After reading the patent, carefuly, there's no magic, but it is clever. They are just using a transformer to realize zeroes.
Its a clever approach to cost reduction, for very high slopes: low parasitic losses. This is harder to achieve with separate inductors.
Since they advertise "infinite slope", by definition they are targeting elliptical, which is characterized by the highest slope transition possible for a given ripple.
Dave
The concepts are very different as are the challenges. The losses and distortions associated with wideband audio transformers are a completely different hurdle from those of standard inductors. Anyone who equates the two has no idea whatsoever of what they are talking about. For anyone who wants an education on the subject - they should contact experts such as those from Lundalh in Norrtalje, Sweden.
LUNDAHL TRANSFORMERS ABOUT US
Btw, still waiting on a "for real" comparison by way of amplitude, phase, and group delay plots from the person who said his ellipticals are the same as the infinite slope crossover patented by Richard Modaferri. If you can't provide proof of your claims by way of measurements - it's not simply a FAIL, but an EPIC FAIL. 😀
LUNDAHL TRANSFORMERS ABOUT US
Btw, still waiting on a "for real" comparison by way of amplitude, phase, and group delay plots from the person who said his ellipticals are the same as the infinite slope crossover patented by Richard Modaferri. If you can't provide proof of your claims by way of measurements - it's not simply a FAIL, but an EPIC FAIL. 😀
fntn,
Wideband audio transformers in the usual conversational context (e.g. tube amp output stages, mic transformers, line driver / balanced input or output transformers for pro audio), need to have a high ratio of magnetizing inductance and low leakage inductance, as well as low non-linear distortion. And the small signal levels require lots of turns and shielding. The lots of turns requirement tends to result in high stray capacitance which will affect the frequency response of the transformer, and so additional attention needs to be paid to winding lay / construction (e.g. use of tape between layers, or using sectioned windings).
The "wideband" part of audio transformers means the high ratio of magnetizing to leakage inductance. I haven't run the numbers, but I imagine such wideband audio transformers require a ratio on the order of 1000:1 (which is the ratio of 20 kHz to 20 Hz) or its inverse, 0.1%.
For the record, some types of SMPS transformers can have a leakage inductance <0.1% of the magnetizing inductance, off the top of my head. As such, they also have a "wide bandwidth".
In contrast, the Modafferi "transformers" are closer to "coupled chokes" wherein there is much lower coupling coefficient aka high leakage inductance, albeit a tuned value. They are not much more complicated to design and build and thus not much more expensive to make than a single passive crossover inductor. They will likely be less expensive than 2 such crossover inductors, and definitely less than the cost of 3 inductors which the single transformer replaces.
Combining 2 (or 3) crossover inductors into a single magnetic structure a la Modafferi doesn't magically make its design challenges similar to wideband audio transformers.
Wideband audio transformers in the usual conversational context (e.g. tube amp output stages, mic transformers, line driver / balanced input or output transformers for pro audio), need to have a high ratio of magnetizing inductance and low leakage inductance, as well as low non-linear distortion. And the small signal levels require lots of turns and shielding. The lots of turns requirement tends to result in high stray capacitance which will affect the frequency response of the transformer, and so additional attention needs to be paid to winding lay / construction (e.g. use of tape between layers, or using sectioned windings).
The "wideband" part of audio transformers means the high ratio of magnetizing to leakage inductance. I haven't run the numbers, but I imagine such wideband audio transformers require a ratio on the order of 1000:1 (which is the ratio of 20 kHz to 20 Hz) or its inverse, 0.1%.
For the record, some types of SMPS transformers can have a leakage inductance <0.1% of the magnetizing inductance, off the top of my head. As such, they also have a "wide bandwidth".
In contrast, the Modafferi "transformers" are closer to "coupled chokes" wherein there is much lower coupling coefficient aka high leakage inductance, albeit a tuned value. They are not much more complicated to design and build and thus not much more expensive to make than a single passive crossover inductor. They will likely be less expensive than 2 such crossover inductors, and definitely less than the cost of 3 inductors which the single transformer replaces.
Combining 2 (or 3) crossover inductors into a single magnetic structure a la Modafferi doesn't magically make its design challenges similar to wideband audio transformers.
Last edited:
fntn,
I never said "the same". I never said my transfer function is the same as Modafferi's. I said, both my crossover, and Modafferi's, use zero pairs (aka notches), which is also what "equiripple", and other, elliptical filters havve.
I never said "my filter's group delay is better or the same as Modafferi's". I said, "here's my filter's group delay, it looks favorable compared to L-R 4, and has a much narrower transition band, for the cost of an additional capacitor, compared to LR-4, and is also roughly in phase in the crossover region, which translates to reduced lobing."
My objective in designing the filter, was for car audio - I wanted a very narrow transition band to reduce driver interference with the widely separated midwoofer and tweeter (this was in the days before digital time alignment), and roughly constant energy in the crossover region (aka flat summation).
And see my reply above re: transformers.
The above is a gross oversimplification. Not only are they required to handle wide bandwidth (high ratio of magnetization to leakage) but they are also required to handle large amounts of current (low pass end) and not sustain unacceptable levels of saturation while at the same time handle small signals without a major compromise in linearity. These are diverging constraints which among other facets, add to the cost of a high fidelity audio transformer. Take a look at the Lundahl website and note the cost of high power, low distortion autotransformers. I'm not pretending to know exactly where Joseph is sourcing its current autotransformers these days but I know from an investigation I conducted several years ago into the matter - a single high performance autotransformer that can dissipate several hundred watts of power cleanly is more expensive than the entire crossovers of most commercial speakers. 🙂
I know they do not have to saturate. I did not want to write an entire treatise on transformer nonlinearities. Suffice to say I understand transformers and magnetic materials intimately.
I still stand by my statement that Modafferi's crossover auto-transformers are not a lot more complicated than crossover inductors, and are cheaper than 2 or the 3 it replaces. That is the reason for his using it (and reduced insertion loss), not some magical "infinite slope" "better than elliptic" nor "magnetic braking".
Neither crossover inductors nor auto-transformers "dissipate several hundred watts of power" - they would melt and that would mean most of the amplifier power goes into the crossover instead of the speaker - your speaker efficiency would plummet.
I still stand by my statement that Modafferi's crossover auto-transformers are not a lot more complicated than crossover inductors, and are cheaper than 2 or the 3 it replaces. That is the reason for his using it (and reduced insertion loss), not some magical "infinite slope" "better than elliptic" nor "magnetic braking".
Neither crossover inductors nor auto-transformers "dissipate several hundred watts of power" - they would melt and that would mean most of the amplifier power goes into the crossover instead of the speaker - your speaker efficiency would plummet.
They do in fact dissipate several hundred watts of power into a speaker driver load - depending on the current drawn by the driver.
For raw data, I recall getting a price of more than $200 for an autotransformer that could handle around 400 watts throughput. At the time (about 5 years ago, you could buy several chokes and caps for that kind of money, especially the cheaper variety).
But hey, you could be right. And Jeff Joseph could be spending a whole lot less and getting a better crossover for it if only he followed your advice. 🙄
So, I'm guessing with all of your knowledge and experience about crossovers, you've built a lot of speakers over the years (maybe more than Joseph) and have won many awards at shows and with magazines. Care to share some of the accolades and praise?
For raw data, I recall getting a price of more than $200 for an autotransformer that could handle around 400 watts throughput. At the time (about 5 years ago, you could buy several chokes and caps for that kind of money, especially the cheaper variety).
But hey, you could be right. And Jeff Joseph could be spending a whole lot less and getting a better crossover for it if only he followed your advice. 🙄
So, I'm guessing with all of your knowledge and experience about crossovers, you've built a lot of speakers over the years (maybe more than Joseph) and have won many awards at shows and with magazines. Care to share some of the accolades and praise?
I see you're going to be another one of those guys that doesn't actually read posts but just posts into the wind ranting about whatever comes to mind. It diminishes your posts.
We are in violent agreement about the benefit of using xformers vs inductors: cost reduction. ie you get use xformers vs expensive low-loss inductors that realize the same transfer function.
Read my post. It referred to the fact that these are ellipticals, or in the family there-of. Even better, read the patent. If you don't believe me, believe the patent authors. Until then, there's no point in discussing further
Dave
Last edited:
For your information, I read the full patent almost 6 years ago. The repeated claims by you and Cuadra about the similarities of typical ellipticals and the infinite slope design of Modaferri are an insult to Modaferri. The slopes are very different. The materials and operating principles are very different and the engineering challenges are very different. Most importantly, the end results are very different. Joseph Audio has fielded several speakers of modest size and cost that have received multiple "best of show" awards at prestigious shows and magazines.
