Rich,
thanks for your input.
Yes I got the definition right, but I should perhaps have
mentioned I was referring to the temp. coeff. of the
transconductance, while you seem to refer to the resistance.
So, we mean the same thing.
As I understand you, the problem usually isn't with the MOSFETs
per se or resistor biasing, but bad design with no margins and/or
people trying to stress their amps to the limit. Why am I not
surprised at those things? 🙂
thanks for your input.
Yes I got the definition right, but I should perhaps have
mentioned I was referring to the temp. coeff. of the
transconductance, while you seem to refer to the resistance.
So, we mean the same thing.
As I understand you, the problem usually isn't with the MOSFETs
per se or resistor biasing, but bad design with no margins and/or
people trying to stress their amps to the limit. Why am I not
surprised at those things? 🙂
Clipping and saturation
If are in VAS used antisaturation diodes ( only 4 - 6 pieces - depend on concrete connection ) you get any problems with saturation and recovery time of output transistors by clipping. This connection is easy possible implement to the every amp, but must be realize one condition - VAS rail voltage must be the same ( or only little bit higher ) than voltage on power rails.
If are in VAS used antisaturation diodes ( only 4 - 6 pieces - depend on concrete connection ) you get any problems with saturation and recovery time of output transistors by clipping. This connection is easy possible implement to the every amp, but must be realize one condition - VAS rail voltage must be the same ( or only little bit higher ) than voltage on power rails.
More to Hugh's point, the speed of the outputs helps
determine the speed of the drive circuit, making the
amp more stable with a wider variety of output devices.
determine the speed of the drive circuit, making the
amp more stable with a wider variety of output devices.
Thanks NP. Sped of amp devices is interesting.
Transistor spec sheets give Ft for the common base configuration. This is comparable to the speed of common collector (emitter follower), but MUCH faster than than the speed of a device operating in common emitter, like the VAS of an audio amplifier.
Consequently, I try to specify the speed of the VAS device(s) at least three times greater than the ouput devices. Normally drivers are not the problem; typically 50-100MHz is easy to find. But the outputs should be around 30MHz, and the VAS around 100MHz.
The diff pair devices are usually TO092s, very small dies, and consequently there are usually very fast, often the fastest in the circuit. Nonetheless they too are in common emitter, so speed is important here, too.
Fortunately there is a cornucopia of good silicon out there, much of it cheap, and suitable choices are not difficult. The amp should be designed so that the VAS is indeed the slowest part of the amp, and in fact forcibly so with appropriate lag compensation.
I'm mindful of Peter Walker's comment, 'There is no problem in audio engineering which cannot be solved by judicious application of Ohm's law and common sense'. Probably a bit simplistic, but close to the mark.......
Cheers,
Hugh
Transistor spec sheets give Ft for the common base configuration. This is comparable to the speed of common collector (emitter follower), but MUCH faster than than the speed of a device operating in common emitter, like the VAS of an audio amplifier.
Consequently, I try to specify the speed of the VAS device(s) at least three times greater than the ouput devices. Normally drivers are not the problem; typically 50-100MHz is easy to find. But the outputs should be around 30MHz, and the VAS around 100MHz.
The diff pair devices are usually TO092s, very small dies, and consequently there are usually very fast, often the fastest in the circuit. Nonetheless they too are in common emitter, so speed is important here, too.
Fortunately there is a cornucopia of good silicon out there, much of it cheap, and suitable choices are not difficult. The amp should be designed so that the VAS is indeed the slowest part of the amp, and in fact forcibly so with appropriate lag compensation.
I'm mindful of Peter Walker's comment, 'There is no problem in audio engineering which cannot be solved by judicious application of Ohm's law and common sense'. Probably a bit simplistic, but close to the mark.......
Cheers,
Hugh
Thanks. If it's supposed to be simple to build, and sound good - I'm buying a couple of them.
-Andy
How would this compare to the sound and ease of construction to the Leach amp?
-Andy
How would this compare to the sound and ease of construction to the Leach amp?
Hi All,
I have gone through this thread with great interest.
In another thread I mentioned distortion of the leading edges of first cycles, and called it 'FCD' = first cycle distortion.
I would not build this amplifier because R1+C1, C3, R10+Q8/C9 and R11+Q9 are a series of first cycle distortion generators. Values and turnovers not known, the delays might be low, but I cannot assume that they are. Stable - yes; accurate - no.
Rod quotes a maximum of 0.02% THD, which is likely to fall to insignificant levels at normal home listening levels.
