PID controller setup issue

[ArmenM]

So - going back to this posting, where you had p=10, I=10, d=240, and refreshing my memory on the inkbird manual .... here is what I am thinking (some of which might be considered wild speculation. Like I said in the thread, I am really not impressed with the clarity of the inkbird manual. Also, like I said in the thread, I really do not trust the autotune algorithms, and suggest that you approach control at first with 1) proporitional band mainly controlling, 2) integral term minimized, and 3) derivative term minimized. My wild speculation is, just based on the units cited in the manual for the integral term and the differential term,that they are using the standard form of the PID equation, and that to minimize the integral term you want to make it really big, and on the other hand to minimize the differential term you need to make it really small. thus, your integral term of 10 out of 9999 means the integral term is highly active, and a differential term of 240, while not being as bad as it could, be, is definitely not zeroed out.

if my guess is right about their equation I would suggest:

• set the integral term to 1 (its minimal value). This does not turn off the integral term, but makes its influence as small as possible.
• set the differential term to zero (which should eliminate its influence).
• then play with the proportional band term "P". I *think* that this, in their usage, is not closely tied to degrees width of the proportional band, but rather that it is an arbitrary number that multiplies the difference from setpoint, and any resulting value greater than 100 turns the output on "full". To test this, I would suggest setting the I and D terms as above, then set the proportional term to some big value .... say 500. What I *think* you will see is the oven being full on until it hits setpoint, then overshooting as you are now seeing. THEN set the P value to something really, really, small, like P=1. What I *think* you will see is a slow rise towards your setpoint, with the temperature leveling out at something less than then setpoint (unfortunately, the integral term can still bite you because the long slow rise to temperature can accumulate a lot of error, and the term just can not be turned off. If you get overshoot followed by oscillations, try bringing the oven up to temperature, then turn the controller off for 30 seconds. this should clear the accumulated error and stop the oscillations.

Let us know how it goes?

(I have definitely seen people with inkbirds that appeared to just be faulty .... struggling with them for a long time only to replace with Auber then having immediate success. But I figure it is worth just a little experimenting with the one you have to see if it can be gotten to work.....?? )

Ok so I just did p-1 I-1 and D-0 and it stopped power at 485ish and overshot to 554

 

[ArmenM]

So - going back to this posting, where you had p=10, I=10, d=240, and refreshing my memory on the inkbird manual .... here is what I am thinking (some of which might be considered wild speculation. Like I said in the thread, I am really not impressed with the clarity of the inkbird manual. Also, like I said in the thread, I really do not trust the autotune algorithms, and suggest that you approach control at first with 1) proporitional band mainly controlling, 2) integral term minimized, and 3) derivative term minimized. My wild speculation is, just based on the units cited in the manual for the integral term and the differential term,that they are using the standard form of the PID equation, and that to minimize the integral term you want to make it really big, and on the other hand to minimize the differential term you need to make it really small. thus, your integral term of 10 out of 9999 means the integral term is highly active, and a differential term of 240, while not being as bad as it could, be, is definitely not zeroed out.

if my guess is right about their equation I would suggest:

• set the integral term to 1 (its minimal value). This does not turn off the integral term, but makes its influence as small as possible.
• set the differential term to zero (which should eliminate its influence).
• then play with the proportional band term "P". I *think* that this, in their usage, is not closely tied to degrees width of the proportional band, but rather that it is an arbitrary number that multiplies the difference from setpoint, and any resulting value greater than 100 turns the output on "full". To test this, I would suggest setting the I and D terms as above, then set the proportional term to some big value .... say 500. What I *think* you will see is the oven being full on until it hits setpoint, then overshooting as you are now seeing. THEN set the P value to something really, really, small, like P=1. What I *think* you will see is a slow rise towards your setpoint, with the temperature leveling out at something less than then setpoint (unfortunately, the integral term can still bite you because the long slow rise to temperature can accumulate a lot of error, and the term just can not be turned off. If you get overshoot followed by oscillations, try bringing the oven up to temperature, then turn the controller off for 30 seconds. this should clear the accumulated error and stop the oscillations.

Let us know how it goes?

(I have definitely seen people with inkbirds that appeared to just be faulty .... struggling with them for a long time only to replace with Auber then having immediate success. But I figure it is worth just a little experimenting with the one you have to see if it can be gotten to work.....?? )

So it dropped down to 476 and then overshot again to 560

 

[Cushing H.]

Ok so I just did p-1 I-1 and D-0 and it stopped power at 485ish and overshot to 554

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

 

[ArmenM]

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

I’m on it!

 

[fitzo]

Very kind of you to share your level of expertise here and your effort on this, C Cushing H.

Very cool to see this is still going on here. Kudos to all who help and teach!

 

[ArmenM]

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

Ok so for #1 it went full on 100% and shut off at set point and overshot by 72.
#2 same thing.

When changing those parameters do I shut the unit off to “reset” it? Cause soon as I did #1 I changed the P while the temp was dropping

 

[ArmenM]

Y

Very kind of you to share your level of expertise here and your effort on this, C Cushing H.

Very cool to see this is still going on here. Kudos to all who help and teach!

Yup! I appreciate everyone’s insight and assistance! I’m going on day 3 of trying to get this to work lol

 

[ArmenM]

Ok

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

Ok so now after a few cycles the elements shut off about 5 degrees before set point and overshoots 35° above set point

 

[ArmenM]

I

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

I’m losing my damn mind with this EFFING thing! Lol it just keeps overshooting. I changed some settings and I saw a difference but it still overshoots and I’m getting fed up with it. Got so many knives backed up lol

 

[ArmenM]

Ugh .... I got screwed up by their documentation again.
to turn off (or minimize) integral, the I value needs to be set to a BIG value. Set I = 9999 and try again....... (sorry - this is the second time I made that mistake - have corrected the above post to reflect the need to set to 9999.

(however, the fact that the thing kept going full bore until basically at setpoint means that 1) the integral term is not really causing you any real problems, and 2) the proportional band is really small.

can you do the following two things for me?
1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.

if you dont see any real difference between #1 and #2, then I would suspect the controller......

Ok, so I bumped up the set point to 1200 and it slowly got there and it o ou overshot by about 10° with pid being set to 10-10-240. Soon as it drops below 1200 elements kick on at full 100 power. However, the heat drops to about 1160-1170

 

[Cushing H.]

Im sorry - I have been unavailable for the last several days. If you got a slower approach to the setpoint at 1200, can you try what I suggested above, but have the setpoint at a higher temp (like your 1200, or even 1500F) (re. your last post .... PLEASE, I recommend you not yet try to fiddle with the integral and differential parameters - they will just cause confusion. I recommend keeping them set to minimize the effect of each. so..... if you could, again, with setpoint at 1200 or preferably 1500 -


1) run again with P=1, I=9999, D=0, and see what happens.
2) then, run again with P=1000, I=9999, D=0 and see if you see the same basic behavior.