Thursday, December 21, 2017

Roasting with Phidgets

An introduction to automatic roast logging using basic hardware from Phidgets in combination with Artisan. This article focuses on standard USB Phidgets that have been supported by Artisan since 2013. The follow-up post More on Phidgets! describes the extended Phidgets support by Artisan v1.2 covering also the new VINT devices that were introduced at the end of 2017 by Phidgets.

Roasting coffee is quite simple. Essentially it is about applying heat to green beans for up to 20min. Often coffee is roasted in a rotating drum that is heated up by some electric or gas burner. In addition to transferring heat from the heat source to the beans via the drum, standard roasting machines also blow a certain amount of hot air through the drum, heated by the same source. Pure air roasters are an alternative that transfer energy solely via a stream of hot air so strong that it can also shuffle the beans around.

The taste of a roasted coffee depends on the green beans used, but also on the total amount of energy applied. Longer and hotter roasting processes result in darker roasts that have more sweetness, bitterness and roast aromas, while shorter roasts may bring out the specific aromatic notes of the green beans along with more acidity. By controlling the heat application over time, the taste of coffee can be influenced in great detail.

Traditionally, heat application is controlled by a roast master, using an analog pyrometer, his nose and ears as well as some magic (aka his "experience"). At different stages into the roast the beans start to smell like toasted bread, sweet or burned. Further, at two moments into the roast the beans start to emit crackling sounds. Those events are named first and second crack.

Profile Roasting

However, that romance in coffee roasting is coming to an end. Mostly, because a manual control of the process makes it hard to replicate a previous successful roast exactly enough to achieve a constant taste. In recent years, driven by home roasters, the traditional rough pen-and-paper based roast logs were replaced by automatically logged roast profiles that detail precisely the temperature development over time as well as the most significant roast events. Combined with real-time plotting, the temperature monitoring allows to anticipate important roast milestones by adjusting the heater or airflow supply at particular temperatures.

The relevant temperatures to document over time are
  • Bean-mass Temperature (BT): an approximation of the coffee bean temperature
  • Environmental Temperature (ET): the temperature of the air above and surrounding the beans

Using temperature sensors and meters in combination with some software allows to gather and record most of that data automatically. The roasting process might still be manually controlled by the roast master or be partially automated.

There are several important milestones in a roast.
  • Bean charging (CHARGE), the moment the beans are inserted
  • Turn-around point (TP), the moment the BT temperature starts to rise after the initial drop
  • Yellow point (DRY), as beans turn yellow
  • Start/end of the first crack (FCs/FCe)
  • Start and end of second crack (SCs/SCe)
  • End of the roast (DROP), where the roasted beans are released

Those events structure a roast profile into specific roast phases.
  • drying phase (CHARGE to DRY), 
  • ramping phase (DRY to FCs) and 
  • development phase (FCs to DROP). 

Note that the naming for those roast events and phases are still under discussion in the roasting community, but the importance of identifying those events and phases is commonly agreed by now.


For real-time plotting of temperatures you need at least

  • one or more temperature sensors
  • a meter to read out the sensors
  • a computing device equipped with some software to connect to the meter as well as read and display the data

The software part is covered at the end of this article. First we focus on the hardware.

Temperature Sensors

There are two main categories of temperature sensors that are used in coffee roasting, Resistive Thermal Devices (RTD) and thermocouples. RTDs are changing resistance with temperature changes, while in the case of thermocouples temperature swings induce small voltages proportional to those changes. In general, RTDs provide a higher accuracy, but a similar resolution than thermocouples (cf. accuracy, precision and resolution). On the other hand, thermocouples usually respond faster at comparable diameters. While thermocouples are usually connected directly to terminal blocks or miniature connectors, RTDs have to be connected to the measuring device via a Wheatstone Bridge or Voltage Divider made from precision resistors (in case the device does not yet provide those internally).

Besides deciding between the two worlds, RTD vs thermocouples, one has to decide on other aspects of probes like specified temperature range, diameter, length, sheathing, junction type (in case of thermocouples), cables and mounting. There are also dual-probes that one might consider to replace an existing single probe connected to some roaster internal electronics (most likely some PID implementing some safety functionality). Those contain two sensors in one sheath and can be used to add an extra probe for roast logging at the very same location as an existing probe, reusing its mounting facility. In any case, sensors need to be electrically isolated from the roaster's metal frame to avoid ground loop problems that may introduce considerable noise into the sensor signal (see below). The Thermocouple Primer and the guide on probes for coffee roasting, published by Cropster gives some hints on the right choice. A recent article in ROAST by Rob Hoos describes experiments comparing different kind of sensor setups. However, a generic advice is to select ø1.8-3mm, sheathed and ungrounded K-Type (or J-Type) thermocouples or PT100 RTDs, with fiberglass braid insulated cables (as most vinyl coverings melt at around 200°C!).

