If I am interpreting what has just been explained, the Quad Class A Amplifier is a very high quality amplifier to which the bridge applies a signal to increase its response time {slew rate}, proportional to the input signal, in the region where hitherto crossover distortion would exist in a Class C amplifier.
To my simple mind, to increase the slew rate from say 5V/uS to 25V/uS, without changing components means the bridge essentially increases the gain of the Class A amp, thus reaching the dumper turn - on voltage much faster. Once dumper conduction has been established and theoretically no error exists between the phase, or amplitude of the class A section and the now conducting dumpers, there is no need for an enhanced slew rate and the slew rate returns to its former figure.
Theoretically; I think, the variable slew should not influence negatively the low level of distortion the amplifier is capable of achieving. However, it is the rate of change of the slew rate which will dictate how quickly both the Class A signal and the Class C signal overlap to become contiguous.
It is in this area improvements could be made despite Walkers assertion that no post - production alignment is needed. That is pure supposition on my part, based on anecdotal evidence that some enthusiasts/engineers have made modifications which improved on the amplifiers already low distortion performance.
If my logic is in anyway unsound, I would appreciate it being corrected.
To my simple mind, to increase the slew rate from say 5V/uS to 25V/uS, without changing components means the bridge essentially increases the gain of the Class A amp, thus reaching the dumper turn - on voltage much faster. Once dumper conduction has been established and theoretically no error exists between the phase, or amplitude of the class A section and the now conducting dumpers, there is no need for an enhanced slew rate and the slew rate returns to its former figure.
Theoretically; I think, the variable slew should not influence negatively the low level of distortion the amplifier is capable of achieving. However, it is the rate of change of the slew rate which will dictate how quickly both the Class A signal and the Class C signal overlap to become contiguous.
It is in this area improvements could be made despite Walkers assertion that no post - production alignment is needed. That is pure supposition on my part, based on anecdotal evidence that some enthusiasts/engineers have made modifications which improved on the amplifiers already low distortion performance.
If my logic is in anyway unsound, I would appreciate it being corrected.
If the simulation shows that it’s wrong. What it should show is what is actually observed at the bases of the dumpers, which is a normal signal in the class C’off’ angle, and then a very flattened curve as the dumpsters switch on and take over.
@delphiplasma, @Dumpster -The Quad's distortion is due to feedback. In open-loop, the Quad's output stage has high distortion compared to an optimally-biased class AB output stage.
@ejp - Unfortunately, I can't find the post.
The "flat" portion with the dumpers conducting is the normal rate of the audio signal.
The "accelerated" portion with the dumpers off is the normal rate multiplied by (Rload + Rfeedforward)/Rload.
Ed
@ejp - Unfortunately, I can't find the post.
The "flat" portion with the dumpers conducting is the normal rate of the audio signal.
The "accelerated" portion with the dumpers off is the normal rate multiplied by (Rload + Rfeedforward)/Rload.
Ed
No matter! we are currently staying with our son whilst the purchase of our apartment goes through. So, not only do I not have either the 405 or 909 before me, even if we were now in our new home, I have virtually no test equipment to look more closely at either.
So, I have to stick to pure hypothesis/logic- or lack of.
The observation made about high distortion of the Class A amp in open loop is pertinent; I believe, as is the observation regarding the dumpers conducting at the normal rate. My logic leads me to conclude, the control exercised by the class A amp upon the class C dumpers does not extend to interfering with their response time, only in the way it speeds up their switch - on by increasing the amplitude of the class A's output signal during the dead - zone where they are off.
Perhaps the the distortion of the class A section does become relevant as, to my way of thinking, the only way this could happen, without physically changing the gain of the class A with components, is to "Inhibit" the inherent feedback loop of the class A, momentarily by way of the error signal.
By forcing the class A into "Open - Loop", while the dumpers are in the dead - zone, the increased amplification that results from it would increase the amplitude of the signal dramatically and cause the dumpers to switch on very quickly indeed.
Once the dumpers are on, theoretically no error signal exists, so the fixed feedback loop re - establishes the gain of the class A section to its former value.
