Question for John Marsh about his crossovers

[Curly Woods]

John,

I am curious why your designs require such large value capacitors as a general rule? I have no experience with crossovers past simplistic designs(basic 2nd order-4th order).

Is there a down side to the large capacitance loads as far as what an amplifier sees? I have no doubt about your expertise, just trying to learn something maybe

[BobEllis]

I'm not Jon, but basically the caps are large because the crossover frequencies Jon uses are low, and the filters high order. If you model the crossovers, you'll see that the filters are pretty well behaved - the impedance phase doesn't go terribly reactive. As long as your amp is happy driving the minimum impedance, it should be fine.

Hope this helps.

[JonMarsh]

Well, my unofficial spokes person Bob Ellis has answered the question quite well. There's a big difference in the size of components you need at 1250 or 1400 Hz versus 2.5 kHz or 3 kHz; low order versus high order networks further that difference.

To amplify only slightly on that, the reasons to cross low are two fold-

1) to stay away if possible from even the early minor cone modes, which contribute to loss of clarity, especially in the upper midrange and presence region

2) AND to keep the integration in the crossover region where the center to center spacing should be no more than a half wavelength.

The disadvantages of that are greater demands on the tweeter, and larger crossover components.

That's also why a good three way speaker is so much more expensive than a two way- the lower crossover uses much larger components.

In the past, it's not uncommon to see a 7" MTM with crossover frequency of 3 kHz or so. Show me a 7" midwoofer that remains pistonic to 4.5 kHz to handle the crossover transistion.... Show me how it's possible to have a driver spread between three drivers that's only 4.5" for a 7" driver MTM... kinda tough, isn't it? In fact, it's kind of tough to do a 5-1/4" MT with that spacing. There's no getting around what happens in the crossover region and the net power response if you ignore such guidelines- and uniform power response and off axis response is even more important, IMO, than ruler flat on axis response.

Just my 0.02.

[Curly Woods]

Thanks Jon and Bob

Now as a follow-up question with regards to xover capacitors,. Often the values are quite large and the best film caps cost is not insignificant. Lets say I need a 100uF cap. I would have to use a Solen for most of this value to keep costs low, but I could use a AuriCap or SoniCap in parallel. Is there a general run as to what the ratio of capacitance of the better film caps should be in a case like this?

I strive for perfection, but know that ican mean expensive.

[BobEllis]

I remember reading a paper on Harman's website that smooth off axis and power response are Revel's main design goals, taking priority over smooth on axis response. they did large scale blind testing that showed speakers with smooth power response vs frequency curves always sounded better to trained listeners, even when on axis frequency response was not as smooth. Not having discontinuity in the curves seemed to be more important than absolute directivity.

So, Jon is not alone

[JohnL]

This guy always seems to have some big value caps for cheap. I'm sure they don't sound like the expensive boutique caps, but you can always bypass with better ones. His prices are pretty close to what you would pay for electrolytics.

John

Cheap Caps on ebay

[capslock]

Show me a 7" midwoofer that remains pistonic to 4.5 kHz to handle the crossover transistion....

Well, I can show you two, the W18 and the old L18 (H1142, straight cone). Looking at the CSDs published in HobbyHifi, they seem to be pistonic up to their first resonance.

(Comparing old and new L18 FR plot from the Seas data sheets, you'll notice the new version begins to droop at 1.8 kHz or so. Björn said at the time of the introduction of the new curved cone that the curvature essentially gave rise to gradual decoupling of the outer cone regions as frequency rises).

The problem with both the W18 and L18 is that harmonics get amplified by the peak in the FR, so you get a huge 3rd harmonic peak at 1.6 kHz and a still seizable 5th at 1 kHz. This and the need to generate enough linear suppression at 4.8 kHZ are the reasons most designers prefer to cut off these drivers by 1.5 kHz.


I've been thinking along two lines to make these great drivers more usable:

1) Reduce motor distortion by using a counterwound coil or inductive feedback (Jon's gonna barf when he reads this...).

2) Reducing the FR peak along the lines of the Vifa NRSC patent. The first resonance of a cone is usually a concentric mode, with bending waves travelling radially outward, getting reflected from the cone edge (impedance mismatch cone/surround) and hence forming a standing wave pattern. The Vifa cones have five linear segments cut from the circumference from the cone, but the depth of these segments is less than the overlap of cone and surround. The idea is that the time and amplitude of the reflection will be different in these regions compared to the uncut regions. Hence, the reflected wave will no longer be circular, and the resulting standing wave pattern will not consist of clearly defined circular nodes, resulting in a cancellation in the far field.

