another_jim wrote:Keep your sensor naked, and don't worry about it moving a few centimeters. The main thing is that it doesn't foul anything, that it reads consistently, and responds fast. In a PID controller, the key variable is how long it takes before a change in the heat shows up as a sensor reading. The longer this interval, the more jittery and overshoot prone the controller has to be. By keeping the sensor is naked as possible, you keep the control as smooth as possible.

gyro wrote:Do you mean I should have used a bare thermocouple wire? ...

A bare thermocouple wire will give a reading that is prompt but discontinuous, thus contributing to the potential instability of a system trying to control the bean temperature directly. In this case, the mass surrounding the sensor will aid stability at the expense of promptness (which will also end up affecting accuracy). I think that some accuracy should be traded for stability in case bean temperature was to be regulated directly using a PID controller (so-called "single-loop" control). Therefore, it may not be a bad idea at all to use the J-type thermocouple you have shown as long as you put in mind that you may end up overshooting bean temperature a little bit.

ADDITION: In the above, I assume that the ability to ensure a thoroughly meaningful measurement is independent from the choice of the sensing element (either be it a bare thermocouple wire or a heavy metallic probe). Failing occasionally to provide a meaningful measurement due to the positioning of the sensor (such as the erratic readings you have at the door) will probably harm the process way more than the choice of the sensing element would, making such choice irrelevant.

(I have no practical experience using PID controllers)

Yeah, I tend to agree about the J type probe. I think its ultimately not a bad thing to have a 'small' amount of lag to stop little temperature fluctuations. The PXG4 seems to have an internal filter setting that smoothes the input into the PID algorithms, currently its set at 5 secs. I think I'll drop it to perhaps 2 secs and see how stable it looks. I'm also going to move the 'decimal point' on the display to see 0.1C readings. Its not going to make any difference to the process, but it will allow me to more easily see the stability of the system.

On my couple of manually controlled roasts with the PXG4 just acting as a temperature display form the J type BT probe, it seemed pretty steady. Didn't see it go down in temperature at all throughout, just a slow and steady rise.

Managed to figure out how to fit everything into the roaster and its now wired the way I want it. Just need to source (or scavenge from inside it) a little more DIN rail to mount stuff on. And of course to borrow my friends Dremel to cut through the plating to mount the controller.

Reading through the 150 page manual now to try and figure out how to get a simple ramp/soak program into it so I can roast some coffee. I now have added motivation, as I just run out of beans!

Cheers, Chris

Hi

This is my PXG4 It is not connected to the roaster yet. I'll probably do it on the weekend. Did you use pin 7 & 8 for push switch input or did you use them to talk to the computer?

Cheers

That was too fast, wasn't it? Arpi . I thought you had a good time "playing" with the knobs as much as I do. Too bad the dimmer switch I use for the element is very non-linear, which limits the fun considerably .

Arpi wrote:This is my PXG4 It is not connected to the roaster yet. I'll probably do it on the weekend. Did you use pin 7 & 8 for push switch input or did you use them to talk to the computer?

In the end I didn't go with the RS485, it can talk to a computer without that option thru RS232, but haven't got there yet.

I did find some interesting stuff on the setup though...

I set the A/M button to run as a toggle between running the set profile and user selectable temperature (CH7, UkEy = 9). This means I can preheat the drum to whatever I want (different depending on bean load) then dump the beans and start the preset profile by just pressing one button.

I've changed it so I can see to .1C which is useful for trying to see whats happening on the setup. (CH6, Pvd=1)

Also changed the internal temperature filter to 2 secs and it still looks very stable, might drop it to 1 sec (CH6, TF=2)

And put the output cycle to 1 sec interval (CH2, TC1=1) to drive the SSR.

I've also set the Nod setting (CH4) to 8 and it seems to all make it work as described above which is really tidy.

On to the tough stuff - I need to figure out how to tune it!!

I found this useful, but I'm still a long way off after 7 roasts today, including two efforts to autotune.

http://www.ambexroasters.com/pages/arti ... ntrol.html

If anyone has suggestions with regards to the following, it would be great...

Autotune at static temp no beans - P5.5, I285, D54.7, AR 29.3
Autotune at static temp (185C) 250 grams of beans. P8.6, I416, D80.0, AR35.1 (anti-reset windup)
Other info I noticed... if a high rate of intercept has built up, and the heat is cut only at the set value, it overshoots by 15C with a static temp. It would be less than this when running the profile, since the profile will also be increasing as the temp overshoots. Similarly, if a decent undershoot rate has built up, it will bottom out around 10C lower if heat is only applied when reaching the set value. This would be worse obviously if the profile was running.

Its not really cutting it though. Initially the bean temp has raced away (higher) from the Set Value (which is constantly changing/displayed along the preset curve). This is probably my fault, for dropping them at too high a charge temp, so I can deal with this. On my last effort, with the above 250g autotune settings, it didnt race ahead much and the intercept from below was quite nicely handled, but then it let the BT slip well behind the curve (up to 30C at one point, about 10C behind at 2nd crack). It just wasn't aggressive enough at keeping it.

Am I right to think that I need to move the P down somewhat to help bump up the aggressiveness?

Some of the other roasts I did with the P down as low as 1 (which seemed too low). But around 2 didn't seem far off, as long as my drop temp was around right. If it wasn't and the controller was playing catchup, it overshot significantly.

In most cases it has lagged behind the Set Value all the way as well, meaning its going full bore to catch up which results in a quick 1C - 2C.

So I think if I reduce P that will help the aggressiveness in heat application to get to the target, but I'm pretty fuzzy (no pun) on the I and D bit. When the temp has overshot, it ends up building a fair bit of downwards momentum before the heater starts amping up significantly. Whats the best way to make this anticipation more?

Cheers, hopefully might get the dremel tomorrow and put it all together,
Chris

orwa wrote:That was too fast, wasn't it? Arpi . I thought you had a good time "playing" with the knobs as much as I do. Too bad the dimmer switch I use for the element is very non-linear, which limits the fun considerably .

Hi Orwa. I still keep the old way with the knobs. I've put a SPDT switch and I can toggle between PID/manual/OFF control. So in case there is a malfunction, I can still use the knobs, etc. Didn't take anything away, only added

Cheers

Hi Gyro.

I haven't read the manual yet

I think that there is a little hole under the PID to connect an RS232 cable. They call it 'loader cable' but I am not sure if it has the same function as the RS485. Just by the name it sounds like it is for 'loading' firmware. Not sure, I'll have to read on it.

I don't know if there is an auto configuration for ramp mode. But since the 400Fs are the most critical, I would set the parameters around that temperature (maybe 410F?) and see what happens.

Thanks for your other tips. I'll be getting some experience and I hope I'll be able to figure it all out in two weekends or so.

Cheers

Its all done and dusted by the time you get to 400, crunch time is well before that in terms of PID settings. And you are correct, there is no autotune while running a ramp program, so I had hoped that the static temp would yield some useful data. Didn't really though, I don't think anyway.

Hi

I connected it today and everything works beautifully. Easy I didn't set anything yet on the PID. The temp is in C but I was unable to set the temperature in F because the parameter is missing from channel 6 (where that parameter should be). Maybe once I connect to the computer I'll be able to change it. I have to get familiar with the software and see what I can or cannot do

Cheers

After a long afternoon and my first play with a Dremel, reassembly is complete!