Depending upon how far the error signal is permitted to force the class A into open - loop, will determine the amount of distortion it introduces into the output signal. Refining the way the error signal is derived, would result in shortening the class A's open - loop period and thus reduce it's potential for increasing distortion.
So, does the Bridge act as a "Integrator", given the error signal is derived from an extremely complex audio signal?
For the life of me, I can see no other way Walkers design can work. If as suggested earlier "the feedback is just feedback", then it would not distinguish the design from any other class AB amplifier. Plus, as the dumpers are off, unlike the output stages of a class AB amp, which are on the cusp of conduction, the dumpers are being operated in class C.
The term bridge may well be a mis - interpretation of what Walker had intended it to mean - meaning it "Bridges the area between Class A and Class C".
So, I have to stick to pure hypothesis/logic- or lack of.
The observation made about high distortion of the Class A amp in open loop is pertinent; I believe, as is the observation regarding the dumpers conducting at the normal rate. My logic leads me to conclude, the control exercised by the class A amp upon the class C dumpers does not extend to interfering with their response time, only in the way it speeds up their switch - on by increasing the amplitude of the class A's output signal during the dead - zone where they are off.
Perhaps the the distortion of the class A section does become relevant as, to my way of thinking, the only way this could happen, without physically changing the gain of the class A with components, is to "Inhibit" the inherent feedback loop of the class A, momentarily by way of the error signal.
By forcing the class A into "Open - Loop", while the dumpers are in the dead - zone, the increased amplification that results from it would increase the amplitude of the signal dramatically and cause the dumpers to switch on very quickly indeed.
Once the dumpers are on, theoretically no error signal exists, so the fixed feedback loop re - establishes the gain of the class A section to its former value.
Depending upon how far the error signal is permitted to force the class A into open - loop, will determine the amount of distortion it introduces into the output signal. Refining the way the error signal is derived, would result in shortening the class A's open - loop period and thus reduce it's potential for increasing distortion.
So, does the Bridge act as a "Integrator", given the error signal is derived from an extremely complex audio signal?
For the life of me, I can see no other way Walkers design can work. If as suggested earlier "the feedback is just feedback", then it would not distinguish the design from any other class AB amplifier. Plus, as the dumpers are off, unlike the output stages of a class AB amp, which are on the cusp of conduction, the dumpers are being operated in class C.
The term bridge may well be a mis - interpretation of what Walker had intended it to mean - meaning it "Bridges the area between Class A and Class C".
Last edited:
Does this help (or not 🙂)
This shows a simple opamp and Class C biased output pair. The opamp takes the place of the high quality Class A amp. As you can imagine, it has a very low ability to drive the 100 ohm load. It needs the dumpers to to achieve decent output levels.
The gain is set by Rf/Rg although Rf is looking at the output node also via R4. So Rf is actually appears as 9000.25 ohm when the transistors are 'off'.
So far so good. If we look at output voltages we don't see what we might expect to see... so lets switch to looking at currents instead a little later. This is the amp fed with a low input voltage. We are going to be looking at the load current and the currents in R4 and R5 as these two current combine at the output node as they flow into the load but first we will look at voltages, just to show it all works after a fashion.
50mv in and 0.5 volts out. Gain is (9000.25/1000)+1 which is essentially 10. Vout and V_Dumper are superimposed on each other. They are the same. The only current in R4 is the feedback current. The transistors are off.
I know its all an imperfect analogy in many ways but this will show in essence what is going on.
Let us now switch to current and see what that looks like.
The load current is clean and the contribution is 100% opamp (via R5) and nothing via R4. Again the load current is the sum of the dumper current (zero) and the opamp output via R5. The opamp is supplying all the load current at these low outputs.
Now we increase the input voltage by a factor of 10 and just as a quick check we switch back to looking at voltage again. Notice we now have 5 volts output.
As before lets look at currents.
