control loops, Feedback, and all the math for SMPS's / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

control loops, Feedback, and all the math for SMPS's

Move Thread

LAN_403

Patrick

Thu Aug 25 2011, 12:36AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

I dont fully understand everything ive read on the subject of control loops and regulation signal needed for a stable SMPS IC, im using both VMC and CMC IC's (for many different PS's).

Heres the problem, i have read alot and from many sources, like unitrode, Dr. Ray Riddley and others, so im famillar with the common Opto-coupler/TL431/compensated loop type, but i have the HCPL 7520 type of Linear opto, in a 8pin dip, its good supposedly to 100khz, but i dont know if/how the HCPL 7520 could be used to replace the conventional optos.

Feedback with the TL431
]15_designing_with_the_tl431.pdf[/file]

HCPL 7520 A/D Opto-Isolator (8 pin dip)
]hcpl7520.pdf[/file]
Note the switching characteristics table, what the propagation delays are for signal in to signalout times. Not sure if these are a problem or not.

Obviously the variable current through the standard opto, changes the brightness and with magnitude/phase and gain roll off, decides error signal return value. but i dont know if the linear optos have different limitations or capabilities. I obviously am wanting to avoid PS instability.

TY, Patrick

cjk2

Sat Aug 27 2011, 04:06PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

I just designed a current mode SMPS based on the UC3825. I would suggest you simulate the open loop response of your feedback section. you want gain to cross 0db at about 1/10th to 1/5th your switching frequency. where gain crosses 0db you want atleast 45 degrees of phase margin. That is a maximum of 135 degrees phase lag when you cross 0db. You also need enough gain margin. That is you need a certin amount of gain when phase finally hits a lag of 180 degrees. I would shoot for -20db of gain at that point, but I think I have seen it done with less.

I read this book and found it very helpful. http://www.amazon.com/Switching-Power-Supply-Design-3rd/dp/0071482725

I can not comment on opto characteristics which is probably what you are really after anyway. I suggest simulating the opto coupler in the loop.

Patrick

Sun Aug 28 2011, 02:24AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Thanks Cjk2 i was starting to think this thread would die without a single reply.

Well, im really trying to get away from the opto-phase-gain-margin thing and just go with a single drop-in solution, with the HCPL-7520. The control loop math has alwatys scared me, as i have seen it done poorly and fear i will too.

Mattski

Sun Aug 28 2011, 05:07AM

Registered Member #1792 Joined: Fri Oct 31 2008, 08:12PM
Location: University of California
Posts: 527

Patrick wrote ...

Well, im really trying to get away from the opto-phase-gain-margin thing and just go with a single drop-in solution, with the HCPL-7520. The control loop math has alwatys scared me, as i have seen it done poorly and fear i will too.

The HCPL-7520 is just a fancy opto-isolator which provides an analog voltage on the voltage proportional to the one at the input, with electrical isolation between the two, though it appears to convert the signal to digital for optical transmission between the two sides of the circuit.

So you'll still need to check that your circuit is stable as the 7520 is just providing an isolated path for your feedback loop, it's not providing any stabilization. Each power converter has a transfer function based on its topology (at least once things are at a steady state operating point), a buck converter for example if I remember correctly is a double pole at the LC corner frequency. To sum up what little I understand about feedback loops is that you want loop gain to go to 1 before phase reaches -180 degrees (and you need to watch out for right half plane zeros in switching power circuits which have negative phase), the closer the phase is to -180 at crossover frequency the more ringing you'll get, higher crossover frequencies (where gain=1=0dB) generally give faster transient response, and you want the crossover frequency well below the switching frequency.

wrote ...
Note the switching characteristics table, what the propagation delays are for signal in to signalout times. Not sure if these are a problem or not.

Propagation delays can cause a problem if they are long enough. A time delay is equivalent to phase shift in frequency domain, so if the delay is long enough it can make a marginally stable system unstable. A microsecond at 1MHz is 360 degrees phase as it's 1 period, so you can figure out how much phase it adds at your frequencies of interest.

Shrad

Sun Aug 28 2011, 07:11AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

there are optocoupler designs with feedback compensation in the audio world which are designed to linearize your brightness over a wide range of frequencies

all you need is another optocoupler which controls the gain of an op amp driving the main opto

Patrick

Mon Aug 29 2011, 05:22AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Shrad wrote ...

there are optocoupler designs with feedback compensation in the audio world which are designed to linearize your brightness over a wide range of frequencies

all you need is another optocoupler which controls the gain of an op amp driving the main opto

im not sure here what you mean, with two opto's, but the HCPL is 4.68$ each and a normal opto is pretty cheap. So if theres no advantage in simplicity, then ill save the expensive opto's for something else.

Shrad

Mon Aug 29 2011, 08:01AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

this document should help a lot...

basically, you use a second optocoupler in reverse, to perform a counter reaction on the op-amp driver of your first optocoupler

that way you have full control of the feedback path and can even add an integrating capacitor or any compensation/comparation you might need

Patrick

Mon Aug 29 2011, 03:48PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Shrad wrote ...

this document should help a lot...

Now i see what you mean, but ill still need the compensation compenents for stability, i think.

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob