LM5030 slope compensation and feedback loop

[zilog]

I want to know how to design the slope compensation for this current mode controller operating in push-pull. Datasheet here: http://www.elfa.se/pdf/73/732/07326778.pdf I have found the formula Se = (Vo/Lo)*(Ns/Np)*Rs*Ai where Vo and Lo are output voltages and inductors respectively, Rs sense resistor and Ai = 1/N for the 1:N CT used. Can I assume this formula to be true?

Further more, the LM5030 implements slope compensation by adding a positive sawtooth ramp of 0->45uA to its CS+ pin, which in turn is connected to the outside world through a 2k resistor. I assume my only means of getting the right amount of slope compensation is by connecting this CS pin through a resistor before it enters the CT? Then, are my assumptions correct, based on the formula above, that this resistor falls somewhere in the 940 ohm region?

Vo=45V, L=50u, Rs=0.5 ohm, 1:100 CT, T=0.4/80000 (40% limited duty cycle for 80kHz frequency push-pull) leading to a Se = 21150 v/s meaning that voltage of Se*0.4/80000 = 106mV should be added to CS+ at the end of a cycle. 0.8*45 uA = 36uA which must terminate in 2940 ohm to equal a voltage addition of 106mV => series resistor of 2940 ohm. Am I correct this far?

Now that I have got this far, could someone please help me with pointer on how the feedback loop should look for this circuit, and also with pointer on how to calculate it? I want to have isolated secondary ground, output of +-45V with sensing across the entire 90V gap.

[zilog]

Can I use zeners in series with an opto coupler to sense the secondary side voltage and do all the compensation using the error amplifier in the control chip to avoid having to supply an op-amp on the secondary side etc? Or is this method non-favored for some reason? I remember reading something about non-linearity of opto couplers being a hinder.

[Eva]

The current transfer ratio of optocouplers changes from part to part and is quite unpredictable, it even changes with aging. Also, standard optocouplers like 4N27 are quite non-linear. That's why it's recommended to put the frequency compensation before the optocoupler. If you don't want to power a conventional op-amp, you can consider a classic IC like TL431, which is a very handy combination of op-amp and 2.5V precision reference.

[zilog]

Originally posted by Eva
The current transfer ratio of optocouplers changes from part to part and is quite unpredictable, it even changes with aging. Also, standard optocouplers like 4N27 are quite non-linear. That's why it's recommended to put the frequency compensation before the optocoupler. If you don't want to power a conventional op-amp, you can consider a classic IC like TL431, which is a very handy combination of op-amp and 2.5V precision reference.

Just to be sure, using zeners to drop 70V or so in series with the TL431 for a 2x rail-voltage of 90V will still work with the feedback loop, right?

Btw. How high feedback-loop bandwidth/phase margin do I need to get acceptable performance to power a set of switching amplifiers from a car electrical system? Just dont want to push things if I dont need to. I plan on using something like 1000-2000 uF per rail on the secondary side for buffering.

[VEC7OR]

If you plan to use this PSU for car why cant you just sense directly, or you want isolation ?
P.S. 1-2mF is just too low for +/- 45V rails, not that it cannot filter the ripple out, its just the electrolytic capacitors cannot take the charge/discharge currents.

[jackinnj]

Originally posted by zilog
Now that I have got this far, could someone please help me with pointer on how the feedback loop should look for this circuit, and also with pointer on how to calculate it? I want to have isolated secondary ground, output of +-45V with sensing across the entire 90V gap.

Hint: One of the great pleasures of working with National Semiconductor is that their technical support people actually answer questions. The semi industry really understands that oaks grow from acorns.

[zilog]

Originally posted by VEC7OR
If you plan to use this PSU for car why cant you just sense directly, or you want isolation ?
P.S. 1-2mF is just too low for +/- 45V rails, not that it cannot filter the ripple out, its just the electrolytic capacitors cannot take the charge/discharge currents.

I dont quite follow you here, the rms rating of caps this size should be more than enough. Or do you mean that the caps will sag under step-load due to slow PSU feedback loop?

[VEC7OR]

No, I just meant the current rating, if the caps take it its fine.