Yes, sure. PV & SP are prescaled to %% from Fahrenheit and CV scaled to an appropriate range of VFDs reference speed (also in %%).
I'm using a 60-95 F range for scale output temperature to PV.
Micro800 PID question
- Thread starter Yashka
- Start date
Yes, sure. PV & SP are prescaled to %% from Fahrenheit and CV scaled to an appropriate range of VFDs reference speed (also in %%).
I'm using a 60-95 F range for scale output temperature to PV.
drbitboy
Lifetime Supporting Member
N.B. there is a typo in Post #8: ".Ti and .FC" in the third paragrpaph should read ".Td and .FC." It is restated correctly in later posts, so that should not be source of @Peter's confusion. Everything else is, AFAICT, correct.
TL;DR
The rest of this post is all about returning @Peter generous trolling in kind; so feel free to ignore it as there is absolutely nothing new in it. I know, because I did the work to reverse-engineer how the CCW/Micro800 PID instruction works, and I made my best effort to ensure that everybody, but one, who bothers to read my posts to this point should be in the same state.
Wrong again. actually I did it before that and recorded the result in my first post cf. post #5.
Wrong again, cf. post #5.
Or maybe, if @Peter Nachtwey doesn't see something it does not exist? Do we all cease to exist every time he blinks? If a Peter blinks in a forest, ...?
which it is not; wrong again: in any instance of the PID_GAINS structure/object in the CCW/Micro800 episode of The Twilight Zone,the .FC member expresses the quantity usually referred to as "derivative gain." I have stated that I don't know how many times; is @Peter blinking every time he skims over it?
Wrong again. "Which is it" implies I have been saying FC is more than one thing, which I have not.
Wrong again. First of all and for the umpteenth time, .FC is not the bloody filter time constant. cf. every post of mine post #6.
Second of all, if either or both of .Td and .FC is/are 0, then there is no derivative action cf. post #5 and post #6 and post #8.
Wrong again. That would be true if the FC was indeed a time constant, but it is not.
Wrong again. By continuing to treat FC as a time constant, which it is not even though that has been stated many times, although, because you have held to your presuppositions like an I don't know what, and refused every vain plea to challenge them for so long, I suspect you are the only one confused.
Prove me wrong.
Caveats
Kc, Ti, Td and FC being the four members of the CCW/Micro800 PID_GAINS object structure, in that order, assuming Ti, Td and FC are positive.
N.B. I did intend, and the data support, putting FC in the numerator, as well as the denominator of the third term, as well as Td in the numerator of the denominator of the third time. I an that makes no sense semantically, but remember that derivative gains and time constants, by any other name, would smell as sweet.
N.B. This only answers the query "What is the equation that the Micro800 uses?" posed above; any relationship of those PID_GAINS members' to the canonical meanings of identical names in control theory is apparently accidental at best.
TL;DR
The rest of this post is all about returning @Peter generous trolling in kind; so feel free to ignore it as there is absolutely nothing new in it. I know, because I did the work to reverse-engineer how the CCW/Micro800 PID instruction works, and I made my best effort to ensure that everybody, but one, who bothers to read my posts to this point should be in the same state.
Wrong again. actually I did it before that and recorded the result in my first post cf. post #5.
Wrong again, cf. post #5.
Or maybe, if @Peter Nachtwey doesn't see something it does not exist? Do we all cease to exist every time he blinks? If a Peter blinks in a forest, ...?
which it is not; wrong again: in any instance of the PID_GAINS structure/object in the CCW/Micro800 episode of The Twilight Zone,the .FC member expresses the quantity usually referred to as "derivative gain." I have stated that I don't know how many times; is @Peter blinking every time he skims over it?
Wrong again. "Which is it" implies I have been saying FC is more than one thing, which I have not.
Wrong again. First of all and for the umpteenth time, .FC is not the bloody filter time constant. cf. every post of mine post #6.
