A little while ago, I was posting about how my omega HH506RA only allowed me to write to a file and then read the file after the roast was done. Omega provides what seems to be a really, really stripped down version of LABVIEW 2009 to provide simple real-time graphs of the temperatures, but they are very basic. I was looking for information on how to get my real-time data into excel (without hand-entering it, as i wanted the 1/sec datapoints my logger provided). Once in excel, there seems to be a whole bunch of good stuff you can do with the data - starting with derivatives. We talk about derivatives here in terms of rate of change - figures thrown about were 20 deg/sec prior to 1st crack and 10 deg/sec after first crack. But what exactly were the beans doing? Looking at derivatives is like holding a magnifying glass up to a smooth roast curve - it brings out the minute details. Here's the first roast I've done with the derivatives plotted in real time. The weak link is actually using this data to adjust the roast - hence many people automate the process with a PID.

Here's the roast.


Here's a look at the derivatives. You can see how the roast slowed down as I approached first crack. It's evident in the above graph as well, but the derivatives really bring it into focus. It's especially nice, because I can't develop the roast graph (above) until the roast is already over. But I can notice that my 20deg/sec has slipped to 15, and then lower. Blue is BT rate of change, red is ET rate of change.

Interesting stuff, thanks. The PID which I am hopefully going to fit to my roaster I believe is run on a ramp/soak profile by setting times and target points, and it then linearly plots the inbetweens. An alternative I considered when I was having trouble sourcing the PXG4 was set differently, by setting the rate of change per unit time. It nice to consider some absolute targets, ie a set time to reach 300F for example, but quite convenient to think of a rate of change thereafter as your latter graph shows.

Was it a good roast?

Cheers, Chris

The rate of change of bean temperature is proportional to the difference between bean and environmentla temperature. If you want an intereting plot take your data out of the time domain, and plot (ET(t)-BT(t)) on the X axis, and BT(t+x)-BT(t), i.e. delta BT, on the Y axis, if you get the time span of the delta right, yoiu'll see a nice straight line relationship.

Jim, I varied the fan speed during the roast, so I don't think the ET/BT relationship that you mentioned would hold over the entire roast. However, I'll give it a shot, maybe it will break down the data by fan speed. I think i'll end up with a PID eventually.

The roast - well, i think i'm too slow post first crack. I also think i need to finish the ramp to first crack a little more strongly -the roast slowed down as it approached first crack. I think that my initial charge temp may be too high too.

The roast cupping - well, i got a pretty good roast, with some "back" flavors (hiding in the back of the cup) of burnt/ashy. I think the drying phase was short enough to preserve moisture, but maybe my first ramp was too long and my finishing ramp too slow.

I just realized what I wrote is a bit cryptic. I'll log some roasts tomorrow and illustrate what I mean.

My advice is to use plots of this sort when doing a PID setup. Then PID the ET, not BT, for greater stability, and use this style plot to work out the resulting bean profile.

Here's what I'm getting: All other things constant (i.e. fan speed) the rate of change of my BT is directly proportional to the difference between the ET and the BT. so, if i PID my roaster for ET, I can roughly work out the BT rate of change by the difference between the ET and BT.

If that's correct, airflow is the other big variable that would affect the results. I could keep it constant, or i could work out relationships for all four Hottop fan speed settings. thanks for the help.

I personally have the ET PIDed, then tweak the airflow to "trim" the bean temperature to conform to the intended profile. This works quite well.

I've contemplated adding a second PID to do this automatically, but in an air roaster, there's an upper and lower limit on airflow that changes a the beans get lighter. I've never figured out a way to stop the roaster from self destructing if it goes outside these moving limits.

There is a gotcha. Suppose you go fast to the first crack, say 390 BT, and you have the ET up at 450 to do this. Now you want to slow down the roast, and the ET should drop to 430 and rise slowly from there. This is a no-no, so you have to go from straight line BT profiles to curved ones to prevent these discontinuities and ET drops

I've been steering my Poppery roasts ** based on this principle ** lately and getting good predictable results. I'm doing it 'seat of the pants' using manual Variac control of heater and fan using BT and MET probes. The popper lends very well to this method as the ET control is very responsive to changes in heater power (or air flow).

Edit: maybe I should clarify 'the principle' I refer is as Jim mentions above:
"The rate of change of bean temperature is proportional to the difference between bean and environmental temperature"

I will be looking forward to what you find with your data processing.
Some experiments I did early on was to establish the lower (BT stalls) and upper (? what happens when you push too hard on the beans). The stall point is easy but the upper limits show slowly more detrimental effects to the roast. Uneven roasting, and harsh flavors seem to be the first sign. It seems there is an optimum window here where the BT moves forward at the maximum rate without undue stress on the bean. (this mostly for the 'bulk heating' ramp to first crack) ...anyway, it seems that as you try to push harder/faster beyond this window, the bean temps do not take on the same proportional heat, but just go uneven and the process seems to go 'unstable'. Ok, this is just seat of the pants so I'm looking forward to your data to see if it may shed some hard light on these ideas.

another_jim wrote:I've contemplated adding a second PID to do this automatically, but in an air roaster, there's an upper and lower limit on airflow that changes a the beans get lighter. I've never figured out a way to stop the roaster from self destructing if it goes outside these moving limits.

Sum the PID control of the fan with a minimum value and a value based on the ET with an opamp and limit the output of the sum with a regulator adjusted by the ET. If you have some idea of the voltages to drive the fan at various stages, it shouldn't be too difficult to come up with the bits.

I've mostly been adjusting fan speed for heat transfer and trying to keep the beans moving. (One thing I would like to automate on the popper is mechanical bean agitation!) Generally high fan at the beginning and lowering as first crack gets started when much less heat and agitation is required.

I think the best approach for PID is to control ET/MET. The relationship of BT to heater output is too complex to model with only PID parameters. The beans are reactive and respond differently during the various phases of the roast. Also, PID control of the air temp in a popper (ET) is a simple and good match for what a PID is designed to control. ET control is fundamental to the profile and roast results as well. It needs a constant stable control.

But how to integrate BT into the equation?

A 'smart controller' sits as 'front end' of the PID to integrate BT, ET, and time/rate of rise. roast state/phase, etc This would allow a simple script program or 'state machine' to tell the PID the ET temperature setpoint based on the 'user specified algorithm'. A separate state for each phase of the roast.

DIY PID-Controller with microcontroller

It looks like a board like this could do it, but I suspect there is not much front end software available to help you get it running and use it easily. Most of this type controller is sold to OEM to integrate into a product.
There are likely other products available that are better supported, but I just recalled this one here on HB a while back and thought hmmmm, a roast process controller.