This influential paper was only presented at a conference in 1975 and never published in the JAES.
I would very much appreciate a copy or link if you, or anyone else, has one.
The USA speaker development scene seems to be as interconnected as you mentioned for the UK, but unusually, Henricksen never worked at JBL, he is known for work mainly at Altec and ElectroVoice.
Still surprised I didn't know of Button
Then there's one more of his papers that may be of interest - "Maximum SPL from Direct Radiators", a sort of follow-up to the "...Parameters and Trade-Offs..." paper.
It's also a convention only paper and once more I can't find a copy.
Best wishes
David
That's another damn convention paper that was never printed in the JAES.
As usual, you or anyone have a link or a reference to a published version?
On the topic of waterfalls, did you see Don Keele's wavelet based method with time normalized to cycles of each respective frequency and log frequency?
To his usual, clever standard.
So, what are the catches that you explained the JA?
The TAD woofers presumably sound OK but seem fairly ordinary, have no technical advances that I can see.
Best wishes
David
David,
At this point in time, pretty much all that can be done has been done by the major players, and any tiny incremental changes done have diminishing effects on the overall response of mid-bass drivers.
Although your "pursuit of excellence" is commendable, so much can be accomplished with little expense in implementation of FIR filters, I'd suggest putting your efforts there, rather on transducers themselves.
After measuring a Mackie HD 1502 back in 2009, and finding both frequency response and phase dead flat, using quite ordinary, inexpensive drivers, it became apparent to me how Dave Gunness (EV, EAW, now Fulcrum Acoustic) was able to drastically improve the response of products that had been around for years.
Art
I'll see if I have a dead tree copy but I'm in the process of preparing for cyclones so my shed is a bit disorganised
It may also be in my sister's shed on the other side of the continent.Have you got a reference for Keele's paper? It sound like a different implementation of a PAFplot which has time normalised to cycles too.
Time-Frequency Display of Electroacoustic Data Using Cycle-Octave Wavelet Transforms ?
I wish you hadn't sent me hunting through Don's stuff. There's loadsa good stuff he's done while I was bush.
The important one is that high Q resonances are much less audible/objectionable than low Q. Max audibility is at about Q=1.So, what are the catches that you explained to JA?
First investigated by Fryer (surprise surprise) in Intermodulation Distortion Listening Tests
Don't let the title fool you. Several other really important listening tests were described in the appendix to this paper.
There's a long story why this is so but I won't bore you with it.
Sean Olive & Floyd have confirmed this seminal 1970's work since then and I think reference Peter.
Nasty & benign Delayed Resonances appear differently on CDS/waterfalls/PAFplots. eg good sounding UFM cones have a rather 'dirty' looking PAFplot with many high Q resonances .. compared to good sounding plastic cones.
There was more stuff with JA ... but I can't remember as it was 30+ yrs ago.
Last edited:
Don't want to throw a wrench in the conversation, but...
If we're discussing improving current transducer design, wouldn't the Acoustic Elegance be a better target than the venerable but aging JBL 2226? The AE drivers have a motor which (I hope this isn't controversial) is superior to JBL, and probably among the SOTA for ferrite motors in general. Many of the benefits of underhung motors without the efficiency penalty, etc.
AE Speakers --- Superb Quality, Unforgettable Performance, Definitely.
Moreover, unless one can improve upon ~that~ driver (which come in different applications for sealed/ported/etc.) this seems more a thought experiment than a pragmatic one (I find it hard to believe that one could outperform the AE drivers for less money than purchasing them, given R&D, etc.
Just my $0.02.
Cheers,
Tal
If we're discussing improving current transducer design, wouldn't the Acoustic Elegance be a better target than the venerable but aging JBL 2226? The AE drivers have a motor which (I hope this isn't controversial) is superior to JBL, and probably among the SOTA for ferrite motors in general. Many of the benefits of underhung motors without the efficiency penalty, etc.
