Hi,
Any idiot can build an amplifier that will blow up.
Building one that doesn't is more difficult.
rgds, sreten.
If you did read correctly , you would realize that i didnt say that they
do not matters , but that with modern devices and up to date designs,
they are no more an issue.
That said, i wonder why you give the preminence to designs that display
THD at -60/-70db with no switching distorsion over designs that can do
-100db in both THD and switching distorsion.
Indeed, what you call switching distorsion is still harmonic distorsion...😉
Hi,
Because I was referencing Self I used his definitions, its all his fault .... 😉
But I agree with BC. I don't like Self's Class C, its nothing like radio class C,
(class C in radio only has one device, other polarity via a reactive component)
and to many seems to be class B (no bias). To differentiate between optimum
Class B bias and deliberate overbias, e.g. 1/4 class A = bias at 0.5 of full class
A so up to 1/4 power is class A the rest AB, i'd prefer to use the terms
class aB for optimum bias and class AB for deliberate overbias.
Anyway the scope is AB as defined and bashed by Self, his examples are 1/4 class A AB.
Self : class B = optimal bias B, class AB = overbiased B, Class A = A.
rgds, sreten.
Last edited:
Agree....🙂
Anyway, for whom has little experience , there s quite enough
knowledge in this forum to help bring a project to a usefull end...
I wish i had such a tool back in my youth , it would have spared me
a lot of burned devices...and fingers..😉
Hi,
the only way to prevent the nasty switching distortion is class A biasing, crossover distortion will still be present due to various transition artifacts between devices, like the gm-doubling distortion. The only way to prevent crossover distortion is using just one device. Simulators will tell you differently, in reality there`s no way to get around.
the only way to prevent the nasty switching distortion is class A biasing, crossover distortion will still be present due to various transition artifacts between devices, like the gm-doubling distortion. The only way to prevent crossover distortion is using just one device. Simulators will tell you differently, in reality there`s no way to get around.
The success or failure of class AB amplifiers depends on how well the match at the transition point between the positive and negative going sides is executed. Early solid state amplifiers had serious defects in this regard however it can happen with tube amplifiers as well. The result is called crossover notch distortion which exhibits itself as a harsh bright sound. This is reflected in high harmonic distortion at low levels. One trick to hear this is to play the amplifier through highly efficient speakers at very low levels. If crossover notch distortion is present, it will be at its most audible. For this reason the specification for rating power amplifers was changed from THD at rated output to THD at rated output and any level below. It took about 5 to 10 years for design engineers to figure out how to get out of this problem during the conversion from tubes to transistors. BTW, I seem to recall that FET transistors don't exhibit the same characteristic of thermal runaway as bipolar transistors.
The advantages of class AB over class A are obvious, higher power output and greater efficiency. Class B biased at cutoff can't be a good choice for a high fidelity power amplifier.
The heart and soul of any good audio power amplifier is IMO its power supply. This is often the limiting factor. Proper use of negative feedback, not necessarily an easy thing to learn how to do is also critical. Criticizing the best available strategies by citing the shortcomings of examples that were poorly executed for one reason or another IMO is an absurd way to make inferior concepts seem better unless that's all you can design and you know that the market does not have the knowledge to know. For this reason, virtually all professional installations use high power negative feedback class AB solid state amplifiers. Audiophiles by contrast seem to be willing to buy almost anything if the advertising is slick enough.
The advantages of class AB over class A are obvious, higher power output and greater efficiency. Class B biased at cutoff can't be a good choice for a high fidelity power amplifier.
The heart and soul of any good audio power amplifier is IMO its power supply. This is often the limiting factor. Proper use of negative feedback, not necessarily an easy thing to learn how to do is also critical. Criticizing the best available strategies by citing the shortcomings of examples that were poorly executed for one reason or another IMO is an absurd way to make inferior concepts seem better unless that's all you can design and you know that the market does not have the knowledge to know. For this reason, virtually all professional installations use high power negative feedback class AB solid state amplifiers. Audiophiles by contrast seem to be willing to buy almost anything if the advertising is slick enough.
