I was wondering about this
Lets say that we want to create from scratch an "entry level" 10-15W PP amplifier.
Do You think it will be useful to put together a set of design/construction rules that will .."Remove all the builder/assembler mistakes
to improve the audio performance to a standard for which the builder/assembler cannot be blamed." ,
i.e in a similar way D. Self do for solid state circuitry?
Or, the valve input-driver stage topologies are simply too many to redact such a set of rules?
Lets say that we want to create from scratch an "entry level" 10-15W PP amplifier.
Do You think it will be useful to put together a set of design/construction rules that will .."Remove all the builder/assembler mistakes
to improve the audio performance to a standard for which the builder/assembler cannot be blamed." ,
i.e in a similar way D. Self do for solid state circuitry?
Or, the valve input-driver stage topologies are simply too many to redact such a set of rules?
Member
Joined 2009
Paid Member
Back in my "Sand Pit" days I built quite a few of Doug Self's "Blameless Amplifiers".
It was very interesting to note which of his blameless mods actually improved my enjoyment of the sound and which did not.
For example the bootstrapped load on the VAS gave the amp a lovely tube like sound which the CCS load "blameless mod" killed stone dead. So that "blameless mod" was deleted.
I went from a version of that amp (Hugh Deans AKSA55N+) to the Baby Huey tube design.
Cheers,
Ian
It was very interesting to note which of his blameless mods actually improved my enjoyment of the sound and which did not.
For example the bootstrapped load on the VAS gave the amp a lovely tube like sound which the CCS load "blameless mod" killed stone dead. So that "blameless mod" was deleted.
I went from a version of that amp (Hugh Deans AKSA55N+) to the Baby Huey tube design.
Cheers,
Ian
Hi,
Its doable for a mullard 5-10 type design, details are in :
30W Push Pull amplifier designed by Claus Byrith, Royal Academy of Music in Aarhus, Denmark. | Lundahl Transformers
rgds, sreten.
FWIW "a lovely tube like sound" is the very antithesis
of the "Blameless" concept and is a nonsense critique.
Its doable for a mullard 5-10 type design, details are in :
30W Push Pull amplifier designed by Claus Byrith, Royal Academy of Music in Aarhus, Denmark. | Lundahl Transformers
rgds, sreten.
FWIW "a lovely tube like sound" is the very antithesis
of the "Blameless" concept and is a nonsense critique.
There's one for the list. I'll add:
The Red light District
The "El Cheapo"
Tubelab Simple PP
Dynaco ST-35
There must be more.....
jeff
Poor old Doug Self was much misunderstood by many, even though his writing style is very clear. And like me he didn't like people dumping unfounded criticism on him. And he really didn't like people building his designs, making mistakes & changes, and then complaining the performance was not good.
His major concept was not that his circuits were especially good (though they ARE pretty good); it was that by scientifically investigating EACH SEPARATE cause of distortion, and properly understanding each cause, you are then in a position to optimise a given amplifier topology for the lowest overall distortion.
His work has its limitations in these ways:-
1) He limited his work to only one type of topology. There are others.
2) He focussed on THD. THD has only a rough correlation to the pleasure in listening, there are other factors, and the THD of almost any amp these days is well below what you can hear.
3) He used SPICE as his primary analytical tool and thereby misled himself on the performance of power FETs.
4) He didn't appreciate that the ear's sensitivity to distortion corresponds to the time average probability density function of the music. What this means is that a step in the transfer characteristic (say due to cross-over distortion) at the quiescent point (instantanous zero volts to the loudspeaker) is a lot more unpleasant than the same size step at say +5 volts or at -5V. That's because music and natural sounds cross zero more often than they cross higher voltages.
It ought to be readily possible to do the same for tube amps - that is, select a topology (say triode input amp, concertina phase splitter, tetrode push pull output, or whatever takes your fancy), and analyse it & test it to understand the various causes of THD, transient intermod, or any other type of unpleasantness, and then produce the best possible performance that particular topology is capable off.
