Hi,
Read Dr Cherry, for ideas on differentiated loops.
For good sound most designers avoid miller compensation (VAS cb)
Also JLH shows; LTP cap bypass on collector load, LTP c -c cap bypass, LTP e-e cap bypass, output Thiel network, VAS to -ve input CR //C
Many show a //cap across the global NFB resistor.
Leach & many others use a very small miller comp, I suspect to minimise its effect on sound and also to boost speed.
Read Dr Cherry, for ideas on differentiated loops.
For good sound most designers avoid miller compensation (VAS cb)
Also JLH shows; LTP cap bypass on collector load, LTP c -c cap bypass, LTP e-e cap bypass, output Thiel network, VAS to -ve input CR //C
Many show a //cap across the global NFB resistor.
Leach & many others use a very small miller comp, I suspect to minimise its effect on sound and also to boost speed.
Hello.
A capacitor from the VAS collector to ground is not a good idea. It needs to be at least 10 nF, as opposed to the 50-200pF required when used collector-to-base (Miller compensation). The VAS and the LTP then need to run at a much higher current in order to drive this larger capacitance. The late Peter Baxandall said that this compensation scheme was "In all respects, sub-optimal"
John
A capacitor from the VAS collector to ground is not a good idea. It needs to be at least 10 nF, as opposed to the 50-200pF required when used collector-to-base (Miller compensation). The VAS and the LTP then need to run at a much higher current in order to drive this larger capacitance. The late Peter Baxandall said that this compensation scheme was "In all respects, sub-optimal"
John
Couldn't have said it better myself, Andrew T!
Often, if you're stuck with a Vas pole (after all it's a high gain stage) it's reasonable to use a small Miller C to 'define' it ,say 10pF and then use feedback comp and/or other pole zero cancellation. A fully complementary Vas can drive a fairly large C to ground without significant distortion AND it doesn't degrade PSRR.
Often, if you're stuck with a Vas pole (after all it's a high gain stage) it's reasonable to use a small Miller C to 'define' it ,say 10pF and then use feedback comp and/or other pole zero cancellation. A fully complementary Vas can drive a fairly large C to ground without significant distortion AND it doesn't degrade PSRR.
Hi Andrew T.
Yes certainly the nested approach is all about juggling poles and zeros - but, of course, we need them to be stable and defined and they in turn need to ensure their negative impact on their local block is minimal in terms of PSRR, THD, slewing rate, ....
Yes I would like to do that but I am also of the view that the forum's strength is it's mulling of ideas rather than offering up complete solutions. I'm still thinking this one out.
Cheers,
Greg
Yes certainly the nested approach is all about juggling poles and zeros - but, of course, we need them to be stable and defined and they in turn need to ensure their negative impact on their local block is minimal in terms of PSRR, THD, slewing rate, ....
Yes I would like to do that but I am also of the view that the forum's strength is it's mulling of ideas rather than offering up complete solutions. I'm still thinking this one out.
Cheers,
Greg
amplifierguru said:
Who said anything about bypassing the OP stage?
The impact is that there is less loop gain to linearize the OP stage.
I have seen this method is employed with the an output stage with local feedback such as a CFP. However that can open it's own can of works if not treated properly.
the vas must be stabilised more than any other stage especially against
p2 the reasons for this is in the details of the spice simulator that i have
programmed with my dos pascal software
phase shifts in this stage could be rendered negligible if only pcb is in
the equation of my spice simulator
cheers
p2 the reasons for this is in the details of the spice simulator that i have
programmed with my dos pascal software
phase shifts in this stage could be rendered negligible if only pcb is in
the equation of my spice simulator
cheers
Hi bscally,
I'll have to watch my phraseology in your presence. I meant that the loop gain was being redirected back to the input stage leaving less global NFB around the output stage for reduction of it's distortion. Same difference.
Hi Mastertech,
Yes - if your design has a Vas ( my latest doesn't - unless you include the front end chip), being a high gain stage it will need consideration of all it's foibles.
P1 - insert your zero here - P2
It cdr data,
I have activated my email button if you wish to contact me.
I'll have to watch my phraseology in your presence. I meant that the loop gain was being redirected back to the input stage leaving less global NFB around the output stage for reduction of it's distortion. Same difference.
Hi Mastertech,
Yes - if your design has a Vas ( my latest doesn't - unless you include the front end chip), being a high gain stage it will need consideration of all it's foibles.
P1 - insert your zero here - P2
It cdr data,
I have activated my email button if you wish to contact me.
darkfenriz said:
Hi Darkfenriz,
I think the disadvantage of this is that it increases the noise gain, and reduces the loop gain available for non-linearity reduction - basically you reduce the open loop gain at higher frequencies and thus the feedback gain. But it will work for sure.
Jan Didden
amplifierguru said:
Apologies if I was TOO pedantic.
IMO .. Now this may be a whole different topic...
Another reason for attempting to linearise the OP stage.. or run substantially in class A is that the high (positive) and low(negative) OP stages gain false off as the devices switch off.
At this point additional feedback is needed to linearise the stage than at other points. No FB is a wonderfull thing but it is only a refuction of the error not a cancellation. (Please I don't want to debate the if it's reduced to the point of imperceptability it doesn't matter argument )
For an AB stage this generates 2nd harmonic distortion that is level dependant. When in the A part of the operating curve this is not present. When in the B part it is.
The concept of AB is that this distortion is no swamped by the main signal AND there is tolerencing to the OPS bias network.
Meandering back to topic...
The aim here is stabalise an unstable thing...
First ask yourself why it is unstable different topoligies have differnt solutions. Each is a set of compromises in it's own right and as such each needs different stabalisation.
Now by definition we start with a nested feedback loop system. So we have to analyze it as such.
Brian
Hi mikeks,
How about some substantive discussion instead of uninformative derision
In nested feedback you have to apply Bode’s multiloop criteria, individual loops are permitted to be “unstable” without the overall amplifier being unstable – if you are referring to conditional Nyqusit stability, Cherry brings this up himself, the existence of physical prototypes with scope probe traces in his articles strongly suggest some range of stable operation
There are only few hundred million counter examples to your claims that output inclusive and nested compensation can’t work
NE5532/4 compensation is described in these terms in Eschausier’s 1995 book “Frequency Compensation Techniques for Low-Power Operational Amplifiers”
The IEEE has more than a few papers in the “nested miller” genra
Other op amps use output stage inclusive compensation
LT1028/1128/1115 has a unity gain buffer type output stage inside the Ccomp loop
all common emitter/source types have to use output connected compensation
the TDA7294 block diagram shows the Cherry stlye output stage inclusive compensation (I would call the output devices “power transistors” in this chip amp):
http://us.st.com/stonline/books/pdf/docs/1057.pdf
“A significant aid in keeping the distortion contributed
by the final stage as low as possible is provided
by the compensation scheme, which exploits
the direct connection of the Miller capacitor
at the amplifier’s output to introduce a local AC
feedback path enclosing the output stage itself.
“
How about some substantive discussion instead of uninformative derision
In nested feedback you have to apply Bode’s multiloop criteria, individual loops are permitted to be “unstable” without the overall amplifier being unstable – if you are referring to conditional Nyqusit stability, Cherry brings this up himself, the existence of physical prototypes with scope probe traces in his articles strongly suggest some range of stable operation
There are only few hundred million counter examples to your claims that output inclusive and nested compensation can’t work
NE5532/4 compensation is described in these terms in Eschausier’s 1995 book “Frequency Compensation Techniques for Low-Power Operational Amplifiers”
The IEEE has more than a few papers in the “nested miller” genra
Other op amps use output stage inclusive compensation
LT1028/1128/1115 has a unity gain buffer type output stage inside the Ccomp loop
all common emitter/source types have to use output connected compensation
the TDA7294 block diagram shows the Cherry stlye output stage inclusive compensation (I would call the output devices “power transistors” in this chip amp):
http://us.st.com/stonline/books/pdf/docs/1057.pdf
“A significant aid in keeping the distortion contributed
by the final stage as low as possible is provided
by the compensation scheme, which exploits
the direct connection of the Miller capacitor
at the amplifier’s output to introduce a local AC
feedback path enclosing the output stage itself.
“
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.