Back to PI(D) school

You can do anti-windup on an integrator (analog or digital) or a counter, but if the prop term is added after the integrator, the sum can exceed the range of the process. So anti-windup is best done as a feedback from the process drive back into the integrator, somehow. It helps to limit the proportional influence to a fraction of the output range.

"Bumpless transfer" (smooth switchover between manual and closed-loop control) shares the same issues as anti-windup. The old mechanical and pneumatic controllers did all of that stuff.

This stuff matters more when a few megabucks or gigabucks of process are involved. I almost ripped a 32,000 HP LASH ship off the docks at Avondale Shipyard, tuning a control loop.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

Worst fat-fingered technical thing I ever did was blow $800 worth of helium-3 backwards through the seals of a mechanical pump in a dilution refrigerator. My prof was Not Happy--he lost three weeks of his summer research.

Second worst was when I sequentially pulled three out of four fibres out of a laboriously handmade variable fibre coupler--borrowed from another student.

Oh, and then there was the time when I was 14 and building a 1500V transmitter power supply in my parents' attic (couldn't afford the transmitter). It was made from a gigantic 750V CT transformer from a big old Admiral colour TV.

I got the full B+ from one hand to the other. The spasm pushed the power supply through the closed window. As it fell it knocked a shingle out of the porch roof and half buried itself in the lawn. I woke up on the floor across the room. So maybe that was the most fat fingered thing I ever did.

Ah, the days when "men were real men, women were real women, and small furry creatures from Alpha Centauri were real small furry creatures from Alpha Centauri." ;)

Cheers

Phil Hobbs

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

e:

n.

and I

.
g

he

avior

duce

r at

ith

The

the

ll the

ime

ill

find

h

rstand

it

the

rom & T..Hagglund

n

't

ded

,

the

loops.

There was mumblings of a bumpless transfer too. I haven't thought that much about it, ... don't I just have to load the right voltage onto the integrating cap? (Where the "right voltage" is determined from the manual setting.)

George H.

Reply to
George Herold

Integral gain is pretty much meaningless when the system is that nonlinear. It's basically acting like relay control into the integrator.

--
Tim Wescott 
Wescott Design Services 
 Click to see the full signature
Reply to
Tim Wescott

At it's simplest, you can usually do just fine by limiting the integrator state to the range of the drive, but you can get fancier than that.

Limiting the integrator such that integrator + proportional = drive limit isn't a bad way to go, but it can slow the system's recovery from a transient. You can often get the best response by allowing the sum of the integrator and the proportional terms to exceed the available drive somewhat.

Yup. I never had to worry about equipment damage, but transferring control from an operator's thumb to a video tracker can be a challenge when you're peering through a telescope with a half a degree wide field of view.

--
Tim Wescott 
Wescott Design Services 
 Click to see the full signature
Reply to
Tim Wescott

:

on.

e
t
f
n
t
n
I
,
o
e
n

it

but there is no point in letting integrator continue integrating in the direction the drive is already maxed out, that is just more to unwind

Reply to
Lasse Langwadt Christensen

I could have killed people, if they were on a gangway, or got hit by a snapped line, or we drifted into traffic and smacked a barge or something. I was sitting on the deck of the engine room, turning trimpots, when the chief ran over and shut things down. I was over 60 RPM on the prop. I think I was about 20, and it was the first closed-loop controller I ever designed. I simulated it in Focal on a PDP-8.

That's too much responsibility. Like you, I now prefer that the worst I can do is fry a few chips.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I

or

e
t
e

the

d

and

& T..Hagglund

One thing to worry about with TEC's is that heat-transferred-per-unit-curre nt does depend on the temperature difference between the two sides of the P eltier junction. Even if the side you are controlling stays close to room t emperature, you can dump enough heat in the heat sink (or the room temperat ure can change enough) that the proportional gain around your feedback loop can change enough to mess up your PID tuning.

If you monitor the temperature of the exhaust side of your TEC (as well as the temperature of the you want to control you can run the system to have a constant amount of heat transferred per unit error signal, and everything stays much better behaved.

This should be obvious, but people have been known to miss it.

Sloman A.W. ?Comment on ?Implementing of a precision fast t hermoelectric cooler controller using a personal computer parallel port con nection and ADV8830 controller?[Rev.Sci. Instrum. 74, 3862 (2003)] ? Review of Scientific Instruments, 75 788-9 (2004).?

--
Bill Sloman, Sydney
Reply to
bill.sloman

Hi all, PI(D) update: (responding to PH, 'cause I want to ask a double loop question later.)

1.) Stuck the sensor on the heater and it's ~10 times faster. Nice (brings up an old debate.)

2.) changing the PI equation to one taking the setpoint(sp) out of the P term*. This worked great the second time**. It gave a nice damped step response to SP changes. (no overshoot) And had more dynamic range (It stayed in the linear region of the system over a bigger input step change.) All discussed elsewhere but it's nice to see it for yourself. Oh, I could also turn the integrating time down by ~2.

The downside to this is that the capacitor holds the setpoint voltage as well as the output. This gave me windup problems, particularly when the temp was lowered and the heater turned off. (It might work better with a TEC. Or I've got to change the windup setpoints.. I usually use the power rails.)

For the normal PI I had to add a diode across the integrating cap to stop windup when the heater shut off. Does anyone use RR opamps for this sort of thing? (yeah, don't tell me this is trivial in code space.)

I'm adding manual control/ PI control. (bounceless transfer) My plan is to drive the whole PI loop with feedback from the current sense resistor and reduce the integrating resistor.

I'm going to save the double loop for another day.

George H.

  • Error(e) = SP-y (y is the process variable, temperature in my case.) so, out= gain(-y + (1/Ti)int(e))

**The first time I tried this I made an oscillator. I got the sign of the feedback wrong.

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.