Like most people, I've heard other systems where everything is in alignment in that particular instance, for the specific recording. But, the aim is for a setup to always be in that groove, every time you decide to play something - the ideal is for all the parameters that make that happen to be fully known, and under your control at all times. I'm steadily getting closer to having that be - not perfectly by a long shot, very much still a learning exercise, but well worth the journey so far.
At the moment I'm exploring getting a high power amp to run, in sim, at sub-ppm distortion at all powers, into impedances of 2 ohms or so ...
At the moment I'm exploring getting a high power amp to run, in sim, at sub-ppm distortion at all powers, into impedances of 2 ohms or so ...
I've never measured this because it doesn't apply to me, but I have a nagging feeling that the Karan, fabulous as it may be, does like to pick and choose speakers it will do best with.
This is because with AR94s its sound is a truly shade different than with my other two pairs of speakers, both of which are nominally 8 Ohms. I'm not even 100% sure I'm not imagining it, but if so, I must have an uncanny ability to imagine something which then comes off the same each and every time. It's just a shade, but after so many years, I hear it. Most people don't, not after a 30 minute session, or some such.
Thus far, the literally only amp which changes 0% with any speaker I own is the HK Citation 24, which is why I am so taken with it. So far, it appears literally bullet proof sound wise, in conjuction with its team mate, Citation 21 preamp. Originally, I was after the bigger model of the premp, but have in the meanwhile discovered that the two are exactly the same, onl the bigger model has a different signal switching setup and more fascilities - this means zilch to me, as I honestly don't need the added extra.
I believe this completely. A snare drum will draw prodigious power levels from the amp just to stay in tune with the rest, so if you're going a bit louder than usually, 50W or more is no wonder.
And if you happen to have a rather large room, and your bass driver is not all that efficient ...
At the moment it's purely an exercise, to understand precisely what factors contribute to optimum technical performance, and what elements subtract, and to what degree. At the moment this performance is achieved in fairyland sim world, of near perfect power supplies, perfect grounding, nil extra parasitics - it will an interesting process, adding in more and more real world imperfections, and seeing if desirable levels of behaviour can still be extracted from the circuit ...
Last edited:
"Fairyland sim world".
My experience is such that my sim is sometimes even wildly conservative. If it says 300 kHz, you can bet your bottom dollar the circuit will do at least 360 kHz. If it says THD is say 0.05%, you can bet it will be more something like 0.04% or less.
In the last 5 years, it has never failed me once, the actual model will perform as it says it will, only a bit better than it says it will. Its older versions, back to 2001, were exactly the same.
I will readily agree that the fact that I use 20+ years of experience PCB artwork designers has something to do with it - I am good, but they are better, so I get them to do the artwork. Their work is more compact and has less spurious capacitance than I'd get, to be sure. I try to never underestimate anyone's expertise, and I never assume friendly favors, I prefer to honestly pay for honest work. Why not, I don't work for free either as a general rule, though I'm not above friendly favors for individuals I deem as good people. That definition covers a surprisingly large number of people from this form, for example (not that many need favors).
Frank, I believe in the wide open loop bandwidth approach. I've stopped discussing this matter with others a long time ago, ever since I became aware that most audio electronics I have encountered and think of as good sounding use this approach. Obviously, it's hardly an automatic process, simply using this approach guarantees nothing but in my view a better chance of getting it right.
Much like yourself, I assume that to get it right, one needs very good power supplies. It's like our own heart, and one can't ask a man with a cranky heart (poor PSU) to run a marathon (go loud without changing its voicing). I am sometimes criticized for being a PSU glutton, but I see nothing wrong in dimensioning my PSU to the highest demands I have from the amp, plus a safety margin.
If your output stage can deal with nasty loads and not run out of breath, I feel most of your problems will never happen.
My only issue with chip amps is twofold, and I must stress I have little experience with them: they typically require what I see as a lot of global NFB (a most arguable point), and they are not very good at delivering high currents. My AR94 speakers, a nasty load with lower efficency of 87 dB/2.83V/1m, say so.
