The interesting thing is that the minimalistic, single stage amps can easily be shown to modify and color the sound, while the multistage ones are generally (not always) far more transparent and modify the sound much less. The minimalistic ones often color the sound in their own way, and many people obviously like that.
I believe it is analogous to photos and paintings. Photos can make far more realistic reproductions from a scene than any painter can, yet many people buy the paintings.
But that's no reason to say that therefor the paintings are "better". It's apples and oranges.
Coming back to audio, if your goal is transparent, uncolored reproduction, there is no 'magic' number of stages, and less stages do not necessarily get you closer to that goal.
jan didden
In this case, I don't have scientific proofs, only a life-story.
Nearly 15 years ago, one really very experienced and respectable designer, proposed quite complicated, multystage schematics, of a power amp in AB class, deep GNFB, excellent in simulations (when simulations became possible) and in measurements, which was assembled and reported by hundreds of people.
Majority of them reported their dissatisfaction, the amp sounded as average mid-fi.
The designer feeled himself obliged to rectify the situation somehow, and after a decade definite improvement was achieved, due to PCB optimization, 3-layer, 5-layer. The schematics did not change, only PCB perfection. He claimed, that such an effect of electromagnetic interactions was not expected by him.
I listened not so many transparently sounding (without sibilance or darkness) SS amps, the MBL 9007 in monoblock mode is one of them. When I have read about it " 7-layer optimized PCB", I concluded this could explain the situation (provided all other important points, including power supplies and parts used, being excellent).
But, honestly saying, who among DIYers can admit several years optimization of multy-layer PCB, in order to achieve perfection with complicated-schematics design? It resembles me the situation with class D, it can sound good, if a big company spend several years for optimization of the device.
Also, stories about coloration of tube amps or not-many-stages designs, I consider "biased by interests". Look for effects of presence, when one feels a guitar just near by, and this feeling usually hidden not by THD, but by those effects refused by nowdays majority.
Designing amps, one must not be limited to using traditional parts, which demand many stages for achieving power. Look at radar-application aimed LDMOSfets, or HF MOSfets, with them some traditional stages become just not needed, and sound becomes more life-like.
Last edited:
Bias is a two-sided sword, my friend. Why not stick to what can be readily heard (and measured).
jan didden
Bob, you are welcome.
I do not have to read it as carefully as if I were reviewing it as no one will rub my nose in my misteaks as they are now private.
Ok so nobody reads my articles showing how PC layout affects measurable distortion. Or pays attention when I show the level of distortion in interconnections. But that was not the issue!
The argument started over triple stacked output transistors having more non linear properties than single or double. This of course is backwards! The drop across a junction increases exponentially with current. A single yielding the same output current will have all of the drive current (and its' variations) across a single junction. As covered this puts more variation on the preceding stage which is typically your voltage amplifier stage. That makes it less linear and it is the stage that has gain of course is the most important particularly when you use feedback either local to it or globally.
Depending on how you design your triple stack there is less variation of drive current on each active device of the stack. Of course it also does what is presented which is reduces the load on the prior stage.
I do not have to read it as carefully as if I were reviewing it as no one will rub my nose in my misteaks as they are now private.
Ok so nobody reads my articles showing how PC layout affects measurable distortion. Or pays attention when I show the level of distortion in interconnections. But that was not the issue!
The argument started over triple stacked output transistors having more non linear properties than single or double. This of course is backwards! The drop across a junction increases exponentially with current. A single yielding the same output current will have all of the drive current (and its' variations) across a single junction. As covered this puts more variation on the preceding stage which is typically your voltage amplifier stage. That makes it less linear and it is the stage that has gain of course is the most important particularly when you use feedback either local to it or globally.
Depending on how you design your triple stack there is less variation of drive current on each active device of the stack. Of course it also does what is presented which is reduces the load on the prior stage.
Last edited:
Bob,
knowing what happens in an amplifier dynamically would have a certain academic interest, but that is far too much to ask for. While functions can be expressed mathematically, the functions are far too many, the derivatives are far too many, the conjugate variables are at least one too many and real signals are far too complex. We will probably never know, anyway, no information is better than false information.