This could also be achieved by gluing little tangential bars of rubber, balsa etc to some regions of the cone edge, or by using radial stiffeners glued to the cone itself, changing the propagation speed as a function of rotational angle.

[Evil Twin]

Now as a follow-up question with regards to xover capacitors,. Often the values are quite large and the best film caps cost is not insignificant. Lets say I need a 100uF cap. I would have to use a Solen for most of this value to keep costs low, but I could use a AuriCap or SoniCap in parallel. Is there a general run as to what the ratio of capacitance of the better film caps should be in a case like this?

I strive for perfection, but know that ican mean expensive.

One way to get lower costs is paralleling good parts that are available at lower prices, like the GE Polypropylene caps which Madisound offers. Check out the Arvo threads, where ThomasW uses those, or some other threads.

Some of the designs I've done recently have been focused on unconventional crossover approaches which work to minimize the cost while still achieving 80-90% of the performance of my more "all out" designs- the Natlie P and Modula MT are examples, with their "unusual" network design.

Understand that designs like the Modula MTM and the M8ta are "aimed" at commercial speakers in the $3000 and up range. Since you can build a set of Modula MTM's, expensive crossover parts an all, for well under $1K, I think that's pretty good value. (remember, my "cut" is even less than the pittance iTunes gets for selling songs online). Compare the cost of buying the Thor Kit at Madisound with the premium componets to building the Modula MTM- it's close to a 2:1 ratio. The performance is not in that ratio at all, though. This is possible becuase of the excellent cost performance of the RS180's, and selecting other parts with similar cost performance ratio's.

IMO, the Natalie P is the best value, becuase the crossover cost is similar to the Modula MT, but the SPL and distortion performance is close to the Modula MTM. I actually like the Modula MT a little better (more extended bottom end, though can't play as loud), but for a lot of folks, especially in an HT setup, the Natlie P would make more sense. If you couple the Natlie P with a decent sub setup, and good placement in the room and source electronics (OK, I'm "cheating" with Ayre electcronics, Cardas cables, etc), some pretty scary performance emerges (example; for the Natlie P, a quick two channel upgrade would be dual bass bins/subs using RS sub drivers and biamped at 100-125 Hz) (say, with something like the Sumo Delilah Crossover I have), then you have sound (for boxes) which plays well out of their price class.

I prefer the RS subs because of their low distortion even at higher frequencies (above 75 Hz), very stiff cone, and low Le (fast rise time). Many sub drivers start rolling off above 70 Hz on their own- this makes integration more difficult, due to excess phase shift and problems with amplitude matching.


~Jon

[JonW]

IMO, the Natalie P is the best value, becuase the crossover cost is similar to the Modula MT, but the SPL and distortion performance is close to the Modula MTM. I actually like the Modula MT a little better (more extended bottom end, though can't play as loud), but for a lot of folks, especially in an HT setup, the Natlie P would make more sense. If you couple the Natlie P with a decent sub setup, and good placement in the room and source electronics (OK, I'm "cheating" with Ayre electcronics, Cardas cables, etc), some pretty scary performance emerges (example; for the Natlie P, a quick two channel upgrade would be dual bass bins/subs using RS sub drivers and biamped at 100-125 Hz) (say, with something like the Sumo Delilah Crossover I have), then you have sound (for boxes) which plays well out of their price class.

Jon-

I appreciate seeing your opinions (on your own designs) of the NatP vs. the Modula MT. I’d like to try and build some speakers. Never tried that before and I don’t know what I’m doing at all. So I’m trying to pick a good first project. No real constraints, I guess. I just want good sound (for music, HT is much less important) and a project I can do. So not the most complicated speakers to start with. Any opinions on what to go for first?

I’d really like to try a clone or modification/improvement (e.g., MTM) of the Totem Forest, the speakers I currently own. But maybe I’d first better get some experience with a more established DIY design? Thanks.

-Jon

[cjd]

JonW:

I think that once you get into the DIY thing, the inclination to try to clone will diminish.

