Hi Bob, Nelson
post #940
I assume that the compensation under discussion is the classical phase-lead but taken not from the output but the VAS stage collector to the feedback point?
If so, it is the traditional compensation scheme discussed in servo loops etc, and was used in Bailey's 1968 amplifiers. John Linsley Hood also used it.
cheers
John
post #940
I assume that the compensation under discussion is the classical phase-lead but taken not from the output but the VAS stage collector to the feedback point?
If so, it is the traditional compensation scheme discussed in servo loops etc, and was used in Bailey's 1968 amplifiers. John Linsley Hood also used it.
cheers
John
estuart said:
yes the only point I was illustrating was a method to establish a DC operating current in the VAS with the current mirrored complementary diff input topology
The VAS collectors are basically AC short circuited by the bias stage (and explicit bypass/damping in the sim), any gain imbalance should be resolved by the usual differntial loop gain operation - worst case one branch has no gain = constant current bias of the other branch of the VAS
pooge said:
Hi Pooge,
I have to disagree a bit with some of what you have said.
On page 338 of his paper, Leach discusses why the output impedance (inductance) of the amplifier goes up (by a factor of 4 or 5) when his feedforward scheme is employed. In his speculation, he states "It was concluded that the change in output impedance was caused by positive shunt-shunt feedback from the V3 node [amplifier output] to the V2 node [VAS output] in Figure 2." He further states, "this positive feedback could be minimized by choosing the V2 node to be a lower impedance point in the circuit.."
Unfortunately, his speculation about the cause of the impedance rise is wrong, and the solution he suggests doesn't change things much.
First, look at the circuit and see that the possible loop gain of any such positive feedback is very low, on the order of less than 0.1, due to the attenuation encountered by the "positive feedback" signal when it hits the input feedback shunt resistor R4.
So, what is the cause of the impedance rise? Simple: reduced negative feedback around the output stage, because it is being bypassed at high frequencies. One needs to understand that just because there is a break frequency at some point, things don't suddenly change. They can begin to change long before the break frequency.
In his amplifier, he has about 26 dB of NFB at 20 kHz, with a closed loop bandwidth of 400 kHz. On page 336, he states that his break frequency of his feedforward transition is at 100 kHz. This means that there is essentially no effective NFB around the output stage at 100 kHz, and that the effective NFB around the output stage will increase at approximately 6 dB per octave as frequency is decreased from 100 kHz. You can see just this in his Figure 5 plot of output impedance as a function of frequency.
This means that, with his feedforward in place, he only has about 14 dB of NFB around the output stage at 20 kHz, rather than the 26 dB he had without the feedforward technique in place. This is thus a reduction of 12 dB in the amount of negative feedback around the output stage. This corresponds to a factor of four. This is about the same as the factor in the rise in output impedance, and will also be about the amount by which his HF distortion due to the output stage (which probably dominates overall) will rise.
Leach is a pretty sharp guy, and I'm surprized that he did not pick up on this.
Cheers,
Bob
Bob Cordell said:
Hi Bob
I have no idea why that should be inferred from what I said.
Bob Cordell said:
Have you ever dabbled with the more complex single-ended fully complementary dual differential pair designs which don't use simple resistive loading? I think you might be surprised.
Cheers,
Glen
john_ellis said:
Hi John,
Yes, that is correct. However, when the capacitor is taken from the VAS back to the input stage (instead of from the output to the input stage), I don't think it is functioning as phase lead feedback or should be called that. I don't know who invented it, and honestly don't know what it has been called.
I just think of it as dominant-pole compensation where the dominant pole is formed by an integrator that just happens to enclose the input stage. Its just a different way to make a Miller integrator. The advantage is that the feedback through this compensation greatly increases the signal-handling capability of the input stage at high frequencies.
I guess this just shows that there is more than one way to look at a given architecture.
Cheers,
Bob
Bob Cordell said:
Well you made a general statement that the input/VAS topology of your 50W amp outperforms those of the full complementary dual differential pair configuration, so how specific do I have to be?
BTW, where have you been the last couple days?
I've been doing other things over the weekend. Is this a problem??
Cheers,
Glen
Hi, Glen,
What do you think of 1 transistor being driven by 1 opamp like in here :
http://www.diyaudio.com/forums/showthread.php?postid=1235226#post1235226
Is it a good approach to get a better output stage?
What do you think of 1 transistor being driven by 1 opamp like in here :
http://www.diyaudio.com/forums/showthread.php?postid=1235226#post1235226
Is it a good approach to get a better output stage?
jcx said:
Hi JCX,
"worst case one branch has no gain = constant current bias of the other branch of the" ??
