romys dset concept

[cressy]

ok. from reading various threads here where the dset concept has popped up and to provide a thread specifically to examine this i have created this thread.
my understanding of the concept is that there are dedicated amplifiers providing a very specific bandwidth for each driver in a multi driver speaker array.
eg, an amplifier optimised for a bandwidth of say 5 hz to 150 hz for a specific bass driver in that array. so the output transformers for example are optimised for this bandwidth, removing the compromise of having to have the output transformer recreate the full spectrum.
at the other end of the scale, one specifically for the selected tweeter, and one for the midrange.
this is probably an over simplification and romy will correct me if i am wrong.
i can see the merit in this approach, but i am sure that there is alot more to it, and can see that the speaker and amplification are inherently coupled together as each part of the amplifier system is tailored to the driver it is powering.
i suppose a very simple iteration of this is a full range speaker optimised for mid and top and a powered sub for the bass end. again, romy please correct me if i am wrong.
this thread is not here to induce a slanging match, or for participants to have a go at this approach, or to have a go at the approach others take, so please dont start an argument.
i actually want to understand why this is such a 'holy grail' for romy. obviously there are many approaches to take, and personally this is not an approach i can take as i am constrained by the speakers i could have. i have neither the space or funds or wife to go this route but as a concept i find it interesting.
so please explain this as fully as possible, not the how, but why you have gone this route, what it was that prompted this concept, (romy must have been doing something before dset) and the benefits you see in this approach. and romy please remember we are not idiots, laymen or beneath your level of intellect. i do not mean this in an argumentative manner, i mean that we are perfectly capable of understanding the ideas and perfectly capable of going this route if we choose to, but i can see that it is a large investment in terms of time and money.
cheers ant

[Nick]

I think its a perfectly valid way of doing things, with clear advantages. It is what Mark intended to do in his plan.

[Steptoe]

Dear Cressy, thank you for your explanation of Romy's ideas which, for the first time,I now understand. This was, in fact, a hobby horse for my late father in law who was for ever banging on about the physical impossibility of designing either speakers or transformers which could perform equally well across all audio frequencies. He would only use multiple speakers, arranged in a most complicated network of different pairs connected in both series and parallel configurations to improve the frequency response. He also stated categorically that anything less than four different o/p transformers, driven by its own amplifier, was a compromise too far. Needless to say, I couldn't hear any difference but it certainly kept him occupied. Yours sincerely, Steptoe.

[shane]

Does this approach go as far as tailoring the response for each amplifier to remove the need for passive "crossover" components?

[Cressy Snr]

Does this approach go as far as tailoring the response for each amplifier to remove the need for passive "crossover" components?

I would have thought so.
I think that's what Mark was doing with his SE Horn amp project.
Tailoring each channel, to its own particular driver characteristics would seem to be the logical thing to do.

So you have to pick your driver carefully, then fit your amp channel to that characteristic, probably four or five channels per side would be needed, each channel set up for its receiving speaker.

Then you'd have to blend all the levels properly to get the seamless integration from one section to the other and position the individual channels appropriately in space - front to back, side to side and up and down - to remove phase distortion effects. The only sensible way to do this would be by ear, ie. you would have to know precisely what you wanted and when you had achieved it, as conventional measuring equipment would be next to useless in this context.

Once this had been done, I reckon the resultant sound would be spectacular, with performers freed from the constraints of the hi-fi system, sitting right there in the room. Stunning I would have thought. The sense of realism would be so great, that the hi-fi considerations of coloration and flat frequency response and the rest of the stuff that conventionally goes with the deal would become the stuff of disdain.

I'm certain that if I had the time, the space and the money to realise such a system, I would find it almost impossible to listen to anything else.
I'd be divorced and sleeping in my bass horns - that's for sure.

[pre65]

I can see the merit of DSET, BUT, however good it might be I can't go there for many reasons.

1) Time. Seeing how long Mark has been experimenting, and still a way to go.

2) Money. A lot of iron would have to be custom made, and I'm a poor person.

3) Space. Not enough of it.

I'm reasonably happy with where I am in life, good sound, loving partner, reasonable health, good friends (including most on this forum),and interesting hobbies (not just this one). So I'll stay where I am, bumbling along building what pleases ME and trying to create even more of a monster with the soon to be rebuilt 833a amp.

I'd love to hear what DSET can do, perhaps Mark will have an open day when everything is sorted ?

[pre65]

As an aside, what would be each members favourite amps for (as an example) a 5 way speaker set up.