BUT. First cycle distortion does not reduce with output level.
An amplifier with first cycle distortion sounds smooth because it is not properly following the leading edges of transient waveforms. It can be pleasing to the ear, but not accurate, and can remove some of the natural harshness that really should be there in the same way that it is in real life - as with violin string harmonics which naturally extend beyond the teens.
Another clue comes from the output impedance figures.
His topology cannot help but introduce a small NFB time delay - like an inductor.
All amps do, but some more than others, and when there is a high level of NFB there is a greater potential for initial load induced overshoot during that propagation time delay, which could cause momentary cross-conduction induced problems
This means that when say a mid-bass driver suffers cone break-up or its driven impedance non-linearity interacts with crossover circuitry, its back emf may not be quickly enough damped at the output terminal, thus a potential suddenly develops wrt input and the tweeter waveform instantly becomes distorted.
With what kind of a phase response and transient error pulse is Rod's damping achieved at the frequencies qouted ?
This has nothing to do with signal path lengths, but to the effective series inductance and parallel capacitance along that path.
I first observed this over thirty years ago when checking out supposedly fantastic amplifiers.
Method - headphone monitor the amplifier output whilst switching its loading between a resistor and a loudspeaker in another room.
Another way of checking is to listen to the tweeter ouput from a 'three-way' that has been isolatied from its composite setting on extended leads, though you do need to have a good reference to start with. I had valve amps and the JLH.
This is how I concluded that the simple JLH 10W class-A had real merit and that it warranted further development. It does not have great damping in its original form, but what there is is so little delayed at mid band frequencies.
I think Rod Elliots amplifier would make a good 'first' home-construction amplifier, and be great for parties.
Purist hi-fi listening with complex high end loudspeakers might be better served with extremely low NFB delay cicuitry, which means low propagation delay, however, there are few simple purist amps for home construction that can thrill in the same way that a powerful chassis can.
There might be some merit in extending the leads of D1+2 to have them in direct contact with output devices.
(Time for the fireproof suit again ?)
Cheers ............... Graham.
I have gone through this thread with great interest.
In another thread I mentioned distortion of the leading edges of first cycles, and called it 'FCD' = first cycle distortion.
I would not build this amplifier because R1+C1, C3, R10+Q8/C9 and R11+Q9 are a series of first cycle distortion generators. Values and turnovers not known, the delays might be low, but I cannot assume that they are. Stable - yes; accurate - no.
Rod quotes a maximum of 0.02% THD, which is likely to fall to insignificant levels at normal home listening levels.
BUT. First cycle distortion does not reduce with output level.
An amplifier with first cycle distortion sounds smooth because it is not properly following the leading edges of transient waveforms. It can be pleasing to the ear, but not accurate, and can remove some of the natural harshness that really should be there in the same way that it is in real life - as with violin string harmonics which naturally extend beyond the teens.
Another clue comes from the output impedance figures.
His topology cannot help but introduce a small NFB time delay - like an inductor.
All amps do, but some more than others, and when there is a high level of NFB there is a greater potential for initial load induced overshoot during that propagation time delay, which could cause momentary cross-conduction induced problems
This means that when say a mid-bass driver suffers cone break-up or its driven impedance non-linearity interacts with crossover circuitry, its back emf may not be quickly enough damped at the output terminal, thus a potential suddenly develops wrt input and the tweeter waveform instantly becomes distorted.
With what kind of a phase response and transient error pulse is Rod's damping achieved at the frequencies qouted ?
This has nothing to do with signal path lengths, but to the effective series inductance and parallel capacitance along that path.
I first observed this over thirty years ago when checking out supposedly fantastic amplifiers.
Method - headphone monitor the amplifier output whilst switching its loading between a resistor and a loudspeaker in another room.
Another way of checking is to listen to the tweeter ouput from a 'three-way' that has been isolatied from its composite setting on extended leads, though you do need to have a good reference to start with. I had valve amps and the JLH.
This is how I concluded that the simple JLH 10W class-A had real merit and that it warranted further development. It does not have great damping in its original form, but what there is is so little delayed at mid band frequencies.
I think Rod Elliots amplifier would make a good 'first' home-construction amplifier, and be great for parties.
Purist hi-fi listening with complex high end loudspeakers might be better served with extremely low NFB delay cicuitry, which means low propagation delay, however, there are few simple purist amps for home construction that can thrill in the same way that a powerful chassis can.
There might be some merit in extending the leads of D1+2 to have them in direct contact with output devices.
(Time for the fireproof suit again ?)
Cheers ............... Graham.
