@Dumster, the major point is that feedback cannot, even in theory, reduce the error to zero. In theory, current dumping can.
Practise is always different from theory of course, but that's the gist of it.
The nice fallout is, that once you have a principle that can reduce error to zero, you don't have to worry anymore about getting the correct bias setting to minimize the error. In fact, Peter decided that he wouldn't need any bias at all, which, again in principle, is perfectly valid.
But theory is not the same as practise, and due to circuit/parts limitations, zeroing out the error is not 100%. Therefor he intruduced minimal bias (1 diode drop) in later models to shave the edges of the remaining erors.
Jan
Practise is always different from theory of course, but that's the gist of it.
The nice fallout is, that once you have a principle that can reduce error to zero, you don't have to worry anymore about getting the correct bias setting to minimize the error. In fact, Peter decided that he wouldn't need any bias at all, which, again in principle, is perfectly valid.
But theory is not the same as practise, and due to circuit/parts limitations, zeroing out the error is not 100%. Therefor he intruduced minimal bias (1 diode drop) in later models to shave the edges of the remaining erors.
Jan
Sorry, I could have been clearer - "The 'Class A'' amplifier has it's own internal feedback for stability and as was attested, will function in Class A operation up to 1 watt. So, it is a Stand alone amplifier in it's own right. Add to this the fact that, as Quad state, ''It controls the Current Dumpers at all times'' - why then would Quad introduce feedback from the ''Non - linear'' region, other than to generate an error signal which it applies to that region to reduce/eliminate that non linearity?
The Standing conditions of the Class A section exists, prior to the invocation of the Dumpers and from Quad's statement is Very High Quality Class A amplifier. In this I assume it is unconditionally stable at all levels.
My observation about feedback, relates to the error signal from the dumpers. That I presume is derived from the voltage drive of the class A section to the dumpers and the post dumper voltage - dumper voltage gain unity? So, for my money, this constitutes the error and is fed back to the class A amplifier which sees it as no more than an addition to the audio input.
''In fact, Peter decided that he wouldn't need any bias at all, which, again in principle, is perfectly valid.'' - did he introduce it (a diode) to reduce the non - linearity that exists in the dumpers before they operate over their linear section? It would reduce some of the non linearity and the time it would take for the dumpers to transit to their linear section - would it not. Again, I am focusing on the speed of operation - rightly or wrongly.
Ps, Thanks for the response - never too old to learn something new.
Last edited:
He introduced a 1 diode drop bias at the dumper bases, indeed to "reduce some of the non linearity and the time it would take for the dumpers to transit to their linear section".
There's alot of documentation on the 'net that explaims it clearly - no need to spend all that time speculating and possibly still get it wrong ;-)
I also don't understand why you are so hung up on the class A amp that is part of this concept. If you are after a class A amp, get that and forget about all that current dumping stuff, it has nothing to do with class A.
Jan
There's alot of documentation on the 'net that explaims it clearly - no need to spend all that time speculating and possibly still get it wrong ;-)
I also don't understand why you are so hung up on the class A amp that is part of this concept. If you are after a class A amp, get that and forget about all that current dumping stuff, it has nothing to do with class A.
Jan
At seventy seven, I am still not ready to study my Navel. Yes, a solution, albeit an expensive one, would be to buy a Sugden. However, this all came about - of late, as we shall be moving to a smaller place where headphones will probably replace my Spendors. So the Class A section of the 405 and subsequent 909 seemed a good means of avoiding the expense of a Sugden - although I have also bought a 'Headmaster'. As for getting it wrong - well no one learns anything by being perpetually right!
The output stages of the 405/909 are emitter followers are they not, and as such have a low output impedance? Plus, it doesn't matter, whether the 405 is running at full output or is working at a sufficiently low level to remain in class A, the conditions it sees remains the same, only the current dumpers add current to its output. True, there is an error signal fed back to it, when the dumpers cut in. However, it is fed in, supplementary to the audio, so how can that alter its voltage gain or be seen as a substantial amount of NFB?
If that signal is derived from the error between the Class A section (Reference) and the Current Dumper output (Variable), then the error will be applied to reduce/minimize the difference. However, the class A sections standing conditions won't change, I would have thought? Plus Negative feed back is not an error, whereas an error can be fed back negatively.