I seem to remember Andrew I has a 6C33s amp at Witham one year and the general comment was that it would make a great BASS amp.

[Romy The Cat]

Cressy, there is no reason to assure me that you are not idiot, laymen or beneath my level of intellect - do not behave like them and I will treat you accordingly.

OK, to understand what DSET does and it's sonic benefits you need to take two-way speaker, only a very good speaker and drive it with two identical very good SETs that you like. This is just a simple multi-amping. You might like result and most likely you will hear "someâ€￾ improvement as suddenly your tweeter does not hears the reactance of you woofer, you have no inner-modulation between the channels and many other things. Then you might discover that the passive filter in your speakers (or active filter before amps) are an impediment and since any amp is filter itself you will find in a given topology of your SET a location to integrate the filter with the amp, in most cases without using extra parts for filtration. Now you will get more improvement as you got rid of filters, begin to use subtraction filter (if you can hear), the filters see constant impedance, you use way small and peter parts and many other things... They you say: I am wasting a lot of my let say HF amp to drive my twitter, so can I optimize at least OPT to lose all unnecessary inductance from the core, to have HF optimized winding, drop the capacitance to absolute minimum and let say to use faster core. Then you do the same for LF amp gaining inductance and dropping DCR let say. As you experience even more gain of performance you might ask yourself if you can optimize PS for separate HF and LF amplification, you play with rectification type and filtration type and you will discover that it make difference. Then you need different power. For your HF you might need let say 3W but for your LF you would need 30W. You get AD1 let say and GM70 and realized that different output tubes in context multi-amping do not work well. In fact different tube drivers do not work well and different topologies of amps could not multi-amp. Well, then you will be trying to make acrobatics with your SETs to get seminal harmonic and dynamic characteristics (across the different volumes) in order to properly integrate them in context of a single acoustic system. You might discover then your speaker drivers are not enough for you and you might update them. Then playing with different loading scenarios you might find a configuration that you feel does what you need. You will lock the success and will start to ask yourself of what will happen if it was 3 channels or perhaps 7 channels....

There is one very important thing in all of it. In audio everything works. If you take a tube and put it to plate voltage it will sound. It will produce a rundle accidental sound that has no meaning. 95% of people in audio do listen exactly this - an accidental sound. They do differentiate the differences between operation point of tubes and quality of resistors and caps but it all juts differential different and it serve no higher objective then to be another pastime. In order to navigate a serious DSET to right direction a person shall have a very well formed sense of cultural and musical identity and equally important to develop a proper objective (subjective) methods of listening assessment. There are very few people out there who have those skills and if one does not have it then I would not advised to go for more than 3 channel DSETs, perhaps this is why most of multi-amps out there sound like shit.

For sure an over 3 channels DSET is a lot of efforts but surprisingly little money. SETs are expensive as they try to be full range but DSETs are not. In my Audio For Dummies section I wrote a few years back the following:

http://www.goodsoundclub.com/Forums/Sho ... =5344#5344

I do not consider that DSET is a holly grains as you put it but I do feel that it is a valid mechanism to make a playbacks to sound how you need it to sound instead of having "as isâ€￾ result.

Rgs, Romy The caT

[Alex Kitic]

Just as Steve puts it, and Cressy explains it, and Philip would like to try but does not have the time or space or enough wish to do it, the DSET concept is basically multi-amplification.

The fact that it is being done with SET (why not SEP?) amplifiers with different power output, and tailored frequency range for each amp powering each speaker in the array - would not warrant it a different name, besides the implied need for "dedicated" output transformers.

The number of speakers necessary might vary between 4 (with a sub-bass?) and "just 2", with maybe 3 ways being an optimum in terms of space.

The multi-amplification concept with tube amplifiers, even on efficient speakers, is not new (I mean, multi-amplification is not new at all). I remember reading an interesting article a very long time ago in Sound Practices. If I recall correctly, it was called "The Island System", or something like that. I think the bass was a PP amp, but the remaining 2 speakers were powered by SET amps. I have found the idea very interesting at the time, but after all these years am not yet convinced in the rationality of such a choice, as well as completely aware of the inherent difficulties, obstacles, and the lurking danger of crossing the line between listening to music, and playing at music production.

As for myself, I could do it rather easily, since I have several SE amps of different power at hand, as well as 3-way speakers that being not efficient would probably benefit by being multi-amplified. All it would require is some rewiring and the change of a few caps/resistors in the amps in order to achieve crossover points.

What I see as a potential benefit is not burning power in resistors, and eventually no need for passive crossovers. On the other hand, I would gladly keep the Zobel networks, and if the crossovers are not particularly complex, the gains achieved by removing them are rather relative.

The sum of the power of the various amps can easily be achieved by a more powerful amplifier, like in the case of my current project, the RH813. At 35W per channel, it is 5x more powerful than the RH-TTA, or 3x more powerful than the RH300B... and it shows. Loudness levels and implicit dynamics are really impressive - but at 88dB/W/m I guess the same might be achieved with 3.5W, or let's say 5W with more efficient speakers, like 98dB/W/m. Thus in a way we are back where we started, it's power vs. sensitivity, where sensitivity has an elegant way of winning, although power may achieve it's own elegance by means of sheer size...

Another issue is the "dedication" of output transformers. This is, in my view, the "weakest link" in the concept. Not because one should not play with transformers at ones own will and expense, but because it is something that, on one hand, goes "normally" with power (if the amplifier is a 15W unit, it must have a larger transformer than a 5W unit, while a 1W amplifier may be built with a rather small output transformer). On the other hand, sheer size is not enough explanation, as it also depends on the permeability of the core material and the geometry of the windings, and other factors, as that all combine in obtaining a certain primary winding induction needed for extended bandwidth at a given power. The sheer fact that you will not need more than probably 3 to 5W for a mid-driver horn, and probably just 1W for a high frequency driver horn, means that those transformers will be small, and no particular care needs to be taken in the intrinsic design of these units (assuming quality units, not recycled console transformers) in order to achieve adequate bandwidth for the task (not forgetting that the input frequency range is already determined). All in all, adequately powered transformers will most probably be "dedicated" per-se, without the need to "dedicate them"... which is borderline mystification.

On the other hand, there are means of achieving adequate bandwidth from an amplifier (as a sum of parts, including the output transformers) and if the concept is not necessarily a "no-feedback", just the adoption of feedback (there are several ways to do it, of which one is what I do in my RH amps) is enough to achieve adequate bandwidth, particularly with high quality transformers... not to mention the fact that frequency response can always be tailored to increase bandwidth beyond the sheer physical limitation of a given output transformer. I do not look at the intelligent adoption of feedback as something to be avoided, but quite the contrary.

And, all the above mentioned physics can be calculated, simulated, measured, and can find objective proof. The final judgement whether it is suitable for the purpose of the highest quality sound reproduction is probably in the ear of the listener.

Once we have adequate bandwidth from a single amplifier, and assuming enough power/sensitivity, it is highly questionable whether improvement might be further achieved by adopting separate amplifiers for each driver unit... and if so, whether it would fall under the category of the ever present law of diminishing returns?

[Nick]

Once we have adequate bandwidth from a single amplifier, and assuming enough power/sensitivity, it is highly questionable whether improvement might be further achieved by adopting separate amplifiers for each driver unit... and if so, whether it would fall under the category of the ever present law of diminishing returns?

Yes, but as Romy says, what you are ignoring is the effect of the drivers on each other. To prevent that requires solid state amps, or at least amps with moderate amounts of feedback, to many either of those choices bring their own problems.

It seems to me the aim of the DSET setup is to gain many of the advantages of a SET + single driver, but to avoid the restrictions of a single drive unit.

[Alex Kitic]

...what you are ignoring is the effect of the drivers on each other. To prevent that requires solid state amps, or at least amps with moderate amounts of feedback, to many either of those choices bring their own problems.

It seems to me the aim of the DSET setup is to gain many of the advantages of a SET + single driver, but to avoid the restrictions of a single drive unit.

Indeed, since my amps employ feedback, the influence of drivers on each other becomes less important. That said, there is no mystical interaction between drivers, it all comes down to combining impedances with other passive elements in a crossover: the result is what we call "loudspeaker", with all its characteristics. If you don't like a loudspeaker, you choose another, or build your own. Let's not forget the box, or lack thereof (open baffle), as well as the various systems for loading the drivers, all interacting obviously with the driver itself.

I understand how feedback introduces problems, and that is why I have devoted a lot of time and effort in designing my amps - around a form of feedback that does not add more problems than it is able to solve (to put it lightly).

I would really like to avoid offending anyone's feelings, but in my view there are very few advantages in a so called no-feedback SET (what is generally expected to match the SET description), which again is the reason why I have devoted time and effort both to feedback mechanisms, and beam-tetrode/pentode tubes. I have recently widened that to applications with triodes, since I feel that triodes can benefit from the same treatment (while the effect on beam tetrode/pentode tubes is much more pronounced and effective, the improvement with triodes is important nevertheless).

