Ok, here's what I've got my 2362 set at right now. I'm still playing around a good bit so consider this a work in progress:
Inty = P10.0
outy = 2
Hy = 0.1
Atdu = 0
Psb = 0
rd = 0
CorF = 1
P = 0.9
I = 420
d = 0
SouF = 0.7
ot = 2
FILt = 0
As "D" is the culprit for most stability issues (engineers who use PIDs in manufacturing often joke the D stands for Death), and as our systems generally don't experience rapid PV changes, I'm currently using my 2362 as a PI only controller. Read on for why as well as a breakdown of the 2362's PID controls...
P is easy to understand. Set to Fahrenheit and P10.0, the temp range of the 2362 is 1100 degrees. A P of .9% x 1100 = roughly a 10 degree proportional band. This means if SV is 150 and PV is 140 (or lower), output should be 100%. At a PV of 145, output should be 50%.
I is tougher to put into simple math terms but the key thing to note about I is it is time based. It boosts the output based on how long PV has been away from SV. If you override your PID's control by killing output to the heating element, it doesn't know this. All it knows is PV is NOT changing based on its output so I becomes a bigger and bigger influence on the PID output. When you turn the element back on, an overshoot can occur because the PID thinks a LOT of output is now necessary to get PV to SV. Every time you override the PID's output, you should reset its time reference by power cycling or by setting a new SV.
The dreaded D... D's job is to influence output based on how quickly PV is diverging from SV. A great example of the need for D would be in an oven. When the door is opened, the hot air rapidly escapes. Instead of waiting for PV to fall to the point where P ramps up output, D will immediately boost output itself based on how rapidly the divergence from SV is occurring. The thing is, our systems don't have rapid PV changes. An incorrect D value can lead to significant oscillations in the system - both overshoot and undershoot of the SV. This is why I've scrapped using D.
Finally, SouF. This is the 2362's "artificial intelligence" control on output. No details on the algorithm are provided by Auber. All we know is it dampens overshoot with a variable level of influence from 0.1 to 1.0. In my tests it does seem to work quite well, as long as one keeps in mind the previously mentioned time factor of I.
So, how to set up your 2362 PID? First, make sure your config is correct. For an RTD probe, set Inty = P10.0, not P100. With an SSR, hysteresis should be small, like Hy = 0.1. Set Atdu = 0 as it's OK if PV overshoots SV during auto-tune. Set CorF to your preferred temperature scale and stick with it. Changing CorF or Inty changes the math and will require a new auto-tune run. Don't mess with Psb at this point.
Now you've got choices...
Option 1: Auto-tune.
Set up your system as it would normally run for brewing. Typical water volume, pump on if you recirculate your mash, etc. Give the temp probe 5 minutes to stabilize before powering up the PID. Now power up the PID and enter a SV of 150F. Press and hold > until the "AT" indicator starts to blink. Go have a few beers. This will take a while. Once "AT" is done blinking, auto-tune is done and PV should match SV (150F). Using your lab thermometer, check the actual system temp. If it's not 150F, now's the time to adjust the 2362's Psb setting to compensate. Now go into the PID parameters and set SouF to at least .5, higher if you experience overshoot in operation.
Option 2: PI operation.
Basically same steps as option 1 except while you're in the PID parameters, change I to be 1.5x the auto-tuned value and set d to 0.
Option 3: P operation.
Screw auto-tune, screw this damn PID trying to out-think you! Set P to the proportional band you want. Assuming Inty = P10.0 and CorF = 1, a P of 1.8 = about 20 deg, 0.9 = about 10 deg, and 0.5 = about 5 deg. Set I and d to 0. Now the PID will strictly set output in proportion to the PV. This actually works EXTREMELY well, assuming one critical point - that your temp probe is very close to the heat source. If it's not, say the probe's in the MLT and you're heating in a RIMS tube, the delay between heat output and probe reading can again lead to overshoot.
Disclaimers:
I ain't an expert in this stuff, but I've been reading quite a bit on PID theory, talking to Auber about the 2362 specifically, and running lots of tests. I'd definitely like feedback from PID gurus out there.
PID values & tuning procedures are not transferrable across models - even Auber's! The 2352 uses an ENTIRELY different value for P. Someone else will need to write the 2352 setup guide.