AE Speakers --- Superb Quality, Unforgettable Performance, Definitely.
Moreover, unless one can improve upon ~that~ driver (which come in different applications for sealed/ported/etc.) this seems more a thought experiment than a pragmatic one (I find it hard to believe that one could outperform the AE drivers for less money than purchasing them, given R&D, etc.
Just my $0.02.
Cheers,
Tal
Hi Art
I don't expect to find any dramatic improvement that the major player have missed.
But the major players are constrained in some ways, they tend to stick with what they are tooled up for, and with what is acceptable for the mass market.
For instance the JBL Double Differential Voice coil is a superior technique to avoid 2nd harmonic flux modulation but most companies stick with less effective methods.
Similarly the Inside/Outside voice coil wind is a better technique but most companies aren't tooled up for it, even JBL don't use it.
I haven't seen any one driver that incorporates all the known best practices. (That's all the mutually compatible ones, of course)
I have done back of envelope calculation and about 3 dB improvement in efficiency and reduced power compression looks possible from a combination of incremental improvements and some minor innovation of my own.
That's sufficient to make it worthwhile to me.
I do think that there are vast benefits to DSP too, but why not start with the best transducer as a base?
Thanks
Yes that's the one but there's better news, it's available on-line
AES Papers -- Official website of D.B.Keele
Sorry if you think it's not better news that it's on-line
I am happy that his work is available, it really is excellent, even his wonky papers are better than most people's.
That's another convention paper not in the JAES, what is his problem!?
I remember the observation is in my Floyd book and he seems to take care with references.
Yes, I didn't expect a transcript of the discussion, just what were the problems to watch for.
Perhaps we can discuss this face to face one day. I'll be up in Queensland in about 6 months, you ever make it down South?
Best wishes
David
Last edited:
Thanks for this David.
It is indeed an implementation of Fryer & Millward (his reference [11]) but with loadsa obfuscating maths
We used much simpler (?!!) means in 1979 and the 'analogue' reference was cos KEF were the only other people showing similar curves (CDS which is the modern 'waterfall') using their 'digital' HP 5454.
Nearly 2 decades later, I helped carry the KEF 5454 to the dump. They led the world in 1975 but their tech. had not moved out of the 70s.
Peter's box could also do a KEFplot using what we called a Slow Fourier Transform (B&K 2010)
___________________________
To get back to Don's paper, his Fig 11 shows a 2 way system with 8" bass.
He plots 100 cycles of decay on his display. We plotted just 2 cycles as anything beyond that would be clearly audible and, IMHO, unacceptable colouration. ie his Fig 11 is of a poor and very coloured sounding speaker
I think I may have a Wharfedale 'tech' brochure with similar pics I wrote for our Marketing Dept. somewhere in my junk to support job hunting in Oz.
The original idea was to spend a year seeing Oz and then get a job with an Australian speaker company. Instead it took me 2 years to get to Cooktown from Perth travelling over the top. I've been here this whole Millenium and will probably never leave.
I go to the Australian Festival of Chamber Music in Townsville and that's about it apart from the odd diving trip from Cairns. But this year, I'm officially a Senior and am entitled to a free annual train ride to Brisbane. I might take this up. Send me an email.
Last edited:
AE apparently make some fine products but there's not much objective data that I can find.
The TD15 series use a ferrite motor with well known, conventional techniques to reduce distortion.
It's actually comparable to a JBL 2227 in structure but the JBL is both more efficient and rated for more power.
So I don't see any evidence the AE motors are superior.
The AE website makes statements about problems with UH motors.
Perhaps they don't know how to optimize a UH speaker or perhaps they present a sub-optimal UH alternative to make their own choice look better.
In either case, their claims are not consistent with my results, or those of Don Keele, to name someone with a bit more industry street-cred.
I believe I can.
Better magnetic circuit structure- DCD.
Better magnetic materials- NdFeB has lower temperature co-efficient and better conductivity.
Probably better voice-coil structure and materials- Inside outside wound ribbon aluminium.
(I expect AE would conspicuously advertise it if they used these techniques so presumably they use lower tech)
And I have a couple of ideas of my own that I have not seen elsewhere.
So do I.
I expect they made reasonable decisions in a price sensitive market, my respect to them.But I think I can outperform them.
And in the meantime it provides motivation to learn more.
Best wishes
David
Last edited:
Hmm. Even his theoretical "perfect" loudspeaker in the first plots takes about 10 cycles to reach a stable peak value and then the same number of cycles to decay.
There is inherent trade-off between time uncertainty and frequency uncertainty - is 2 cycles adequate?
Alternatively, does Don's paper show implementation issues?
Somewhat related.
I started to look at JBL's own effort at a more efficient 2226, called the 2227 naturally.
More magnet so BL**2/Re is increased but the response becomes less flat, rises more in the mid but not much in the bass.
I suspect the Thiele/Small assumptions start to breakdown, radiation resistance starts to matter.
This is related because it's nicely covered in another Don Keele paper "Maximum Efficiency of Direct Radiator Loudspeakers" from 1991 convention (on his website, linked above).
The result is that the 2227 is not really an "improved" 2226, it's not simply more efficient.
I wonder is this part of the reason there are almost no woofers much more efficient than 98-99 dB 1W/1m ?
Any more efficient than that and then a tilted frequency response becomes an issue?
One option is to make the speaker as efficient as possible and who cares about T/S, just EQ it to what you need.
A bit heretical but I have started to wonder why not.
Of course, once an idea occurs then one notices other people with the same idea.
So I notice Earl Geddes seems to be of the same opinion.
Best wishes
David
Last edited:
A PAFplot shows just the decay but for single units (and most systems) this is more than adequate.
2 cycles is adequate if you are dealing with SOTA midrange & treble colourations.
No. This is well covered by TS assumptions. If you want more mid efficiency and the same LF response, you need a bigger box.
This is what Garner & Jackson formalised. For each type of loading, eg closed box, B4 ported bla bla there is a combination of Efficiency, LF response & Box size. They introduced a Figure of Merit for the various tunings.
One non-intuitive result is the B4 ported has a higher FoM than horns. I think Keele has a paper investigating this.
This is an important part of my Powered Integrated Super Sub tech. The Wharfedale Force 10 and a sub to do 32' ranks on organs were the most successful commercial implementations.
I'll have to get my patent lawyers to see Geddes
____________
Which speaker design programme are you using?
In Keele's plots even a perfect speaker takes 10 cycles to decay 80db.
So 2 cycles would surely not be adequate on his plots.
Presumably this is related to his analysis technique, I need to compare with Fryer.
How many dB did you measure over?
I understand the requirement for more box.
But I think that what happens with the 2227 is more than that.
Don does a more detailed analysis in his paper and I suspect it's applicable in this case.
Have you read that paper yet so we can discuss it?
I don't need no stinkin' computer
This is all a study of the equations and physics at the moment.
Not sure what I will use once I want actual numerical results.
Maybe just simulate in LTSpice.
You have any comments on software?
Best wishes
David
I know some of your Powered I.S.S is a P.I.S.S.take but I don't know what is serious.
Care to elucidate?
Can
Last edited:
Sorry. I lied. A PAFplot did 5 cycles and the vertical scale was 50dB.
In 1980, I would have said if a system showed a 20dB 'cliff' over all frequencies when the signal switched off, it would easily meet all the 'Audibility' criteria tested by Fryer & others. From what I've managed to see from Floyd & Co., this still holds. So Don's Fig 11 is OK but not Button's examples.
The speaker closest to meeting this was the ESL63 which has very good Mid & LF. What spoils its performance in DBLTs is the treble which is sorta blah. [*] There are many MC trebles including some cheap ones which do far better.