I wish you had put that in post1. I certainly read it differently and judging by the following discussion many others read it differently as well.
ClassA by Self's defn. is a sub-set of Self's defn ClassAB since it is an overbiased ClassB, (by Self's defn) and simply transitions out of ClassA into Class AB/B when the output current exceeds ~ twice the output stage bias current.
Self does not rubbish Push-Pull ClassAB, he shows how he designs for it and claims it is better than "His ClassB", when the output current is <2*Ib.
He is in effect telling us to overbias the ClassB topology and keep output current within the ClassA limit.
Hi again,
we are talking about power amplifier classification by the position of the operating point on the transfer characteristic of the output device.
In low biased class AB, gm varies widely with the constantly changing current and temperature, the best way to keep it stable is class A biasing. Switching devices starved of current perform insignificant or no amplification creating high level of odd and high order harmonics. The rapid change of control voltage triggers high slew rates and peak currents. Switching transients consist of 100% intermodulation products standing in sharp contrast to the pleasing simulated THD. No amount feedback can help. In comparison, the gm-doubling distortion is rather benign.
we are talking about power amplifier classification by the position of the operating point on the transfer characteristic of the output device.
In low biased class AB, gm varies widely with the constantly changing current and temperature, the best way to keep it stable is class A biasing. Switching devices starved of current perform insignificant or no amplification creating high level of odd and high order harmonics. The rapid change of control voltage triggers high slew rates and peak currents. Switching transients consist of 100% intermodulation products standing in sharp contrast to the pleasing simulated THD. No amount feedback can help. In comparison, the gm-doubling distortion is rather benign.
sreten, Bob
I hope you consider my questions on topic and somewhat helpful (if not just say so and I'll stop).
Bob, Thank you for your detailed and insightful response (like having my own personalized section of your book). I do understand the issues you bring up including the small signal optimization of the bias setting. For the most part my question was why these issues no longer seem to matter with newer better components available, as well as the dependance of the VAS output impedance.
From what I'm gathering it maybe that I was overly concentrating on strictly the output stage, and that some of the replies may have been inferring that with more feedback the output stage's deviations are no longer significant.
I have satisfied my understanding of Douglas Self's figure 6.43 (sorry to those who don't have the book) as I needed to remind myself of the polarities associated with the plateaus. Even within this book's chapter on output stages many of the associated plots rely on the overall Blameless amp's characteristic's, not sure why with an emitter follower one couldn't perform explicit testing on just that stage and then fold in the other effects slowly (but I suppose that's the whole point to the Blameless amp in the first place).
Thanks again to all, I enjoy picking up these pointers.
-Antonio
I hope you consider my questions on topic and somewhat helpful (if not just say so and I'll stop).
Bob, Thank you for your detailed and insightful response (like having my own personalized section of your book). I do understand the issues you bring up including the small signal optimization of the bias setting. For the most part my question was why these issues no longer seem to matter with newer better components available, as well as the dependance of the VAS output impedance.
From what I'm gathering it maybe that I was overly concentrating on strictly the output stage, and that some of the replies may have been inferring that with more feedback the output stage's deviations are no longer significant.
I have satisfied my understanding of Douglas Self's figure 6.43 (sorry to those who don't have the book) as I needed to remind myself of the polarities associated with the plateaus. Even within this book's chapter on output stages many of the associated plots rely on the overall Blameless amp's characteristic's, not sure why with an emitter follower one couldn't perform explicit testing on just that stage and then fold in the other effects slowly (but I suppose that's the whole point to the Blameless amp in the first place).
Thanks again to all, I enjoy picking up these pointers.
-Antonio
The stability of quiescent operating point of a tube or transistor depends on several factors. One is the regulation of the power supply. A poorly regulated supply whose voltage changes with output current demand, input voltage, or temperature will not be stable. The nominal design point is not necessarily reflective of actual operating conditions at any moment. Regulation of power supplies is a form of negative feedback in the supply that keeps the output voltage constant.