But I have never seen it done in one series of articles or in one book like D Self did for his selected solid state topology. It most definitely HAS been done bit by bit in a considerable body of 1950's literature - professional journals, tube manufacturer applications lab reports,etc. Trouble is these are not "accessible" to non-professionals without a lot of work.
I've given serious thought of producing a couple of Doug Self-style "blameless" article series for topologies for tubes and FETs. Four things inhibit me:
a) kit sellers see no money in it. And kit sellers have a considerable influence on what magazines publish;
b) The editors of the few remaining electronics magazines like Silicon Chip don't like detailed engineering. What they do like is projects that newbies can succesfully put together.
c) It would be a lot of work with little scope to bring in something really new
d) The example of Doug Self. He did truely outstanding work and got pilloried for it.
I've built Class A solid state amps. The D Self style topology used by Silicon Chip for ss Class A is good, very good, but it is not the best for Class A.
I've engineered and built FET based amps for professional test purposes. A reference amp for evaluating other amps must have perfomance comfortably exceeding the amps under test. And I got that performace with FET's, depite Doug repeatedly asserting, with some good arguments, why that should not be so.
Perhpas people should tell the editors of Silicon Chip, Elektor, and the like, that blameless amp articles for different topologies is what they want. Lobby for what you want. If editors see a strong need, they find someone to do the job.
His major concept was not that his circuits were especially good (though they ARE pretty good); it was that by scientifically investigating EACH SEPARATE cause of distortion, and properly understanding each cause, you are then in a position to optimise a given amplifier topology for the lowest overall distortion.
His work has its limitations in these ways:-
1) He limited his work to only one type of topology. There are others.
2) He focussed on THD. THD has only a rough correlation to the pleasure in listening, there are other factors, and the THD of almost any amp these days is well below what you can hear.
3) He used SPICE as his primary analytical tool and thereby misled himself on the performance of power FETs.
4) He didn't appreciate that the ear's sensitivity to distortion corresponds to the time average probability density function of the music. What this means is that a step in the transfer characteristic (say due to cross-over distortion) at the quiescent point (instantanous zero volts to the loudspeaker) is a lot more unpleasant than the same size step at say +5 volts or at -5V. That's because music and natural sounds cross zero more often than they cross higher voltages.
It ought to be readily possible to do the same for tube amps - that is, select a topology (say triode input amp, concertina phase splitter, tetrode push pull output, or whatever takes your fancy), and analyse it & test it to understand the various causes of THD, transient intermod, or any other type of unpleasantness, and then produce the best possible performance that particular topology is capable off.
But I have never seen it done in one series of articles or in one book like D Self did for his selected solid state topology. It most definitely HAS been done bit by bit in a considerable body of 1950's literature - professional journals, tube manufacturer applications lab reports,etc. Trouble is these are not "accessible" to non-professionals without a lot of work.
I've given serious thought of producing a couple of Doug Self-style "blameless" article series for topologies for tubes and FETs. Four things inhibit me:
a) kit sellers see no money in it. And kit sellers have a considerable influence on what magazines publish;
b) The editors of the few remaining electronics magazines like Silicon Chip don't like detailed engineering. What they do like is projects that newbies can succesfully put together.
c) It would be a lot of work with little scope to bring in something really new
d) The example of Doug Self. He did truely outstanding work and got pilloried for it.
I've built Class A solid state amps. The D Self style topology used by Silicon Chip for ss Class A is good, very good, but it is not the best for Class A.
I've engineered and built FET based amps for professional test purposes. A reference amp for evaluating other amps must have perfomance comfortably exceeding the amps under test. And I got that performace with FET's, depite Doug repeatedly asserting, with some good arguments, why that should not be so.
Perhpas people should tell the editors of Silicon Chip, Elektor, and the like, that blameless amp articles for different topologies is what they want. Lobby for what you want. If editors see a strong need, they find someone to do the job.
Last edited:
I was aware of Linear Audio.