Second of all, if either or both of .Td and .FC is/are 0, then there is no derivative action cf. post #5 and post #6 and post #8.
Wrong again. That would be true if the FC was indeed a time constant, but it is not.
Wrong again. By continuing to treat FC as a time constant, which it is not even though that has been stated many times, although, because you have held to your presuppositions like an I don't know what, and refused every vain plea to challenge them for so long, I suspect you are the only one confused.
Code:
/ s FC * s \
/ 1 + -- + ------------------- \
CV = Kc * ( Ti / Td \ )
\ ( -- * s ) + 1 /
\ \ FC / /Caveats
Kc, Ti, Td and FC being the four members of the CCW/Micro800 PID_GAINS object structure, in that order, assuming Ti, Td and FC are positive.
N.B. I did intend, and the data support, putting FC in the numerator, as well as the denominator of the third term, as well as Td in the numerator of the denominator of the third time. I an that makes no sense semantically, but remember that derivative gains and time constants, by any other name, would smell as sweet.
N.B. This only answers the query "What is the equation that the Micro800 uses?" posed above; any relationship of those PID_GAINS members' to the canonical meanings of identical names in control theory is apparently accidental at best.
drbitboy
Lifetime Supporting Member
Thx!
Set CCW up like the image in post #8, with SP=PV=0, push the [Play] button on the trend, tweak the tuning parameters as desired, then type in -100000 for the output_of_model/pv_to controller quantity, and hit enter. the CV will jump and trigger the trend.
If Ti is large (e.g. 1e+30), assign 0 Td or FC, then assign 0 to PV, and CV will return to 0.
If a non-zero Ti was in play, CV will have an offset from 0 when PV is returned to 0l there was a trick to getting CV back near 0; I forget the details, but I think it was setting Ti to a small number and putting the PID into manual mode by assigning 1 BOOL named pid_in_auto or summat like that.
Peter Nachtwey
Member
Last edited:
drbitboy
Lifetime Supporting Member
You [@drbitboy i.e. I] screwed up the integrator term. It should be Ti/s. s is not seconds. It is the Laplace operator. Dividing by s integrates.
d'Oh! Thanks, yes, my bad, my typo.
Aaaaaaahhhhh ... the penny drops.
In a word: yes.
No, but maybe they do serve to communicate the facts to those who cannot otherwise see.
Last edited:
drbitboy
Lifetime Supporting Member
Unit analysis, part 1
Not quite; since CCW mixed up the names, they also mixed up the units ...
Let's assume that, in CCW, there is a tag Micro820 of data type PID_GAINS, so it has members Micro820.Kc, Micro820.Ti, Micro820.Td, and Micro820.FC.
In CCW/Micro800 PID only,
From that, it should be obvious that Micro820.FC has units of seconds.
Not quite; since CCW mixed up the names, they also mixed up the units ...
Let's assume that, in CCW, there is a tag Micro820 of data type PID_GAINS, so it has members Micro820.Kc, Micro820.Ti, Micro820.Td, and Micro820.FC.
In CCW/Micro800 PID only,
- The value of member Micro820.FC is the derivative time
- PLEASE remember: a rose by any other name ...!!!
- So the quantity (Micro820.Kc * Micro820.FC) is the factor, which I think is called derivative gain, that is multiplied by the error rate (∆Error/∆t) to calculate the CV derivative term.
- the CV derivative term has units of CV, and
- ∆Error has units of PV, and
- ∆t (PID update time) has units of seconds, and
- Micro820.Kc has units of CV/PV
Code:
terms: [CV derivative term] = Micro820.Kc * Micro820.FC * ∆Error * ∆t**-1
units: CV = CV/PV * ??? * PV * s**-1
Last edited:
drbitboy
Lifetime Supporting Member
Unit analysis, part 2
Again, Micro820 is a tag for an object of data type CCW/Micro800 PIC_GAINS with members Micro820.Kc, etc.
