Would that not defeat the point of the resistor? It's not there to establish a DC condition only, but to stabilize by gain reduction.
Best wishes
David
OK, in that case I can be less tactful
. It seems to me that in a conventionalcircuit it's a bad idea with no obvious pro and some cons.
Best wishes
David
I found that degeneration on a common-emitter VAS (or TIS?) stabilized with Miller comp could decrease stability. My reasoning was that the current going to the output stage through the miller capacitor is an element of positive gain in the stage, which is added to the negative gain of the transistor which usually is larger (except at UHF where the output stage gain is sufficiently low - we hope - to keep the positive gain from dominating feedback). As frequency increases, there is a point where gain cancels to zero and then becomes positive as it is dominated by the current in the miller cap, which at this frequency has a conductance higher than the transistor's transconductance. Degeneration lowers the transistor's negative gain which lowers the frequency at which total stage gain becomes positive, thus decreasing stability margin.
Another thing which worsens this is an overly large miller capacitor. The important factor it seems is Gm/Cmiller.
Another thing that may be important is that degeneration adds a resistance in series with Cbe of the transistor, but I haven't thought about this recently. It would cause a decrease in transconductance at RF because it limits the drive current available through Cbe.
I don't think degeneration is useful with the type of stage I describe here because it actually worsens all the characteristics of the transistor, including linearity as it increases Vcb variation.
I think this should also apply to a differential VAS.
Incidentally on third harmonic, for simple double EF BJT output amps, where the distortion contribution of the input and VAS stages is insignificant, the third harmonic can be reduced at 3A output with just a resistor from the output stage input to output. The reason is that the Vbe distortion imposed across this resistor creates a gain increase at high currents that cancels the Hfe droop of the outputs. This actually lowers total distortion even though open-loop gain is decreased. IE, open-loop linearity is increased. However IIRC this only applies to power levels where the 3rd harmonics dominates; below this point 2nd harmonic is increased.
Another thing which worsens this is an overly large miller capacitor. The important factor it seems is Gm/Cmiller.
Another thing that may be important is that degeneration adds a resistance in series with Cbe of the transistor, but I haven't thought about this recently. It would cause a decrease in transconductance at RF because it limits the drive current available through Cbe.
I don't think degeneration is useful with the type of stage I describe here because it actually worsens all the characteristics of the transistor, including linearity as it increases Vcb variation.
I think this should also apply to a differential VAS.
Incidentally on third harmonic, for simple double EF BJT output amps, where the distortion contribution of the input and VAS stages is insignificant, the third harmonic can be reduced at 3A output with just a resistor from the output stage input to output. The reason is that the Vbe distortion imposed across this resistor creates a gain increase at high currents that cancels the Hfe droop of the outputs. This actually lowers total distortion even though open-loop gain is decreased. IE, open-loop linearity is increased. However IIRC this only applies to power levels where the 3rd harmonics dominates; below this point 2nd harmonic is increased.
Last edited:
I think it is a valid naming convention. Two notes however:
1. Not all front end (VAS) designs can be easily split into TCS/TIS - if there is no cascode-like final stage in the VAS there is no clearly defined TCS/TIS division.
2. Only the upper (GB) transistor(s) in a final cascode stage can be properly called TIS, the rest is really a part of TCS.
Hi Forr,
I like mine even more.
That was a long time ago, about input inclusive compensation.
Nice you still remember the letter to the editor.
Cheers,
E.
Yes, it is funny to see the step response go the wrong direction initially. It is called a 'right half plane zero transfer function' behaviour in academic nerds' jargon.
What are they talking about Rocket Science or WHAT?
I hope the trickle down osmosis theory works for me, but it is great to watch for the folks that have expertise in other areas and talk over the heads of non experts. I am not being critical just blatantly honest
Carry on, I can hardly wait for the build of this symposium, I just hope it sounds better than what everybody expects...
The Right Half Plane zero is non-minimum phase behaviour, so the phase is more disturbed that it needs to be.
Bode discusses this in his book, in summary it is practically always undesirable in a feedback system.
Bob Cordell discusses practical examples on p.176 of his book. For a typical Cm of 30 pF and Re of about 22 ohms the zero is about 200 Mhz so not a real problem. The zero can easily be removed with a resistor in series with Cm if so desired.
Not a real expert, but hope that summary helps.
Best wishes
David
I should clarify that I think a small capacitor across the VAS transistor Re could add a bit of phase lead and potentially be helpful to stability. What I would call compensation, it's not to decouple.
Best wishes
David
Cherry claims the degen aids stability – he used a more detailed Admittance Matrix model of all of the amp stages together
with a high Z diff pair output ( = mirror ) the degen increases open loop VAS input Z by about the same as it reduces open loop gain so you have nearly a wash - the degen in this position doesn't necessarily "cost" much
with a high Z diff pair output ( = mirror ) the degen increases open loop VAS input Z by about the same as it reduces open loop gain so you have nearly a wash - the degen in this position doesn't necessarily "cost" much
Or possibly a series R with Cm to reduce the forward gain and counter-act the effect?
Thanks
-Antonio
For conventional Miller compensation my first impression is that the series R would be equivalent, and since post #393 says it's theoretically desirable anyway, and an R is cheaper than a C, then yes!
For Input Inclusive Miller Compensation as in Edmond's front end and some others then I think they are probably not equivalent. A question of some interest and I need to sim this.
I don't know about the impact of second order effects like collector/base capacitance modulation. I'd appreciate any educational comments.
Best wishes
David
Hi Edmound have you considered a different input vas interface doing "the same thing" like the attachement? Omitting the "mirror".
Attachments
With no degeneration the miller local loop is itself "degenerated".
The emitter degeneration will allow for the base/emitter capacitance
to be bootstrapped in respect of the base.
Without this degeneration the said Cbe will be seen as a short to ground
at high frequency , drasticaly reducing the local NFB where it is most needed.
A simple schematic help to understand more easily.
The emitter degeneration will allow for the base/emitter capacitance
to be bootstrapped in respect of the base.
Without this degeneration the said Cbe will be seen as a short to ground
at high frequency , drasticaly reducing the local NFB where it is most needed.
A simple schematic help to understand more easily.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.