If the ~$250 budget for a pair is particularly tempting, Jon's Modula MT is probably the way to go if you don't need the sensitivity. If you want more sensitivity, my MTM may fit the bill. The crossover in mine is probably simpler to follow and implement for a first-time effort, but it's also bigger - though it can be built as a tower or a "bookshelf" variant. I think it may be a wash as to which delivers more bass due to differences in XMax and sensitivity. And I think my MTM is a slightly less refined crossover, though the recent addition of the 1800 crossover point variant with the Seas tweeter may make that statement inaccurate.

C

[JonMarsh]

Well, I can show you two, the W18 and the old L18 (H1142, straight cone). Looking at the CSDs published in HobbyHifi, they seem to be pistonic up to their first resonance.

(Comparing old and new L18 FR plot from the Seas data sheets, you'll notice the new version begins to droop at 1.8 kHz or so. Björn said at the time of the introduction of the new curved cone that the curvature essentially gave rise to gradual decoupling of the outer cone regions as frequency rises).

The problem with both the W18 and L18 is that harmonics get amplified by the peak in the FR, so you get a huge 3rd harmonic peak at 1.6 kHz and a still seizable 5th at 1 kHz. This and the need to generate enough linear suppression at 4.8 kHZ are the reasons most designers prefer to cut off these drivers by 1.5 kHz.


I've been thinking along two lines to make these great drivers more usable:

1) Reduce motor distortion by using a counterwound coil or inductive feedback (Jon's gonna barf when he reads this...).

2) Reducing the FR peak along the lines of the Vifa NRSC patent. The first resonance of a cone is usually a concentric mode, with bending waves travelling radially outward, getting reflected from the cone edge (impedance mismatch cone/surround) and hence forming a standing wave pattern. The Vifa cones have five linear segments cut from the circumference from the cone, but the depth of these segments is less than the overlap of cone and surround. The idea is that the time and amplitude of the reflection will be different in these regions compared to the uncut regions. Hence, the reflected wave will no longer be circular, and the resulting standing wave pattern will not consist of clearly defined circular nodes, resulting in a cancellation in the far field.

This could also be achieved by gluing little tangential bars of rubber, balsa etc to some regions of the cone edge, or by using radial stiffeners glued to the cone itself, changing the propagation speed as a function of rotational angle.

You know me well, Eric, using relying on farfield cancellation of cone modes does tend to get me a bit queasy- probably why I get all preach about a variety of drivers some others think are ok- and why Dan Wiggins and I don't see eye to eye about the Extremis 6.8.

Maybe I should have said earlier, show me a 7" driver under $100 that is smoothly pistonic to 4.5 kHz AND doesn't have an upper end peak that will amplifier the 2nd or 3rd harmonic distortion components. As nice as the W18 is, I still think it works best crossed no higher than 1.8 kHz. I'm hoping to get away with a clean 2 kHz with the C90/T6. But I'll be watching those distortion harmonics closely....

~Jon

[Brian Walter]

Jon, where can you buy the C90/T6? I don't see it listed on Madisounds website.

Brian

[JonMarsh]

Hello Brian,

Not sure where you're looking, but if you go to the Madisound home page, click on the Accuton logo, it takes you to online ordering, then click on the Accuton text link (underlined), and that will take you to the main online Accuton page. (I've had drivers sometimes not show up in Peerless page depending on which route you take to get there, too).

On this page should be 13 Accuton driver models, including the C90-T6 at the left side of the last full row of drivers (below that is the C20 diamond tweeter). When I ordered mine they were out of stock, would have them soon in Europe; the last update indicates shipping from Europe late this week or early next week (I'm paying for expeditied shipment on top of the "healthy" base price. Ouch. Needless to say, I'm going to treat them pretty carefully. As efficient as they are, they won't normally need much power.

I'm ordering a few other bits from Madisound this weekend, some XG18WH00, to round out the selection for the midwoofer shootout, and probably a pair of C13-6 (I'm still hemming and hawing a little bit over the cost). With it's main breakup mode at about 34 kHz, and what may be moderately low distortion above 2 kHz, it looks like a possible candidate. I'm just a little nervous about what I may want to do with it...

[JonW]

JonW:

I think that once you get into the DIY thing, the inclination to try to clone will diminish.

Yeah, and from what I gather, the inclination to build more and more speakers seems to take hold as well.

