E61 Integrated Temperature and Pressure Sensor
E61 Integrated Temperature and Pressure Sensor
Embedded Files
A caffeinated engineer goes on an espresso machine deep dive.
A caffeinated engineer goes on an espresso machine deep dive.
The (ab)use of stimulants in grad school is a time-honored tradition. As an engineer, I needed an efficient (and legal) way to ingest these stimulants and found espresso as a tasty, interesting, (and once again legal) way to do that. But simply having an espresso machine for caffeine consumption would not be enough. I'd need to make it into a project.
The (ab)use of stimulants in grad school is a time-honored tradition. As an engineer, I needed an efficient (and legal) way to ingest these stimulants and found espresso as a tasty, interesting, (and once again legal) way to do that. But simply having an espresso machine for caffeine consumption would not be enough. I'd need to make it into a project.
After getting my hands on a totaled, but entirely fixable, Rocket Appartamento espresso machine and fixing it up to (mostly) working, I wanted to go further. The Appartamento is an E61-style heat exchange-style (HX) machine. To cut to crux of it, the thermal stability is bad, and the lack of automation makes repeatable, tasty espresso just barely within my skill range. If I could measure temperature, pressure, and shot time, I could easily quantify and correct for the key parameters that divide the ok-ish espresso from the terrible espresso.
After getting my hands on a totaled, but entirely fixable, Rocket Appartamento espresso machine and fixing it up to (mostly) working, I wanted to go further. The Appartamento is an E61-style heat exchange-style (HX) machine. To cut to crux of it, the thermal stability is bad, and the lack of automation makes repeatable, tasty espresso just barely within my skill range. If I could measure temperature, pressure, and shot time, I could easily quantify and correct for the key parameters that divide the ok-ish espresso from the terrible espresso.
A screenshot from Lance Hedrick's Lelit Mara X review. His grouphead has both the Smart Espresso Profiler (pressure) and a thermocouple probe (temperature) installed using a wye fitting. It's quite bulky and not very aesthetically pleasing. Furthermore, what's not shown is the separate displays for each device (also bulky and not aesthetically pleasing).
Making Things Simple
Making Things Simple
E61 mods are quite common, as there's a plugged hole left from the manufacturing process that the brew water passes through immediately before reaching the coffee puck. Removing the plugged screw provides a threaded attachment point with direct access to the brew water. Typically, a mechanical pressure gauge or temperature probe is installed in this spot. Since there's only one spot, the only way to measure both temperature and pressure is with a wye fitting attaching the two sensors to the grouphead. This is ugly (see left).
E61 mods are quite common, as there's a plugged hole left from the manufacturing process that the brew water passes through immediately before reaching the coffee puck. Removing the plugged screw provides a threaded attachment point with direct access to the brew water. Typically, a mechanical pressure gauge or temperature probe is installed in this spot. Since there's only one spot, the only way to measure both temperature and pressure is with a wye fitting attaching the two sensors to the grouphead. This is ugly (see left).
I wanted to combine these in some way to create a compact, elegant solution.
I wanted to combine these in some way to create a compact, elegant solution.
The First Iteration
The First Iteration
Originally, I wanted to create a more compact version of the wye design using smaller sensors and a smaller, custom wye or other similar shape with an integrated electronics package that could completely mount to just the grouphead. I'd still use a temperature probe that could stick into the brew water, but I'd use a pancake-shaped pressure sensor that I could stick in from the side without protruding out too much.
Originally, I wanted to create a more compact version of the wye design using smaller sensors and a smaller, custom wye or other similar shape with an integrated electronics package that could completely mount to just the grouphead. I'd still use a temperature probe that could stick into the brew water, but I'd use a pancake-shaped pressure sensor that I could stick in from the side without protruding out too much.
This quickly started to feel like not a great path forward. I was primarily not super confident in getting a good seal with the pancake sensor and not having a leak with boiling, 10 bar water, and also with having to integrate two very different sensors in a very small envelope. I headed back to the drawing board.
This quickly started to feel like not a great path forward. I was primarily not super confident in getting a good seal with the pancake sensor and not having a leak with boiling, 10 bar water, and also with having to integrate two very different sensors in a very small envelope. I headed back to the drawing board.
My frankly heinous first iteration mock-up. I didn't make it to the enclosure design stage with this version.
There's only one sensor here! Look at how neat I could make things!
Going Even Smaller and Simpler
Going Even Smaller and Simpler
It turns out that most temperature and pressure sensors aren't super happy in boiling water at ten atmospheres. After searching sensor suppliers for a veryyy long time, I finally found a sensor that can do both, the Sensata 112CP! It was small(ish), operated in the temperature and pressure range I needed, and was (relatively) inexpensive. The only (large) issue with it is that the temperature probe does not extend into the group head, meaning that the temperature measurement has a slower response time. As a result, I had to accept the compromise that the sensor would give me a rough idea of the grouphead/water temperature, i.e. if the group was up to temperature while heating up or if the thermosiphon had stalled, rather than the exact temperature of the brew water. This was acceptable as precisely knowing the temperature would probably give me decision fatigue and another variable while dialing in a shot that would probably only be somewhat helpful.
