Everyone, I would like to point out that different people like to put their time and energies into different things. This is important to note.
For example, some people find it a wonderful challenge to see how cheaply they can get something built. I, once, had to make a phase locked tape capstan servo for essentially the same price as the precision AC motor that it replaced. In this case, every resistor counted. I still wince at some of my tradeoffs, but it worked and went into production, for years.
Others, are in competition with a bunch of metal 'artistes' who just love to make metal masterpieces of beauty and impressiveness. In this case, at least 1/2 of ones time, money and energy has to be devoted to external and internal appearances of the product. Look at a high price sports car or sedan. It is the same in audio.
Well, when criticism comes up, often it is one camp against another, slinging the comments.
In my case, while I don't like to bother with metal fabrication, I have grown used to quality examples, and it is hard to go back. Does this mean that an equivalent sounding preamp, for example could not be made in a pie tin, or cake box? No, but it would be difficult to sell one, made that way.
For example, some people find it a wonderful challenge to see how cheaply they can get something built. I, once, had to make a phase locked tape capstan servo for essentially the same price as the precision AC motor that it replaced. In this case, every resistor counted. I still wince at some of my tradeoffs, but it worked and went into production, for years.
Others, are in competition with a bunch of metal 'artistes' who just love to make metal masterpieces of beauty and impressiveness. In this case, at least 1/2 of ones time, money and energy has to be devoted to external and internal appearances of the product. Look at a high price sports car or sedan. It is the same in audio.
Well, when criticism comes up, often it is one camp against another, slinging the comments.
In my case, while I don't like to bother with metal fabrication, I have grown used to quality examples, and it is hard to go back. Does this mean that an equivalent sounding preamp, for example could not be made in a pie tin, or cake box? No, but it would be difficult to sell one, made that way.
Housekeeping.
SY
Thanks for the kind words, are you sure you are into the spirit of this thread!
ikoflexer,
I think you asked why unbalanced power supplies were an advantage. Since most of the traffic here really runs on misunderstanding, I will start at the basics.
Music when produced is a complex 4 dimensional modulation of air pressure. We reduce this to two undulating voltages (Bell's term) for most of the purposes discussed here. As we are not immortal (I apologize in advance if I am wrong about this) there are only frequency components above 0 cycles per second. This means that the voltage goes above and below zero. The rate at which it does that is the slew rate. Louder passages have higher voltages and slew more for the same time period between zero crossings.
A typical music signal spends most of it's time around the zero crossings in terms of voltage level. I am not concerned here with the large level limitations, that is another issue.
In a Class B output stage one device is used to pull the signal voltage up and another pulls it down. (John uses current output in the Blowtorch as mentioned elsewhere). There is a dead zone around zero when neither device is on. Since much of the music is around zero crossings this distortion is easier to hear that many other kinds. Note that timing errors from either digital or analog time delay causes also cause problems.
In a class AB output both upper and lower devices are on at the zero crossing. This gives double the gain for the signal which is not as bad, but just past the zero point the gain shifts back to normal, again increasing nasty things.
In a class A circuit much of this is avoided, but there still are gain modulation issues, due to other factors.
If we offset the power supply voltages then the hand off between output devices does not occur as close to the zero crossings, so it actually happens less often and is more masked by the desired signal.
In the Blowtorch I have gathered that there are two class A devices opposing each other at each stage. If they were perfectly match the total distortion would be less, but it would occur at the zero crossings. Since the N's and P's are not perfectly matched the total distortion is greater but does not happen where the problem is most easily heard.
I expect you can now begin to understand why tight tracking of the power supplies is not always a good idea. Also if you carry the thought a bit more you can see some of the advantages of a current output.
The biggest advantage of current out has more to do with other factors, but that is also a different issue.
SY
Thanks for the kind words, are you sure you are into the spirit of this thread!
ikoflexer,
I think you asked why unbalanced power supplies were an advantage. Since most of the traffic here really runs on misunderstanding, I will start at the basics.
Music when produced is a complex 4 dimensional modulation of air pressure. We reduce this to two undulating voltages (Bell's term) for most of the purposes discussed here. As we are not immortal (I apologize in advance if I am wrong about this) there are only frequency components above 0 cycles per second. This means that the voltage goes above and below zero. The rate at which it does that is the slew rate. Louder passages have higher voltages and slew more for the same time period between zero crossings.
A typical music signal spends most of it's time around the zero crossings in terms of voltage level. I am not concerned here with the large level limitations, that is another issue.
In a Class B output stage one device is used to pull the signal voltage up and another pulls it down. (John uses current output in the Blowtorch as mentioned elsewhere). There is a dead zone around zero when neither device is on. Since much of the music is around zero crossings this distortion is easier to hear that many other kinds. Note that timing errors from either digital or analog time delay causes also cause problems.
In a class AB output both upper and lower devices are on at the zero crossing. This gives double the gain for the signal which is not as bad, but just past the zero point the gain shifts back to normal, again increasing nasty things.
In a class A circuit much of this is avoided, but there still are gain modulation issues, due to other factors.
If we offset the power supply voltages then the hand off between output devices does not occur as close to the zero crossings, so it actually happens less often and is more masked by the desired signal.
In the Blowtorch I have gathered that there are two class A devices opposing each other at each stage. If they were perfectly match the total distortion would be less, but it would occur at the zero crossings. Since the N's and P's are not perfectly matched the total distortion is greater but does not happen where the problem is most easily heard.