The only chip amp I was pleased with is the now long gone LM 391-100 (I think?), which was in fact a driver chip which required outside discrete driver and output tranistors, obviously because I could drop in things like MJ21193/21194, and more than one pair, too.
"Fairyland" in that insufficient modelling has been done, that it is necessary to add in, as conventional Spice elements, devices that model more closely all aspects of the environment - the closer the Spice circuit is to the real world, the less the subtle things will bite you.
Hefty supplies are good, but I part ways on FB philosophy - I'm with Putzeys, either don't do global feedback at all, or use lots and lots of it; a halfway house is a poor compromise, IMO. Where chip amps can falter is that the PSRRs are not good enough, so very stiff, stable power supplies are necessary to get the best from them - I wouldn't try and drive very current hungry speakers with them - or only do so with heavy paralleling of units.
Hefty supplies are good, but I part ways on FB philosophy - I'm with Putzeys, either don't do global feedback at all, or use lots and lots of it; a halfway house is a poor compromise, IMO. Where chip amps can falter is that the PSRRs are not good enough, so very stiff, stable power supplies are necessary to get the best from them - I wouldn't try and drive very current hungry speakers with them - or only do so with heavy paralleling of units.
MC2 Audio. Two from their now discontinued MC series and one from the T series.
The audio circuitry is the same for all of them, they differ in the protection circuitry only.
For the super tweeter I use an Omniphonics/BeyerDynamic which is also a BJT amp but with my advancing age (50 later this year) I hear less and less from this one as it only kicks in at 15kHz.
Does half a chip amp count ?
(the Karan looks like one, two of them tied in bridged mode to reach the power level)
What are you talking about, Jacco?
True, Karan uses BB 2604 as power amp inputs, but after that, everything else is discrete, What you may have seen were the wide, 2SA/2SC LAPT Sanken transistors in MT-200 package. There are two pairs of them per channel, a SEPP output per each half of the output signal waveform.
His biggest offering, rated at 1,200/2,200/3,600 W into 8/4/2 Ohms, uses (if memory serves) 20 pairs of these transistors per channel, weighing in at just over 80 kilos (app. 180 lbs).
He is in fact totally partisan about these Sanken devices, they and none other are good enough.
Gummel-poon spice model does not model :
- quasi-saturation (ie, ugly stuff that happens to BJTs when Vce is low but not low enough to saturate)
- self-heating
- and some other "details" (approx. 1 truckloads)
Mextram BJT models fixes all the above. Interesting. Where can I get mextram models for my favorite BJT ?....... hm, nope. Can't find any. Duh. Can I build a model for my favourite BJT ? (sure, need 6 months work for an individual like me, $$$ gear, life is short).
I've torrented microcap 10 and tested default mextram models for BJTs... works as it's supposed to do, self heating included, but there is only one BJT model included, the default one... which bjt is it, discrete or IC, which process, dunno. This looks like the modeling software that would be useful to big boys IC manufacturers, and totally useless to us normal guys....... duh.
Keep (dare i say ) chipping away , the truth will reveal itself ..
Sanken makes Sanken transistors. They have been at it for over 60 years now. They were the first to introduce RET (Ring Emitter Transistor) technology, where a power transistor consists of several smaller transitors in a ring. This allows for rather high FT of 60 MHz and more, without sacrificing current capability and therefore power capability.
An odd way to categorize, in my view. Using that logic, even Phipils' behemoth Lab series amp, good for over 200W/8 Ohms, also using an op amp for its output stage would be a "half chip" amp. As would quite a few other amps, such as some Aragon models, designed on commission by Dan d'Agostino, and so forth.
I don't much like that approach, but I have to admit that if judged by the Karan, it sure can work, because after all these years, it's still the best sounding integrated amp I have ever heard. Not perfect, but very close. Of the ones I've heard, of course.
Frank, I find your logic regarding global NFB, boiling down to all or nothing, to be rather strange.
If regarded as a oersonal vuew, I can live with it, but overall, it seems bizzare to me.
I have always thought of global NFB not as a panacea here to solve all problems, but simply as a way to improve something that should be fairly clean even without it. In other words, global NFB to me is much like icing on a cake; no icing will ever make a poorly baked cake a great one, but it can improve a mediocre cake. But on a good cake, it can be the ultimate refinement.
In my experience, every zero global NFB audio device I have ever come across always seemed that little bit out of absolute focus, that little bit without the small detail, no matter what its price point and land of origin. So while it may sound fine as it is, I am always left feeling a little bit shortchanged. And, let's be honest here, with zero global NFB overall staility is a much bigger issue that it needs to be.
I bought into Matti Otala's concept that we should make it as clean as possible, and then only use as much global NFB as we need to sort of iron thing out. I have no preset must-do limit or else, but I find that in most cases, something between 20 and 26 dB should suffice. I find this is not very hard to do if one controls local stage (input and VAS) gain, both by setting limits and using local stage NFB by local degeneration. For example, using local degeneration on the input stage, to keep its overall gain down to 20 dB or less.
How can this be bad? How is that worse than using much more? Do you seriously think that an amp hitting all of 400 Hz open loop and needing say 80 dB of global NFB will sound better than another amp, hitting 50 kHz open loop and using just 26 dB of global NFB?
If regarded as a oersonal vuew, I can live with it, but overall, it seems bizzare to me.
I have always thought of global NFB not as a panacea here to solve all problems, but simply as a way to improve something that should be fairly clean even without it. In other words, global NFB to me is much like icing on a cake; no icing will ever make a poorly baked cake a great one, but it can improve a mediocre cake. But on a good cake, it can be the ultimate refinement.
In my experience, every zero global NFB audio device I have ever come across always seemed that little bit out of absolute focus, that little bit without the small detail, no matter what its price point and land of origin. So while it may sound fine as it is, I am always left feeling a little bit shortchanged. And, let's be honest here, with zero global NFB overall staility is a much bigger issue that it needs to be.
I bought into Matti Otala's concept that we should make it as clean as possible, and then only use as much global NFB as we need to sort of iron thing out. I have no preset must-do limit or else, but I find that in most cases, something between 20 and 26 dB should suffice. I find this is not very hard to do if one controls local stage (input and VAS) gain, both by setting limits and using local stage NFB by local degeneration. For example, using local degeneration on the input stage, to keep its overall gain down to 20 dB or less.
How can this be bad? How is that worse than using much more? Do you seriously think that an amp hitting all of 400 Hz open loop and needing say 80 dB of global NFB will sound better than another amp, hitting 50 kHz open loop and using just 26 dB of global NFB?
Pierre, there's no doubt that SPICE modelling leaves much to be desired, and that we all could do with more down-to-real-life data incorporated, but I must admit I am a little surprised at all this wailing at the sims.
It would be hard to believe that I am the only one so lucky as to use a sim which corresponds well to reality, and is in fact even a little conservative. Since it allws me to create my own SPICE models, this is what I did, based on manufacturer's data sheets, and, if I use the same transistor from different sources, I make as many models as I have sources (e.g. BD 139/140, models for Siemens, Telefunken, Motorola), since differences are small, but sometimes evident even from data sheets.
The only factory supplied SPICE models I use are those from Motorola/ON Semi, as I discovered that their Group II SPICE models work well for me (there used to be 4 SPICE versions available, don't know about now).
dvv, I won't get into a heavy technical discussion on the ins and outs of it, but I have always seen the value of full blown FB to be self-evident. I note that you were already recommended to read the article by Putzeys, "The F-word - or, why there is no such thing as too much feedback", which states in a far more authorative manner than I ever could, what "it's all about".
Subjectively, I've never heard a problem with FB correctly applied, and the vast majority of recorded material has already been 'heavily contaminated' from passing through countless active devices using 'full' GFB with gay abandon - it doesn't sound too crippled to me ...
I'm sure the solutions you've applied work extremely well - but I would always be happier to use gung-ho FB, to 'guarantee' technical correctness ...
- Status
- Not open for further replies.