The load current in green is undistorted but its how that current is made up that is interesting. We now have both the dumpers and the opamp now contributing. The dumpers are contributing the typical Class C biased distorted output signal and the opamp is adding a current that perfectly adds to that distorted signal and sums to the clean load current.
This final shot shows an input voltage of just enough that the opamp can't quite manage and the dumpers contribute a little. You can see the dumpers adding just that little bit of load current that the opamp can't manage. Again it all sums to a clean load current.
This shows a simple opamp and Class C biased output pair. The opamp takes the place of the high quality Class A amp. As you can imagine, it has a very low ability to drive the 100 ohm load. It needs the dumpers to to achieve decent output levels.
The gain is set by Rf/Rg although Rf is looking at the output node also via R4. So Rf is actually appears as 9000.25 ohm when the transistors are 'off'.
So far so good. If we look at output voltages we don't see what we might expect to see... so lets switch to looking at currents instead a little later. This is the amp fed with a low input voltage. We are going to be looking at the load current and the currents in R4 and R5 as these two current combine at the output node as they flow into the load but first we will look at voltages, just to show it all works after a fashion.
50mv in and 0.5 volts out. Gain is (9000.25/1000)+1 which is essentially 10. Vout and V_Dumper are superimposed on each other. They are the same. The only current in R4 is the feedback current. The transistors are off.
I know its all an imperfect analogy in many ways but this will show in essence what is going on.
Let us now switch to current and see what that looks like.
The load current is clean and the contribution is 100% opamp (via R5) and nothing via R4. Again the load current is the sum of the dumper current (zero) and the opamp output via R5. The opamp is supplying all the load current at these low outputs.
Now we increase the input voltage by a factor of 10 and just as a quick check we switch back to looking at voltage again. Notice we now have 5 volts output.
As before lets look at currents.
The load current in green is undistorted but its how that current is made up that is interesting. We now have both the dumpers and the opamp now contributing. The dumpers are contributing the typical Class C biased distorted output signal and the opamp is adding a current that perfectly adds to that distorted signal and sums to the clean load current.
This final shot shows an input voltage of just enough that the opamp can't quite manage and the dumpers contribute a little. You can see the dumpers adding just that little bit of load current that the opamp can't manage. Again it all sums to a clean load current.
@Mooly - Thanks for the waveforms. We are interested in the voltage at the output of the op-amp with Rload=4 or 8 ohms. Its shape will look somewhat similar to I(R5).
Ed
Ed
Here you go. First up is at the 50mv input level. Notice the opamp output is much larger than the wanted output voltage because of the effect of R5 being in series with the load.
And now with 0.5 volt applied.
The opamp models are more like behavioral models and the distortion of the opamp alone is spectacularly good as a simulation, its nothing like a real discrete component design. This is at the 500mv output level into the 100 ohm load. The transistors are snipped out.
I've attached the sim with some default models if anyone wants to play.
This is with 300 ohm for R5 which shows the opamp slew quickly across the dead zone.
An opamp swap makes a big difference surprisingly. LT1001 at the left and LT1056 at the right. The LT1056 is a bit like a TL071, quite fast and FET input.
An opamp swap makes a big difference surprisingly. LT1001 at the left and LT1056 at the right. The LT1056 is a bit like a TL071, quite fast and FET input.
This is all great stuff. I don't suppose there is any chance of calculating/guesstimating from the plot the change in Slew Rate. This has really nailed it for me. And all this to overcome the non-linearity of the dumpers at turn -on. Impossible to completely overcome, but the improvement in slew is really marked.
I had previously been viewing the transition in the dead zone as distortion, when in truth it is simply a vastly improved slew - rate which dramatically increases the speed at which the dumpers start to conduct - all but eradicating the non - linearity of the dumpers junctions.
I have owned a 303 power amp since 1971 and Quad have regularly maintained it for me. In the Nineties, I picked - up a 405 for a good price as it had never been serviced - I took it to Quad within a week of purchase where it had a full service. On its return, I played some music, firstly through the 303 and then the 405. The 303 went back in its packing case. Even with the Spendors, the improvement in the bass region was a revelation.
For Christmas 2023, I treated myself to a new/old stock 909 with the intention of exploring the possibility of using it at low level, thinking I could keep it operating in the Class A zone, whilst driving my Stax headphones.
What a delightful "Can of Worms" that theory opened up.
A/B comparisons between the 405 and the 909 followed and it became obvious to me - I have commented frequently else where about the "Attack Time" of the 909, it is akin to a rifle shot - as I described it. Phenomenally quick with complex music. I also commented it was unforgiving with anything less than high quality recordings or singing. Some artists were made to sound dreadful.
Having only just fitted a pair of brand new, but stored bass/mid-range drivers to my BC1's - stored since 1996, the difference was frankly alarming and I fretted over the punishment they could suffer at the hands of the 909. It is so "lightning-quick", hence my thought of building a pair of low frequency enclosures to take some of the bass load of the Spendors.
I set about comparing again, the 303, 405 and 909.
The 303 remains sweet sounding, quite mellow and although there isn't any drama at bass frequencies, what there is, is natural. Still a very good amplifier as long as one doesn't press it to deliver too much. The Spendors help being relatively low power.
The 405 however, even with the Sponsors extend the bass beautifully, which surprised me as I had not expected them to perform so well in the lower registers. I didn't get much further with that as it had to be quickly packed in anticipation of our move.
However, I have spent over a year living with the 909 as my mainstay, with a Sugden Headmaster Headphone Amplifier and a pair of hifiman headphones. I can't take much of them at higher levels - a certain brittleness with some music hurts my ears.
Once I get an idea in my head I cannot let it go until it has been thoroughly thought through and proven or disproved. So I have spent countless hours listening to music with a passive preamplifier which comprises of a variable loudness control, which I modified to give only bass lift, whilst everything above 800hz - ish remains flat. I listen in the dining room, where I can keep the volume level low and the linear progressive lift to bass frequencies compensates for my failing hearing in that region. This has long been my target level, hence my obsession with the 909 and in particular that Class A section.
I may sound like a heathen in admitting to using this little loudness "Gizmo", but the benefits extend beyond just lifting the bass - the input level of all my quad gear is too sensitive for modern devices such as my streamer and Yamaha CD. So the insertion loss it adds to the signal path ensures that I don't over-drive the input of any of my amps and blast the speaker cones against the wall.
A/B comparisons with this set up and the Sugden/headphone combination aren't entirely fair although I am pretty damned focused on the performance of them at low levels. Even so, I find myself drawn to the 909 principally, not just because of the purity of the music - fabulous solo singers sound even more fabulous and so too many poor artists and recordings have gone either in the bin or to the charity shops - but because the attack time is maintained.
It is just as I described it, rifle - shot quick - lightning fast. I always regarded my Spendors as something special. But the grip the 909 exercises over them makes them outstanding. And, because of the preceding exercise and the contribution made by you guys, I at last, I have and understanding as to why. I don't think even Peter Walker fully appreciated the impact his design would have and that is not being unfair as today we have much better equipment to drive his amplifiers into. I am not joking when I say, with the 909, I can detect any slippage in streamed music which would go completely unnoticed in any other equipment, albeit it when using my pc via the wireless link to my hub. I have to use the ethernet cable to avoid this and just to qualify my point, this is over a fibre optic broadband connection.
Ok, I'm convinced! I can't use my 909 at low levels and expect it to remain in Class A as a means of avoiding buying a Sugden Class A Power Amplifier. However, I no longer feel the need to. I am not going to spend a King's ransom only to discover, the amplifier does not respond to musical transients as quickly as the Quad 909.
If there is a finer amplifier to be had, I don't know what it would be. However, please don't say valves. I have lived with them throughout my professional life and am quite happy never to do so again.
I have added this further paragraph after further thought, relating to the grip of the 909 over my speakers, particularly at bass frequencies but also noticeable - to me at least - in the clarity of the mid range {the Spender crossovers are 3.4khz. and 13khz.} Damping factor was always quoted as an important characteristic of power amplifiers. This, if my understanding is correct, is the ability of the power amplifier to "Arrest" speaker cone excursions in the absence of a drive.