Sure, if the compensating cap between base and collector of the VAS is very low (or zero),
but otherwise you might have a problem. Bod Cordell explains why:
http://www.diyaudio.com/forums/showthread.php?s=&postid=1119659&highlight=#post1119659 post #145
Cheers, Edmond.
Bob Cordell said:
Bob, I asked a rather simple straight forward question:
"Have you ever dabbled with the more complex single-ended fully complementary dual differential pair designs which don't use simple resistive loading?"
Since you are asking for schematics, the answer is obviously "no".
Cheers,
Glen
I think it's more likely that Bob (along with me), has no idea what topology you're reffering to. And since no one knows what topology you're are reffering to no one here knows whether or not they have tried it before.
As far as I know "single-ended fully complementary dual differential pair designs which don't use simple resistive loading" is not a standard term,
As far as I know "single-ended fully complementary dual differential pair designs which don't use simple resistive loading" is not a standard term,
Tim__x said:
Well I think that at least Bob should know what I am talking about, since we were discussing the virtues and vices of “fully complementary dual differential pair” (Bob’s terminology) input stages previously.
Bob pointed out his dislike for such an input arrangement because the LTP’s are typically loaded with a fixed resistor (eg: the Leach amp and the previously discussed Hafler amps), which in turn biases each VAS transistor. Fair enough. I pointed out the fact that the performance of the “fully complementary dual differential pair” can be improved with the inclusion of current mirrors, and at least one other contributor here elaborated a bit on this also, with the presentation of schematics a lot fancier than mine. I guess that this was largely ignored.
By “single-ended” I simply meant single-ended as in Bob’s 50W amp, as opposed to differential (bridged designs) of which I provided examples of in a previous post.
Bob stated that he thinks that his input arrangement performs better. With regards to the Leach / Hafler designs, I agree fully. With regards to some of the more advanced “fully complementary dual differential pair” variations, I do not think that such a general statement can be made.
G.Kleinschmidt said:
Hi Glen,
I thought you were interested in my opinion and experience with such a design. Dabbling can certainly include evaluation and simulation. To be clear, I have never built an amplifier with dual complementary input pairs, because from Day One I did not think that they were the best way to go. Others will disagree with that opinion, and that is fine. Many very good amplifiers have been built with them.
Glen, I am not a mind reader and am losing patience with this game of cat and mouse you are playing on this. I still do not know what topology you are specifically referring to. God bless John Curl if he really knows exactly what you are talking about.
However, if you are referring to the dual complementary differential input stage arrangement you use in your 12W amplifier, I do not think it is a very good design, but I have not bothered to simulate it, since it appears to be sub-optimal by inspection.
If, on the other hand, you are referring to servo schemes that make it possible to use current mirror loads and still achieve VAS standing current stability, I am familiar with the discussions on those that have taken place. They can probably be made to work, and good performance is probably possible, but I just don't think it is worth the complexity and the addition of yet another feedback loop when the day is over.
Bob
Glen, it has been my experience that all else being equal, that complementary differential has about 1/2 the distortion compared to a single differential design. This can be up to 5 times compared to a single diff pair with a single second stage device and a current source as a load.
I have and do build both, but I personally prefer to make overall simpler circuits than you or Bob prefer, if and when possible.
When necessary, such as making a line driver for Sound Technology 25 years ago, that had to have a distortion at 100KHz below .001%, I used an input circuit similar to Bob's, with cascodes everywhere to reduce even small traces of distortion, but I maintained a 100KHz open loop bandwidth, so I used resistive loading on my inputs and 2 pole compensation.
In this case I chose the single diff. pair design to have less nonlinear input capacitance to worry about.
I have and do build both, but I personally prefer to make overall simpler circuits than you or Bob prefer, if and when possible.
When necessary, such as making a line driver for Sound Technology 25 years ago, that had to have a distortion at 100KHz below .001%, I used an input circuit similar to Bob's, with cascodes everywhere to reduce even small traces of distortion, but I maintained a 100KHz open loop bandwidth, so I used resistive loading on my inputs and 2 pole compensation.
In this case I chose the single diff. pair design to have less nonlinear input capacitance to worry about.
john curl said:
Hi John,
This sounds reasonable to me. With different designers and philosophies, of course, all things are often not equal. One question I do have is this: in your experience cited above, are you referring to a single differential pair driving a single-ended VAS (e.g., as loaded with a current source), or are you referring to a single differential pair driving a push-pull VAS (of which my approach is only one of many ways of doing this)?
Thanks,
Bob