Well... it would be IFF we assume the units of Micro820.Td are seconds. But we have dispensed with assumptions are are instead focusing on reverse-engineering the CCW PID instruction as a black box ...
We know, from the behavior of the derivative term of that black box, that the units of Micro820.FC are seconds (cf. previous post).
And if we agree that seconds are indeed the units of the time constant term,
in the filter
to cancel out the s**-1 units of s, then we have the following:
From which we can see, quite plainly, that the units of Micro820.Td must be s**2 i.e. seconds squared.
QED from Post #10.
Agreed.
Again, Micro820 is a tag for an object of data type CCW/Micro800 PIC_GAINS with members Micro820.Kc, etc.
Well... it would be IFF we assume the units of Micro820.Td are seconds. But we have dispensed with assumptions are are instead focusing on reverse-engineering the CCW PID instruction as a black box ...
We know, from the behavior of the derivative term of that black box, that the units of Micro820.FC are seconds (cf. previous post).
And if we agree that seconds are indeed the units of the time constant term,
Code:
(Micro820.Td/Micro920.FC)
Code:
1/((Micro820.Td/Micro820.FC)*[COLOR=blue][I][B]s[/B][/I][/COLOR]+1)
Code:
[time constant] = Micro820.Td * Micro820.FC**-1
s ??? s**-1QED from Post #10.
Last edited:
drbitboy
Lifetime Supporting Member
Background: in this post, pidgains is a tag for an object of data type CCW/Micro800 PIC_GAINS with members pidgains.Kc, pidgains.Ti, pidgains.Td, pidgains.FC; I used the tag name Micro820 for the same object in previous posts; that one thing you must remember, or nothing that follows will seem wondrous*
.
Hol' on there, Baba Louie! I actually did address it: three days ago. Read on; apparently there are yet more pennies to drop.
@Peter Nachtwey's usual .signature is so poignant in the context of this thread:
...
In addition to everything else, @Peter asks great questions (the fact that they come from confusion only points out that @Peter immediately sees the source of that confusion):
That would be true IFF pidgains.FC, with a value of 4s, was indeed the time constant. But from the frequency domain formula, which you were the first to correctly transcribe, pidgains.FC is the derivative time, and it is the ratio [pidgains.Td/pidgains.FC] that is the filter time constant.
So, in the trend (below) shown in Post #8, the time constant is
with units of seconds.
So after 2 seconds (cf. event (4) in the image), the initial, post-step-change 40k (40,000) derivative term of the CV (cf. event (1) in the image) should decay by 3 (= 2s/(2/3)s)) time constants, which is the same as saying it should drop by a factor of e**-3, which means it will be reduced by the amount [40000 *(1 - e**-3)] = 38009. And that is exactly what we see in the CCW trend plot from Post #8 (look for ∆-38009 in the image), where this was originally addressed:
* From The Muppet Christmas Carol:
Gonzo : Once again, I must ask you to remember that the Marleys were dead, and decaying in their graves.
Rizzo the Rat : Yuck!
Gonzo : [whispering] That one thing you must remember, or nothing that follows will seem wondrous.
Rizzo the Rat : Why are you whispering?
Gonzo : It's for dramatic emphasis.
.Hol' on there, Baba Louie! I actually did address it: three days ago. Read on; apparently there are yet more pennies to drop.
@Peter Nachtwey's usual .signature is so poignant in the context of this thread:
But I digress/troll
...In addition to everything else, @Peter asks great questions (the fact that they come from confusion only points out that @Peter immediately sees the source of that confusion):
That would be true IFF pidgains.FC, with a value of 4s, was indeed the time constant. But from the frequency domain formula, which you were the first to correctly transcribe, pidgains.FC is the derivative time, and it is the ratio [pidgains.Td/pidgains.FC] that is the filter time constant.