Because Spender designed the BC1 with bass/midrange speaker, the fast "Attack" of the 909, ensures the period when the loudspeaker is not under the control of the amplifier output stages is very, very short, a function of the improved "Slew Rate". So not only does this account for the perceived tightness of the bass, but also, because of the 3.4khz crossover, there is a sharpness to the clarity of midrange frequencies I hitherto have not heard.
For Christmas 2023, I treated myself to a new/old stock 909 with the intention of exploring the possibility of using it at low level, thinking I could keep it operating in the Class A zone, whilst driving my Stax headphones.
What a delightful "Can of Worms" that theory opened up.
A/B comparisons between the 405 and the 909 followed and it became obvious to me - I have commented frequently else where about the "Attack Time" of the 909, it is akin to a rifle shot - as I described it. Phenomenally quick with complex music. I also commented it was unforgiving with anything less than high quality recordings or singing. Some artists were made to sound dreadful.
Having only just fitted a pair of brand new, but stored bass/mid-range drivers to my BC1's - stored since 1996, the difference was frankly alarming and I fretted over the punishment they could suffer at the hands of the 909. It is so "lightning-quick", hence my thought of building a pair of low frequency enclosures to take some of the bass load of the Spendors.
I set about comparing again, the 303, 405 and 909.
The 303 remains sweet sounding, quite mellow and although there isn't any drama at bass frequencies, what there is, is natural. Still a very good amplifier as long as one doesn't press it to deliver too much. The Spendors help being relatively low power.
The 405 however, even with the Sponsors extend the bass beautifully, which surprised me as I had not expected them to perform so well in the lower registers. I didn't get much further with that as it had to be quickly packed in anticipation of our move.
However, I have spent over a year living with the 909 as my mainstay, with a Sugden Headmaster Headphone Amplifier and a pair of hifiman headphones. I can't take much of them at higher levels - a certain brittleness with some music hurts my ears.
Once I get an idea in my head I cannot let it go until it has been thoroughly thought through and proven or disproved. So I have spent countless hours listening to music with a passive preamplifier which comprises of a variable loudness control, which I modified to give only bass lift, whilst everything above 800hz - ish remains flat. I listen in the dining room, where I can keep the volume level low and the linear progressive lift to bass frequencies compensates for my failing hearing in that region. This has long been my target level, hence my obsession with the 909 and in particular that Class A section.
I may sound like a heathen in admitting to using this little loudness "Gizmo", but the benefits extend beyond just lifting the bass - the input level of all my quad gear is too sensitive for modern devices such as my streamer and Yamaha CD. So the insertion loss it adds to the signal path ensures that I don't over-drive the input of any of my amps and blast the speaker cones against the wall.
A/B comparisons with this set up and the Sugden/headphone combination aren't entirely fair although I am pretty damned focused on the performance of them at low levels. Even so, I find myself drawn to the 909 principally, not just because of the purity of the music - fabulous solo singers sound even more fabulous and so too many poor artists and recordings have gone either in the bin or to the charity shops - but because the attack time is maintained.
It is just as I described it, rifle - shot quick - lightning fast. I always regarded my Spendors as something special. But the grip the 909 exercises over them makes them outstanding. And, because of the preceding exercise and the contribution made by you guys, I at last, I have and understanding as to why. I don't think even Peter Walker fully appreciated the impact his design would have and that is not being unfair as today we have much better equipment to drive his amplifiers into. I am not joking when I say, with the 909, I can detect any slippage in streamed music which would go completely unnoticed in any other equipment, albeit it when using my pc via the wireless link to my hub. I have to use the ethernet cable to avoid this and just to qualify my point, this is over a fibre optic broadband connection.
Ok, I'm convinced! I can't use my 909 at low levels and expect it to remain in Class A as a means of avoiding buying a Sugden Class A Power Amplifier. However, I no longer feel the need to. I am not going to spend a King's ransom only to discover, the amplifier does not respond to musical transients as quickly as the Quad 909.
If there is a finer amplifier to be had, I don't know what it would be. However, please don't say valves. I have lived with them throughout my professional life and am quite happy never to do so again.
I have added this further paragraph after further thought, relating to the grip of the 909 over my speakers, particularly at bass frequencies but also noticeable - to me at least - in the clarity of the mid range {the Spender crossovers are 3.4khz. and 13khz.} Damping factor was always quoted as an important characteristic of power amplifiers. This, if my understanding is correct, is the ability of the power amplifier to "Arrest" speaker cone excursions in the absence of a drive.
Because Spender designed the BC1 with bass/midrange speaker, the fast "Attack" of the 909, ensures the period when the loudspeaker is not under the control of the amplifier output stages is very, very short, a function of the improved "Slew Rate". So not only does this account for the perceived tightness of the bass, but also, because of the 3.4khz crossover, there is a sharpness to the clarity of midrange frequencies I hitherto have not heard.
Last edited:
This the opamp output in the dead zone.
Using the two cursors and their info and if I've done the maths correctly it is slewing at about 0.652 volts in 0.00523ms which works out to approx 0.13v/uS which is very slow but then this is a only a 1kHz sine that is traversing the dead zone of two B/E junctions.
Can you trigger it using a 1mS pulse train with short duration - if that makes sense? I.e., start/stop pulse gen., looking at delay across the amp? Jeez - I regret not being hands on. Rise time being calculated as from 10% low to 90% high value of pulse - rudimentary I know.
Not sure if this is what you mean. A very fast rise and fall time squarewave applied. Second image is zoomed into the rise time. Remember that the bandwidth of any amp determines how 'fast' it is and that wide bandwidth amps will pass a signal 'quicker' than a low bandwidth one. In other words the output appears phase shifted or delayed from the input and that figure depends on the bandwidth.
The output signal here is brought back down to the input level by feeding it through an attenuator equal to the feedback factor of the amp. The two signals superimpose on each other.
The output signal here is brought back down to the input level by feeding it through an attenuator equal to the feedback factor of the amp. The two signals superimpose on each other.
Thanks for the effort - Not sure how, or if my approach is valid. Ideally, if there was instantaneous response, then the plot would be linear. Deviation from could be due to both the delay due to the Slew factor + non - linearity in the normalising network. No matter, the variation is not dramatically different, the central "Kink" being product of slew?
It has been many years since analysing circuits this way and test procedures have advanced? But my idea was to inject a very fast rise time pulse into the amplifier and look at the delay at its output, with dumpers disconnected. A start/stop pulse generator could be connected with the pulse introduced to the input and the stop trigger at the output. This would display the delay, which would be a function of the Class A's slew rate.
The amps overall performance across the full spectrum could be checked by changing the pulse repetition frequency. Obviously it is necessary to keep the pulse width small. However, one could then determine just how effective the amplifier is in dealing with complex audio over its full spectrum - if this makes sense?
Although, this is purely academic, given I am already persuaded. Still it would be nice to wrap it up in a nice gift box! Looking back, it all seems so incredibly simple. But isn't this always the case with the individual who firsts explores such new territory with no reference library to act as a guide?
It has been many years since analysing circuits this way and test procedures have advanced? But my idea was to inject a very fast rise time pulse into the amplifier and look at the delay at its output, with dumpers disconnected. A start/stop pulse generator could be connected with the pulse introduced to the input and the stop trigger at the output. This would display the delay, which would be a function of the Class A's slew rate.
The amps overall performance across the full spectrum could be checked by changing the pulse repetition frequency. Obviously it is necessary to keep the pulse width small. However, one could then determine just how effective the amplifier is in dealing with complex audio over its full spectrum - if this makes sense?
Although, this is purely academic, given I am already persuaded. Still it would be nice to wrap it up in a nice gift box! Looking back, it all seems so incredibly simple. But isn't this always the case with the individual who firsts explores such new territory with no reference library to act as a guide?