The speakers that did best in DBLTs have a balance of strengths rather than excel in any particular aspect.
We found looking at the 5 cycle plot the easiest to interpret though there are caveats too.
I found some plots and will scan & post in a day or 2.
Which one? Do you mean Comparison of Direct-Radiator Loudspeaker System Nominal Power Efficiency vs. True Efficiency with High- Bl Drivers
If you are after equations, the first accurate equation was by Novak (No. I don't remember the paper but it wasn't AES. Maybe IEEE Trans). What Thiele did was to realise you could use this equation to design formal filters. It was Peter Fryer you told me they were equivalent and I laboriously confirmed this in da days wen I kud reed en rite.
I converted & corrected Garner & Jackson (a Fortran programme running via a terminal to an IBM mainframe in London) to a Commodore PET programme with graphics. It was the first PC
When you've written your own, you are fussy about what you use but I don't have the source code now
I like Unibox which is Excel based but I can't save stuff cos I only have Excel 97.
Not cos its great .. but cos it does stuff like I used to do.Oh Powered Integrated Super Stuff was very real and it all serious. It was the R&D Depts' name for exotic LF schemes but Marketing never liked the moniker.
It started with Don Keele's 6th order alignment (Actually this is in Thiele's original but Don did a good analysis) and included both Analogue & experimental Digital schemes at the end. The commercial examples were all analogue. You can do loadsa good stuff simply once you decide to have a black box.
I forgot there's a couple of Celestion subs too.
You won't mind if I keep this to myself. If I ever abandon the beach, this is my most marketable skill. Very few people are interested in lower distortion bla bla ... but everyone wants a better boom box
[*] Peter Walker told me this was due to the dustcovers. The response in his AES paper was without dustcovers. I did do one measurement without the dustcovers which confirmed this. But its unacceptable in use as the speaker would not last a week cos electrostatic attraction of muck.
You'll have to excuse me if I'm tardy in replying over the next week or so. I've loadsa things happening over Christmas. Prod me in the New Year as this is a subject dear to me.
That sounds more believable. I try hard to develop some sense of how the maths is connected to reality, nice to have even a minor confirmation.
The more useful one is "Maximum Efficiency of Direct Radiator Loudspeakers", which I did reference in the prior post.
There is a Novak IRE transactions paper but also an AES paper, published more or less simultaneously.
I have noticed a certain lack of input on this thread.
A pity that "SpeakerDave" seems to be absent recently.
You would know him I think?
No problem, nice to have some helpful discussion.
Enjoy the season and New Year.
Best wishes
David
Last edited:
Hope you had a nice break, I had some ideas in the interim, one is for a carbon fibre voice coil former.
Have you any information on how they do "sandwich" voice coils where the former is in the middle, with a layer of wire on each side?
Seems a smart construction but not sure it's compatible with my idea.
Also what do you think of ATC's claims? http://test.atcloudspeakers.co.uk/wp-content/uploads/2014/10/ATC-CORP-BROCHURE-FINAL-10.pdf
I have some comments but I will wait, so as not to prejudice the discussion.
Best wishes
David
Is no one else interested in low distortion, low compression driver tech?
Last edited:
I recall reading a paper a decade or so ago (the main public library here used to carry JAES but stopped about then) mentioning the eddy current minimising technique ATC uses. I don't recall who the paper author worked for, but it may have been Skaaning or ATC. Seemed a reasonable theory at the time, but I don't recall any of the data.
I'm more interested in box design than driver design, although I shared a spreadsheet here where I toyed with driver design fundamentals here.
As far as box design goes the best source for info on that is probably the papers by J.E. Benson, which may be easy to find in Australia. (I'd love color scans, BTW
) The stuff from Small and Thiele is available on the internet from various places. Small's work has a nice section on alignments, which may be what you were after in Novak's work.
David,
1) As far as I can see, it's done with epoxy adhesives. B&C (and likely others) do two voice coil layers on either side of the former.