Vacuum tubes are especially vulnerable in changes to temperature. Thermionic emission of cathodes varies with the square of the temperature (in degrees Kelvin) so tubes that are not gain stabalized by feedback will drift all over the place. They are inherently unstable and are a very bad choice for a high fidelity amplifier IMO.
The correct use of negative feedback is determined by using just the right amount to cancel out distortion products, not too much or too little. The classic example of negative feedback is the relationship between steering a car and learning by seeing the effect of turning the wheel what the effect on the car's direction is. Oversteer and understeer are example of excessive and inadequate negative feedback in that example. Too much feedback in an audio amplifier can increase distortion, even cause the amplifier to become unstable and turn into an oscillator, in fact that 's how you build an oscillator. The equations for negative feedback are extremely complex and are frequency sensitive. However, there are designs which have proven very successful in reduding distortion to barely measurable levels.
Class A amplifiers are those in which the quiescent operating point is chosen so that current always flows on both halves of each cycle. In class B, the devices are biased such that the quiescent operating points are at cutoff. In class AB the bias point is chosen near cutoff but ideally not so close to cutoff that the point is on the curved non linear part of the characteristic curves but instead on the linear region. This allows a smooth transition from the positive to negative sides. Where there are bias adjustment pots, any drift in the value of those pots or of fixed resistors over time or with temperature will alter the quiescent operating points. This has to be take into consideration in the design and servicing of these amplifiers. Bias adjustments used to be routine maintenance in the days of vacuum tube amplifiers but are rarely considered anymore. Perhaps they are generally more stable than they used to be or other tricky circuits are used to keep bias current constant.
Vacuum tubes are especially vulnerable in changes to temperature. Thermionic emission of cathodes varies with the square of the temperature (in degrees Kelvin) so tubes that are not gain stabalized by feedback will drift all over the place. They are inherently unstable and are a very bad choice for a high fidelity amplifier IMO.
The correct use of negative feedback is determined by using just the right amount to cancel out distortion products, not too much or too little. The classic example of negative feedback is the relationship between steering a car and learning by seeing the effect of turning the wheel what the effect on the car's direction is. Oversteer and understeer are example of excessive and inadequate negative feedback in that example. Too much feedback in an audio amplifier can increase distortion, even cause the amplifier to become unstable and turn into an oscillator, in fact that 's how you build an oscillator. The equations for negative feedback are extremely complex and are frequency sensitive. However, there are designs which have proven very successful in reduding distortion to barely measurable levels.
Class A amplifiers are those in which the quiescent operating point is chosen so that current always flows on both halves of each cycle. In class B, the devices are biased such that the quiescent operating points are at cutoff. In class AB the bias point is chosen near cutoff but ideally not so close to cutoff that the point is on the curved non linear part of the characteristic curves but instead on the linear region. This allows a smooth transition from the positive to negative sides. Where there are bias adjustment pots, any drift in the value of those pots or of fixed resistors over time or with temperature will alter the quiescent operating points. This has to be take into consideration in the design and servicing of these amplifiers. Bias adjustments used to be routine maintenance in the days of vacuum tube amplifiers but are rarely considered anymore. Perhaps they are generally more stable than they used to be or other tricky circuits are used to keep bias current constant.
Hi,
With hindsight I should have been very specific.
I don't know he you work out Self does not
continually state AB is worse than A or his B.
Its stated all over the place, whilst I think only
once does he acknowledge that the class A
region of his AB should be better than his B.
He does show how to design for AB if you "must".
He also suggests class A for class G,
where the issues are similar to his AB.
Nevertheless, I think EF would work better than
CFP for his AB, with a less abrupt changeover.
I'm still investigating what the intrinsic Re of
output transistors is, seems hard to pin down
rgds, sreten..
This one isn't difficult: as others have said earlier (perhaps not in such an unambiguous way) Re~=26mV/Ie.
This is an approximate relationship, in real transistors, it might be something like (26mV*Ie^-0.9)+re, re being the ohmic resistances of the emitter connection, but at low currents, it is a good approximation.
Anyway it isn't possible to have a perfect crossover, because of the exponential nature of the I to V characteristic.
