Base stopper resistors are necessary for output stage emitter followers, especially in output stages with several complementary emitter followers in parallel. Ideally, an emitter follower should not be driven from a low source impedance without either a ferrite bead or resistor in its base -- or both.
Read the following:
https://www.analog.com/media/en/technical-documentation/application-notes/an47fa.pdf
Read the following:
https://www.analog.com/media/en/technical-documentation/application-notes/an47fa.pdf
Last edited:
Emitter wire-wound power resistors are used to provide a voltage proportional to the output but in series opposing the driving voltage. This means, if the output current were to increase on its own, it will create a higher voltage across the emitter resistors which opposes spontaneous collector current increases. I do not think base resistors can help much in this situation apart from when a transistor's transfer ratio increases due to heat. However, even in this case, the voltage increase across an emitter resistor would oppose further collector current increase.
If I remember well, a reason for base resistors is to force the output transistor to share current equally. Forgive me if I claim this is already being done with wire-wound emitter resistors. Base resistors would slightly disrupt equal current sharing because they assume the transfer ratio for all output pairs is the same. In reality this is never guaranteed unless one pays very dearly to have transistors cut from the same silicon disc.
If I remember well, a reason for base resistors is to force the output transistor to share current equally. Forgive me if I claim this is already being done with wire-wound emitter resistors. Base resistors would slightly disrupt equal current sharing because they assume the transfer ratio for all output pairs is the same. In reality this is never guaranteed unless one pays very dearly to have transistors cut from the same silicon disc.
It's capacitive loads that emitter-followers don't like - you end up with a Colpitts oscillator.
The usual R-C Zobel and R-L series decoupling components usually present at the output of power amplifiers probably go a long way to avoiding spurious oscillation - even without base-stoppers. Though I always hate to see emitter-followers without them...
The usual R-C Zobel and R-L series decoupling components usually present at the output of power amplifiers probably go a long way to avoiding spurious oscillation - even without base-stoppers. Though I always hate to see emitter-followers without them...
Gate resistors are used to stop HF oscillation JLH was an advocate of them as was an Italian design engineer who designed mosfet amplifiers and had articles in EW .
Yes Local oscillations can be caused by a "wrong " value of compensation capacitor ----any changes made to a design must be checked for stability under load.
When designing my own circuits getting the value of the compensation capacitor right involved re-check after re-check to confirm overall stability .
You knew you did a good job when you only needed a low value of compensation capacitor ( no I am not a follower of 100pf comp. capacitors --its down to design + Layout ).
Yes Local oscillations can be caused by a "wrong " value of compensation capacitor ----any changes made to a design must be checked for stability under load.
When designing my own circuits getting the value of the compensation capacitor right involved re-check after re-check to confirm overall stability .
You knew you did a good job when you only needed a low value of compensation capacitor ( no I am not a follower of 100pf comp. capacitors --its down to design + Layout ).
PCB's designed & shipped with 200 khz Ft 2n3773 or 400 khz 2n3055, can oscillate with 4 mhz epitaxial transistors. As nigel7557 demonstrated with his mapes, 10 ohm base stopper resistors can stop this. Proper layout is key, but re-equipping a 50 year old amp with new pcb from ***** is not economic. those transformers & cases never wear out, though. Many of us reworking old units don't have layout software compatible PC's & Op systems, anyway.
I think the mechanism is that, toward the transition frequency, beta is imaginary. Capacitive reactance (to ground) in the emitter connection combines with the already-at-90-degrees gain, and you therefore get an overall phase of 180 degrees - so input resistance now looks negative (like a tunnel diode).
Cured by adding a base resistor to counter the negative and make the input circuitry "see" a positive resistance. Thus preventing oscillation.
Similar mechanism for a valve (tube), and similar cure.
Cured by adding a base resistor to counter the negative and make the input circuitry "see" a positive resistance. Thus preventing oscillation.
Similar mechanism for a valve (tube), and similar cure.
The choice of the frequency compensation capacitor or network is certainly not a matter of layout, and it should not be a matter of trial and error. It is just a matter of causing the beneficial migration of the dominant poles in the forward-path of the amplifier away from each other, and these poles have nothing to do with layout. I am assuming, of course, that Miller compensation or its derivatives is used.
Did I not say I DON'T use D.Self,s 100pf comp.capacitor ?
I am not a D.Self follower and layout is very important its obvious you use virtual design ---I don't -----experience over the years in Practical--real life building not looking at virtual reality has proved to my own satisfaction that I am right.
I am not a D.Self follower and layout is very important its obvious you use virtual design ---I don't -----experience over the years in Practical--real life building not looking at virtual reality has proved to my own satisfaction that I am right.
Considering beta of output stage decreasing with frequency its open loop impedance will rise with frequency - btw this effect is documented in the data sheets of most opamps. Impedance rising with frequency is by nature inductive - and if you load the output with a capacitor the resonant circuit is complete. To gain stability this resonant circuit must be dampened using some non-inductive resistors. For instance by a boucherot, or gate resistors, or non-inductive source resistors.
And yes, this is a local instability and as such should be treated locally instead of fiddling with the global loop compensation.
And yes, this is a local instability and as such should be treated locally instead of fiddling with the global loop compensation.
I apologise for assuming you use the superior Miller compensation or its derivatives. If your choice of compensation in your designs is sensitive to layout, then it must be a very eccentric form of compensation indeed. I would certainly have nothing to do with such a design, but to each his own.
Amazing how this "eccentricity " allows me to design circuits with compensation of 20pf/30pf/40pf and very low noise floors and also using JLH ,s feedback where it doesn't affect the slew rate.
I do admit to being "eccentric " as a non conformist but that's just me , eccentricity is a positive personality feature in England .
I do admit to being "eccentric " as a non conformist but that's just me , eccentricity is a positive personality feature in England .
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.