2) They seem reasonable, though the reduction in third harmonic seems similar to B&C's "Klipple optimized" magnetic structures, which result in extremely linear performance (and very low distortion) on drivers like the 18SW115 (which uses a four layer voice coil).
3) Few are, but I am. Was recently surprised to find a 3.5" Tymphony TC9FD-18-08 had far lower distortion on the same horn than the lowest distortion 3" diaphragm compression driver I have tested. The TC9 uses a copper shorting ring, but only a single layer voice coil.
Art
There was a JAES paper by ATC in July 2000.
But it seems a bit odd to me, when they minimize eddy currents they maximise inductance.
That's essentially the opposite of the usual technique of copper sleeves, pole caps and the like.
It's nice to minimize power wasted in eddy currents but not at the cost of flux modulation.
I have a suspicion that this area is not well understood even by professionals, the ATC brochure linked above is a bit wonky.
Epoxy is sub-optimal for this, polyimide is better.
And two round layers per side also seems a poor choice, I plan ribbon wound.
But my question was on how it's done.
For a normal VC I assume the former is held on a mandrel.
How is this done for a former material that comes in flat sheets?
Then the wire is wound on and this binds the former.
Polyimide pre-preg must then be cured, presumably best if it's still on the mandrel to maintain roundness but this would be less convenient than to remove the VC and cook in a batch.
So there's lots of production details I don't know even for a simple VC.
For a sandwich VC I assume they must wind the inner layer on a mandrel, place the former and then wind the outer layer.
Interested to learn the inside secrets from a professional.
As commented above, why do ATC do the opposite of practically everyone else?
I wrote unclearly, I meant drivers with low distortion and low [power] compression, not compression drivers.
But I am interested in compression drivers too and your results are of interest. Your Tymphany has a copper Faraday coil, the reverse of ATC's recommendation.
I have a definite suspicion about the ATC explanation, even if it probably is a fine driver.
Best wishes
David
Last edited:
Not sure what the advantage is .. or for Carbon Fibre VC formers
The former can be the major route for heat if its Aluminium. It needs to be very close to something for the heat to go somewhere.
Sounds like seriously liquid BS ...
... but their SLMM is almost certainly NdFeB which has saturation levels well above the 'best' steels.
I did toy with the idea to get above the 2T barrier with steels in treble units.
A Cu cap would get most of what they want except for the increased flux. But you don't need such high flux levels in a bass unit.
BTW, if you saturate the steel, you both reduce inductance as well as inductance modulation. It's in Gilbert Brigg's book "Loudspeakers". I wish I still had my signed copy.
Now that I want.
My most successful design, the 2nd best selling speaker in Europe for more than 7 yrs had a 4 layer coil. It's one of the best ways to get good thermal compression performance and power handling in a small unit but THD is pretty ropy.
It's a good thing THD comes very low on the priority list in DBLTs.
Our 'high temperature' coils were wound with polyimide pre-coated wire. The advantage is that you can partially 'set' the coil by passing a high current through it while its on the winding machine.
Flat former material is held onto Aluminium mandrels by end caps.
The mandrels are removed with the coils in situ and baked for final cure.
But our highest temperature coils were wet-wound with epoxy. This is a pain cos the glue is applied wet during the wind. Then baked on the mandrels as above.
I'm not at home yet so my contribution will still be sporadic
Last edited:
Motors aren't as hard as people like to think. The JBL LE8, the first versions of which were made in the 40s, are underhung with a shorting sleeve (even though it's only a plating and so less effective at lower freqs).
More important is the full scale modeling of suspension components and cones, IMO.
BTW, kgrlee, I understand wanting anonymity but there's something to be said for knowing a pro's body of work for context in these discussions.
More important is the full scale modeling of suspension components and cones, IMO.
BTW, kgrlee, I understand wanting anonymity but there's something to be said for knowing a pro's body of work for context in these discussions.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.