But in an experimental set-up, with no emitter resistors and a minimal quiescent current, say 2mA, you can get an amazing linearity at low levels.
In practice, it's suicidal: as soon as the things heat up differentially by 2 or 3°C, the thermal runaway begins, and if you're on anything else than a current-limited lab PSU, it ends up with molten silicon.
There are solutions: one is to put in series with the output transistors a negative resistance of equal and opposite value, another is to actively control the crossover characteristic.
An example of the former is Unigabuf, examples of the latter include the Renardson circuit and the Circlophone.
But the gm doubling problem does exist, there isn't the shade of a doubt about it.
It is true that overbiasing is certainly less risky than underbiasing: a severely overbiased OP stage might sound slightly unpleasant, whereas a moderately underbiased one is unbearable.
An interesting scheme in this context is the class AC from J. Broskie. Much hotter than class AB, but mainly without the inconvenients, and cooler than pure class A.
Last edited:
havent read the total thread is getting way too long and too fast ...
lets see a point please ..
in post #2 there is a comment regarding 0R0 emmiter resistors and how nice an amp will sound without them while preserving a stable bias will be almost impossible
lets suppose that we use carefully matched and high quality output and a "double" vbe multiplier with a pair of BD135 for example mounted one on the PNP transistor and one on the npn
will a procedure like that preserve stable bias ?
then again why SAP familly that includes all the parts inside for thermal monitoring still cannot be used with 0R0 Re ?
then the coclusion might be that at DC conditions might work but in real life PNP behaviour versus thermals might be diferent than NPN ?
then by that we have to go back make again quasi amps to make sure that thermals are the same use 0R0 Re and "thermals" located in both transistors
now some real experts have to help here ...please
kind regards sakis
PS i could try that see how it sounds IE skip the emmiter resistors while try to preserve a steady bias with any available way
lets see a point please ..
in post #2 there is a comment regarding 0R0 emmiter resistors and how nice an amp will sound without them while preserving a stable bias will be almost impossible
lets suppose that we use carefully matched and high quality output and a "double" vbe multiplier with a pair of BD135 for example mounted one on the PNP transistor and one on the npn
will a procedure like that preserve stable bias ?
then again why SAP familly that includes all the parts inside for thermal monitoring still cannot be used with 0R0 Re ?
then the coclusion might be that at DC conditions might work but in real life PNP behaviour versus thermals might be diferent than NPN ?
then by that we have to go back make again quasi amps to make sure that thermals are the same use 0R0 Re and "thermals" located in both transistors
now some real experts have to help here ...please
kind regards sakis
PS i could try that see how it sounds IE skip the emmiter resistors while try to preserve a steady bias with any available way
Nevertheless, were you to read the opening post and the next few posts you would hopefully understand what the thread was intended to be about and the confusion from some contributors that took it in a different direction entirely.
Read from #1 and enjoy
Hi,
I think I'm struggling with terminology again. Perhaps the right word is extrinsic.
Whatever, what I meant is the Re value the DXamp would appear to have even
though it has no Re's. Apparently the Dxamp's 4.7R input resistors on the output
devices as well as as the device Rb will translate to an effective Re value.
(Apparently depending on Beta device Rb can dominate device Re.)
The "intrinsic" Re is as you state above, it always reduces with more current.
B.C. states ideally this Re = external Re, but in the case of no exterrnal Re
then I presume intrinsic Re = extrinsic Re for the device and its input resistor
becomes an interesting point to identify .....
For class B (Self) the Vbe thermal tracker is an estimator, but IMO for high
bias AB it is genuine thermal feedback, noting for the 1/4 class A case that
device temperature variations are greatly reduced compared to class B (Self).
For class B (Self) dissipation peaks at ~ 1/3 power.
For Class A it drops to 50% at full power (push pull).
Wonder what the 1/4 A AB case actually looks like ....
A problem might be the initial bias current on turn on, AFAICT
seems this will be quite a bit higher than the equilibrium value.
rgds, sreten.
- Status
- Not open for further replies.