It turns out that most temperature and pressure sensors aren't super happy in boiling water at ten atmospheres. After searching sensor suppliers for a veryyy long time, I finally found a sensor that can do both, the Sensata 112CP! It was small(ish), operated in the temperature and pressure range I needed, and was (relatively) inexpensive. The only (large) issue with it is that the temperature probe does not extend into the group head, meaning that the temperature measurement has a slower response time. As a result, I had to accept the compromise that the sensor would give me a rough idea of the grouphead/water temperature, i.e. if the group was up to temperature while heating up or if the thermosiphon had stalled, rather than the exact temperature of the brew water. This was acceptable as precisely knowing the temperature would probably give me decision fatigue and another variable while dialing in a shot that would probably only be somewhat helpful.
The Build
The Build
With the sensor selected, I could select a microcontroller and settled on Adafruit's ESP32 Reverse TFT Feather which used the familiar, cheap, and over-powered ESP32-S3 chip in a small form-factor with integrated buttons, display, and battery charging/management.
With the sensor selected, I could select a microcontroller and settled on Adafruit's ESP32 Reverse TFT Feather which used the familiar, cheap, and over-powered ESP32-S3 chip in a small form-factor with integrated buttons, display, and battery charging/management.
Connecting the microcontroller to the 112CP was relatively straightforward: I powered the 112CP with a 5V boost convertor connected to the 3.3V power rail of the ESP32. Since the 4.5V max analog temperature and pressure signals from the 112CP exceeded the maximum 3.3V ADC voltage on the ESP32, I used a simple voltage divider to half each signal's voltage and doubled the measurement in software while converting to temperature and pressure. For pressure, this worked flawlessly, for temperature, there were some issues I'll discuss later.
Connecting the microcontroller to the 112CP was relatively straightforward: I powered the 112CP with a 5V boost convertor connected to the 3.3V power rail of the ESP32. Since the 4.5V max analog temperature and pressure signals from the 112CP exceeded the maximum 3.3V ADC voltage on the ESP32, I used a simple voltage divider to half each signal's voltage and doubled the measurement in software while converting to temperature and pressure. For pressure, this worked flawlessly, for temperature, there were some issues I'll discuss later.
In Fusion 360, I designed an enclosure that mounted to the barrel of the 112CP and held all the electronics. The enclosure was printed with my resin printer instead of an FDM printer to hide the layer lines better.
In Fusion 360, I designed an enclosure that mounted to the barrel of the 112CP and held all the electronics. The enclosure was printed with my resin printer instead of an FDM printer to hide the layer lines better.
The microcontroller is the biggest part of the electronics system. The two voltage dividers used for reading the temperature and pressure values and the 5V boost convertor which powers the sensor are all contained in the top left quarter of the controller.
There's no better way to check sizing than to build it and assess in person! One of my big concerns was making sure the whole system didn't protrude beyond the grouphead significantly as that would create an easy, breakable target for my clumsy, swinging hands.
The adapter, fully machined. Shown: the scraggly threads I somehow managed to single-point thread. Not shown: the backside covered in battle scars from trying to grind out the broken NPT tap.
I Hate Stainless Steel and Stainless Steel Hates Me
I Hate Stainless Steel and Stainless Steel Hates Me
To mate the 112CP to the grouphead, I had to machine an adaptor that would thread into the M6 E61 bolt hole and that the 112CP could thread into. I would need to drill a hole through the center of the part so the brew water could reach the sensor, as the sensor would have to sit a little outside the grouphead -- no one makes small M6 process thread temperature and pressure sensors; I looked really hard for one.
To mate the 112CP to the grouphead, I had to machine an adaptor that would thread into the M6 E61 bolt hole and that the 112CP could thread into. I would need to drill a hole through the center of the part so the brew water could reach the sensor, as the sensor would have to sit a little outside the grouphead -- no one makes small M6 process thread temperature and pressure sensors; I looked really hard for one.
Material-wise, both the sensor and grouphead are unplated and plated brass, respectively. Brass is really easy to machine, because it contains lead. Personally, I don't like ingesting lead and so I went with 304 stainless steel which is food-safe, strong (remember, the M6 portion of the adapter has a small hole drilled through it), and corrosion resistant (coffee is a hot, brown acid). Unfortunately, stainless sucks to machine.
Material-wise, both the sensor and grouphead are unplated and plated brass, respectively. Brass is really easy to machine, because it contains lead. Personally, I don't like ingesting lead and so I went with 304 stainless steel which is food-safe, strong (remember, the M6 portion of the adapter has a small hole drilled through it), and corrosion resistant (coffee is a hot, brown acid). Unfortunately, stainless sucks to machine.
Although the 112CP comes in both a metric (with a gasket) and NPT process thread, the only in-stock sensors I could find had the NPT thread. This became a learning experience as I tried to machining the corresponding thread in my stainless adapter. Since NPT threads are tapered (the T in NPT), the pilot hole for the tap should also be tapered. I did not do that, and as I tried to cut the thread, the sides of the tap quickly jammed against the sides of the hole. Miraculously, I was able to fully tap the hole, but not before suffering immensely: I managed to break my 1/4 NPT HSS tap and had to drill it out with a carbide center drill, then still had to tap futher down through significant work-hardened areas with my single remaining tap, stressful!