I know of no acclaimed speaker designs that have achieved anywhere near the success that Joseph has that employ elliptical filters. Elliptical filters provide very little benefit to offset the high levels of group delay. You can't cite any data that supports the claimed similarity to ellipticals. You can't cite any speakers that Mr. Cuadra is responsible for that have even been reviewed let alone received widespread acclaim for performance. Sounds a lot like speaker envy. 😀
I know of no acclaimed speaker designs that have achieved anywhere near the success that Joseph has that employ elliptical filters. Elliptical filters provide very little benefit to offset the high levels of group delay. You can't cite any data that supports the claimed similarity to ellipticals. You can't cite any speakers that Mr. Cuadra is responsible for that have even been reviewed let alone received widespread acclaim for performance. Sounds a lot like speaker envy. 😀
Six years is a long time. I think a refresher is worthwhile:
Quoted out of the patent:
Infinite slope loudspeaker crossover filter - US Patent 7085389 Description
"FIG. 8a is a schematic diagram, 136, FIG. 8b is the positive frequency axis poles and zeros (p-z), and FIG. 8c is a typical frequency response plot of a prior art low-pass woofer crossover transfer function. Note that the mirror-image finite p-zon negative frequency axis are not shown. Three zeros are shown at infinity. If, however, a zero could be moved from infinity nearer to the positive frequency pole cluster by simple means, a practical way to increase amplitude function slope would berealized. Note that a mirror image zero would then also move on the negative frequency axis but is not shown. The inventor discovered a new method (i.e., replacing coils L1 138 and L2 140 of FIG. 8a with a transformer) for accomplishing thiszero migration. This invention forms the basis for his U.S. Pat. No. 4,771,466 United States Patent filed in 1987 and included herein by reference."
He's just creating finite zeros, but using xformers. You don't have to use xformers to create finite zeroes.
I discussed these filters directly via email with Jeff Joseph in 1993 (I still have the email exchange). I think they are very clever but I have never used one because I personally don't like their trade offs (especially the insertion loss). This isn't an exercise in man-hood comparison.
Dave
Quoted out of the patent:
Infinite slope loudspeaker crossover filter - US Patent 7085389 Description
"FIG. 8a is a schematic diagram, 136, FIG. 8b is the positive frequency axis poles and zeros (p-z), and FIG. 8c is a typical frequency response plot of a prior art low-pass woofer crossover transfer function. Note that the mirror-image finite p-zon negative frequency axis are not shown. Three zeros are shown at infinity. If, however, a zero could be moved from infinity nearer to the positive frequency pole cluster by simple means, a practical way to increase amplitude function slope would berealized. Note that a mirror image zero would then also move on the negative frequency axis but is not shown. The inventor discovered a new method (i.e., replacing coils L1 138 and L2 140 of FIG. 8a with a transformer) for accomplishing thiszero migration. This invention forms the basis for his U.S. Pat. No. 4,771,466 United States Patent filed in 1987 and included herein by reference."
He's just creating finite zeros, but using xformers. You don't have to use xformers to create finite zeroes.
I discussed these filters directly via email with Jeff Joseph in 1993 (I still have the email exchange). I think they are very clever but I have never used one because I personally don't like their trade offs (especially the insertion loss). This isn't an exercise in man-hood comparison.
Dave
I just showed graphs that showed my ellipticals have better group delay and a smaller overlap region than an LR-6. I have a spreadsheet showing the same thing, for NTM elliptical crossovers, which are similar. I also explained that my original design goals had car audio in mind. I cannot understand why you are being a fanboy for Joseph Audio. Nobody is saying "this is better than Joseph". I merely said "my crossover uses notches like Modafferi, but doesn't use transformers". I came up with this crossover 12 years ago and decided to share it. What have *you* shared with the world, without entering into a pi$$ing contest about?fntn said:
You are confusing the terms "dissipate" and "throughput". If a crossover has 1 dB of insertion loss (typical), it means 11% of the amplifier power is dissipated in the crossover, and 89% reaches the driver.
