CatZ Tweeter Shoot Out to come...

[JonMarsh]

Well, while GF was visiting one of her friends and taking pictures of the humongous holiday decorating job they did, I got the measuring gear up and running (finally- always seems to be some issue with a cable or the preamp or something else going on that makes getting it actually working about an hour long proposition...) and got some data on the 9900 mounted on the good old 1 cu ft PE test cab baffle- new baffle routed yesterday just for testing this part. As a picture is worth a thousand words, lets just segue to the pictures:









As regards the on axis SPL, I wouldn't say I saw what I expected, given the published ScanSpeak data. It shows more evidence of waveguide like behavior.




OTOH, the axis curves look pretty nice as soft dome tweeters go, clearly it will sound best if one stays within 15 - 20 degrees off axis, and the HD3 distortion is excellent, these should have a very dynamic and clear sound, I'd say.

Certainly worth a complete work down and demo crossover for the CatherineZ. This and the TL N26 will definitely get full test setups done.

What will that leave for the budget build? ... I'd say both the SB26ADC and the D2608-9130 are contenders. And both available in Europe.

What about the 71000? High HD2, but the rest looks pretty good. Worth a test pass, too. Hmm, that's going to be 5 designs if I stop there...

[---k---]

For the budget build, the SB26ADC is the same price as the venerable RS28. So I have to wonder if it is better. And in what ways? You're already buying the ES drivers from PE.

I'd be interested in a comparison to the SB29RDC dimple dome that CJD used in his Ansonica. Same price. I thought it sounded great. So many options....

[JonMarsh]

TOO many options. At the moment, I'm not sure if I've got an RS28a lurking around here or not. I will compare specs, but since we're overall shooting for an upgrade, I'll have to give this some thought. Best thing would be to find one of my RS28a and run it through the current test setup. That will give the best correlation, I think.

[JonMarsh]

OK, here's something that's not an option, but is interesting for comparison, and I'm reposting this here because it might not be seen where it was originally posted- not specifically in a tweeter test thread. But it's certainly relevant for comparison.

To avoid any possibility of damage, the RAAL 70-10D was only tested from 1kHz on up- the impedance curve shows the effect of the primary magnetizing inductance set to about 1.3 mH, in order to act as the final shunt inductor in a 4th order LR crossover. More about that later, and how to game it. If you drive the tweeter below 1 kHz, the power consumed will start to go up because of the drop in impedance, until it becomes nearly a dead short.

I'm planning to do a basic analysis and modeling of the ribbon impedance, and compare a theoretical model with what's measured- I want to be sure that things interact properly when modeling a crossover.



The foam deflector pads are truly an integral part of the design, in my opinion- note the on axis SPL with (magenta) and without (red)



With the pads in place, the response is pretty smooth from 3 kHz on up.


How does the standard family of curves look? Pretty dang decent... especially at high frequencies.



There's very little drop off to beyond 20 kHz, even at 45 degrees... and overall, this is a pretty tight window of response. In fact, it's got me scratching my head a bit, because it's better than RAAL's published curves. But I know my horizontal protractor setup is pretty accurate for angle of incidence. I don't have an accurate way to do vertical dispersion yet- will need to work on that.

What's more impressive, to me, is the behavior from 15 degrees to 30 degrees, which really defines a lot of the forward listening power response.



Now, when I see some of the forum trolls talking about how any number of dome tweeters are just as good as a RAAL, well, I have to laugh... but then, that's why they're called trolls, right?

OK, next up is distortion- this was done at 2.83 VRMS like all the rest, which is ~92 dB output.




It's not super low, but it's respectable, and what's particularly interesting is that it doesn't rise in the top end above 6-8 kHz like so many domes due (inductivity modulation of the voice coils?). It's certainly not a trash generator, like the "cheap" or not so cheap ribbons. I think 3.2 to 3.5 kHz may be fine with this one.

[Jonasz]

Raals are fantastic drivers, listening to 140-15 as we speak! :P

The offaxis is very very good, this is an onaxis measurement of a Philharmonic Audio system wich uses the small Raal 64-10 OEM.