Trouble is, it is not available in newsagents. It is a closed circulation magazine available online from his website. These sorts of restricted distribution ezines usually don't pay their authors. Otherwise I would publish in the formerly mass-distribution ELectronics World whose editor Svetlana Josifovska told me she's desperate for authors. I decided to to see how well Linear Audio went, and if it continued. Then I would find out what his rates were. Unfortunately I'm not THAT atruistic to do hundreds of hours work for no pay. It certainly has continued - he's up to issue 8 now.
Thanks anyway.
Last edited:
Music is much like random noise mathematically. Imagine a white noise signal. The instantaneous voltage jigs about sometimes postive by various amounts and sometimes be negative by various amounts. A few properties apply to such a signal, and they pretty much apply to music and natural sounds as well:-
a) the voltage measured over a long time interval averages to zero.
b) the time spent positive averages to equal time spent negative.
c) if the voltage is sampled at any particular instant the probabity of it being any particular voltage when plotted on a graph against voltage is the famous "bell curve" peaking at the average voltage value (which is zero) and having long tails tapering off towards zero the more postive or negative you look.
In theory for pure white noise, the "arms" on either side of the probability peak don't get to zero probability until infinite neg or pos voltage.
Of course in any real physically built system, the bell curve doesn't go to infinite neg or pos voltage - there is a limit, hard or soft - the overload point.
As noise and music has maximum probability at zero voltage, noise and music has the maximum probability of CROSSING any given voltage when that voltage is zero.
If an amplifier has a change in slope, or singularity as mathematitions call it, somewhere in its transfer curve, then by definition distortion occurs at that point, where the signal passes through, whether than point be zero volts, +1V, -10V, or whatever.
But if the music does not cross that particular voltage where the change in slope occurs, then clearly you will not be able to hear it.
It happens that the ear finds it extremely difficult to hear a distortion that occurs once, for less than 10 to 20 milliseconds or so. It appears that the brain does not have the ability to process on such short timeframes.
But if the distortion, lasting at time for even only 100 microseconds or even somewhat less, keeps on happening, then the more often it happens, the more obvious to the ear it is. The brain can detect it if happens all the time.
It is interesting that amp amp showing a distortion in its transfer curve, say like cross-over distortion does by definition have the same THD regardless of where in the transfer it occurs - at zero volts, at +1 volt, or whatever. But to the ear the amp having the distortion at zero volts will sound worse (assuming this is the loudspeaker voltage, much worse actually), because that is where the signal traverses most often.
That is why crossover distortion in class B push pull sounds much worse that a Class A having the same measured THD.
It is why most people are perfectly happy listening to music even when their amp is occaisonally overloading on peaks.
In the 1950's the BBC used to write into the contract specifcation when purchasing new broadcast transmitters a specification that read in part like this (I've simplified it somewhat, and I can't remember the exact percentages):-
x: The tramsittter shall have a total harmonic distortion at 90% modulation of less than 2%
Y: The transmitter shall have a THD at 80% modulation, and at all lower levels of modulation, of less than 1%
z: The sound as assessed by BBC engineers by ear shall be aceptable.
They included Clause z because experince had taught them that THD measurement does not tell the full story!
This is hardly a satisfactory spec on legal grounds in a contract as any good lawyer will tell you. In the early 1960's, E R Wigan of the BBC Reserach Department investigated why THD measurements don't tell the full story, and other researchers built on Wigan's work. The outcome was the realisation that percieved distortion depends on the music probability density function as I have explained.
In AM radio trasmitters, it is conventional to modulate the final RF amplifier with a high power Class B push pull audio amplifier. Thus crossover distortion is possible, but at the high powers involved, an adequate amount of standing bias will easily bring the THD within spec, without neg feedback, which was not that common in old broadcast transmitter modulators. Not necessarily the preceived distortion however.
In 1970, the Japanese started selling high power amplifiers having Class-H topology. Class H exhibits crossover distortion when the output crosses the intermediate rail voltages, the intermediate rails being set at some fraction of the main supply rail voltage. Instead of cross-over at just zero volts, you get cross-over at various voltages. Experience shows, and I demonstrated this to an audience with a speciallly built amp when I was at university, these sets of extra crossover points are damm near inaudible, even when they were clearly visible on an oscilloscope!
