs b******s, cause I've heard many of them and find they sound horrible. I prefer analogue amps just cause I do. If I had the skills to make a bespoke active analogue crossover, like Linkwitz - who is worthy of being called a genius, RIP - that would be ideal, but alas. I haven't tried the latest PC based crossover software, but I have owned and heard Hypex amps, the cheapo TI things, and older versions of minidsp and I just don't like them as much as say ATC actives or even chipamp ABs. I'm not going to get 6 SMSL power amps for my dsp 3 ways cause they've been objectively proven in a double blind test to be indistinguishable from a Benchmark. Forget that.
I agree measurement is indisputably the way to go, but that does not make passive crossovers, simulation or construction, meaningless.
I think I'm done with modelling for now. Would you mind giving your perspective on some of the objections to the OP's design intentions here? Specifically, the worthiness of replicating transfer functions in simulation?
@Turbowatch2 - you make some fair points I agree with on but I also don't care for your unilateral and didactic view on this fun exploration. I also don't like the use of words like fundamentalism and referring to Allah in an audio forum discussion thread. You make some contradictory points, citing Troels as the genius-level expert (yeah, with LSPCAD optimizers) + decrying boutique passive parts and shiz -- who even brought that up? And what does Troels himself like to use in his kits? Dayton and Bennic caps? Don 't tihnk so.
I personally don't like the Alliexpress type / "ASR recommended" digital amps, DACs and associated codswallop. They're not for me, and the OP already said they wanted to use their Hafler/Mac + passive. I couldn't give a fig if the ASR crowd thinks Revel speakers are the dogs b******s, cause I've heard many of them and find they sound horrible. I prefer analogue amps just cause I do. If I had the skills to make a bespoke active analogue crossover, like Linkwitz - who is worthy of being called a genius, RIP - that would be ideal, but alas. I haven't tried the latest PC based crossover software, but I have owned and heard Hypex amps, the cheapo TI things, and older versions of minidsp and I just don't like them as much as say ATC actives or even chipamp ABs. I'm not going to get 6 SMSL power amps for my dsp 3 ways cause they've been objectively proven in a double blind test to be indistinguishable from a Benchmark. Forget that.
I agree measurement is indisputably the way to go, but that does not make passive crossovers, simulation or construction, meaningless.
Last edited:
I'd add that the more important part of a crossover is the acoustic design of the speaker. Once you do this, the job of the electronics is already defined.. regardless of how you do it.
Primarily you have two drivers which have similar directivities and which are operating within their pistonic region. Once you set the levels to be the same across the band, you should have the same result.
There can be differences, likely to be minor, but usually fixable with a little EQ or other similar tweaking. Small response differences can change the weight of the sound, amplifier output impedance can interact with speaker impedance.. both of these can be successfully equalised/tweaked. Harmonic distortion can vary between speakers but speaker distortion is not particularly audible or problematic.
I like to have access to global EQ with any kind of speaker. Any kind of useable/minimum phase EQ will be helpful in theory.
I may have missed where there was an objection?
Primarily you have two drivers which have similar directivities and which are operating within their pistonic region. Once you set the levels to be the same across the band, you should have the same result.
There can be differences, likely to be minor, but usually fixable with a little EQ or other similar tweaking. Small response differences can change the weight of the sound, amplifier output impedance can interact with speaker impedance.. both of these can be successfully equalised/tweaked. Harmonic distortion can vary between speakers but speaker distortion is not particularly audible or problematic.
I like to have access to global EQ with any kind of speaker. Any kind of useable/minimum phase EQ will be helpful in theory.
Thanks for that. Well, there have been many posts poo-pooing the whole endeavor for various reasons - don't use these drivers, this is pointless without measurements in baffle, go dsp, buy a kit, don't try to tweak the crossover with different drivers, don't make this design if not identical, don't bother doing sims. Harumph.
They may be right without knowing what the OP wants from his speakers. It can be a fun exercise to work out on paper what might be done with expensive drivers with a small amount of relevant information but it is unlikely to lead to a particularly good speaker without knowing how to assess good. Why use such expensive drivers with a hack job rather than a proper design? Why drop from a 10" to an 8" woofer unbalancing the speaker with respect to low frequency cone area for many cases? Why drop the sealed cabinet for a ported which, with adequate low frequency cone area, may be a better fit for how the speaker is used in the room? Or perhaps not. Perhaps the OP wants the prestige of expensive drivers or already possesses them, the looks of the KLH speaker in a smaller package with cost and technical performance being of less significance. Or perhaps not.