Thanks for your input anyway.
Brian.
The output stages of the 405/909 are emitter followers are they not, and as such have a low output impedance? Plus, it doesn't matter, whether the 405 is running at full output or is working at a sufficiently low level to remain in class A, the conditions it sees remains the same, only the current dumpers add current to its output. True, there is an error signal fed back to it, when the dumpers cut in. However, it is fed in, supplementary to the audio, so how can that alter its voltage gain or be seen as a substantial amount of NFB?
If that signal is derived from the error between the Class A section (Reference) and the Current Dumper output (Variable), then the error will be applied to reduce/minimize the difference. However, the class A sections standing conditions won't change, I would have thought? Plus Negative feed back is not an error, whereas an error can be fed back negatively.
Thanks for your input anyway.
Brian.
Perhaps I ought to have said ''The speed at which the Current Dumping is applied, is a reflection of how short the amplifier's transit time is''.Morning,
In my subjective experience, "speed" is a low frequency phenomenon, probably related to phase in the the low Hz.
Foot note to the above. I got to a point where I realised that to constructively proceed with the analysis, I need to fully understand the complex imaginary maths. It is difficult trying to dust off that part of my brain, and resurrect the rusty half understood from 30 years ago, so got stalled.... easily distracted by designing / building (high quality) class A valve amps.
In more positive mathematical news, I have remembered / worked out how to do long division, fourth time lucky doing primary school home work... Small person 5 is pretty bright and will probably not benefit form my newly rediscovered talent, sigh.
aka dave
Again, thanks for your contribution.
Brian, those class A stages in the dumper amps are specifically designed for the drive requirements of the dumper stage, including compensation for strongly varying feedback factors over a signal cycle.
If you are after a good class A amp for headphones, it's about the worst choice you can make.
A good class A amp for cans doesn't cost you an arm and a leg and will provide incredible sound reproduction quality. Just my € 0.02.
Jan
If you are after a good class A amp for headphones, it's about the worst choice you can make.
A good class A amp for cans doesn't cost you an arm and a leg and will provide incredible sound reproduction quality. Just my € 0.02.
Jan
The Quad and the Sugden are poles apart in their design goals... I know someone who used to have a Quad 405 set up and who went over to Sugden. I've had the pleasure of visiting the factory in Heckmondwike some years ago... quite an experience.
Quad 405 vs Sugden A21SE at 1 watt into 8 ohm.
Thanks for that Jan. However, rightly or wrongly, ''Tenacity'' is my middle name. It may not be a mathematical analysis, but there has to be a more meaningful way of explaining the ''Current Dumping Principal'' Michael Faraday is said to be the Father of the Electricity - in England at least. However, it was James Watt that provided the scientific understanding. Not that I am drawing any parallels.
The bulk of the projects I worked on were prototype systems, which should never have left the factory. But they did, and in the real world all the theory in the world would not make them work. Individually, components appeared to do what they were designed for, but when systems were configured, they just did not perform. So a need to understand the mechanism of systems in the broadest sense, helped to pin point the areas which prevented systems from working correctly.
After decades in some of the most inhospitable regions, in many countries, where little or no factory support was on hand, a means of moving things forward, in a manner which would convince management of the need for changes had to be found. When projects left the Factory, budgets were closed and engineering support/development moved on to the next project. It was a devil of a job to get things done as there was a firm belief that issues no longer related to ''Under-development'', or Production, but Maintenance!
So, there is a need for me to understand the principal of ''Current Dumping'' - better.
Now, rightly or wrongly, this is how I believe it functions.
''Negative feedback will reduce output impedance, increase the bandwidth and reduce gain. However, if feedback reduces gain, it will reduce the amplitude of the ''Error'' signal, which will increase the time it takes for the dumpers to switch on.'' - the opposite of what needs to be achieved!
The Quad 405 is a dual supply amplifier where the output swings +/-, obviously, so it is essential the error signal is applied to both positive going transients. From Quad's circuit diagram, it can be seen that the error is fed, from the ''Bridge'', to the Class A amplifier's input. I am persuaded by the text, the class A amplifier is high quality, is stable under all load conditions and has a fixed gain. Having got rid of all my test equipment, there is little I can do but make an assumption as to what its voltage gain is, but for the sake of argument I shall say 30db.