And here is the 85° offaxis... 8O



Dennis Murphys measurements of the 70-10 mirrors yours fine.

http://philharmonicaudio.com/phil3.html

[5th element]

That 9900 from SS certainly measures like a champ, it's HD is superb and could really be crossed any way you want it to.

The RAAL isn't half bad, but the HD really isn't anything special and is borderline on what I would actually considered acceptable performance. I don't suppose that it's much of a surprise that the small foil within it can't offer distortion as low as its bigger brother. The Aviawave on the other hand seems to offer considerably better performance, if those other measurements above are anything to go by, but then we still don't know what drive level those were done at. I wonder how the RAAL would perform when measured at lower drive levels.

I know we tend to do measurements at 2.83vrms as a standard, but the way Zaph did his measurements is far better as a way of doing things because it standardises the SPL, rather than drive level, so that it brings every design of driver, be they 85dB or 95dB are brought into line with one another.

[Jonasz]

I'm starting to get the opinion that when you're under a certain level of distortion it really doesn't matter anymore. It can still tell how loud you can go or of possible problems with the driver construction and where to cross it and so on but for the pure listening enjoyment there must be more important data to look at. Wich ones I don't know but linear distortion, dispersion and so on comes to mind. The Hiquphon OW1 for example is no king of low distortion but still sounds great. The Vifa XT25 on the other hand with super low high order distortion (admittedly high 2nd though) sound as boring as watching paint dry... The dispersion in the highs are on the other hand much better in OW1. Does this have anything to do with the IMHO better sound of the OW1 over the XT25, or the Raal over a SS6600 for that matter? :P

OW1


OW1


Aviawave now has a homepage: http://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http%3A%2F%2Fviawave .ru%2F&sandbox=1

[5th element]

I love the way the XT25 sounds, it's probably my favourite tweeter out of anything with a wave guide. I also despise the XT19 with its superior dispersion. I haven't listened to a true 19mm dome but from the trend I've tested with going from XT25>standard dome>XT19 (as in each design offering 'superior' dispersion), I'd expect that I'd hate the 19mm worse than the XT19. And going backwards the other way, from the sound I prefer from the XT25, a wave guide tweeter, even a small one, such as the DXT, just gives me more of and then bigger guides take it further.

I also believe that once you're beyond a certain level of distortion that it does not matter any more, but the RAAL is pushing this limit imo. I have played with drivers that sit at around -40dB for their 2nd and 3rd order products and stepping up the ante, to drivers that display at least -50dB has always resulting in a cleaner, clearer presentation. Certainly the RAALs distortion is only going to go up as the drive level is increased and as we're already pushing the limit on what I think is acceptable, I would only really want to use the RAAL from around 6kHz and up.

Everyone I've seen talk about the 140 says that it sounds far better when used around 3kHz, rather than down to the ~1.5kHz, stated by RAAL, as to how low it can be taken. Looking at the graphs for the 70s distortion it certainly comes as no surprise that it too would probably work better used considerably higher than RAAL say it should be.

Not to speak unfairly of a product, but if this had been a standard dome tweeter that Jon had tested, it would have been struck off the list of potentials if it had measured as poorly as this. I don't see why one should change their criteria just because it's a fabled RAAL and costs a lot of money. Yes, its performance from 6k+ is nice, but below that things go from merely acceptable to worse. Clearly this is no fault of the fact that the driver is a ribbon as the Aviawave above measures very well indeed and would seem far more suited to the designs that Jon has in mind. Maybe this is a result of the Aviawave being tested at a lower drive level, but until we know otherwise we cannot say.

I am just looking at this objectively, which is how I design pretty much all my stuff and from that standpoint the RAAL doesn't really seem to make sense.

And according to Zaph's measurements the OW1 has pretty slick distortion performance, it is nothing to be sniffed at. Certainly not the best, but far better than the RAAL here.

[Face]

I would have said the same thing about RAAL until I heard it used in a few applications. It really is the Bees Knees, IMO of course.

[5th element]

Indeed, which is why I am interested in seeing how it performs at lower drive levels. It is 92dB sensitive for 2.83vrms being applied. This is loud. Even the estimated '85dB' average listening level that is often quoted as what people like to listen at is too loud for me. The RAAL could test really rather well some 10dB lower.

[Face]

Other RAAL tweeters do test better. Too bad this one is OEM only.

[5th element]

That one is excellent.

[JonMarsh]

I think Matt has a lot of good points, and in fact, I agree (mostly) with most of them... but...

I find that often in the past I've talked myself out of experiments which might have been interesting and worthwhile learning experiences.

I think the 9900 has some great potential, and plan to pair it with the Esoteric ES140ti for the first pass of the CatherineZ or CatZ. That should be a very good pairing, and if you will, will represent the first soft cone design I've done in several years (the last being the Modula MT XE). I have high expectations for that.

But while the nonlinear distortion of the 70-10D doesn't completely charm, the linear distortion (or lack there of) does. As does the dispersion. So, for me, this is an experiment, to make a tone sample for low volume playback that might give some idea of what a larger system based on the ZA14 and RAAL tweeters might work like. And while I have as much engineer's disease as the next guy, with a focus on measured characteristics and specs, I also know that what I'm mainly going to be doing is listening to these, so I figure this is a part of the evaluation, too.



The HD3 resonance amplification level on the ZA14 looks low enough that I think it's worth the experiment to see if one can live with it- or even notices it.

If this small experiment sounds promising, I'll probably try building a slightly bigger 2.5 way based on this concept- the "Quattro", which I've already modeled (with two PR's per cab). At this point, the main Sd becomes a little more interesting compared with a 7" MTM. How it will sound, well, who knows, that's why I have to build it. I've got lots of BB ply and LBL bamboo around here, so I'm not lacking for materials.

And before long I should have a set of CatZ's in one configuration to compare to. OTOH, the CatZ concept isn't scalable to a line array, though it should make a very nice 1 cu ft MTM. Still need to look a bit more at the LF alignment for it.

And last, about that spec thing- on paper, a JL1 Signure tube mono block should not sound better than, say, a Boulder Sound 2060, based on nonlinear distortion (at full and 1/2 output)



Versus




But having heard both for hours at a time in a colleague's system in Munich, I have to concur with his choice to selling the 2060 and going with the JL1, even though I recommended the 2060 to him, and we both liked it's sound in his system for several years...

[JonMarsh]

Other RAAL tweeters do test better. Too bad this one is OEM only.

Ah, yes... the RAAL OEM. 70-20XR. Even if I could get my hands on a pair or two (I can), other typical builders wouldn't be able to.

[JonMarsh]

Raals are fantastic drivers, listening to 140-15 as we speak! :P

The offaxis is very very good, this is an onaxis measurement of a Philharmonic Audio system wich uses the small Raal 64-10 OEM.



And here is the 85° offaxis... 8O



Dennis Murphys measurements of the 70-10 mirrors yours fine.
http://philharmonicaudio.com/phil3.html

Thanks for the plots and the link to Dennis's project! Which does look pretty nice! :T

I have a pair of those revelator 8's that I was thinking of doing something similar with at one point. Now I'm a Wavecor kind of guy...

[bvbellomo]

I think the 9900 has some great potential, and plan to pair it with the Esoteric ES140ti for the first pass of the CatherineZ or CatZ. That should be a very good pairing, and if you will, will represent the first soft cone design I've done in several years (the last being the Modula MT XE). I have high expectations for that.

...

And before long I should have a set of CatZ's in one configuration to compare to. OTOH, the CatZ concept isn't scalable to a line array, though it should make a very nice 1 cu ft MTM.

So you're planning an MTM with the ES140ti and 9900? Sounds very similar to my plan for my 9700s. I'd really appreciate anything you post on these. What frequency and order do you plan to cross?

[JonMarsh]

So you're planning an MTM with the ES140ti and 9900? Sounds very similar to my plan for my 9700s. I'd really appreciate anything you post on these. What frequency and order do you plan to cross?

They'll be using the same technology approach as the NatalieP; a quasi series crossover, using a quasi LR3 alignment (LR3 doesn't exist as a defined network alignment by Linkwitz/Riley, but using the mathematical concept of -6 dB at crossover, and no phase quadrature (unlike B3) it can be a very useful alignment for specific drivers with the right time offset for the midwoofer- see the Wavecor Ardent thread or the Three Way Design study thread for more details about the theory). I expect the crossover to be in the 1800-2200 range, depending on how the driver time offset looks and which point produced the most optimized transition region behavior.

[peterpan]

They'll be using the same technology approach as the NatalieP; a quasi series crossover, using a quasi LR3 alignment (LR3 doesn't exist as a defined network alignment by Linkwitz/Riley, but using the mathematical concept of -6 dB at crossover, and no phase quadrature (unlike B3) it can be a very useful alignment for specific drivers with the right time offset for the midwoofer- see the Wavecor Ardent thread or the Three Way Design study thread for more details about the theory). I expect the crossover to be in the 1800-2200 range, depending on how the driver time offset looks and which point produced the most optimized transition region behavior.

What happened to the CATZ, is it off the board?

[Hdale85]

It's just a very full board lol.

[JonMarsh]

It's stacked up in the queue - it's a deep one at this point! Not gone, not forgotten, but no parallel processing going on here!

[---k---]

Maybe we should start a kickstarter campaign to pay Jon to quit his day job and become the forum's full-time speaker tester and designer.

[Hdale85]

Well there is Fundme.com or whatever it is lol.

[JonMarsh]

Maybe we should start a kickstarter campaign to pay Jon to quit his day job and become the forum's full-time speaker tester and designer.

That would be attractive were it not for the fact that it looks like the next project I'll get to work on will be a high power resonant switching power supply for Class D applications- to support the higher power IR reference designs. It will be a dual use configuration, also applicable to telecom applications with a change in the board stuffing.

[JonMarsh]

BTW, anyone watching this thread, I'm trying to gear up to do a whole mess of measurements in about 2 weeks- could be a week longer depending on how long it takes to get the concentric drivers in. Also, there may be more enclosure work than I realize- just preparing baffles.

Worst case, I figure weekend after next to get baffles and test enclosures ready, then the following weekend for measurements.

[meb46]

Jon - what's the chances of adding the Scan-Speak D2908/7140 to this mix? I'm happy to send a pair of mine to you for the sake of test results? I can then arrange shipping back to me once your finished with them.

Let me know,

[JonMarsh]

I don't see any problem doing that- do you want data that's quasi IB, or something resembling the baffle and location on it for the Osiris? I'm guessing you want the FRD files, and you probably might not object to a tweeter crossover design integrating with your current build?

[5th element]

the next project I'll get to work on will be a high power resonant switching power supply for Class D applications

Interesting, my next big DIY project is going to be something similar depending on how high, high power actually is. My idea is to use one of the newer interleaved PFC front ends followed by this from OnSemi. It's a new chip, so new in fact that it isn't even out yet. I emailed them a couple of weeks ago asking about when it was going to be released and the answer was in a few weeks. Compared to all the others parts from the usual manufacturers the NCP1399 looks to combine the best feature set from some chips with the best feature set from another.

I've done low voltage SMPS design before in the shape of LED drivers and a couple of off line LED controllers too, but this will be my first time actually winding my own magnetic parts. It's going to be a decent learning experience.

[JonMarsh]

By high power, I mean a DC supply suitable for about 600W RMS of audio, which is about 1kW DC. The On Semi part seems to be several variants of an LLC controller, specifically with interface support that indicates a focus on LCD TV applications.

I'm in the middle of re-writing an application note for an eval board done by some of our HQ guys who wanted to keep it a "surface discussion"; and it was. Problem is, LLC's are fairly complicated beasties, and there are a lot of little details that need to be done right to make a robust design, particularly to avoid capacitive mode switching and FET body diode stress, and to design a tank configuration that provides the right regulation characteristics and tank Q for good efficiency and ZVS switching. We have a patent applied for for some of the concepts.

now, another point, is that the commonly used FHA approximation used for analyzing and designing the tank circuits does have accuracy issues, and unfortunately, they are worst for the tank Q ranges you actually find useful for a power supply design. If you would like a copy of this when finished, or if, even better, you'd like to be a alpha/beta tester on the copy, let me know- the biggest hazard about righting about something one is familiar with, is not realizing where others are not familiar, and leaving out important stuff in the explanations...

We have both analog and digital controllers, and I'm first re-writing the analog one- they haven't made an attempt yet at the digital one (designed by a 3rd party company- the firmware that is- the controller is ours, developed by our Automotive/Industrial micro division).

[5th element]

That sounds like fun Jon, I'm always up for broadening my knowledge on such topics and it seems that there is a lack of well written and easily digestible stuff out there.

Resonant supplies seem to be the optimal choice for audio applications, well any application really, that requires lower EMI due to the absence of hard switching, high current transitions, but as you say, they have the added disadvantage of designing the tank correctly.

I got the impression that the OnSemi part recommendation was more written just to be relevant to todays consumer market more than anything specific. The part auto manages dead time so that should help with the efficiency around the ZV transitions (if I am reading that correctly) and it also detects if the device is working in capacitive mode and shuts it down accordingly.

What controllers are you going to be using or are they in house designs?

[JonMarsh]

There are two controllers that can be used in this 600W demo board - our analog ICE2HS01, (on it's own plug in board) and our XMC4200 ARM based power controller. (I work for Infineon, the German semiconductor company- used to be Siemens Semiconductor, I've been with them 30 years this September. To switch, you just plug in a different card with the different controller. Firmware has to be loaded on the micro once. The XMC family is designed for power conversion and has a variety of peripherals for A/D, D/A, high precision PWM, capture/compare units, etc. It has more flexility about how things like start up and the synchronous rectifier outputs can be programmed. Both the analog controller and the XMC controller are set up for controlling secondary side synchronous rectification, to optimize the efficiency, particularly for lower voltage, high current applications.

That On Semi series doesn't have synchronous rectification built in; it's designed for lower power applications, and does have a boundary mode or CrCM PFC control function built in. Because of the current shift in the resonant behavior when you're operating in boost up mode below the primary resonance, you can't fixed self driven synchronous rectifier chips on the output the way you can for say, DCM or QR flyback. We have some digital based chips that have similar functionality- one that is fully firmware programmable, another series derived from a lighting part, which I actually like better, that is "programmed" by resistors on pins. We're working on a another spin on that part, I just finished doing a concept engineering review for that one, to be known as the ICL5102. The ICL5101 is in production, and there's a nice 100W demo board for it, 54V DC output, say for LED lighting, which is about 94% efficient.

Many parts either have automatic dead time adjustment or it's programable by a resistor pin or capacitor; the former is better because then you have a chance for dynamic adjustment from an external signal. Worst case situation for dead time is when the supply is running at Fmax with high line and no-load; then the only current from the resonant transition is from the magnetizing inductance of the transformer. This makes selecting the MOSFET a factor based on and interrelated with the tank design.

[meb46]

Consider the Scan-Speak D2908/7140 now officially added to the list... units on-route to Jon

Looking forward to these results... the D2908/7140 is a seriously nice looking tweeter... I just hope it performs as well as it looks!

[JonMarsh]

the D2908/7140 is a seriously nice looking tweeter... I just hope it performs as well as it looks!

Don't we all! :B but especially you, as you have skin in the game!

that's how I feel about a lot of parts waiting for testing...

[5th element]

That On Semi series doesn't have synchronous rectification built in; it's designed for lower power applications, and does have a boundary mode or CrCM PFC control function built in.

Most LLC controllers do not have sync rect built in which is why there are a plethora of chips specifically targeted at adding this functionality (that are designed for LLC power supplies). From what I can see though adding sync rect into a high-power power amplifier isn't really necessary where the voltage rails will be high and the average current draw pretty low. Pretty much all LLC controllers have a boundary/burst mode to keep their efficiency high at low loads, the NCP1399 has additional functionality built in so that it can tell the preceeding PFC stage to shut down, in such a situation, to help improve efficiency.

Many parts either have automatic dead time adjustment or it's programable by a resistor pin or capacitor; the former is better because then you have a chance for dynamic adjustment from an external signal.

From my browsing I could only come up with a single part that had automatic dead time adjustment, the L6699 from ST micro, the rest were programmable.

Worst case situation for dead time is when the supply is running at Fmax with high line and no-load, this makes selecting the MOSFET a factor based on and interrelated with the tank design.

I guess it's a good thing then that all of my targeted applications will never be used in a no-load situation. But it still adds in an additional complication that I didn't think was an issue with sensible MOSFET choices.

[JonMarsh]

well, the way I see it, is an LLC power supply is a lot like Spiderman...

with great power, comes great responsibility... and complexity.

For example, it's my preference to avoid using burst mode, but this places more requirements on care for the design of the tank system with regards to minimum and maximum gain, how that fits with the range of input voltage, and being able to do no-load under worst case conditions without burst mode. I have almost never seen an LLC supply yet that didn't have issues with going into capacitive mode and hard commutation of the MOSFET body diode after coming in to operation from burst mode; this is especially true if you don't use a split cap resonant architecture, but try to get by with a single cap, as most do, including our eval boards from the Singapore applications team. As you might guess, I'm the OCD guy, but it's because of the expectations of my customer base here in the USA.

BTW, for some purposes, people have to define no load as anything from 10% load on down- such as 5A load for a 50A 12V output.

PM me if you'd like to see the App note when it's nearing completion. You can tell me if it's clear and concise, or a bunch of rubbish! :W

[5th element]

I have almost never seen an LLC supply yet that didn't have issues with going into capacitive mode and hard commutation of the MOSFET body diode after coming in to operation from burst mode;

It would be interesting then to see what would happen with a controller that stops the system from working in capacitive mode. The L6699 apparently shuts down if it enters capacitive switching. ST actually recommend that you use their other pin compatible part, without capacitive switching protection, to get your converter working as it can prohibitively towards troubleshooting as your converter simply will not start.

My concern would be if the converter ran in burst mode when the amplifier was idling but then needed to switch temporarily into full operation for transients. It could be jumping in and out of burst mode a lot of the time.

this is especially true if you don't use a split cap resonant architecture, but try to get by with a single cap, as most do, including our eval boards from the Singapore applications team. As you might guess, I'm the OCD guy, but it's because of the expectations of my customer base here in the USA.

It's times like this where the OCD does come in handy. Without me googling it what is a split cap resonant architecture? I don't think I've seen this used in any of the eval boards or datasheets I've looked at and I've looked at a fair few.

BTW, for some purposes, people have to define no load as anything from 10% load on down- such as 5A load for a 50A 12V output.

Well I guess if you are designing a supply for an application that does have quiescent demands, of say 10% the full load, like a power amplifier, that no load would basically equate to the device being on but not actually do its proper job, then you could get away with something like this. Figuratively it would simplify power supply design but you could come undone if the load became faulty and that in turn made it harder to fix/troubleshoot. Then again this isn't usually a major concern for consumer products these days.

PM me if you'd like to see the App note when it's nearing completion. You can tell me if it's clear and concise, or a bunch of rubbish! :W

That sounds like a plan, I hope I have enough fundamental knowledge to understand what it is you'll be writing about!

[Evil Twin]

Everything going as I have foreseen...

Yesterday we had our development concept workshop for the Class D power supply program, and as my master would frequently say, "Things have gone exactly as I have foreseen..."

Or as my nemesis would say, "It's like herding Banta's with an electric prod- in the end, you know they're going to go where you direct, but you get a lot of confusion and noise in the middle..."

The program target is quite aggressive, to have fully functioning reference design modules ready by the 2016 CES show... and a sound booth setup has already been reserved. Impressive... no pressure at all....


Both low and high power designs will be prepared, based on a common concept that delivers a low output impedance, high energy storage, and intrinsically low EMI, with a circuit concept that is forgiving of inexpensive MOSFETs...

Of course, there are some possible embellishments, and these will probably be implemented in the scaled up high power version, but the path is open to some different development concepts, depending on how difficult certain innovative control concepts prove to be...

Even the IR TM manager for Class D, who was skeptical that we understood the application, is now quite enthusiastic...

Of course, I will insist also that we only use the highest quality black metal work for the prototype demonstrators...

[5th element]

Well the IR guy should have known he was in safe hands. I guess it isn't every day that the engineer guy assigned to your project is a serious high quality audio connoisseur with a penchant for making sure that everything that can be done right will be done right. Ie no corner cutting. Are their any real cost constraints placed on a job like this? I mean are you given the freedom to go a little OCD towards the end goal of making something a little bit special rather than run of the mill?

[JonMarsh]

Are their any real cost constraints placed on a job like this? I mean are you given the freedom to go a little OCD towards the end goal of making something a little bit special rather than run of the mill?

Well, yes there are.. and that's where one must be a bit innovative. This will not be a conventional loop regulated power supply. Instead, think of it as a high quality primary bus rectifier, with substantial high voltage energy storage, connected to an IBC- Intermediate Bus Converter. (used in distributed DC-DC power systems, usually at lower voltage on the input, though Vicor has some resonant ones operating with high input voltage buses). An IBC has a fixed input to output conversion ratio- running at very high duty cycle, typically 96-98% for a typical square wave or resonant switching IBC.

Now, let's think about audio issues... we might like to have graceful power limiting... we must have short circuit protection... we must have low EMI- preferably very low EMI. So, imagine using an LLC as an IBC, but never running it in boost up mode below Fo frequency. Sine wave current pattern instead of square wave- much lower EMI, but there is about a 20% RMS conduction loss penalty, haversine versus trapezoidal current, but then we make the lowest Ron and best FOM FETs in the world (not bashful, either) so we can cover that. (do you know anyone else with 19 mOhm 650V MOSFETS in standard TO247, not the so called super 247?)

Then, there should be virtually no possibility of going into capacitive mode, or hard commutation of the high voltage MOSFET body diode. Current limiting and protection can be achieved by scaling up the operating frequency out of Fo, and soft start is handled the way you normally do with an LLC, starting at Fmax and bringing it down in frequency to Fo. At Fo, the gain is fixed, regardless of the Q and effective Rac presented to the tank network- that's the operating point for the "IBC/DC transformer".

Then, the place to spend some money is on the input caps- fortunately, 250V input caps are available with very good lifetime and power density and ESR and ripple current capability- to the extent that they will out perform conventional 450V caps used in server and Telecom applications.

Another trick that will be investigate for upping the power and energy storage, is using a special configuration of PC, with 3rd harmonic injection, which either cuts the ripple voltage on the bulk caps by 40%, or allows you to use 40% less caps for the same performance. Using PFC for an amp is not needed so much for regulatory issues, but could allow one to raise the nominal bulk bus voltage from ~311-325V up to 450V, and as energy storage is proportional to capacitance, but to the square of voltage, this gives a huge boost in energy storage. Also, with 3rd harmonic injection (used in high power 3 phase systems) I calculate one could still achieve this improvement while maintaining a power factor of 0.9, which is much better than a direct bridge rectifier (0.55 typically).

So, there's a lot of cool stuff to work on and figure out. The intent is to develop related systems for two power levels, but to have scale up and scale down covered in an application note, for which we'll have to confirmed the design concept and results. I'll post a few foils later explaining some of this- pictures help compared with words...

[5th element]

That sounds a lot like an interleaved PFC AC>DC converter followed on by a precisely tuned LLC. Increasing the bulk capacitance after the PFC should help with providing good transient capabilities if the main LLC transformer is sized accordingly.

[JonMarsh]

That's probably how it's going to work out, because it should be easier to do the current shaping using a boundary mode or DCM mode PFC, rather than continuous conduction mode. Sizing the LLC transformer and getting low impedance is a fairly easy part, IMO- the IR Technical Marketing guy wants to see a simulation before we start building hardware, verifying load regulation. It's always my working plan for new SMPS to develop a complete simulation model (which usually means coming up with a decent controller model, too), so that doesn't result in any change to my development roadmap.

BTW, the Hypex SMPS supplies are square wave IBC's; concept from an early 90's PhD thesis at MIT. I just like the idea of lower EMI and improved controllability with an LLC based equivalent; probably will set the Fo at something like 90 or 100 kHz for the first transformer design, at least for the 600W-1kW version. There will be a small one, and that will be derived from I design I've worked on with guys from Munich, which uses a modified digital lighting IC to implement the LLC, and has a built in HB coreless transformer driver, and a PFC section if one wants to use it. With PFC enabled, the 100W demo board right now runs about 94% efficient, which is nothing to complain about...

So, I'm jazzed, but need to finish up this application note thing for HQ first.... then on to the fun stuff!

[JonMarsh]

Proof of Concept with N26a

Well, a little insomnia led to a little revisiting this design and working up a "proof of concept" solution with the N26. Though I want to hear the 9900 in this configuration, the engineer in me keeps coming back to the lovely distortion performance of the N26.



And having six pairs of these around and still not having built a speaker with them makes me feel guilty or stupid, take your pick.



OTOH, the ES140ti is not without it's issues, according to some reports, consistency being one of them, (variability in HF breakup mode Q) and theres that peak and bump down in the 1200-1600 Hz area that is of concern, falling right before the tweets crossover. Still, though all is not sweetness and light, you've got to admit they look easier to work with than RS180's, at least in some regards.

So let's go to work....



OK, it's not as weird as a NatalieP- though I suppose I could make it that way if so inclined. But taking care of the stuff in the 1-2K area does take a bit of effort, as well as the upper range breakups.




Voiced a little bit mellow,



With a solid reverse null at the roughly 1800 Hz crossover point (there's more similarities to the acoustic target of the NatP than not...)

A very slight bump in the crossover region, the 15 degree off vertical axis is tuned pretty much flat, as is my preference:




Impedance curve is about what you expect from a dual 8 ohm in parallel MTM:



I think I'm going to need to model this at higher frequencies, and probably insert a damping resistor in the LF network somewhere, because there's a solid capacitive path to ground across the output of the amplifier in the LF section. If you don't drive it with frequencies above 20-25 kHz, all is probably well....

Gonna have to think about this... I suppose worst case I just take out the HF trap, OR, go back to the quasi-series/parallel of the NatalieP- there must have been several reasons I used that topology! Since it has an LF trap, too!

Well, it shouldn't be to hard to reformulate the crossover in to that topology... will just leave that to my next sleepless night...

At least I can say that I've got a good plan for dealing with the necessary electrical means to hit the desired acoustical target and deal with the ES140's issues.

[SpeakerGuy]

...You can get a replacement dome/coil assembly from here.

http://www.rumoh.eu/en/spare-parts/3...2904-6000.html

Oh, neat store - I particularly liked the "Een jig voor het meten van de impedantie van de luidspreker/driv". Would like to see a guide to making one in English - though I suppose I can figure it out. I like their implementation very much.

[mkc]

Hi Jon and Jonas,

I will probably get bashed for hijacking the thread for being off-topic. I just want to share this with you guys. This album, Covered by Robert Glasper, is getting some rotation in the office at the moment, while I'm programming. Here is a sample:

https://www.youtube.com/watch?v=GS2Y_CkaXP0

P.S. Jon, I have seen your DAC's, just wanted to add that I have an Ayre QB-9. Very happy with it. It was the only DAC I would buy if I didn't have time to build one myself.

Cheers,
Mogens

[JonMarsh]

You're always welcome to go off topic with good music recommendations- as far as I'm concerned, a good music recommendation is ALWAYS on topic around here, as that's what we do this stuff for! :T

[JonMarsh]

I stocked up on this one and a couple of others by him from Amazon just now. Instant gratification... Well, almost.