I expect you can now begin to understand why tight tracking of the power supplies is not always a good idea. Also if you carry the thought a bit more you can see some of the advantages of a current output.
The biggest advantage of current out has more to do with other factors, but that is also a different issue.
Thank you!
I just don't understand what you mean by tight tracking of the power supplies.
I am curious what the biggest advantage of current out is.
I just don't understand what you mean by tight tracking of the power supplies.
I am curious what the biggest advantage of current out is.
Simon, not to be picky, but in an optimally-biased class AB stage the gain (actually total gm) is the same at crossover as it is far away from crossover. Unfortuantely it wiggles around crossover. This is basically what Barney Oliver pointed out long ago. What you pointed out above is the so-called gm-doubling situation that can occur when a class AB stage is strongly biased above the optimal bias situation.
Because junction temperatures move around a lot in an output stage, Vbe moves around quite a bit. The optimal bias is usually said to occur when about 26 mV is present at idle across each output emitter resistor. Now consider that Vbe changes at 2.2 mV/C. Even a 10C change in temperature (which will not be instantaneously be compensated) will introduce a change of 22 mV into Vbe of the output transistor. For this reason it is difficult to stay near optimum bias in the face of real-world program signal swings. For that reason, many designers prefer to err on the side of higher bias if they are willing to take care of the slightly higher idle temperature. The slight increase in distortion under mild gm-doubling conditions is usually considered to be far less sonically degrading that the crossover distortion that results when an output stage is under-biased.
Separately, the offset you mention is not an unreasonable thing to do, and people have been doing it with op amps in audio applications for years. Doug Self's Crossover Displacement (XD) output stage is not much more than applying this principle to an output stage.
However, achieving such offsets by mis-tracking of the power supplies is not the way to go. I'm not necessarily saying that precise tracking of the positive and negative rails is essential, but in any properly designed circuit if they happen to track perfectly that should never be a bad thing.
Cheers,
Bob
Simon,
Before anything else, I am not sure why you think de-balancing the power supply would translate (e.g.) the crossover distortions from zero. To me, it's clipping which would be de-balanced rather than the crossover distortions.
Secondly, strictly about zero crossing distortions, I am having some trouble following your thinking on this. So, are you claiming that e.g. crossover distortions are less "damaging" (to what?) if they do not occur around zero?
Looking in the frequency domain there seem to be a contradiction here. The sum of harmonics of a signal with a dead zone crossover distortion (like in the extreme case of a P/N complementary stage with zero bias) has a minimum precisely when the dead zone is around zero. That's because the time domain signal is symmetrical and an odd function of time, therefore the signal would have mostly odd harmonics. If the dead zone is not around zero, then the function is no longer odd (or even) and the harmonics are both even and odd. It is easy to show that the sum of harmonics is now larger.
I'm attaching two pictures. The first is the spectrum of a sine voltage having a dead zone of about +/0.6V around zero. The second is the spectrum of exactly the same signal, shifted by +0.25V. Which signal has overall larger harmonics?
Now, strictly from a Fourier transform perspective, the only impact of adding/shifting a signal by a constant is a Dirac delta function at zero frequency, while the signal spectrum itself is the same. So even in the perfect world of mathematics, perfect shifting from a perfect zero crossing would not have any positive impact on the signal spectrum itself.
So I still wonder why you think de-balancing/shifting the power supplies is a good idea. Is this some empirical result, based on perception rather than signal analysis?
Attachments
Last edited:
Bob,
My understanding of XD is that it actually avoids crossover distortions, by sinking/sourcing a current into the output, rather than displacing the crossover from zero to some +/- value.
This is also what people are trying to do (with dubious results) by connecting a resistor from an opamp output to one of the power supply lines, in an attempt to "linearize" the open loop response of the opamp output stage.
It doesn't necessary do any good. Sinking (or sourcing) only is not a consistent or guaranteed solution. Not that there's anything else that you can do, other than using the opamp as designed.
Crossover distortion is a function of output current, not output voltage. The XD circuit does exactly what you describe, but it does not avoid crossover distortion; it merely shifts its occurrence to a point where the net output current of the amplifier is non-zero (in the simple case by the amount of the pull-down current).
For example, if the pull-down current is a constant 1 amp, the crossover will still occur, but only if the (sinking) output current gets to 1A. Most of the time, with low-level program, the output signal current may not get this far, and if it does, the resulting crossover distortion occurs in the midst of higher-level program material. Notice also that the approach introduces an asymmetry into the circuit in that there is no crossover distortion for positive signal current swings.
The patent mainly covers circuit elaborations that make the process more efficient than just using a current source for pull-down.
Cheers,
Bob
Hi Bob
That is exactly how I understand the XD principle as well.
If somebody is interested I can scan the EW article from November 2006 where Douglas describes the principle.
Cheers
Exactly, but the power supply has no bearing with this mechanism. The power supply is a voltage source, delivering current on demand, not enforcing a current. My understanding from Simon's post was that he's claiming a voltage zero translation as being helpful.
Absolutely right syn08 and in my opinion it was a lot of other things that he stated in his post that doesn’t correspond to (at least) my view. In fact it was a rather confusing post.
- Status
- Not open for further replies.