As for single driver systems (like Lowther), the comparison with multi-amplification is very superficial. This is not about amplifiers, since you can use a 200W class D with a single driver system, but rather about the drivers. Single driver is not necessarily about just being crossover-less, rather about avoiding the need to match the physical and electrical characteristics of several drivers. This issue is much more difficult to solve, as it is not only about avoiding passive crossovers, and while it may be solved with time and effort (and at a cost, obviously), it seems much easier to "tame" a single driver speaker by adequate tailoring of frequency response on it's "dedicated" SE amplifier (I could easily iron our most frequency response issues tailoring the response of an RH amp, to the point allowed by the physical limitations of the driver or its loading, i.e. horn type or length).

With this approach (single driver plus tailored response RH type amplifier) it would be much easier, faster, even cheaper - if you wish - to achieve the promise of "natural dynamics" promised by passive-crossover-less systems, within the physical limitations of the driver, the loading horn, and of course the listening room. Yes, it would include some careful planning and calculation, but isn't that what we like to do?

[Paul Barker]

Question for Romey and definately not a trick question a trap or a pretext for an argument. A genuine question.

you refer to "fast" material for HF output transformer.

I do know that Chris Vry always wanted to get some air cored wound for the purpose. Don't know if anyone ever achieved it.

What would you consider the right materials approach for the tweeter transformer? Are we talking as much percentage Nickel as we can afford? or some sexy modern fancy power transformer material?

[Cressy Snr]

In order to navigate a serious DSET to right direction a person shall have a very well formed sense of cultural and musical identity and equally important to develop a proper objective (subjective) methods of listening assessment. There are very few people out there who have those skills and if one does not have it then I would not advised to go for more than 3 channel DSETs, perhaps this is why most of multi-amps out there sound like shit.

Rgs, Romy The caT

Multi amps sounding like shit is something I have experienced.
Twenty five years ago I went to a hi-fi show in Sheffield.
I had the dubious pleasure of hearing a Naim 135 six-pack driving Naim DBL speakers. Front end was a Linn LP12/Ekos/Troika and a Naim CDS CD player into NAC52 preamp. Must have been £20,000 at least.

It was offensive, but a lot of the customers were bopping along, lapping it up, feet tapping and boogieing; encouraged by the Naim sales staff.

My mate Clive and I fled to the QUAD room where Ross Walker was playing ESL63s driven by the then new QUAD 606 power amp. It was lovely.

For me Naim had had a golden opportunity to make multi amplification blow everything else out of the water. They failed, instead, making a godawful racket.
That Romy has succeeded with this method is obviously a testament to the skill, and the vision required to pull it off, because on the evidence of that Naim demonstration, multi-amplification is not something you want to be doing badly.

Now I can design and build the amps, but I doubt very much if I have the skills needed to blend it all into a coherent whole, so I stay away from it. But that's my problem, nobody else's.

Those that are cable of getting the cultural/artistic side of the equation right, I say good luck to them.

[Romy The Cat]

I forced myself to read two of Alex Kitic post and it is self-evident to me that the guys is clueless. Nope it is not a personal attack, but rather this is an unfortunate circumstance when a person use word but has no comprehending of real meaning. This common "angry electricianâ€￾ class is very common in audio, regrettably completely impotent from sonic perspective. The few things that Alex Kitic has expressed "between the linesâ€￾ identify him mistakably.

Nick, there is no such a thing as "adequate bandwidth from a single amplifierâ€￾. There is so many reasons why if you actually do not run theories but rather practice applied construction of sound. For instance the presents of amp bandwidth is a fictional metaphor that does not exists in DSET word. If you are a speaker driver then what for you would be a DSET bandwidth? What would be difference in case of a full range SET? They are very simple question but even they portray fundamental difference between SET and DSET. Let look into this. In case of SET bandwidth is in most cases is a bandwidth of the OPT at a given power. In case of DSET bandwidth is not limited by anything but an explicit filter. There is a huge different between OPT role off and a filter roll off. In case of a DSET filter you use a constant prefect load and high impedance, you write a unadulterated minimal phase roll off. In case of SET only god know how magnetics roll off. It is never linear, it is with variable slopes, it is not stable across dynamic rage. The transformer in the roll off zone has huge phase problem, so to be save your speaker channel need to be let say 2 octaves from any OPT roll off zone. This is very simple and very naturally could be accomplished by DSET and it is impossible by regular SET

[Romy The Cat]