If the ~$250 budget for a pair is particularly tempting, Jon's Modula MT is probably the way to go if you don't need the sensitivity. If you want more sensitivity, my MTM may fit the bill. The crossover in mine is probably simpler to follow and implement for a first-time effort, but it's also bigger - though it can be built as a tower or a "bookshelf" variant. I think it may be a wash as to which delivers more bass due to differences in XMax and sensitivity. And I think my MTM is a slightly less refined crossover, though the recent addition of the 1800 crossover point variant with the Seas tweeter may make that statement inaccurate.

No idea if I'd need sensitivity. I'd have to buy an amp for the speakers anyway. You prefer the Modula MT over the NatP?

Is this your speaker?

http://www.eldamar.net/audio/rs150mtm/

Given that this will be my first project, I'll need a lot more hand holding than the detail on that web site (e.g., crossover photos, etc.). And I figure it might be best to make something that other people have made as well. So that biases (but by no means restricts) me to the stickied speakers.

Before buying those Totems, I listened to many speakers from many companies. The Forests had just the sound I wanted. So if speaker project #1 goes well...

[cjd]

JonW: that's my site.

I have plans to update and improve what's there. Not sure I want to dig in and snap piccies of my crossovers since they're a royal pain to get to given how I mounted them (my own fault, of course). I think I got more of the goods on the 3-way construction though. If you decide that's the project for you, I can always dig in and provide additional info as needed - it will only help others that might wish to build this.

I haven't heard the Modula MT. I was basing off Jon's comments.

As far as sensitivity, who knows. That's your call. I have a preference for higher sensitivity, but sometimes form-factor is more important. I'm considering a pair of the Modula MT's for a spot in our house actually...

C

[Brian Walter]

Jon M,

I was looking for the C90-T6 by going from Madisounds home page to Catolog, then Accuton. If you go at it that way, the C90-T6 isn't listed. I see what you mean by expensive though, it quite a bit higher than most of their other drivers.

Brian Walter

[JonMarsh]

Yes, but positively cheap compared with their Diamon tweeters. I hear the new "Black Diamond" model is even MORE expensive. What's surprising is not being able to find good measurements on Accuton's web site.

[capslock]

You know me well, Eric, using relying on farfield cancellation of cone modes does tend to get me a bit queasy- probably why I get all preach about a variety of drivers some others think are ok- and why Dan Wiggins and I don't see eye to eye about the Extremis 6.8.

Well, if I succeed in using inductive feedback, I won't be relying on far field cancellation, I simply make the motor behave so the frequency response has nothing to unduly amplify.

As for cone modes, I'm pretty sure that a flexing paper cone that is damped with some goo will generate lots of hysteresis distortion. Metal has much less damping, so having it flex not necessarily a bad thing, you just have to make the standing wave pattern chaotic enough that it does not get back at you by radiating resonantly under 25.8°

By the way, the Peerless sandwich cones also flex like mad. Still, they are capable of delivering very low distortion if used with the right (phase plugged and shorted) motor.

[jj2stack]

DAYTON 7" Reference 3-Way Crossover Design

Hello John! I'm new to the site, so please excuse my bad manners if I exhibit any! I'm looking for recommendations on a 3-way crossover design! I'm looking at Daytons 7" Reference Driver #295-364 (8 ohm), Dayton Reference 2" Alum. Dome Mid Range #285-020 And Vifa's D26TG-35 Tweeters! I have one recommendation from PE, but would like to look at all my options before I get into a build. They will be a bit on the pricey side I know, hence be sure about a good crossover..Any Help or pointers would be appreciated...Jerry

[---k---]

Jerry,
You saw this: http://htguide.com/forum/showthread.php4?t=27218
Right?

[jj2stack]

7" Ref. Crossover!

Pretty Close! I got the inspiration from PE's DIY Showcase! The Tubes, I beleive they were called. Instead of two drivers, I just want to use one driver! I got a good deal on a new pair of the 7" drivers awhile back. One thing leads to another!....Jerry

[JimS]

This guy always seems to have some big value caps for cheap. I'm sure they don't sound like the expensive boutique caps, but you can always bypass with better ones. His prices are pretty close to what you would pay for electrolytics.

John

When I measured a bunch of the Aerovox caps for a similar project they all came out to within +/- 1% typically on the high side. No comment on their "sound" :B