Re the above post - I have one word to say that covers the topic completely ...
Rubbish
Cheers, Rod
Rubbish
Cheers, Rod
rode said:
I have built P3A from Rode and it is an outstanding amp. and I suppose the new one is as good if not better than the widely built P3A.
I am sure there are superhumans out there that can hear the so-called, little documented and yet undefined "first cycle distortion", I am certainly not one of them.
and I agree completely with Rode on his assessement.
Hi jam,
No just that loudspeaker induced back emf does not cause the amplifier to generate more error.
Rod, I'm surprised that you come back with that kind of reply. Rubbish ? Time will tell.
Your amplifier is similar to the D Self topology, and it distorts 'audio' waveforms much more than steady state sinewaves too.
Why not try looking at first cycle distortion and reverse impedance characteristics; these are the real reasons why amplifiers that measure the same on the test bench still sound different in real life.
Cheers ........... Graham
No just that loudspeaker induced back emf does not cause the amplifier to generate more error.
Rod, I'm surprised that you come back with that kind of reply. Rubbish ? Time will tell.
Your amplifier is similar to the D Self topology, and it distorts 'audio' waveforms much more than steady state sinewaves too.
Why not try looking at first cycle distortion and reverse impedance characteristics; these are the real reasons why amplifiers that measure the same on the test bench still sound different in real life.
Cheers ........... Graham
Graham Maynard said:
Graham: rather than making baseless assertions and presenting them as facts, why don't you build the amp and measure how it distorts "audio" waveforms much more than steady state sinewaves and show us facts rather than your assertions (please make absolutely sure that you don't call simulation waveforms as traces, BTW)?
you should not apply if you don't have facts.
Grahm,
We have an ancient proverb "I'm from Missouri, you'll have to show me".
Ive seen and built amps based on schematics by Rod Elliot, D. Self and R. Slone among others. Rather than a lot of words how about posting a schematic of one of your designs/ Perhaps a photo of the completed project, too.😉
We have an ancient proverb "I'm from Missouri, you'll have to show me".
Ive seen and built amps based on schematics by Rod Elliot, D. Self and R. Slone among others. Rather than a lot of words how about posting a schematic of one of your designs/ Perhaps a photo of the completed project, too.😉
Graham has an upcoming article or series in Wireless World, if I'm not mistaken. I guess you'll get your chance.
Missouri? Ancient? Hmm.
Missouri? Ancient? Hmm.
Hi Millwood, More digs!
and Hi Sam,
I have told you how to examine the performance of an amplifier I the post above. Use headphones to see how the loudspeaker affects the amp when compared to the linear resistor method we all trust.
Listen to the tweeter in isolation, but then of course you need a better amp to compare it with.
Look at the phase of the amplifier's response when you are determining its output impedance - if it is not in phase then it generates additional errors.
You have to try this for yourself, and there is no other way around that. I cannot do it for you, especially over the internet.
Cheers .......... graham.
and Hi Sam,
I have told you how to examine the performance of an amplifier I the post above. Use headphones to see how the loudspeaker affects the amp when compared to the linear resistor method we all trust.
Listen to the tweeter in isolation, but then of course you need a better amp to compare it with.
Look at the phase of the amplifier's response when you are determining its output impedance - if it is not in phase then it generates additional errors.
You have to try this for yourself, and there is no other way around that. I cannot do it for you, especially over the internet.
Cheers .......... graham.
paulb said:
if it is published in (any reputable journal of your choosing), it must be good.
🙂
for those interested, may I mention one name?
Jan Hendrik Schön
Ok, let's look at this in a sensible manner for a moment. If I use an accurate differential input circuit to monitor the input and feedback points of an amplifier, then any discrepancy between the two is shown. With a trimpot to obtain very good nulling of the two signals, any residual is distortion (time, phase or amplitude). For example, if the output signal starts to clip (even ever so slightly), the output of the differential amp increases dramatically - as one would expect.
Now, note what I pointed out above ... any difference - time, phase or amplitude differences are shown very clearly. One does need an oscilloscope for this, but that is not a major issue. At low signal levels (anything below clipping), the output of the differential amp can be nulled for amplitude only - time and phase are not balanced out (and I deliberately made no attempt to do so), so show up as a residual signal.
Every amplifier I have tested this way behaves slightly differently, but the discrepancies only become apparent at (or near) clipping, or if the input signal is too fast for the amplifier (e.g. a fast risetime squarewave). The critical test is (of course) music. Testing my P3A or P101 designs shows (almost) zero output from the differential amplifier, and this applies whether the load is connected or not (resistive and speaker loads alike). Naturally, as the amp starts to clip there is plenty of output, but this is to be expected if the feedback and input signals cannot be equalised by the power amplifier.