For me what is interesting is that the OP sorta/kinda/seems to know a hack with expensive drivers in a vaguely similar cabinet from a designer with leanings towards subjectivism and away from engineering is a bit doubtful but will believe with a bit of encouragement. Then again perhaps I am spending too much time reading things into forum posts.
Yes, that's good modelling work by motokok but I think unfortunately the underlying assumption behind it is flawed.
In matching the responses from 1 set of drivers with another we are assuming that the differences between the measurements in the spec sheets and the measurements in the real world for each set of drivers is more or less the same. Or to put it another way, we are assuming that the 2nd set of drivers has exactly the same discrepancies between spec sheets and real measurements in their FR's and impedances as the original ones which, even though the drivers are all from the same manufacturer, is an idea that may or may not be true but more than likely not. So still a bit of a shot in the dark in other words.
Staying solely in the realm of modeling, you might be better off if you can find pre-existing builds with on baffle measurements using the upgraded drivers and then work backwards from there, subtracting out the baffle effects to obtain more accurate real world measured FR's and maybe Z's too that can then be
used to model the new design after the original.
As another suggestion, if the OP might be willing to learn how to take accurate measurements, then there are any number of competent people here who can take those and do the xo simulation work for you which should give you the best results.
In matching the responses from 1 set of drivers with another we are assuming that the differences between the measurements in the spec sheets and the measurements in the real world for each set of drivers is more or less the same. Or to put it another way, we are assuming that the 2nd set of drivers has exactly the same discrepancies between spec sheets and real measurements in their FR's and impedances as the original ones which, even though the drivers are all from the same manufacturer, is an idea that may or may not be true but more than likely not. So still a bit of a shot in the dark in other words.
Staying solely in the realm of modeling, you might be better off if you can find pre-existing builds with on baffle measurements using the upgraded drivers and then work backwards from there, subtracting out the baffle effects to obtain more accurate real world measured FR's and maybe Z's too that can then be
used to model the new design after the original.
As another suggestion, if the OP might be willing to learn how to take accurate measurements, then there are any number of competent people here who can take those and do the xo simulation work for you which should give you the best results.
Motokok & AllenB - I'm impressed with your enthusiasm to share knowledge in this forum - exactly what these town halls really should be about. I'm sold on the CF001 tweeter substitution from my original selection and it appears the W22 woofer may also be a sound pick. The midrange is what looks to have driven complexity - perhaps I should rethink using the one in the current Seas Classic and also search for one with published characteristics more closely aligned with that starting point to see if the Xover could be simplified in that area.
Turbowatch2 - while your passion for active shows through clearly, I'm simply not interested in the route - and do not consider my opinion "wrong" in this, just different from your. Let's mutually respect that difference here.
To others on what am I thinking, I consider it fairly basic. It's not about the prestige of expensive drivers; I've simply spent a career in mechanical design observing that there is often - not exclusively, but mostly - a reason for increased cost that results from materials and tolerances impacting performance. It's not been discussed, but I'll take that same route with the Cover components. So I consider it reasonable to look at the current market selections and how they may have evolved from the original design point. I've also worked with enough control experts [mechanical, not electrical granted] to see how a knowledge model is absolutely the right guidance on concept, but almost always becomes directionally correct while requiring tuning after build. MotoKok's work [Thanks once more!] here seems to show promise in the idea - enough to build it in test level cabinets - and also directed me to keep looking at the midrange options before doing so.
I mentioned the KLH 5 as an example - not because I think sealed systems are the singular path. They were good speakers when introduced and if still made in the US, I'd likely have just ordered a set. But, ported is good also - my JBL 4311's are with my daughter, but I enjoyed them for many years. I recently picked up a used pair of B&W 685s to try out.
Another perspective, that's just mine but since "what is he thinking / why create a "hack" came up - in the 50's - early 70's there were less tools, more cookbooks if you will [EV & Altec come to mind] and projects like this were fun. Tech advancement is awesome - I'd not want my designers to give up CAD and get out pens as I started with - but build it, try it and learn should still be that way. And yes, I'm somewhat motivated to learn more on measurement software from this discussion, so perhaps that's a learning path for me.
Turbowatch2 - while your passion for active shows through clearly, I'm simply not interested in the route - and do not consider my opinion "wrong" in this, just different from your. Let's mutually respect that difference here.