I am also convinced that it requires no negative feedback from the output stages as it's working conditions need to be stable, to ensure ''Accurate Control'' of the output stages. I believe what feedback there may be, is is simply supplementary to the input signal and affects only the dumpers and not the Class A amplifier.
''He introduced a 1 diode drop bias at the dumper bases, indeed to "reduce some of the non linearity and the time it would take for the dumpers to transit to their linear section".
Quad state, once the signal reaches the ''Target'' figure, the dumpers switch on. The diode will reduce some of the non linearity , but still it will not be fast enough to avoid distortion at the point where the dumpers start to conduct, either on the +/- excursions. And so a small transients will appear at the bridge, when the + going dumper comes into play and then goes off again, and likewise when the - going dumper comes into play and goes off again. These small transients will be shaped by the ''Bridge'' and applied to the input of the Class A amplifier. They are also applied at a time when the input audio is at a level which triggers the error - obviously 'Post', but how 'Post' will dictate how swiftly the transients can induce the dumpers to switch on, thus minimizing ''Phase Errors''.
These Errors, which I conclude are small, yet very short sharp transients, being applied to the input of the Class A amplifier, will be subject to the same gain as the input signal. Thus, being amplified by the hypothetical 30db, will switch on the dumpers; and off, in an incredibly short time. The net effect is to ''Superimpose the current, supplied by the dumpers, upon the Class A signal, at virtually the same amplitude, and in phase.''
Regardless of whether the Current Dumpers are operating, the Class A amplifier signal will always appear at the output.
Brian.
Thanks for the plots Mooly.
That is a graphic expression of the differences. However, unless there is a rather large lottery win coming in my direction, I shall be living with my 909 for some time to come. Having has an early Sugden, I know how good they are. But, second hand prices are ridiculous despite, in some cases, despite their age and the obvious thermal decay. From what I have read, folks have been making ''Improvements'' to the 405 for quite some time now, probably focusing on the ''Feed - forward'' signal, via the Bridge. However, I doubt that Quad have been resting on their laurels. So it would be interesting to compare the 909 under the same conditions.
I have to say, the 909 driving my Stax SR - 3's is brilliant and they certainly can take a lot more than the BC1's. However, with the Spendors, there is an incredible sweet spot at modest levels which I cannot believe was there when driving them with the 303. The delicacy is quite extraordinary and the musical Attack superb. However, a lot is lost at lower levels, to a degree that I miss the tone controls and filter of the 33. So much so, I have been experimenting with some modest addition of Loudness Control, just to lift the top and bottom a little when listening at levels just above a whisper.
In all honesty, I cannot foresee me parting with the 909, or for that matter my 405, or 303. Noise figures and distortion aside, it really is the speed at which the 909 responds to the signal that stands out for me - the beautiful brittleness it brings out from the likes of piano and saxophone that are just spell binding. Listening to anyone that can actually sing is pretty riveting as well. I got it for a good price - as I did my Suggie Headmaster, but there are demands for expenditure elsewhere. Such Items as a pair of Quad ELS's - now we are back to pipe dreams!
That is a graphic expression of the differences. However, unless there is a rather large lottery win coming in my direction, I shall be living with my 909 for some time to come. Having has an early Sugden, I know how good they are. But, second hand prices are ridiculous despite, in some cases, despite their age and the obvious thermal decay. From what I have read, folks have been making ''Improvements'' to the 405 for quite some time now, probably focusing on the ''Feed - forward'' signal, via the Bridge. However, I doubt that Quad have been resting on their laurels. So it would be interesting to compare the 909 under the same conditions.
I have to say, the 909 driving my Stax SR - 3's is brilliant and they certainly can take a lot more than the BC1's. However, with the Spendors, there is an incredible sweet spot at modest levels which I cannot believe was there when driving them with the 303. The delicacy is quite extraordinary and the musical Attack superb. However, a lot is lost at lower levels, to a degree that I miss the tone controls and filter of the 33. So much so, I have been experimenting with some modest addition of Loudness Control, just to lift the top and bottom a little when listening at levels just above a whisper.
