Bricolo said:
Connecting the compensation capacitor from the input of the second stage to the output, (instead of from the input of the second stage to the second stage collector), ensures that as foward path gain rolls off as a result of the compensation, local feedback through the compensation capacitor increases across the second stage (TIS), and the output stage.
This in principal greatly reduces the usual rise in THD from the output stage that occurs as foward path gain is rolled off by a compensation capacitor that is otherwise connected to the TIS collector.
In practice however, including the output stage in the compensation loop causes instability within this loop, particularly if a low-ft power BJT output stage is so enclosed.....
Power MOSFETs on the other hand usually have much wider banwidth than power BJTs....and local loop transmission within the minor loop tends to fall well below unity before non-dominant singularities introduce significant excess phase shifts....
Bricolo said:
....splitting the cap as described to accomodate a power BJT output stage, results in less loop transmission for that part of the minor loop enclosing the output stage...
While this gaurantees local stability, it deprives the output stage of the local loop transmission required to effect a significant reduction in HF THD....
AKSA said:
I think you are the first to bring PSPICE up in this discussion..
...and ....No......PSPICE is extemely usefull within its limitations, at getting the designer through first-base...viz:the 'what-if' phase of design....
AKSA said:
Viz: phase lead compensation...works well in respect of stability, but increases HF THD dramatically if the phase lead capacitor is refered to the TIS collector rather than the output.....
However, connecting the lead capacitor to the TIS collector does effectively isolate the input stage from RF pickup from output leads......
Usually, contrary to popular belief, phase lead compensation is desirable, but does not entirely gaurantee stability if used exclusively, .....
.....may need to be used together with miller compensation for 'unconditional' stability with all possible loads....
Electron said:
First of all, you need to use a type of power FET called a V-FET if you want to get linear operation. HEXFET's are very nonlinear, by way of it construction.
speaking of construction, miller capacitance is inherent to the PN junctions of BJT's.
MosFETs have capacitance between the gate - source channel. This can make them harder to drive at high freq. V-FET's are good for low Z sub woofer amps; MOSFETs have a negative temp. coefficient.
Re: Re: PWR stage in miller loop
I am aware of lateral fets (L-fets) used typically in amplifier circuits.,but what are the V-fets?
Please, could you explain how the Cgs and temp, coefficient effect to the idea of including power stage into miller loop ?
Or do you mean that there need to be more driving current for fets when they are in the Miller loop, to prevent loop internal overdrive (causing local TIM distortion) ?
I have used IRF641/9641 in my own designs with good results. I used them in hybrid CFP power stage which hides their unlinearity and thermal (problem) well. I am going to include the power stage into miller loop and see if I can keep some how the amp stable.
cunningham said:
I am aware of lateral fets (L-fets) used typically in amplifier circuits.,but what are the V-fets?
Please, could you explain how the Cgs and temp, coefficient effect to the idea of including power stage into miller loop ?
Or do you mean that there need to be more driving current for fets when they are in the Miller loop, to prevent loop internal overdrive (causing local TIM distortion) ?
I have used IRF641/9641 in my own designs with good results. I used them in hybrid CFP power stage which hides their unlinearity and thermal (problem) well. I am going to include the power stage into miller loop and see if I can keep some how the amp stable.
Re: Re: Re: PWR stage in miller loop
Hybrid MOSFET-BJT CFP power stage unlikely to be stable with inclusive single-pole miller compensation.....No harm in trying though....
keeeping it simple works best methinks....
try cross-coupled complementary BJT-follower drivers...for complementary MOSFET output stage.....stability more predictable...
Electron said:
Hybrid MOSFET-BJT CFP power stage unlikely to be stable with inclusive single-pole miller compensation.....No harm in trying though....
keeeping it simple works best methinks....
try cross-coupled complementary BJT-follower drivers...for complementary MOSFET output stage.....stability more predictable...
Hi Everyone,
I've just found Electrons thread.
I cannot agree with Sreten's Post 4.
A Miller connected VAS C.dom degrades a VAS stage's capability for driving an output stage because it slew limits its correction capabilities; it not only *increases* output-stage output-impedance but renders the entire amplifier inductive, typically above 500Hz. It also increases high audio frequency distortion, this being due to the 90 degree *leading* VAS charging current flow, typically through what were previously balanced input transistors !!!
____________________________________________________
Hi Bricolo.
Watch out for what you read in textbooks.
Beginners have to start somewhere and unfortunately many textbooks mislead with regard to Miller connected VAS C.dom enforced power amplifier stability because they apply op-amp theory.
Op-amps are rarely called upon to drive highly reactive loudspeaker loads, and are rarely as slow as an audio power amp. This is where AKSA's 'hands/ears-on' work has relevence in Post 18, as do Mikeks stability comments in Post 21.
Search out Nelson Pass's A40.
This is relevant with regard to the Miller connected C.dom because it does not have one - instead it has the arrangement suggested by AKSA, with the flat and resistive output impedance that all amplifiers should have. A more lowly biased class-B circuit is likely to need Mikek's treatment though.
____________________________________________________
I was not aware that Mr Miller discovered and wrote about the *Miller Loop* ?
____________________________________________________
Is it not the case that any *Miller Capacitance* is going to always be a problem due to the way it delays/distorts the voltage correction of a global NFB voltage sensed output terminal error which is caused by loudspeaker back EMF generated current flow, when that NFB loop encloses a Miller degenerated sub-circuit because this causes amplifier correction current to lag the error, especially at high audio frequencies ?
