Today the rate meter really helped. I usually roast batches around 1 1/4 lbs. and today I did a couple 1/2lb batches. With this batch size difference the heat transfer ability made a difference and caused me to adjust the ET along with the variac. The rate meter made this much easier to do.

That's great.

Are you dialing in for the HomeRoast competition?

No beans yet. But did try a Sumatra 1/2lb. Rode the brakes. Was on the low edge of the Kill a Watt meter voltage readings at the end and still ended up with a couple divots when they hit the cooling tray.

Now that I have more roasts under my belt with the rise-o-meter, I'm beginning to learn how to use the additional information with my PID-controlled Hottop.

A little background: with my setup, I (indirectly) control the BT profile by manually adjusting the PID setpoint for the roaster ET. Usually, my ET setpoint is between 420F and 430F for the run up to FC. By design, this causes the BT profile to slow down some as FC is developing.

I like to maintain a rate of rise of ~10F per minute for the period between FC and end of roast. This requires gradually bumping up the ET setpoint once FC has started. Using only a BT display, it is difficult to know when to make the ET setpoint adjustments. Make them too late and the roast will stall. Make them too soon and you'll rush into SC too quickly.

But using the rise-o-meter, I get instant feedback during this part of the roast. When the rate begins dropping down below 10F/min, it shows up right away on the meter. So I bump up the ET setpoint by 5F and keep watching. I normally need to make 2 setpoint adjustments, typically ending up with the ET at between 430F and 440F at the end of the roast.

Jim

That's very slick.

I am impressed by the circuit you have designed. Here are some comments on it (I'm an electronics nut. Please don't interpret anything as criticism, I just like to explore different ways of doing things):

The differentiator capacitor in your circuit does not need to be a bipolar type - it can be an ordinary polarised electrolytic with the negative terminal connected to the op amp input - provided that the temperature signal from the thermocouple is always positive, and with roasting I would expect that to be the case.

Here is my reasoning, but you can verify by measuring the voltage on the capacitor in your unit using a DMM (measure both when the temperature is rising and when it is falling to be fully convinced):

The inverting input of the op amp, to which the negative terminal of the capacitor should be connected, is a virtual earth, since the non-inverting input is connected to earth and there is negative feedback. The voltage at the other end of the capacitor is always positive if the temperature signal is positive. In fact, the voltage across the capacitor is exactly the temperature signal (it only deviates from this at power-up when the capacitor is initially discharged and the op amp output will swing to the negative rail in a bid to bring the inverting input of the op amp back to earth potential, an action which results in the capacitor charging fairly quickly to the temperature signal voltage).

Your design needs the op amp to be offset-nulled because the meaningful outputs are of the same order (a few millivolts) as the op amp's offset error. You can reduce this problem by a factor of 10 (or more - subject to having sufficient output voltage swing from the op amp, and also subject to the differentiator resistor and capacitor not becoming too large) by having the differentiator output 10mV per degree per minute (rather than 1mV/degree/min) then use a potential divider to divide the op amp output by 10 so that the DMM can still display the rate-of-change directly in mV/degree/min.

Of course, this probably means reverting to a 1000uF differentiator capacitor, so the start-up problem rears its ugly head again. I have designed a solution for that - years ago I designed a temperature rate of change meter for environmental temperatures, which required a 0.047F supercapacitor in the differentiator, and the start-up problem was so severe that I incorporated circuitry to force extra current into the capacitor, bypassing the differentiating resistor, when the op amp output went near the rails. The extra circuitry involved a handful of cheap transistors, diodes, and resistors (I hope you have large hands, though )

Another trick to reduce the start-up problem is to offset the temperature signal (by subtracting a fixed voltage from it) so that it is zero at the lowest temperature of interest. This reduces the amount that the differentiator capacitor has to charge at power-up.

Also, subject to still having enough voltage swing available from the reduced supply voltage, you may be able to power the whole thing off a single 9V battery without the inverter i.c. if you use another op amp to generate the earth rail. Connect a pair of 100k resistors across the 9V supply, wire an op amp as a voltage follower (i.e. a non-inverting buffer), and connect the buffer input to the junction of the two resistors - the op amp output becomes your ground (0V) potential, the battery positive is then +4.5V and battery negative is -4.5V.

The final blow-everything-out-of-the-water-and-start-from-scratch idea is to go digital. I have designed a temperature rate-of-change meter based on a PIC 16F628A microcontroller, a 16x2 LCD, a 32kHz watch crystal, and a DS18B20 temperature sensor that does everything you need (it even displays the temperature!) except that the DS18B20 only works up to 125 degrees C. If there is enough interest from this forum I might modify it for a thermocouple input.

Jim, you say that the nominal value for the three resistors in the op amp feedback is 600k, but this is impossible with the values shown - even with the 100k variable resistor at max resistance the total resistance will be 1 / (1/10M + 1/470k) + 100k = 549k.

Hi, Andrew -

I appreciate and agree with your comments. Prior to reading your posts above, I had already added a few of your suggested changes to the circuit.

Presently, my breadboarded version is based around 10mV per F output from the t/c amp. And there is now an active filter between the t/c amp and the differentiator section.

I am also using a 1000u cap and 6k resistor in the differentiator section (the initial ramp up time issue was solved). I found through modeling that the signal currents in the RC section were running a little too close to the bias currents needed for the low noise amp I was evaluating.

The nominal 600k of the resistors in the schematic posted above is correct for a 100u cap (the product should be 60). But there are a couple of things that affect this: the +/-20% tolerance on the electrolytic cap, and those pesky bias currents into the op amp (probably not as much of a problem on the LF411, though). The R values in the schematic were determined experimentally.

You are probably correct about using a polarized cap. The only scenario I can think of where that might be a problem would be when I carry the warm rise-o-meter from the house into the cold garage and connect it up. Rare occurrence, and probably would still not cause any problems.

I've eliminated the inverter that provided the split rail supply. Right now I am using a single 9V battery with couple of LED's to give -4V for the op amps, but plan to use a real voltage splitter as soon as the Digikey box arrives

Jim

Jim,

If you are using your own op amp circuit to connect the thermocouple to the differentiator rather than the Fluke 80TK, which offsets the thermocouple output so that temperatures less than 0 degrees produce a negative output, then you are safe to use a polarized capacitor regardless of the temperature of the thermocouple - the output of your front-end is presumably directly proportional to the voltage produced by the thermocouple, which relates to its absolute temperature (and there are no negative absolute temperatures).

I have ordered a Maxim MAX6675, which does all the messy stuff needed to interface a type K thermocouple to a microcontroller, including cold junction compensation, and a 12-bit ADC.

Cheers,

Andrew

AndrewPartridge wrote:... the output of your front-end is presumably directly proportional to the voltage produced by the thermocouple, which relates to its absolute temperature (and there are no negative absolute temperatures).

Unlike semiconductor and resistance sensors, thermocouple output is not related to absolute temperature. Instead, T/C's measure temperature differences between the sensing end and the measuring end.

Jim