Using UC3825 and the function of EAOut

Hi everyone;

I am planning to design 100kHZ pwm circuit for half bridge converter but there is some confusion for me in this circuit that I showed below. As I understand from this circuit, Vref=5.1v and two 4k99 resitor divide it to 2.55V and it connect to noninverting input. When the inverting input(pin1) is higher than noninverting input(pin2), duty cycle will be decreased means that ic will not give pwm out.

But is it right? I do not understand function of eout(pin3). Can anybody explain it?

Rout is the error amplifier output, in a typical 3825 application used for feedback frequency compensation. The present circuit has an error amplifier on the secondary side and thus cuts down the internal error amp to -1 gain by resistors.

FvM said:

Rout is the error amplifier output, in a typical 3825 application used for feedback frequency compensation. The present circuit has an error amplifier on the secondary side and thus cuts down the internal error amp to -1 gain by resistors.

Click to expand...

But how it works? When the transistor of optocoupler is triggered, the collector current is 2.43mA(5.1V-0.24V=4.86V and 4.86V/2K=2.43mA). But what will be INV input? It must be higher than 2.55V to turn off PWM output or decreased duty cycle. If the Vce=0.2V, 5.1V-0.2V=4.86V/2=2.43V(INV input). Is it right? But how can I determine EAout voltage or relatiion?

The error amplifier has a certain amount of gain. To make it work properly you need to give it enough gain so it returns output to the desired level, but not so much gain that it responds so abruptly as to overshoot.

The optocoupler is operating in linear range. It's not "triggered", the LED is supplied by IC4 with the amount of current required to set the right operation point. It's a feedback loop, you don't need to worry about individual node voltage. Just that IC4 is able to control Eout voltage (the actual 3825 current setpoint) through opto coupler with appropriate gain.

ive just noticed that the "error amp" LT1006 is not set up as an integrator, and will in fact be operating kind of in on/off mode.
LT1006 has just a 249k resistor connected as negative feedback.
Your circuit is not a typical feedback circuit involving an error amplifier set up as an integrator.

Don't fully agree. The feedback circuit is configured as P-controller, depending on the loop gain it can be stable. The advantage of a PI-controller is to cancel the static error.

Thanks, i see it is a P controller indeed, and i noted that on the schem the C21 and R18 are not fitted.
I think it would be creating unecessary difficulty for the OP to try and stabilise the circuit using just an error amp configured as a simple inverting amplifier, since already the OP is expressing difficulty.
If there was a capacitance at C21 then at least the OP would have a pole to play with, with which to help get stability.

This type of compensation, as you know, compared to integrator type feedback, has low gain at DC, and therefore exhibits relatively poor load regulation.
Its nice to have a pole in that error amp, so that you can roll the (open) loop gain off and get your crossover frequency down to a point where you can get a nice gain and phase margin.

- - - Updated - - -

As is known, in some power converters, there are right half plane zeros, and for these, we need the crossover frequency to be less than one fifth of the RHPZ. In these cases, its very useful to have an error amplifier with a pole there so you can roll your gain off well short of the dreaded RHPZ. The OP may struggle to do that with their chosen compensation method.
Also, there is very little literature available for SMPS’s compensated with anything other than type 1 , type 2, or type 3 compensators…so the OP, will be struggling and without the comfort of some tutorials.
I would think the OP would want to move away from the shown method of feedback compensation.

Surely it's no elaborated design, just wanted to mention it might work. Replace zener and LT1006 by TL431 and you have the most popular isolated SMPS voltage regulator. 3825 side stays essentially the same, compensation problems too.

Yes, with TL431/opto compensation, OP could do a lot worse than looking through Basso's books for details on it, as he gives a great coverage of it.