I elected to call this circuit a 'SIM' (Sound Impairment Monitor), but the basic idea has been around for a long time - it was originally suggested by Peter Baxandall, but in slightly different form because valve amps were the order of the day at the time. For this reason, it was not possible to simply monitor the input and feedback signals, since they were often radically different - a valve amp is massively different from a transistor amp.
If (and that is a very large if indeed) the 'First Cycle Distortion' proposed by Graham Maynard exists, then the SIM would show it - it must. Naturally, there is scope for the idea to be poo-poohed because the opamp will allegedly have the same problem as the power amplifier, but thousands of instrumentation systems used throughout the world that use opamps would therefore be wrong, oscilloscopes will be unable to resolve the first cycle properly (they all use feedback amplifiers internally), and analogue science as we know it will be forever changed.
Likely? I think not. Regrettably, Mr. Maynard appears to be caught up in the pseudoscience that pervades audio, ascribing magical properties to various topologies, and demonic (or worse) to others.
FCD and TIM are 'birds of a feather' as it were, and as TIM (Transient Intermodulation Distortion) was discredited many years ago, so shall FCD be discredited - hopefully, the information in this post will go some way to accomplishing the demise of FCD as a valid premise.
For those who wish to experiment with the SIM idea, I recommend that you read the article and project details ...
SIM Article
Simple SIM project
Build one, test the results. You will be astonished at the resolution of the system, and it works with sinewaves, band limited squarewaves (to limit the slew rate to something within the amp's capabilities), music, pink noise, etc.
If anyone should take the time to do some real tests, I'm sure that the readership of this forum would be very interested to hear about it. There is no point my doing any further tests, as I would be seen to be biased (a not unreasonable assumption).
Cheers, Rod
Now, note what I pointed out above ... any difference - time, phase or amplitude differences are shown very clearly. One does need an oscilloscope for this, but that is not a major issue. At low signal levels (anything below clipping), the output of the differential amp can be nulled for amplitude only - time and phase are not balanced out (and I deliberately made no attempt to do so), so show up as a residual signal.
Every amplifier I have tested this way behaves slightly differently, but the discrepancies only become apparent at (or near) clipping, or if the input signal is too fast for the amplifier (e.g. a fast risetime squarewave). The critical test is (of course) music. Testing my P3A or P101 designs shows (almost) zero output from the differential amplifier, and this applies whether the load is connected or not (resistive and speaker loads alike). Naturally, as the amp starts to clip there is plenty of output, but this is to be expected if the feedback and input signals cannot be equalised by the power amplifier.
I elected to call this circuit a 'SIM' (Sound Impairment Monitor), but the basic idea has been around for a long time - it was originally suggested by Peter Baxandall, but in slightly different form because valve amps were the order of the day at the time. For this reason, it was not possible to simply monitor the input and feedback signals, since they were often radically different - a valve amp is massively different from a transistor amp.
If (and that is a very large if indeed) the 'First Cycle Distortion' proposed by Graham Maynard exists, then the SIM would show it - it must. Naturally, there is scope for the idea to be poo-poohed because the opamp will allegedly have the same problem as the power amplifier, but thousands of instrumentation systems used throughout the world that use opamps would therefore be wrong, oscilloscopes will be unable to resolve the first cycle properly (they all use feedback amplifiers internally), and analogue science as we know it will be forever changed.
Likely? I think not. Regrettably, Mr. Maynard appears to be caught up in the pseudoscience that pervades audio, ascribing magical properties to various topologies, and demonic (or worse) to others.
FCD and TIM are 'birds of a feather' as it were, and as TIM (Transient Intermodulation Distortion) was discredited many years ago, so shall FCD be discredited - hopefully, the information in this post will go some way to accomplishing the demise of FCD as a valid premise.
For those who wish to experiment with the SIM idea, I recommend that you read the article and project details ...
SIM Article
Simple SIM project
Build one, test the results. You will be astonished at the resolution of the system, and it works with sinewaves, band limited squarewaves (to limit the slew rate to something within the amp's capabilities), music, pink noise, etc.
If anyone should take the time to do some real tests, I'm sure that the readership of this forum would be very interested to hear about it. There is no point my doing any further tests, as I would be seen to be biased (a not unreasonable assumption).
Cheers, Rod
- Status
- Not open for further replies.