In all honesty, I cannot foresee me parting with the 909, or for that matter my 405, or 303. Noise figures and distortion aside, it really is the speed at which the 909 responds to the signal that stands out for me - the beautiful brittleness it brings out from the likes of piano and saxophone that are just spell binding. Listening to anyone that can actually sing is pretty riveting as well. I got it for a good price - as I did my Suggie Headmaster, but there are demands for expenditure elsewhere. Such Items as a pair of Quad ELS's - now we are back to pipe dreams!
The 405 is not an easy circuit to analyse. 1 watt into 8 ohms is achievable without the dumpers in circuit but the circuit still displays all the distortion artefacts at those levels.
No dumpers fitted:
With the dumpers fitted they actually take over at this low level as can be seen by the emitter currents of the dumpers:
Output levels need to be lower than around 200mv peak to peak before the dumpers cut off. At this point the FFT is more Class A like but no match for the Sugden.
Voltage gain from base of first stage to output node:
Giving a gain of 3.7 or 11.4db (by looking at the rms levels at those points in the sim) which agrees very closely with the expected gain set by the feedback resistors:
500/180 + 1 gives a gain of 3.77
No dumpers fitted:
With the dumpers fitted they actually take over at this low level as can be seen by the emitter currents of the dumpers:
Output levels need to be lower than around 200mv peak to peak before the dumpers cut off. At this point the FFT is more Class A like but no match for the Sugden.
Voltage gain from base of first stage to output node:
Giving a gain of 3.7 or 11.4db (by looking at the rms levels at those points in the sim) which agrees very closely with the expected gain set by the feedback resistors:
500/180 + 1 gives a gain of 3.77
The dumpers are slow slow slow in comparison to the other circuitry.
If you understand the operation of the bridge, it is clear that any 'slowness' of the dumpers is not very important.
With the bridge in balance, any voltage difference between the bases (input) of the dumpers and the output voltage (the 'error voltage') will not appear at the output. So the speed at which the dumpers come on or go off is immaterial, as long as the class A stage can keep the bridge balanced.
Jan
With the bridge in balance, any voltage difference between the bases (input) of the dumpers and the output voltage (the 'error voltage') will not appear at the output. So the speed at which the dumpers come on or go off is immaterial, as long as the class A stage can keep the bridge balanced.
Jan
That's right
When I mentioned them being 'slow slow slow' it was in response to the comment about them having to switch on and off very quickly... they don't.Quite a short but good and readable analysis of the 405 was given by John Linsley Hood in his book "Valve and Transistor Audio Amplifiers".
I must be a minority but I believe the whole "bridge" concept in CD amplifiers is rather confusing and hides, rather than explains, how a CD amplifier functions. IMO a CD amp is a normal feedback amp where the load is driven by both the very linear amplifier A and the not-so-linear current dumper Tr1 Tr2:
A little analysis (qualitative or with some algebra) shows that the perfect cancellation of the current dumper distortion happens when the ratio of impedances of R2 and L1 mimics the loop gain of the integrator formed by A-R1-C1. While not contradicting Walker's original results, this opens way for a better real-life CD amplifiers, where e.g. L1 has some DC resistance, and the integrator has a finite DC gain and/or non-6db/octave-single-slope transfer function.
A little analysis (qualitative or with some algebra) shows that the perfect cancellation of the current dumper distortion happens when the ratio of impedances of R2 and L1 mimics the loop gain of the integrator formed by A-R1-C1. While not contradicting Walker's original results, this opens way for a better real-life CD amplifiers, where e.g. L1 has some DC resistance, and the integrator has a finite DC gain and/or non-6db/octave-single-slope transfer function.
Last edited:
Without a shadow of a doubt. Walker did make the DC loop gain of the Quad high (to obtain a "good virtual ground") and mentions a resistor in parallel with the bridge capacitor, which would compensate for other parts being non-ideal.
Walker's analysis only works, at least precisely, for zero-DC-impedance inductors and infinite-DC-gain, single-pole-compensated amplifiers with the time constant set by the "bridge" capacitance. The insight about the ratio of impedances mimicking the loop gain allows for e.g two-pole-compensated amplifiers, which would not be so easy to incorporate into the "bridge" concept. I will show a (simulated) example, but it will take some time for me to post it.
Still, instead of perfecting the distortion compensation, it may be better to just increase the loop gain, which would take a better care of the current dumper's distortion.
Last edited: