Has anyone seen this front-end before?

[Waly]

Hi wahab,

I wasn't suggesting that you had changed your mind, I was just trying to get straight in my mind what it was that you were asserting when you stated that TMC was equivalent to TPC if a lead-lag network was added to TPC.

If I have got it straight now, you are just saying TPC then has no bump, but you are NOT saying that the closed loop frequency responses are exactly the same. Right?

Who wants to add a bandaid to TPC to make it almost the same as TMC? Why are you so stuck on TPC?

I'll look forward to seeing your simulations.

Cheers,
Bob

1. I am not wahab.

2. No. I am saying that TMC and TPC plus lead-lag are mathematically identical as much as (a+b)^2=a^2+2ab+b^2.

3. I am not stuck with TPC. I was (and still am) only irritated by the magic properties (jcx called it "oversold") attributed to TMC, when in fact it is just another form of two pole compensation, with an embedded additional phase correction.

It is though obviously more intuitive to speak about linearizing the output stage, then transitioning to Miller compensation at HF, than to speak in terms of maximum Bode feedback, linear(izing) phase, or Lurie integrals. I'll give TMC network the benefit of being more intuitive.

If Mike claims that TPC is better than TMC, or that optimal TMC and TPC may have identical element values, then you may call him about. I strongly disagree with both claims.

Over and out.

 

[PB2]

Found it on wayback, but not sure how to get to a specific thread:
DIY Audio Engineering - Index

I don't think anything other than the index is available there.

 

[vladn]

I've checked those books (of course) but nothing comes close. It's not about the double (i.e. complementary) input stage, rather the Baxandall super pair in the TIS (or VAS if you like, a misnomer btw)

Hi Edmond,

Few years ago I have experimented with the SE front end that used a Baxandall pair as the top "transistor" in a cascode TIS/VAS. (I believe I've sent you some version of the schematics at that time). I came up independently with that idea, but later found references in some old thread here discussing a similar concept. I guess nothing is completely new under the sun. See post #5 and later:
http://www.diyaudio.com/forums/solid-state/25172-baxandall-super-pair.html

I got very good linearity from a SE+servo arrangement (well below 0.1ppm at 20k and mostly in the 2nd harmonic), so decided not to explore the symmetric form. However such linearity required (i) running VAS CCS load from a bootstrapped rail and (ii) using Baxandall pairs in the follower stage after VAS to avoid effects of non-linear load at high frequencies.

I do like your symmetric arrangement though. I suggest trying a follower stage made of baxandall pairs. I think you will like the result

 

[Edmond Stuart]

Hi Edmond,

Few years ago I have experimented with the SE front end that used a Baxandall pair as the top "transistor" in a cascode TIS/VAS. (I believe I've sent you some version of the schematics at that time).

Hi Vladimir,

Although I couldn't find that schematic right away, I think it was about about a VAS with a kind of built in error correction, right?
I also remember our fruitful discussions, where we exchanged many variations on the AB2-OPS.

I came up independently with that idea, but later found references in some old thread here discussing a similar concept. I guess nothing is completely new under the sun. See post #5 and later:
http://www.diyaudio.com/forums/solid-state/25172-baxandall-super-pair.html

There is indeed nothing new under the sun, that is, at first sight. There is, however, a small though important difference: The TIS has no (current) gain. BTW, in this respect it might be more appropriate to call it a differential current mirror. Anyhow, the lack of gain eases the implementation of input inclusive compensation (less prone to oscillations) and controlling the quiescent current of the TISes, because no CMCL or whatsoever is needed.

I got very good linearity from a SE+servo arrangement (well below 0.1ppm at 20k and mostly in the 2nd harmonic), so decided not to explore the symmetric form. However such linearity required (i) running VAS CCS load from a bootstrapped rail and (ii) using Baxandall pairs in the follower stage after VAS to avoid effects of non-linear load at high frequencies.

I do like your symmetric arrangement though. I suggest trying a follower stage made of Baxandall pairs. I think you will like the result

Um... I think I already did just that (or at least something which comes close to that), "see fig.17. SuperTIS combined with the Hawksford-Cordell error correction output stage" on my website , where I've put a diamond buffer between the front end and the HEC-OPS. If you mean something else, please correct me.

Cheers,
E.

 

[Edmond Stuart]

TMC vs TPC

Hi Waly,
Have you simulated any amplifiers with both TMC and TPC compensation approaches?
Cheers,
Bob

Hi Bob,

It already has been done. Moreover, I believe it was a fair apples to apples comparison. Please look here , here and here.
Regarding TPC, notice that I have drawn two additional compensations in the schematic: C7 & R29, respectively C8 & R30. Either one of the two is needed to compensate for the 'bump'.
For simulation, however, I've used only the latter (because I know you don't like lead compensation between the output and the non-inverting input of the IPS ).

Although the differences in results were practically zero (at least in this case), it doesn't mean that I'm indifferent on using either TPC or TMC. I still tend to prefer TMC. Why? just read my website.

Cheers,
E.

 

[Bob Cordell]

1. I am not wahab.

2. No. I am saying that TMC and TPC plus lead-lag are mathematically identical as much as (a+b)^2=a^2+2ab+b^2.

3. I am not stuck with TPC. I was (and still am) only irritated by the magic properties (jcx called it "oversold") attributed to TMC, when in fact it is just another form of two pole compensation, with an embedded additional phase correction.

It is though obviously more intuitive to speak about linearizing the output stage, then transitioning to Miller compensation at HF, than to speak in terms of maximum Bode feedback, linear(izing) phase, or Lurie integrals. I'll give TMC network the benefit of being more intuitive.

If Mike claims that TPC is better than TMC, or that optimal TMC and TPC may have identical element values, then you may call him about. I strongly disagree with both claims.

Over and out.

Hi Waly,

I apologize to you for getting your name wrong. I apologize to wahab as well. It was late.

As I said above, I'll look forward to seeing your simulations.

Cheers,
Bob

 

[vladn]

Um... I think I already did just that (or at least something which comes close to that), "see fig.17.

Sorry, I missed that circuit diagram, you are right.

There is, however, a small though important difference: The TIS has no (current) gain.

Correct me if I am wrong but to me it seems more like a semantics question. The trans-impedance stage in your circuit (as well as mine and the one in the early thread I referenced) is strictly a Baxandall pair providing unity current gain. There are indeed differences in the current output amplifier preceeding the TIS (and yours is very clever in many respects), but conceptually I think it does not matter if it is a folded cascode or straight cascode. The general topology is a current output amplifier followed by the unity current gain Baxandall TIS.

 

[Bonsai]

I would agree TIS is a better term for what we generally call the VAS

 

[Edmond Stuart]

unity current gain

[..]
Correct me if I am wrong but to me it seems more like a semantics question. The trans-impedance stage in your circuit (as well as mine and the one in the early thread I referenced) is strictly a Baxandall pair providing unity current gain. There are indeed differences in the current output amplifier preceeding the TIS (and yours is very clever in many respects), but conceptually I think it does not matter if it is a folded cascode or straight cascode. The general topology is a current output amplifier followed by the unity current gain Baxandall TIS.

Hi Vladimir,

I agree with your remarks and it wasn't my intention to start a semantic debate. As I don't have your schematic at hand*, I didn't know your circuit also has a unity current gain.
Regarding the circuits of other thread (on the Baxandall Super Pair), to me it seems that they are intended to provide current gain, hence my remark
about differences in gain. I'm sorry if I wasn't clear enough.

Cheers,
E.

PS: * could you mail your circuit once again, because I couldn't find it any more.

 

[Edmond Stuart]

TCS * TIS = VAS

I would agree TIS is a better term for what we generally call the VAS

Someone else here (sorry, I forgot your name) remarked that it is actually the whole front-end (voltage in, voltage out) that should be called the 'VAS'.
Since the front-end comprises two stages and the 1st one is a TCS, then, by definition, the 2nd one must be a TIS.

Cheers,
E.

 

[Dave Zan]

Someone ... remarked that it is actually the whole front-end (voltage in, voltage out) that should be called the 'VAS'.
Since the front-end comprises two stages and the 1st one is a TCS, then, by definition, the 2nd one must be a TIS.

Cheers,
E.

That would be me in post #172.
That re-raises the question that I asked at the time but perhaps didn't make clear. - In Fig.1 why is the TIS characterized by a current gain and Gm. Should it not just be a transimpedance?
That's how it's treated by, say Hawksford in the improved cascode paper, or Dymond and Mellor in the two-pole compensation paper.

Best wishes
David

 

[Edmond Stuart]

That would be me in post #172.#

That was you indeed. (I couldn't find your posting that fast)

That re-raises the question that I asked at the time but perhaps didn't make clear. - In Fig.1 why is the TIS characterized by a current gain and Gm. Should it not just be a transimpedance?
That's how it's treated by, say Hawksford in the improved cascode paper, or Dymond and Mellor in the two-pole compensation paper.

Best wishes
David

The point is that a typical VAS does have and must have a degen. resistor* in series with the emitter. As result, it has a fairly high input impedance and without Cdom it behaves like a VCCS, i.e. a TCS. Since gm does matter, I also have specified it. But together with Cdom it becomes a TIS, because now (mainly) Cdom defines the I/O properties (current in, voltage out) . Rather confusing, isn't it?

The SuperTIS (fig.3), on the other hand, has -intentionally- a low input impedance, with or without Cdom. How low exactly, doesn't matter, so I didn't specify gm for this case.

* as pointed out by Ed Cherry

Cheers,
E.

 

[wahab]

The point is that a typical VAS does have and must have a degen. resistor* in series with the emitter. As result, it has a fairly high input impedance and without Cdom it behaves like a VCCS, i.e. a TCS. Since gm does matter, I also have specified it. But together with Cdom it becomes a TIS, because now (mainly) Cdom defines the I/O properties (current in, voltage out) . Rather confusing, isn't it?

Wether to use a said function or rather its reciprocal in describing
a behaviour is just a matter of convenience...

 

[Dave Zan]

...to use a said function or rather its reciprocal... is just a matter of convenience...

The Gm specified (25milliS) is not the reciprocal of the transimpedance.
A reasonable reciprocal DC transimpedance should be a few microS at most, very much less for this circuit I think.

Best wishes
David

 

[Edmond Stuart]

Wether to use a said function or rather its reciprocal in describing
a behaviour is just a matter of convenience...

Hi Wahab,

I'm afraid you missed the point.
Said VAS has a transconductance (without Cdom) as well a transimpedance (Cdom included). But the first is not the reciprocal of the latter.

Cheers,
E.

 

[wahab]

Hi Dan and Edmond,

Agree on the matter strictly speaking , i somewhat streched
the words to point that there s different possible modelisations.

 

[Bob Cordell]

That was you indeed. (I couldn't find your posting that fast)



The point is that a typical VAS does have and must have a degen. resistor* in series with the emitter. As result, it has a fairly high input impedance and without Cdom it behaves like a VCCS, i.e. a TCS. Since gm does matter, I also have specified it. But together with Cdom it becomes a TIS, because now (mainly) Cdom defines the I/O properties (current in, voltage out) . Rather confusing, isn't it?

The SuperTIS (fig.3), on the other hand, has -intentionally- a low input impedance, with or without Cdom. How low exactly, doesn't matter, so I didn't specify gm for this case.

* as pointed out by Ed Cherry

Cheers,
E.

Hi Edmond,

Not to further confuse semantics, but I guess I will anyway. I like to call the "middle" stage the VAS because it is where the large voltage for driving the output is usually developed. I don't call it a Voltage Amplifier Stage because it is strictly a voltage-in, voltage-out stage. As we know, virtually any VAS is more complex than that, especially when we put a compensation capacitor around it. Over frequency and over architecture, a VAS may transition among various operating modes, depending on what dominates the impedances at the input and output.

Why did Ed Cherry say a VAS MUST have a degeneration resistor? Although I always use one, I don't think it is strictly necessary for the stage to function. Or did he mean that it must have a degeneration resistor to qualify to be called a voltage amplifier stage (but even that would be wrong, since a VAS could have its voltage gain defined by re against a physical collector load resistor).

I like IPS, VAS, OPS, even though none of these is semantically perfect. At least people usually know what we mean by using these common terms. Of course, then there is the folded cascode that doesn't quite fit this model. Even with the folded cascode, my preference is to define the amplifier as having three stages, where the first stage is transconductance into the emitters of the folded cascode, the "VAS" comprising the folded cascode and its associated current mirror. That is a somewhat imperfect fit to my model that the VAS is where the voltage is made.

Cheers,
Bob

 

[Edmond Stuart]

from VAS fighting to semantic fighting

Hi Edmond,

Not to further confuse semantics, but I guess I will anyway. I like to call the "middle" stage the VAS because it is where the large voltage for driving the output is usually developed. I don't call it a Voltage Amplifier Stage because it is strictly a voltage-in, voltage-out stage. As we know, virtually any VAS is more complex than that, especially when we put a compensation capacitor around it. Over frequency and over architecture, a VAS may transition among various operating modes, depending on what dominates the impedances at the input and output.
Why did Ed Cherry say a VAS MUST have a degeneration resistor? Although I always use one, I don't think it is strictly necessary for the stage to function. Or did he mean that it must have a degeneration resistor to qualify to be called a voltage amplifier stage (but even that would be wrong, since a VAS could have its voltage gain defined by re against a physical collector load resistor).

Hi Bob,

Maybe you have missed it, but somewhere else I wrote:

According to Ed Cherry(1), this emitter series resistor is essential to maintain stability, as opposed to what D. Self(2) says about it.
[..]
(1) "Feedback, Sensitivity, and Stability of Audio Power Amplifies", JAES, Vol.30,No.5, 1982 May, pp. 282-294.
(2) "Audio Power Amplifier Design Handbook", 5th edition, pp. 124-125: Note that the emitter resistor R2 is not present to introduce local negative feedback; it is normally put there to allow current-sensing and also over-current protection of the VAS transistor when output stage overload circuitry is operating.

Also my simulations confirm that a degen. R is indispensable to avoid local instability of the Miller loop.

I like IPS, VAS, OPS, even though none of these is semantically perfect. At least people usually know what we mean by using these common terms. Of course, then there is the folded cascode that doesn't quite fit this model. Even with the folded cascode, my preference is to define the amplifier as having three stages, where the first stage is transconductance into the emitters of the folded cascode, the "VAS" comprising the folded cascode and its associated current mirror. That is a somewhat imperfect fit to my model that the VAS is where the voltage is made.

Cheers,
Bob

Sure, that's where the voltage is made, but still the input is a current, so......
BTW, any idea who was the first who dubbed this stage a 'VAS'?

Cheers,
E.
__________________

 

[wahab]

Also my simulations confirm that a degen. R is indispensable to avoid local instability of the Miller loop.

I also observed the thing in sims.
The drawback , negligible , of degenerating the VAS is higher
third harmonic , at least in the so called blameless.

BTW, any idea who was the first who dubbed this stage a 'VAS'?

I guess it date back to the tube era when this stage was
actually a voltage amplifier stage...

 

[Bonsai]

You can always decouple the emitter degn resistor with a big cap.