Basically, in this thread , I have presented three ways of connecting up an opto in an SMPS feedback circuit.
1. "common emitter"
2. "common collector"
3. "casc0ded"
When running each of these on the simulator, (and I have done this time and again over several years), it is *always* noticeable that getting a fast Vout response, as well as maintaining stability, is massively more difficult with the "common emitter" arrangement than the other two.
I can almost guarantee that nobody here (or anywhere else ) will be able to adjust the "common emitter" feedback SMPS simulation (the one attached earlier) to make it stop its vout from overshooting at start-up.
I think you're probably wrong about that. The emitter output and collector output configurations are fundamentally different in how the feedback is applied to the control IC, and I believe this is the reason for the different feedback loop response (i.e. the overshoot).
Have a look at the connections to pins 2 (voltage feedback) and 1 (compensation) in the two designs in your "Flyback_opto_CE_CC_compare" files. I'll use "VFB" and "COMP" to refer to these pins here.
In both versions, feedback through the optocoupler acts to reduce the output voltage. That is, the current in the optocoupler's LED is increased in response to increased output voltage, and the corresponding conduction in the optocoupler's transistor acts to reduce the output voltage. But the way the feedback is applied to the control IC is quite different.
In the emitter output ("CC") configuration, increasing conduction in the opto's transistor causes an increasing voltage at its emitter, which is coupled through a resistor into VFB, the inverting input of the error amplifier. Negative feedback is applied to the error amplifier by the paralleled resistor and capacitor from COMP to VFB. The main effect of this is to reduce the speed of the error amplifier's response, i.e. it slows down the control IC's response. Given that there are delays in the rest of the loop (from the control IC to the output voltage, and from the optocoupler's LED to its phototransistor output), I think this delay is enough of an explanation for the absence of overshoot.
You should also see that at startup, the output voltage rises more slowly than with the other design; this is an important clue. It will also respond more slowly to load current changes.
In the collector output ("CE") configuration, VFB is simply grounded, and the error amplifier has no effect. Feedback is applied to COMP, which is actually the output of the error amp, but this pin is not a standard op-amp output, as you can see in the diagram at the bottom of page 5 in the TI data sheet. The error amplifier effectively has an open collector output which can only pull down; with VFB grounded, this transistor is turned OFF and has no effect. A 500 µA current source pulls the compensation pin positive.
There are two diode drops between COMP and the following circuitry, so anything connected to COMP to provide feedback doesn't need to pull all the way down to 0V. This is quite a clever design, because it allows the feedback to be applied to either pin, according to its polarity. But when feedback is applied to COMP, instead of VFB, there is no way to slow down the control IC's response, because the error amp isn't used. The delays in the loop that I mentioned earlier are the reason for the overshoot; the control IC responds too quickly to the feedback when it is applied to COMP.
Of the "common collector" and "casc0ded" connection methods, the casc0ded method *always* offers more ease in getting fast transient response in vout whilst maintaining stability.
I'm sure that's true. The reason is the reduced change in the voltage across the transistor in the optocoupler, which reduces the Miller effect greatly and causes much faster response in the optocoupler. This is the reason for the cascode design in the first place.
BTW why are you putting a "0" in "cascode"? It's just a normal word.
The Renesas and Vishay say that the Miller effect is not relevant when we don't have an actual base connection......the Linear.com app note of post #7 definitely does point to the Miller Effect.
The Miller effect will be present whether or not the base is connected, and is the reason for the improved response with the cascode configuration. I can't see any reason why anyone would say otherwise. Can you link to the documents where they make this claim, and give me section or page references?
Christophe Basso, one of the greatest SMPS engineers that ever lived, refers to the connection methods as "common emitter" and "common collector" in pages 288 -308 of his book "switch" mode power supplys".
Back in post #17 on this thread I said:
If a document refers to the collector output topology as "common emitter" and the emitter output topology as "common collector" or "emitter follower", it's because the author hasn't realised that those names they aren't appropriate, or the author is using them as (misleading) shorthand for collector output and emitter output, because they are easily recognised.
I meant that the names "common emitter" and "emitter follower" (or "common collector") are well-known. But they are misleading, which is why I'm using "collector output" and "emitter output" here.
The really strange thing is, that Basso makes no mention of the fact that the "common collector" connection method offers greater ease in achieving fast transient response in vout, whilst simultaneously maintaining stability.
That's because it doesn't, as I explained earlier. The difference is because the feedback is applied to the control IC at a different point, and in the collector output configuration, the error amplifier is bypassed so the control IC's response cannot be slowed down.
Basso also does not speak of the Miller Effect being prevalent at all. (He doesn't even mention the Miller Effect.)
*shrug*
Adding a physical base resistor has been put forward, but this reduces CTR and CTR is low enough at the low photo diode currents used in modern SMPS, and also CTR degrades over time and temperature, so anything that reduces CTR is not welcome really..........
Optocouplers really aren't a very good way to pass feedback. The variations in CTR due to temperature, age, phase of the moon etc will also affect the response of the control loop significantly, hence the overcompensation used in the emitter output design. I'm sure someone who knows something about control loops could elaborate here
...so given that we can't use a base resistor, in the field of opto-coupled SMPS feedback, can we correctly make the sweeping statement that....
*'common collector' connection of the opto facilitates faster transient response and improved phase and gain margins over 'common emitter' connection of the optocoupler.* ?
No, I don't believe that's the reason for the difference.