"unity gain stable" is the op amp description - I don't consider the feedback factor a "condition" that varies - it is set by the designer, doesn't change short of magic smoke escaping
few audio power amp gain block circuits are "unity gain" stable - we often use gains of >20, it would be a waste of perfectly safe loop gain not to use circuits with some min stable gain > 1
generic control theory practice suggests 60 degree phase margin, 12 dB gain margin are OK for many feedback systems - at the feedback and amp gain defined unity loop gain intercept
we see in practice that many audio power chip amps are speced with min stable gain of ~10, Av 20 recommended
trivial examples show "uncondionally stable with any feedback" is meaningless - unity gain stable op amps oscillate fine in a Wien Bridge
so immediately you have to restrict the feedback to constant with frequency - but we see lead caps all the time...
few audio power amp gain block circuits are "unity gain" stable - we often use gains of >20, it would be a waste of perfectly safe loop gain not to use circuits with some min stable gain > 1
generic control theory practice suggests 60 degree phase margin, 12 dB gain margin are OK for many feedback systems - at the feedback and amp gain defined unity loop gain intercept
we see in practice that many audio power chip amps are speced with min stable gain of ~10, Av 20 recommended
trivial examples show "uncondionally stable with any feedback" is meaningless - unity gain stable op amps oscillate fine in a Wien Bridge
so immediately you have to restrict the feedback to constant with frequency - but we see lead caps all the time...
Last edited:
Historically 
conditional stable refers to a circuit that 'normally' is stable but could become unstable with certain loads, or certain source impedances, or temperature changes or combinations thereof.
There's a well-established mathematical derivation involving Smith charts and the like but that's for another post
For instance, power amps are often conditionally stable with respect to reactive loads, and that is cured by using an output inductor and/or a zobel to make it unconditionally stable with regards to reactive loads.
jan
conditional stable refers to a circuit that 'normally' is stable but could become unstable with certain loads, or certain source impedances, or temperature changes or combinations thereof.There's a well-established mathematical derivation involving Smith charts and the like but that's for another post
For instance, power amps are often conditionally stable with respect to reactive loads, and that is cured by using an output inductor and/or a zobel to make it unconditionally stable with regards to reactive loads.
jan
My working interpretation of unconditional stability is for the loopgain to have less than 180 degrees phase shift for frequencies lower than the crossover frequency. This says nothing of the phase and gain margins at the crossover frequency. Therefore I typically would consider load effects to directly pertube gain and phase margin.
Different points of view may extend this to require “unity gain stable” if for example you are designing a typical opamp where one may need to consider the open loop crossover frequency as the upper limit.
Thanks
-Antonio
Different points of view may extend this to require “unity gain stable” if for example you are designing a typical opamp where one may need to consider the open loop crossover frequency as the upper limit.
Thanks
-Antonio
Antonio,
It's probably best to put this in your sig line, so that others know what you mean when you use the term. Even if everyone puts his own interpretation on something, we still try to communicate here, no?
jan
its only useful if the amp gain can actually be reduced to 1 at the frequency the phase margin dip - by some practical operating "condition"
for SS the amp circuits can reach operating gain much faster than a single cycle of the potential oscillation frequency at turn on - not like tubes for which the criteria was developed
and the excess gain can be >30 dB at the phase margin dip - clipping can't get you that much "describing function" gain reduction without more input V than output for many audio power amps
for SS the amp circuits can reach operating gain much faster than a single cycle of the potential oscillation frequency at turn on - not like tubes for which the criteria was developed
and the excess gain can be >30 dB at the phase margin dip - clipping can't get you that much "describing function" gain reduction without more input V than output for many audio power amps
Last edited:
I am referring to the claim the phase margin at frequencies substantially below the unity gain intercept is important - I don't think it has much meaning with SS
this old definition of “unconditional stability” isn't especially useful unless you have tubes with slow heating cathodes that take seconds to go from zero gain to operating gm when turned on
higher order compensation has to still achieve the necessary margins at/near the UGLF, and with loading
but the phase margin say a decade or more lower in frequency can drop to 0 with 2-pole compendsations like TMC, and amps with 3rd order loop gains can be made to work despite negative phase margins at some range of lower frequency
higher order loop agin roll off compensations can make clipping recovery complicated, current limiters, clamps, windup control likely mandatory – but self sustaining large signal oscillation from “describing function” gain reduction can be avoided
this old definition of “unconditional stability” isn't especially useful unless you have tubes with slow heating cathodes that take seconds to go from zero gain to operating gm when turned on
higher order compensation has to still achieve the necessary margins at/near the UGLF, and with loading
but the phase margin say a decade or more lower in frequency can drop to 0 with 2-pole compendsations like TMC, and amps with 3rd order loop gains can be made to work despite negative phase margins at some range of lower frequency
higher order loop agin roll off compensations can make clipping recovery complicated, current limiters, clamps, windup control likely mandatory – but self sustaining large signal oscillation from “describing function” gain reduction can be avoided
jcx, post #611:
Nevertheless, during turn-on and without additional circuitry, it may burst into oszillation if small-signal gain-stages get into saturation and thus do not at all behave as simulated under steady-state conditions. This instability is also replicable in a transient simulation, if all supply/filtering circuitry is included into the simulation model and the two main DC voltage sources are suitably controlled (PWL option in LTSpice).
I wouldn't generally deny that such a thing might happen with a TMC/TPC design, for instance with a cascoded darlington 'VAS'.
Matze
In my experience, stability during turning on or off certain circuits may be a problem. I have designed an amplifier with nested Miller compensation (see elsewhere in the forum) that has nearly ideal phase and gain margins in all nested loops [and sounds marvelous ...].
Nevertheless, during turn-on and without additional circuitry, it may burst into oszillation if small-signal gain-stages get into saturation and thus do not at all behave as simulated under steady-state conditions. This instability is also replicable in a transient simulation, if all supply/filtering circuitry is included into the simulation model and the two main DC voltage sources are suitably controlled (PWL option in LTSpice).
I wouldn't generally deny that such a thing might happen with a TMC/TPC design, for instance with a cascoded darlington 'VAS'.
Matze
I thought so to earlier, but now I'm thinking that there is more benefit to Mr Self by his absence. Would seem there is potentially less bias this way.
Certainly a few topics have shown a lot of interest, higher order or sophisticated feedback loops leading the pack.
Just getting individual suggestions doesn't give him the same feedback or show general interest as much watching how these topics can take off on there own.
(so, yes I think Mr Self is still monitoring)
Thanks
-Antonio
Hi Damir,
I think your simulation of the loop around the OPS, that is created due to TMC, is incorrect. When connecting node "A" to the TMC resistor R30, you should break the global loop e.g. by connecting a 1F capacity from the LPT's inverting input to ground (C27 in the picture below). If you do not do that, then you see a mixed response from global loop and TMC loop.
The result shows a unity gain crossover of around 1.5MHz together with good stability margins.
My concerns about the VAS loop on its own (post #537, low phase margin) remain. I think it is a topic all too often neglected. People only consider it, when the VAS really starts to oscillate.
BR, Matze
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.