Although the 112CP comes in both a metric (with a gasket) and NPT process thread, the only in-stock sensors I could find had the NPT thread. This became a learning experience as I tried to machining the corresponding thread in my stainless adapter. Since NPT threads are tapered (the T in NPT), the pilot hole for the tap should also be tapered. I did not do that, and as I tried to cut the thread, the sides of the tap quickly jammed against the sides of the hole. Miraculously, I was able to fully tap the hole, but not before suffering immensely: I managed to break my 1/4 NPT HSS tap and had to drill it out with a carbide center drill, then still had to tap futher down through significant work-hardened areas with my single remaining tap, stressful!
Not Terrible Software!
Not Terrible Software!
Coding turned out to not actually be that bad, once I figured out interrupts, button debouncing, and the Adafruit graphics library.
Coding turned out to not actually be that bad, once I figured out interrupts, button debouncing, and the Adafruit graphics library.
The software gives real-time readings of the analog outputs of the 112CP with simple analog read commands. The raw voltages are converted to their respective temperature and pressure measurements using simple transfer functions. These values are then output to the integrated display for easy reading. Besides these simple functions, there's also an integrated shot timer (blessed be straightforward ESP32 hardware timer implementations), a mode to allow a user-defined pressure threshold to start and stop the timer for hands-free timing, and an automatic sleep mode since the whole system is battery powered.
The software gives real-time readings of the analog outputs of the 112CP with simple analog read commands. The raw voltages are converted to their respective temperature and pressure measurements using simple transfer functions. These values are then output to the integrated display for easy reading. Besides these simple functions, there's also an integrated shot timer (blessed be straightforward ESP32 hardware timer implementations), a mode to allow a user-defined pressure threshold to start and stop the timer for hands-free timing, and an automatic sleep mode since the whole system is battery powered.
Software testing. The floating pins are giving some very peculiar values.
Thermistor characterization. Since the water is boiling, the temperature known and the corresponding resistance can be measured. I used a double boiler out of an abundance of caution since conduction between the sensor body and the over 100 C pan could slightly alter readings.
Integration Hell
Integration Hell
Do you ever wonder where all the data on the datasheet went? I do.
Do you ever wonder where all the data on the datasheet went? I do.
After picking back up the project after a hiatus, I swore I had a datasheet for the 112CP that had lovely simple transfer function curves and equations for the temperature and pressure outputs. However hard I looked, I couldn't find it, and I had to resort to the barren "datasheet" that felt more like a very basic brochure for the sensor. While I had gotten pressure readings working just fine (yay!), my temperature readings were completely wrong (see previous section). After combing the datasheet again, I realized it was a thermistor, and of course there was next to no info on it. A quick bit of research later led me to a method for characterizing the sensor, and I did so using a multimeter and boiling water (left).
After picking back up the project after a hiatus, I swore I had a datasheet for the 112CP that had lovely simple transfer function curves and equations for the temperature and pressure outputs. However hard I looked, I couldn't find it, and I had to resort to the barren "datasheet" that felt more like a very basic brochure for the sensor. While I had gotten pressure readings working just fine (yay!), my temperature readings were completely wrong (see previous section). After combing the datasheet again, I realized it was a thermistor, and of course there was next to no info on it. A quick bit of research later led me to a method for characterizing the sensor, and I did so using a multimeter and boiling water (left).
I also made a weird discovery: normally I think it would be safe to assume that the common practice for a voltage divider for measurement would put the measurement resistor between the input voltage and the sensing voltage, with a known resistor between the sensing voltage and ground. In the case of this sensor, there's no built-in voltage divider and the thermistor is connected to ground. In other words, the circuit is flipped from my expectation: the measurement load is on the ground side and the remainder of the voltage divider has to be built as an external circuit which also supplies the reference voltage and reference resistor. Irritating!
I also made a weird discovery: normally I think it would be safe to assume that the common practice for a voltage divider for measurement would put the measurement resistor between the input voltage and the sensing voltage, with a known resistor between the sensing voltage and ground. In the case of this sensor, there's no built-in voltage divider and the thermistor is connected to ground. In other words, the circuit is flipped from my expectation: the measurement load is on the ground side and the remainder of the voltage divider has to be built as an external circuit which also supplies the reference voltage and reference resistor. Irritating!
Blissful rest.
Current Progress
Current Progress
I'm taking a quick break on this project, but don't worry, this project is close to completion.
I'm taking a quick break on this project, but don't worry, this project is close to completion.
There's a few more bugs and gremlins that need to be ironed out. When completed, this last section will be fully updated. In the meantime, I'm gonna take a nap!
There's a few more bugs and gremlins that need to be ironed out. When completed, this last section will be fully updated. In the meantime, I'm gonna take a nap!
Google Sites
Report abuse