I hope this is all clear - some people nmot accustomed to statistics find probability density functions hard to visualise.
Last edited:
Thanks Keit
that's exactly the kind of insight I was looking for. Nevertheless I fail to consider the huge amount of work that will take to get this job done. Nor the necessity to establish a blameless criteria for each one of the most common topologies. A mindblowing amount of work.
But I think one thing must be common to any of them, a "blameless" amp can't be speaker picky, it must be able to drive almost everything (reasonably well engineered), properly, or you can not consider it blameless. No matter how good it sound with "certain" types of systems and music.
that's exactly the kind of insight I was looking for. Nevertheless I fail to consider the huge amount of work that will take to get this job done. Nor the necessity to establish a blameless criteria for each one of the most common topologies. A mindblowing amount of work.
But I think one thing must be common to any of them, a "blameless" amp can't be speaker picky, it must be able to drive almost everything (reasonably well engineered), properly, or you can not consider it blameless. No matter how good it sound with "certain" types of systems and music.
Mosquito,
Thank you for your compliment!
And you comments are spot on. Speaker loads present some engineering challenges, but that is not an excuse for not doing the work and solving them.
You may be aware that Doug Self addressed the "drive any decent speaker" issue in subsequent work. He called it "load invariant" design. As I recall the full work wasn't published in Electronics World like his blameless series was.
Thank you for your compliment!
And you comments are spot on. Speaker loads present some engineering challenges, but that is not an excuse for not doing the work and solving them.
You may be aware that Doug Self addressed the "drive any decent speaker" issue in subsequent work. He called it "load invariant" design. As I recall the full work wasn't published in Electronics World like his blameless series was.
I'd argue that Williamson, Hafler and Keroes did for tube amplifiers what Doug Self did for solid state. Therefore, the tube equivalent of a Blameless would obviously be a Williamson amp with a UL output stage.
To a large extent Williamson did. His design addressed almost all sources of distortion input to output that others before him had not. He didn't publish to anywhere near level of analysis and detail that D Self did though. In essense Self provided a wealth of engineering detail that other designers can use. Williamson just said "here is a solution."
Halfler & Keroes, while they did the World an immense favour by publishing a definitive work on ultra linear, do not really count as a Doug Self equivalent as they only address the output stage in their detailed paper.
The tube equivalent of a Self Blameless design is not necessarily a Williamson amp with or without UL. Outstanding performance can be achieved with other topologies. Examples that can perform at least as good include a) the Quad-style distributed load + differential input tube topology also used by HMV in up-market models and by Trimax (a one-time manufacturer of broadcast station monitoring amps); and b) variations of the Murray single ended tube topology, which achieves 0.02% distortion without global feedback; c) Sandman-style feedforward topology, particulary suitable for tube amps, and d) tube equivalents of the Matshushita/National Panasonic (now known as Technics) leader-and-slave topology.
Last edited:
OK, but would you agree that a tube Blameless is the exact opposite of a single ended triode with no global feedback?
Scopeboy,
You may have missed Doug Self's and my point. That's ok - Doug is a much misundestood guy.
There are two concepts in Doug's work:-
1) Any topology can be carefully analysed so that the design engineer understands all the separate causes of distortion, and when he does that, he is then in a position to extract the best possible performance that that particular topology is capable of.
2) Doug chose one particular topology (you can reasonably infer he thinks it is a good one) and he did the detail analysis for that particular topology, and he published the resulting design as a worked example.
He used the term "blameless" to describe a circuit that was optimal - that is it contains no design errors that would make it perform not as well as its potential. Concept 1.
His published design is (very reasonably) free of design errors and performs as well as that topology can, and better than many well regarded alternatives. Concept 2.
Thus a single ended triode with no global feedback if properly engineered will perform as well as a triode with no feedback can. There are no engineering/design errors. So we can call it a Blameless design. Not that there's much to get wrong with a triode perhaps....
But that SE triode will not perform as good as say a mere tetrode amp with global feedback. So in that sense it is certainly not the tube equivalent of the Self solid state design.