So, in the trend (below) shown in Post #8, the time constant is
Code:
[I][B]pidgains.Td[/B][/I],s**2 / [I][B]pidgains.FC[/B][/I],s = 2.66667,s**2 / 4.0,s = 2/3,sSo after 2 seconds (cf. event (4) in the image), the initial, post-step-change 40k (40,000) derivative term of the CV (cf. event (1) in the image) should decay by 3 (= 2s/(2/3)s)) time constants, which is the same as saying it should drop by a factor of e**-3, which means it will be reduced by the amount [40000 *(1 - e**-3)] = 38009. And that is exactly what we see in the CCW trend plot from Post #8 (look for ∆-38009 in the image), where this was originally addressed:
* From The Muppet Christmas Carol:
Gonzo : Once again, I must ask you to remember that the Marleys were dead, and decaying in their graves.
Rizzo the Rat : Yuck!
Gonzo : [whispering] That one thing you must remember, or nothing that follows will seem wondrous.
Rizzo the Rat : Why are you whispering?
Gonzo : It's for dramatic emphasis.
Last edited:
drbitboy
Lifetime Supporting Member
Again, I already addressed that; Calvin and Hobbes images are more useful, IMNSHO, than another boring trend plot. I perhaps merit the benefit of the doubt, at least in the absence of any contradictory data, notwithstanding all other, to-date baseless, claims.
CCW has a freebie version, complete with a Micro850 emulator that runs in Windows 10. I supplied the code to test the PID; anyone are welcome to trivially replicate my experiment (and prove me wrong). To be fair, I will warn you that, if you have done stuff with Logix, CCW will feel like Fisher-Price in comparison, and can be quite annoying in general.
That said, we don't know if the emulator mis-implements the PID instruction in the same incorrect way; for the record, I did my experiment with a CCW,v13 and a Micro820,v12.
Thank you so much for so detailed explanations here. Could you confirm or correct my understanding of the equality of members between PID's PID_GAINS data type and IPIDCONTROLLER's GAIN_PID data type:
PID_GAINS.Kc = GAIN_PID.ProportionalGain
PID_GAINS.Ti = GAIN_PID.TimeIntegral
PID_GAINS.Td = GAIN_PID.DerivativeGain
PID_GAINS.FC = GAIN_PID.TimeDerivative
And I would be very appreciative if you could say me appropriate corresponded Cohen-Coon coefficients.
Thanks in advance!
drbitboy
Lifetime Supporting Member
I did not do any reverse engineering of the IPIDCONTROLLER or GAIN_PID, so I cannot say anything about the parameters in the GAIN_PID object.
First of all, @Peter and others have warned/mentioned that the parameters from the Cohen & Coon (1953) paper are quarter-wave damping; there is no guarantee that is acceptable for your process, so consider any coefficients derived from that paper to be a starting point at best; if anything maybe set the parameters a bit less aggressively if this will be controlling the actual process.
Second of all, I am not going to do the work for you, at least not without a purchase order, and in that case you should get someone else for whom this is not a hobby. Also, if I gave you some numbers and you used them, and then your process was damaged or hurt someone or worse, that would be a problem I am not willing to risk for goodwill on the 'net. And not to put too fine a point on it, if you are not able to calculate the values from the information below, then that risk seems even greater to me. You are certainly welcome to make the calculations and post them here, showing your work, and I am fairly certain someone will do a sanity check. Note that @Mispeld, someone with far more experience than me, has already suggested numbers, so I suspect if you get something not too different from those, then they are at least close.
The Cohen-Coon paper is here. The relevant quantities and formulae, with other conventions' symbols and descriptions in blue, are
Also, I found what appear to be equivalent versions of those tuning rules in other places e.g. this
and this (g_p = M = K and t_d (not Td) = L = θ):
First of all, @Peter and others have warned/mentioned that the parameters from the Cohen & Coon (1953) paper are quarter-wave damping; there is no guarantee that is acceptable for your process, so consider any coefficients derived from that paper to be a starting point at best; if anything maybe set the parameters a bit less aggressively if this will be controlling the actual process.