Cheers ............. Graham.
I've just found Electrons thread.
I cannot agree with Sreten's Post 4.
A Miller connected VAS C.dom degrades a VAS stage's capability for driving an output stage because it slew limits its correction capabilities; it not only *increases* output-stage output-impedance but renders the entire amplifier inductive, typically above 500Hz. It also increases high audio frequency distortion, this being due to the 90 degree *leading* VAS charging current flow, typically through what were previously balanced input transistors !!!
____________________________________________________
Hi Bricolo.
Watch out for what you read in textbooks.
Beginners have to start somewhere and unfortunately many textbooks mislead with regard to Miller connected VAS C.dom enforced power amplifier stability because they apply op-amp theory.
Op-amps are rarely called upon to drive highly reactive loudspeaker loads, and are rarely as slow as an audio power amp. This is where AKSA's 'hands/ears-on' work has relevence in Post 18, as do Mikeks stability comments in Post 21.
Search out Nelson Pass's A40.
This is relevant with regard to the Miller connected C.dom because it does not have one - instead it has the arrangement suggested by AKSA, with the flat and resistive output impedance that all amplifiers should have. A more lowly biased class-B circuit is likely to need Mikek's treatment though.
____________________________________________________
I was not aware that Mr Miller discovered and wrote about the *Miller Loop* ?
____________________________________________________
Is it not the case that any *Miller Capacitance* is going to always be a problem due to the way it delays/distorts the voltage correction of a global NFB voltage sensed output terminal error which is caused by loudspeaker back EMF generated current flow, when that NFB loop encloses a Miller degenerated sub-circuit because this causes amplifier correction current to lag the error, especially at high audio frequencies ?
Cheers ............. Graham.
I got some questions about this topic.
1. Doug self write in his book that a parrarel R with Cdom will linearize the open loop. He uses 220k in pararel with Cdom. How important is this open loop linearization to the sound? Audible?
2. What is the phase shift effect of using Cdom?
3. I've seen a design that uses 1k resistor in place of Cdom, not using Cdom. Is this R have the same effect as Cdom?
4. In designs with K1058 and J pairs, usually the K1058 has small cap, like 22pf attached from gate to source or drain. Is this cap for local loop or what?
5. We use complementary output stage. Why is that the N part of audio power amp breakdown more often than its P part?
1. Doug self write in his book that a parrarel R with Cdom will linearize the open loop. He uses 220k in pararel with Cdom. How important is this open loop linearization to the sound? Audible?
2. What is the phase shift effect of using Cdom?
3. I've seen a design that uses 1k resistor in place of Cdom, not using Cdom. Is this R have the same effect as Cdom?
4. In designs with K1058 and J pairs, usually the K1058 has small cap, like 22pf attached from gate to source or drain. Is this cap for local loop or what?
5. We use complementary output stage. Why is that the N part of audio power amp breakdown more often than its P part?
MILLER.
Hi Lumanauw.
Re Q1
If you put 220k in parallel with typical C.dom this might possibly reduce THD at frequencies around 10-20kHz, because this can set up a natural cancellation of oppositely acting input stage and VAS stage characteristics when a steady sinewave is being amplified.
The THD at lower frequencies will be worse, and the dynamic response at 10-20kHz will also be slightly worse due to increased input stage loading, but as for audible changes I would not expect to hear any due to the leading Miller cap current flow still acting with leading dominance upon the input transistors.
I was seriously impressed by Andy C's coverage of this aspect under A-B-C in;-
http://www.diyaudio.com/forums/showthread.php?postid=419336#post419336
(He must have taken a real big breath.)
I wonder if D Self had noted a valueless drop in the steady sine-wave THD when he added 220k, and put the improvement down to his flat topping of the open loop frequency response that Andy has explained ?
Q2.
The phase shift is negligible with steady sinewave and resitor loaded investigation. BUT, the shift is within the NFB loop and it allows a bipolar output pairing to be loudspeaker back EMF driven through a portion of their conduction crossover bias potential before the NFB loop sensed difference can generate correction.
This does not happen with Mosfets.
Q3.
Not a C.dom if 1k.
Q4 and 5 - - not my territory.
Cheers ............ Graham.
Hi Lumanauw.
Re Q1
If you put 220k in parallel with typical C.dom this might possibly reduce THD at frequencies around 10-20kHz, because this can set up a natural cancellation of oppositely acting input stage and VAS stage characteristics when a steady sinewave is being amplified.
The THD at lower frequencies will be worse, and the dynamic response at 10-20kHz will also be slightly worse due to increased input stage loading, but as for audible changes I would not expect to hear any due to the leading Miller cap current flow still acting with leading dominance upon the input transistors.
I was seriously impressed by Andy C's coverage of this aspect under A-B-C in;-
http://www.diyaudio.com/forums/showthread.php?postid=419336#post419336
(He must have taken a real big breath.)
I wonder if D Self had noted a valueless drop in the steady sine-wave THD when he added 220k, and put the improvement down to his flat topping of the open loop frequency response that Andy has explained ?
Q2.
The phase shift is negligible with steady sinewave and resitor loaded investigation. BUT, the shift is within the NFB loop and it allows a bipolar output pairing to be loudspeaker back EMF driven through a portion of their conduction crossover bias potential before the NFB loop sensed difference can generate correction.
This does not happen with Mosfets.
Q3.
Not a C.dom if 1k.
Q4 and 5 - - not my territory.
Cheers ............ Graham.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.