To others on what am I thinking, I consider it fairly basic. It's not about the prestige of expensive drivers; I've simply spent a career in mechanical design observing that there is often - not exclusively, but mostly - a reason for increased cost that results from materials and tolerances impacting performance. It's not been discussed, but I'll take that same route with the Cover components. So I consider it reasonable to look at the current market selections and how they may have evolved from the original design point. I've also worked with enough control experts [mechanical, not electrical granted] to see how a knowledge model is absolutely the right guidance on concept, but almost always becomes directionally correct while requiring tuning after build. MotoKok's work [Thanks once more!] here seems to show promise in the idea - enough to build it in test level cabinets - and also directed me to keep looking at the midrange options before doing so.
I mentioned the KLH 5 as an example - not because I think sealed systems are the singular path. They were good speakers when introduced and if still made in the US, I'd likely have just ordered a set. But, ported is good also - my JBL 4311's are with my daughter, but I enjoyed them for many years. I recently picked up a used pair of B&W 685s to try out.
Another perspective, that's just mine but since "what is he thinking / why create a "hack" came up - in the 50's - early 70's there were less tools, more cookbooks if you will [EV & Altec come to mind] and projects like this were fun. Tech advancement is awesome - I'd not want my designers to give up CAD and get out pens as I started with - but build it, try it and learn should still be that way. And yes, I'm somewhat motivated to learn more on measurement software from this discussion, so perhaps that's a learning path for me.
That's all fair @andy19191 and @jReave's comments are totally on point.
An issue I wanted to restate if going ahead with this; the box volume should be reexamined for the W22. I think 37L is too small vs the RNX, so depth should be increased. I think Troels used a Q of 15 suggesting an F3 of 48hz for the RNX whereas the W22 doesn't seem to lend itself to this tuning. That might be the stronger reason to reconsider its suitability.
Now with the W15, I looked at the plots again and kind of have some doubts similar to what Troels mentioned about the M15 which is a weird dip at 1k in the response. So hmm, if the W22 isn't particularly strong at bass and the W15 isn't as pure a mid as the MCA12 then there are some issues here. Thing is, having heard the Nextel drivers, they do sound quite lovely, a slightly romantic fingerprint to them that I like.
@TroutBear you're very welcome for any help I gave. You've stated you have a cognitive value placed on cost-quality, which isn't always the case when it comes to drivers. If you wanted more assurance on quality then perhaps also look at Troels' Audio Technology 3WC variant (insanely expensive but nobody doubts the driver quality - more the sticker price) and the newer Revelator 851, though I find that looks a bit ugly. There's also the forum design Open Source Monkey Box which uses top quality drivers and a Volt midrange dome that is not cheap at all.
An issue I wanted to restate if going ahead with this; the box volume should be reexamined for the W22. I think 37L is too small vs the RNX, so depth should be increased. I think Troels used a Q of 15 suggesting an F3 of 48hz for the RNX whereas the W22 doesn't seem to lend itself to this tuning. That might be the stronger reason to reconsider its suitability.
Now with the W15, I looked at the plots again and kind of have some doubts similar to what Troels mentioned about the M15 which is a weird dip at 1k in the response. So hmm, if the W22 isn't particularly strong at bass and the W15 isn't as pure a mid as the MCA12 then there are some issues here. Thing is, having heard the Nextel drivers, they do sound quite lovely, a slightly romantic fingerprint to them that I like.
@TroutBear you're very welcome for any help I gave. You've stated you have a cognitive value placed on cost-quality, which isn't always the case when it comes to drivers. If you wanted more assurance on quality then perhaps also look at Troels' Audio Technology 3WC variant (insanely expensive but nobody doubts the driver quality - more the sticker price) and the newer Revelator 851, though I find that looks a bit ugly. There's also the forum design Open Source Monkey Box which uses top quality drivers and a Volt midrange dome that is not cheap at all.
Last edited:
Perhaps.. One goal of using Troels' design in this case as a basis, is that he has listened and tuned the speaker. Not that I am familiar with his judgement, but listening verification potentially covers a multitude of sins. You then move on to putting it in the room... If we start measuring cabinets on one axis then the diffraction contribution may not be easy to determine by level, its timing and audibility information will be unavailable, the in-room crossover lobing would remain untested.. and we'd be back to listening and tuning.