I used to be active on audio fora and quit for several years, and now I started looking around to see what's new... your calling my little contribution an "EPIC FAIL" without any depth of understanding, reminds me of the typical forum blowhard that made me quit posting years ago.
Last edited:
Why, too much riff raff? 😉
The class D section here seems to have a lot of power electronics expertise though, right up my alley. 🙂
How did the bass list expire? Did it just fade away like an old soldier, or ride off into the sunset?
The class D section here seems to have a lot of power electronics expertise though, right up my alley. 🙂
How did the bass list expire? Did it just fade away like an old soldier, or ride off into the sunset?
I was going to ask you the same. 🙂 I left the list probably about 12yrs ago and didn't post a thing anywhere for over 5yrs. Came back and was surprised by all the web forums that had popped up.
Looks like the bass list kept chugging along till a few years ago:
The DIYspeakers Archives
You'll probably like it here too, tons of expertise on tap, (not too many haters)
Looks like the bass list kept chugging along till a few years ago:
The DIYspeakers Archives
You'll probably like it here too, tons of expertise on tap, (not too many haters)
The OP started the thread in 2006 or did you miss that, too? I doubt seriously that he cares one whit about where it goes now. It should go where anyone wants it to go.
DDF (as you noticed) and Jason, be aware that no matter what you present, you'll get demands for more and more regardless of what you write. You won't get agreement on much of anything.
Dave
I'll tell what really happened, when I was searching for links to my page, I found a few here. I also opened a bunch of windows about elliptical and NTM crossovers, then I decided to post the group delay graphs which I had promised to post 12 years ago. 🙂 I happened to post it in the wrong window, in the wrong thread, this one, instead of a thread that actually discussed my page. 😛 This is what got Mr. Congeniality's panties in a bunch. And I didn't notice until well into the pi$$ing match. So my apologies posting in the wrong thread. 😛 Now if a moderator could move my posts to the right one. 😛
This crossover does definitley have some interesting properties.
The steep slopes help to use drivers that would otherwise be impossible to tame. The ones used in the actual Joseph models seem to be particularly difficult in this respect. You can still see some of this resonant behaviour in the response measurements - despite the steep filters. Vertical lobing behaviour behaviour is also much better than with any other crossover I have seen so far.
This all comes at the cost of terrible time-response.
As always: Speaker design means making compromises and this crossover is no exception in this respect. I don't doubt one second that this is a fine-sounding speaker but there are many other fine-sonding speakers out there as well that use different crossover topologies
Regarding the filter coils:
I don't think that they are that difficult to manufacture. Just make an ordinary crossover-coil with a tap - simple as that. It doesn't autotransform the full audio power from one side to the other as soemone misunderstood - far from that. It is basically a coil that BLOCKS high frequencies (like a coil usually does). When the resonant (i.e. notch) frequency of the series-resonant circuit part is reached, the whole arrangement blocks even more - or in other words: no power is transferred - actually the contrary is taking place.
BTW: Coupled inductors for passive notched filters were nothing new when this was filed for patent - the application to speaker crossovers was the novel part IMO.
Regards
Charles
The steep slopes help to use drivers that would otherwise be impossible to tame. The ones used in the actual Joseph models seem to be particularly difficult in this respect. You can still see some of this resonant behaviour in the response measurements - despite the steep filters. Vertical lobing behaviour behaviour is also much better than with any other crossover I have seen so far.
This all comes at the cost of terrible time-response.
As always: Speaker design means making compromises and this crossover is no exception in this respect. I don't doubt one second that this is a fine-sounding speaker but there are many other fine-sonding speakers out there as well that use different crossover topologies
Regarding the filter coils:
I don't think that they are that difficult to manufacture. Just make an ordinary crossover-coil with a tap - simple as that. It doesn't autotransform the full audio power from one side to the other as soemone misunderstood - far from that. It is basically a coil that BLOCKS high frequencies (like a coil usually does). When the resonant (i.e. notch) frequency of the series-resonant circuit part is reached, the whole arrangement blocks even more - or in other words: no power is transferred - actually the contrary is taking place.
BTW: Coupled inductors for passive notched filters were nothing new when this was filed for patent - the application to speaker crossovers was the novel part IMO.
Regards
Charles
Last edited:
