"Automatic Control Systems" R.M. Phelan (PDF control)

Hi all, I=92ve taken R.M. Phelan=92s =93Automatic Control Systems=94, with= me on vacation. (Gotta have something to read on the beach besides trashy novels.) Half way through the book he finally gets to his idea of pseudo differential feedback (PDF) control. Perhaps best visualized at this website,

formatting link

This looks interesting! And I wonder if any of you have tried it (in practice)? The PI(D) controllers that I=92ve built always have an overshoot. (Which I now understand comes from the zero in the transfer function.) It strikes me that I should be able to tune the PDF system for a slightly underdamped system and get an even nicer* step response.

Are there any gotcha=92s in PDF control? Thanks for reading,

George H.

*nicer for me would be shortest time to ~1% error (or something like that.)
Reply to
George Herold
Loading thread data ...

Phelan's a good read, even though his circuits are junk.

AIUI his shtick is basically to wrap a local first-order control loop around the plant, which reduces sensitivity to forcing, and then wrap a fancier loop around that.

The general strategy does wonders for laser diode temperature control, IME.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Yes, and it works. I'd claim that I came up with it independently, but it's pretty freaking obvious. I think it's also been done long before Phelan.

You can still achieve overshoot with this system, but it's harder. You'll lose some snappiness of control: a better way to do this is to divvy up the proportional feedback between the forward compensator and the feedback-only compensator.

(An easier way to express the above is to do it Phelan's way, but to put in a summing junction after the integrator, into which you shove a proportion of the command).

You need to do your analysis more carefully. It'll respond slower. Integrator windup is still with you no matter what you do, and you may find that it's integrator windup that dominates the overshoot for large excursions.

--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
 Click to see the full signature
Reply to
Tim Wescott

Works in lots of IC applications. Local feedback makes that local element nearly ideal and boosts its bandwidth.

Of course you can turn around and walk yourself into head-room disasters, too ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Interesting. I'm in the late stages of designing a digitally tweaked laser noise canceller, which SPICE says can achieve about 80 dB of cancellation out to 10 MHz over a wide range of conditions. (Of course, that's using a brain-dead photodiode model--just a current source in parallel with a capacitor.)

Using local feedback to do stuff like that is key to the tweaking strategy too--for instance, with a FB loop wrapped round a current mirror to reduce its input resistance, I can get a useful tweak for quadratic frequency error [proportional to (j omega)**2)] by using a varactor to adjust its frequency compensation.

The Darlington Vbe log conformance problem looks like it can be fixed with local feedback as well.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

I initially missed the point about headroom. Care to elaborate?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

  • What you want is called critical damping; neither over-damped (slow), nor under-damped (overshoot / ringing). I was never into that, but read articles here and there; the factor of 1.2 comes to mind as being that critical setting.
Reply to
Robert Baer

Critical damping is fine when you have a plant with stable characteristics and not a lot of coupling between multiple loops. If you lack either of those things, you're better off with lower aggressiveness than critical. In particular, multiple interacting loops all tuned individually for some overshoot can easily take off together.

Reply to
Bruce Varley

Mixed thoughts. When it's a single element with local feedback, no problems. But I've encountered issues with feedback-triples with reactive feedback end up clipping in that middle stage.

Also can be an issue in state-variable filters... though there's a fix I keep in my private bag of tricks ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

There's a current mirror design that looks a lot like a Wilson except that the cascode stage is swapped around (PNP version shown).

0 | *-------------* | | | | \ Q3 Q4 / >| |< |---*---| /| | |\ / | \ | | | | *------* | | | | \ Q1 Q2 / >| |< |---*---| /| | |\ / | \ | | | *------* | | | 0 Iin 0 Iout

This one cancels all the base currents to really good accuracy, whereas the Wilson loses you three copies of the base current. This also has three feedback loops in it, vs. just one in the Wilson, or four vs. two if you count the cascode as one.

With a fast matched quad, it works very well out to where the bond wires start limiting things.

It has to be an old design, but I've never seen it given a name. Is it one of yours?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Well he is a mechanical engineer.

Yeah he gets to the local negative feedback in a weird way. He starts with just an integrator in the big loop, adds a differentiator in front of that, and then observes that differentiation followed by integration is the same as doing nothing to the control signal.

OK, I was thinking more about temp control where you just have a heater and cold bath. Then the over shoot can be a real PITA.

George H.

Reply to
LIN WEBBER

, with me

Ahh, well it wasn't obvious to me. (Till I saw it.)

h I

Yeah that's what he ends up doing, He should have give it a different name.

I've always dealt with integrator windup by having the 'plant'* and the electronics before it hit the rail at the same 'time'.

I guess it's about time I bought a copy of your book.

George H.

*I hope I'm using the right terminology. For me the 'plant' is usually a heater or TEC.

ttdesign.com

Reply to
LIN WEBBER

e
n
r

Critical damping is not all that 'critical'. A little underdamping can be just fine. It doesn't make that much difference.

I'm mostly stuck (at the moment) thinking about this in terms of simple harmonic motion. Which I think I understand.

George H.

George H.

Reply to
LIN WEBBER

=94, with me

lf

al

in

ut

e

ich I

t

cottdesign.com

Oops... still on vacation and not looking closely at who I'm signed on as...

Lin Weber , my aunt, does some nice paintings... but knows nothing about control loops.

George H.

Reply to
George Herold

It _does_ have a name, but I can't remember it right now. I'll look it up.

Most bipolar stuff has not enough headroom for such shenanigans... so my stuff is more subtle... and I should write a book someday when I stop making money at it... CMOS hauls ass ;-)

I knew Wilson when he was at Moto... a real prick... even nasty to his pregnant wife... so nasty and egotistical I'd equate him to a combo of NymNoNuts and Slowman. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

I'm using it with +-50 volt supplies, with another loop wrapped around it to reduce its input resistance at low current. In simulation it's a little underdamped at high frequency, but that can go either way due to board strays. A bit of series resistance in the emitters of the mirror pair will cool it off if it wants to oscillate.

It's really a pretty circuit--kudos to whoever it was who came up with it first.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Software

formatting link

With the steady hand that painting requires do you really think she knows nothing? :-)

--
Thanks,
Fred.
Reply to
Fred Bartoli

Systems=94, with me

novels.)

(in

a

like

loop

wrap a

control,

problems.

feedback end

fix I keep

=20

=20

wires=20

=20

Wow. A full on 0th class jerk.

?-)

Reply to
josephkk

th me

Phelan=B4s PDF works very well as an engine controller. Much better than th= e usual PID design. Turns a Cummins Diesel engine into a fire breathing mon= ster that could not be achieved with the usual PID control. Stable and very= responsive with minimal tuning. Recommended. I=B4ve used it in a variety o= f applications with good results. Plus it=B4s a lot simpler to implement.

Reply to
xv1942

, with me

the usual PID design. Turns a Cummins Diesel engine into a fire breathing m= onster that could not be achieved with the usual PID control. Stable and ve= ry responsive with minimal tuning. Recommended. I=B4ve used it in a variety= of applications with good results. Plus it=B4s a lot simpler to implement.=

- Hide quoted text -

Interesting, Thanks. Engine control sounds a lot more involved than the loops I do. So the input is the accelerator position, and the output is (perhaps?) the engine RPM. But the 'plant' is complicated!

George H.

Reply to
George Herold

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.