View Thread

JimG wrote:
Hi, Rama -

The slightly larger resistor values should work fine. The value on those is not critical at the low output levels expected to Ot1 and Ot2.

You may have to angle the 1/2W resistors upward to get them to fit the hole spacing. Hopefully, the leads on the 1/2W units will still be skinny enough to fit through the holes on the board.

If you have trouble with the ones you ordered, I'd be happy to mail you a few of the resistors I've been using.

Jim

JimG wrote:
Hi, folks -

All 25 of the initial TC-4 shields have been sold.

After deducting PayPal fees and mailing costs, the net proceeds from all sales was $115.65. This includes $18.00 that I charged myself for the 3 boards that I kept.

The total bill from GoldPhoenix was $92.03 for the boards.

So there was a whopping total of $23.61 profit from this endeavor!! I sent the good folks at Coffekids.org a payment of $25 tonight to even things out.

Jim

farmroast wrote:


:Clap: This is just another of many examples of why this community is sooo great.

JimG wrote:
Hi, folks -

All 25 of the initial TC-4 shields have been sold.

After deducting PayPal fees and mailing costs, the net proceeds from all sales was $115.65. This includes $18.00 that I charged myself for the 3 boards that I kept.

The total bill from GoldPhoenix was $92.03 for the boards.

So there was a whopping total of $23.61 profit from this endeavor!! I sent the good folks at Coffekids.org a payment of $25 tonight to even things out.

Jim

JimG wrote:
Hi, Randy -

I browsed your PID code and was impressed by the customization you've added for fine tuning your roasts! I doubt you could ever accomplish the same level of control using an off-the-shelf controller.

You and I are working on parallel efforts (although my PID library is being written for espresso machines).

FWIW, I find/found it easier to express the P, I, and D parameters this way:

For proportional term, I use "Pb." This has units of degrees (can be F or C). Proportional gain, Kp, is 1 / Pb. This is nice because it demonstrates that proportional output goes to 100% when the error is larger than (i.e. outside of) Pb.

For the integral term, I use rE (resets) which has units of 1/minutes. Integral gain becomes Ki = Kp * T * rE (where T = cycle period, minutes). This is nice because rE = 0 corresponds with no integral action (as compared to having to use a giant value for I to accomplish the same thing).

For the derivative term, I use rA (rate). This has units of minutes. Derivative gain becomes Kd = Kp * rA / T.

Obviously, it makes no difference internal to your program how you define the P, I, and D parameters. But the above is used in the Watlow controllers, as well as several others. Tuning is a lot easier for me when using parameters defined this way because I have a pretty good feel for the ranges of Pb, rE, and rA that work on different heating systems.

Jim

randytsuch wrote:
You show Kp in all three terms, you really use it this way? That is different from my implementation, based on my limited research into PIDs.

JimG wrote:

Yeah. That's the way I have my code set up.

Proportional band (Pb) is the reciprocal of proportional gain. So if you have a Pb value equal to 15 DEG, and the error is 3 DEG, then the power output becomes 0.20 (or 20%) for the P term. Same result as if you defined Kp = 0.0667 per DEG, and multiplied the error times Kp.

The nice thing about using Pb is that it has physical significance. And you know that the proportional term will be 100% when the error is larger than Pb.

The sum of the past errors (used in the integral term) has units of DEG*TIME. You want to end up with a dimensionless quantity for the output, so the constant multiplying the errors sum has to have units of 1 / (DEG*TIME). So it makes sense to define Ki as having units of 1 / TIME, and then divide by Pb (which has units of DEG).

Ki is a constant, so you can choose it any way you want. The references I'm using define Ki as having units of 1 / TIME, and then you divide it by Pb to get the final coefficient.

Similarly for Kd. The derivative quantity has units of DEG/TIME. So if you choose your Kd as having TIME units, then Kd/Pb gives you the right units (i.e. none) for the output.

Finally, if you use some of the standard tuning guides (ZN, for instance), I think they are set up expecting Ki and Kd to be in purely TIME units.

Jim

randytsuch wrote:
My end goal is to create a PID roasting program that other people can use and tweak for their setups, so I want something easy to understand, but that is tweakable enough to work in different setups.

JimG wrote:

:Clap:
Which roaster(s) are your targets?
I've added PID control to a P1 and to a Hottop. The popper was a lot easier to hack than the HT. But once I'd put a switch on the HT that toggles between OEM control and external control, the results were good.

Jim

allenb wrote:Glad to hear we're getting closer to a PID-capable Arduino/shield.

JimG wrote:

Which roaster(s) are your targets?

I've added PID control to a P1 and to a Hottop. The popper was a lot easier to hack than the HT. But once I'd put a switch on the HT that toggles between OEM control and external control, the results were good.

Jim

randytsuch wrote:
I would think fluid air roasters, like a popper would be the hardest to control, since the only thermal mass is in the beans. Since I have never worked with a drum roaster, I may not appreciate the issues a drum would have.
For things like supporting different parameters after 1st crack, I have the program set up to support different parameters at the end, versus the beginning of the roast. It is easy to change where the switch occurs. Right now, only a couple of parameters are switched, but that would be easy to expand.
Randy

randytsuch wrote:I'm not sure I really need to use an eeprom, I have about 1/2 the program space left the arduino.

JimG wrote:

That should work, too. I guess you'd create an array of struct's, and then insert the values in a series of initialization statements in your PID sketch for the Arduino.

OTOH, the EEPROM turned out to be fairly painless to get up and running on the Adafruit prototyping board. This board has pads for soldering a surface mount chip. Then the connections needed are +5V, GND, AIN4, and AIN5. I used fly wires and it went together pretty quick.

If you used an EEPROM, then you could store your profiles in a PC text file and write a Processing sketch that would read the file(s) and send them over the serial line to the Arduino. A simple (I think) Arduino sketch could read the serial port and write the profiles to the EEPROM.

I like dreaming up projects for someone else


Jim

bvwelch wrote:
Randy - I like your idea of being able to choose from several pre-compiled profiles.

If you don't want to use the external eeprom, you could try this:

Develop your profiles in Excel, write out as CSV, and then use a little script to convert that into a C "include file". Then you just recompile your 'sketch' and reload into your Arduino anytime you want to tweak the profiles.