There are many good reasons not to refer too much to measurements.
knowing what happens in an amplifier dynamically would have a certain academic interest, but that is far too much to ask for. While functions can be expressed mathematically, the functions are far too many, the derivatives are far too many, the conjugate variables are at least one too many and real signals are far too complex. We will probably never know, anyway, no information is better than false information.
There are many good reasons not to refer too much to measurements.
>The drop across a junction increases exponentially with current. ??
Probably, you mean logarithmically. Never mind.
oops add inverse
I believe everybody would like to read these articles. Very interesting indeed. Could you post them here, or, please, post references?
They are in Audioxpress and Linear Audio. AX has some stuff at their website.
Hi WuYit,
I'm afraid that you are just dealing in generalities. You are throwing the baby out with the bathwater. It is certainly true that there are many things in audio that we may not understand, or may be controversial, but that is no reason to ignore good engineering practice. You would merely have us throw up our hands, throw out engineering, measurement and simulation, and just do a bunch of aimless experimentation followed by listening tests. I believe, design, simulate, build, listen, iterate.
BTW, there are many good-sounding minimalist designs out there, but most of them were arrived at with at least some directed scientific method and measurement.
Cheers,
Bob
Hello Bob,
I'd like to thank you for all the effort you put into your book. I've recently finished it and really enjoyed and appreciated the insights you provided, very well done.
I also like your comment above, except for the "at least some" aspect.
Years ago I had one of the high paid engineers tell me that designing without requirements is like throwing an arrow and painting a Bull Eye around it, I disagreed then (to some extent) and still do. I've always thought you design to please someone.
My home amps are to please my lovely wife (a linguist and music teacher “good ear”) who has always liked a good triode sound. Even with tubes she is discriminating, with my solid state versions I get the feeling she is just flattering me. Its really quite difficult, especially if you cant hear the same differences easily. Relating theory, designs, predictions, measurements to the satisfaction of the person you are designing for is in my opinion more than “at least some” worthy (sorry not meant to imply that a minimalist design is related to a tube sound).
Again, many thanks for your writings.
-Antonio
Yes, I see, and would like just to thank you for these efforts, since they proove the things, that sometimes estimated as being arised by fraudsters.
Thats a pitty for designers, that we easily recognize such kind of distortions near the thermal basic noise floor, since the problem shifts more to materials science and electro-magnetic pollutions, rather than to circuit design.
This is from the review of your paper: "Now, he (Ed Simon), being a circuit hack from way back, built what I can only politely describe as an extremely sensitive test device. I saw it, and the solder joints it was testing, and the cable that was, well, "kinda" bidirectional. Now, the artifacts in question are extremely small (when you read the article, try to digest the noise level he's working at, please), but he's making measurements very near the thermal noise level, and dealt with that quite carefully as well."
Hard work , actually.
First you have to do that with the prototype. Then try to go "minimalist" with any of the building blocks, trying hard to make the right compromises in the performance vs. parts count arena. So you start at "simulate" again to make the final version (it's also nice to have the working prototype to confirm) , then you have to do something that only experience can define .. layout (below).
For DIY , one might not want to overwhelm the builder (121 parts is enough) , an OEM can make a 200mm X 400mm main board and trim the parasitic's after the fact (lucky them !) , 600 parts w . daughterboards is not uncommon.
PS - that's what the final "DBT" will be ( below). MUCH harder than a EF2 or Lateral OP.
OS
First you have to do that with the prototype. Then try to go "minimalist" with any of the building blocks, trying hard to make the right compromises in the performance vs. parts count arena. So you start at "simulate" again to make the final version (it's also nice to have the working prototype to confirm) , then you have to do something that only experience can define .. layout (below).
For DIY , one might not want to overwhelm the builder (121 parts is enough) , an OEM can make a 200mm X 400mm main board and trim the parasitic's after the fact (lucky them !) , 600 parts w . daughterboards is not uncommon.
And they are easier to refine , layout , thermally compensate. Most are lower power class A or bootstrapped LIN AB's ... hard to go wrong. For 120v p-p with acceptable or excellent linearity , you have to throw a few more parts in there.
PS - that's what the final "DBT" will be ( below). MUCH harder than a EF2 or Lateral OP.
OS