See the difference?
Doug is damm good design engineer, but his choice of terminology was unfortunate. But then terms such as "error free" and "competent" don't have a nice ring to them.
You may have missed Doug Self's and my point. That's ok - Doug is a much misundestood guy.
There are two concepts in Doug's work:-
1) Any topology can be carefully analysed so that the design engineer understands all the separate causes of distortion, and when he does that, he is then in a position to extract the best possible performance that that particular topology is capable of.
2) Doug chose one particular topology (you can reasonably infer he thinks it is a good one) and he did the detail analysis for that particular topology, and he published the resulting design as a worked example.
He used the term "blameless" to describe a circuit that was optimal - that is it contains no design errors that would make it perform not as well as its potential. Concept 1.
His published design is (very reasonably) free of design errors and performs as well as that topology can, and better than many well regarded alternatives. Concept 2.
Thus a single ended triode with no global feedback if properly engineered will perform as well as a triode with no feedback can. There are no engineering/design errors. So we can call it a Blameless design. Not that there's much to get wrong with a triode perhaps....
But that SE triode will not perform as good as say a mere tetrode amp with global feedback. So in that sense it is certainly not the tube equivalent of the Self solid state design.
See the difference?
Doug is damm good design engineer, but his choice of terminology was unfortunate. But then terms such as "error free" and "competent" don't have a nice ring to them.
Last edited:
The fundamental points underlying Doug Self's work are, as I understand it:
All distortion is bad, the ultimate goal of audio amplifier design is to reduce distortion to unmeasurable levels under the assumption that unmeasurable means inaudible.
Clipping behaviour is not important, all amplifiers are assumed to be operated with power reserves.
It is very hard to see how these concepts map onto tube amplifier design, where distortion can never vanish and soft clipping of underpowered output stages is everywhere. But the meaning of "Blameless" in a tube context must depend on this mapping of concepts. By my questions I'm trying to clarify what mappings the contributors to this thread are using.
All distortion is bad, the ultimate goal of audio amplifier design is to reduce distortion to unmeasurable levels under the assumption that unmeasurable means inaudible.
Clipping behaviour is not important, all amplifiers are assumed to be operated with power reserves.
It is very hard to see how these concepts map onto tube amplifier design, where distortion can never vanish and soft clipping of underpowered output stages is everywhere. But the meaning of "Blameless" in a tube context must depend on this mapping of concepts. By my questions I'm trying to clarify what mappings the contributors to this thread are using.
Last edited:
You're right, he didn't dicuss clipping. And even solid state amps if incompetently designed can clip in a way that sounds dreadful. I learnt that the hard way with an early design of mine!
However, top class amplifiers, be they Williamson, GEC 88-50 (which is the UL version of the Williamson), RCA Orthophonic, etc all have large amounts of negative feeback, and that makes them clip hard just like a solid state amp. The Murray does as well distortion-wise without global feedback, but it clips hard as well.
If you want a quality amp to clip softly, be it tube or ss, and there are grounds for doing that if you are restricted in power output, you basically have two choices:-
1) don't use negative feedback
-or-
2) Use a special low level circuit that clips in a controlled soft manner, so that the signal stays at all times within the amp's power rating.
In practice, a low level soft clip circuit will cause harmonic distortion much like having no neg feedback. I'm not aware than anyone has solved that one.
The time honoured way used by pros is to have an amplifier sufficiently powerfull that it hardly ever clips. Then you can use all the feedabck you like, or not, without worrying about how it clips. The advent of CD's with their wider dynamic range and extended bass response changed the power needed to do this up by a factor of about 10. And today's loudspeakers are smoother and distort less, but they are much less efficient that speakers made in the 1950's and 60's. So now you need a 100 Watt amp just to be as good as a 4 watt amp was before.
I have a copy of the Buddy Holly Story (London Cast) CD, and certain tracks sound bad on anything less than 100 watt. Inserting a filter cutting of below 50 Hz improves things at the cost of some realism. It well may also sound better if I tried an amp with soft clipping. But all mine clip hard.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.