Second of all, I am not going to do the work for you, at least not without a purchase order, and in that case you should get someone else for whom this is not a hobby. Also, if I gave you some numbers and you used them, and then your process was damaged or hurt someone or worse, that would be a problem I am not willing to risk for goodwill on the 'net. And not to put too fine a point on it, if you are not able to calculate the values from the information below, then that risk seems even greater to me. You are certainly welcome to make the calculations and post them here, showing your work, and I am fairly certain someone will do a sanity check. Note that @Mispeld, someone with far more experience than me, has already suggested numbers, so I suspect if you get something not too different from those, then they are at least close.
The Cohen-Coon paper is here. The relevant quantities and formulae, with other conventions' symbols and descriptions in blue, are
- L = dead-period lag = θ = deadtime
- Z = process time constant = τ
- M = process sensitivity = K = ∆PV/∆CV at steady state after CV disturbance
- R = unit reaction rate = M/Z = K/τ
- μ = RL/M = L/Z = θ/τ
- S = Controller proportional sensitivity = υ2/(RL) = Kc = ∆CV/∆Error = (τ/(Kθ))υ2
- U = Reset rate = υ1/(Lυ2) = 1/Ti
- Ti = 1/U = integral time = Lυ2/υ1 = θυ2/υ1
- T = derivative time = Lυ3/υ2 = Td = CCW/Micro800 PID_GAINS.FC = θυ3/υ2
- P-only tuning (no I, no D)
- υ2 = 1.03 + 0.35μ
- PD tuning (no I)
- υ2 = 1.24 + 0.35μ
- υ3 = 0.34 - 0.11μ
- PI tuning (no D)
- υ2 = 0.9 + 0.083μ
- υ1 = 0.27 + 0.6μ
- PID tuning
- υ2 = 1.35 + 0.25μ
- υ1 = 0.54 + 0.33μ
- υ3 = 0.5
Also, I found what appear to be equivalent versions of those tuning rules in other places e.g. this
and this (g_p = M = K and t_d (not Td) = L = θ):
Peter Nachtwey
Member
You should have just said in post #8 that Rockwell switched the how Td and FC were used. From what was written it was hard to tell if it was your mistake or Rockwell's mistake. I was giving the benefit of the doubt to Rockwell. OK, bad.Aaaaaaahhhhh ... the penny drops.
In a word: yes.
No, but maybe they do serve to communicate the facts to those who cannot otherwise see.
Wow, did Rockwell ever screw up on this one.
Peter Nachtwey
Member
Why use C-C tuning? There are much better methods out there. C-C tuning only has a small advantage over Z-N. The quarter wave response is awful.
I have tuning formulas where the dead time can be a much as 10 times the plant time constant and there will be no or little overshoot.
I have tuning formulas where the dead time can be a much as 10 times the plant time constant and there will be no or little overshoot.
drbitboy
Lifetime Supporting Member
drbitboy
Lifetime Supporting Member
That's an acceptable explanation for your response to post #8, but I have to ask why you continued to do so for so many posts after that.
You've had a successful career; one thing I know about a successful career is that it involves paying attention. So what happened here? Why did I have to say that .FC is the derivative time in six different posts and yet you still asked:
Go back and read what I wrote. Was it somehow unclear? Seriously, if I can't communicate a simple thing like that, I need to rethink how I write.
Similar Topics
Cf. https://github.com/drbitboy/CCW_Micro8xx_PID_reverse_engineering
So I am using a Micro850 and I downloaded the RA_PID UDFB for tuning the temperature of my Thermocouples(TC). I am not sure if I am using enough...
Can someone tell me what i am doing wrong here? I have set the limits of the CV value for 5min and 60max. The output of the PID is not staying...
Hi there,
I have a small heating system with a micro 830 PLC to assist with the learning process of automation and PLC's etc. The system is a...
My company built a small test machine using a Micro800 PLC and CCW software. We chose the Micro800 because the machine is very simple.
We are...