Sensor Positioning

Compression fittings or mounting nuts that are specified for the expected temperatures are commonly used to mount the probe to the roasting machine. The probe to measure the bean temperature (BT) has to be immersed into the coffee bean mass during roasting (also at low charges weights) without touching the agitators. Moreover, the BT probe has to be as far away from the front plate and the drum surface as possible, to minimize the influence of the temperatures of those. The Environmental Temperature (ET) has to be placed so that it may measure the air temperature above the beans, again away from the front plate and drum surface. Luckily most probes can be bended carefully. See the excellent article on probe placement by Terry Davis and Paul Ribich (free text). Note further, that positioning and probe type, as well as the physics of the roasting machine, are influencing the readings and may result into significant differences in the gathered temperatures on an absolute scale. This is usually not an issue for roasting coffee, as it is generally not advised to apply roast profiles to different setups without adoption, but rather use the roast logging method to ensure exact replication of a roast on the very same equipment. Rob's article linked in the previous section also compares the effect of different sensor positions.

Temperature Meters

To read temperature values from sensors, those need to be connected to a meter that is compatible to the probe type and the software to be used (see Device Selection for a list of devices supported by Artisan). There are plenty of temperature meters available on the market supporting a wide range of probe types, often in the format of a handheld device featuring a battery, a display and in some cases a memory that can be used to log the gathered data over time. Some of those devices also feature a data connection that can be used to transfer logged data or the current readings to a connected computer by means of some vendor specific software. This software usually provides only some basic logging and graphic features, but usually comes without any specific support for the coffee roasting process and is often restricted to certain computing platforms (like Microsoft Windows).

It turned out over the years, that those handheld meters that have been developed for field and lab workers are mostly in the way if mounted around the roasting machine: their batteries tend to be empty at unexpected moments, they are often rather costly, their data connection is based on slow, out-dated and undocumented serial communication protocols, and the provided driver and software does usually not fit all the needs of serious coffee roasters.

The Phidgets 1046 (featuring four RTD inputs, requiring Wheatstone bridges or Voltage Divider wirings) and the Phidgets 1048 (featuring 4x J, K, E and T-type thermocouple inputs) overcome all the issues that traditional handheld meters are associated with. They come in a small box that can be easily attached on a roasting machine, out of the way of the operator. Instead of a display they offer a high-speed direct USB interface, supported by a multi-platform driver and feature a powerful software interface based on open APIs.

The Phidgets APIs make integration into any software easy as they are provided for virtual all programming languages. For this and because the Phidgets hardware turned out to provide a flexible and reliable solution, the 1048 and 1046 Phidgets have been put forward by several providers of roast logging software, like the commercial iOS app RoastMaster, the Java-based cloud solution Cropster’s Roasting Intelligence (Java/Cloud-based solution), and the cross-platform open-source roast loggers Typica as well as the visual scope for coffee roasters, Artisan (see next chapter for details on this) and several roasting machine producers are equipping their machines with Phidgets on request or by default, like the US-based Mill City Roasters, Buckeye Coffee Roasting Equipment, and many other vendors. Also many local roasting operations with some engineering background, like Tico Coffee Roasters and Phil & Sebastian,  have started to come up with individual Phidgets-based roast logging solutions. Even the Speciality Coffee Association (SCA) based some of its roasting research initiative, headed by Morten Münchow in Denmark, on a Phidgets setup.


As with all technical sensing tasks, roast logging can be affected by noise. There are various sources of noise. A certain amount of noise results from technical limitations of the equipment, its rated precision. Noise can also be induced by electro magnetic interferences (EMI) such as those generated by the motor turning the roasting drum. Note that a temperature delta of 1C corresponds roughly to a voltage difference of only 0.04mV induced in a thermocouple at standard roasting temperatures. EMI-suppressing zinc ferrite cores located at the sensor wires close to the metering device can help to minimize this type of noise. Another major source of noise that can render signals completely useless stems from the so called ground loops, the situation where two or more electrical ground points in the system are at different potentials. The best way to avoid trouble like this is to completely isolate the measurement system from the power electrics. The easiest way to achieve this is to run the meter and the directly connected computer from batteries. If the USB connection has to be made via an USB hub, don't power the hub with an external power source. If this does not help or is not possible consider connecting the meter via some optical USB isolator like the 3060 USB Isolator. See also the article How To Avoid Grounding a Thermocouple for further suggestions and the thread Troubleshooting Electrical Noise with Roaster Thermometer [Solved].

More Hardware

Besides the Phidgets 1046 and 1048, there are other Phidgets that can be applied to coffee roasting. There is the Phidget 1045 IR. It integrates an infra-red temperature sensor with a corresponding meter that can be directly connected via USB. This one can be used to monitor the cooling of the roasted beans in the cooling bin to ensure proper cooling down to room temperature within about 4 minutes to preserve sweetness and aroma. Another application is to observe the roasting machines drum temperature by pointing the 1045 sensor (well covered against the high temperatures) from the outside towards the drum surface.

The various Phidget Interface Kits (e.g. the 8-channel Phidget Interface Kit 1018) are generic I/O boards that allow to connect various sensors and actors. Most interesting for coffee roasting are the differential gas pressure sensors (e.g. the Phidget 1136) that directly connect to the analog inputs provided by the Interface Kits. In combination to a pitot tube, to avoid overheating of that sensor, it can measure the (negative) air pressure in the roasting chamber or the duct. But the analog inputs of the Phidget Interface Kits can also be used to gather the damper setting (a mechanical flap on some roasting machines that allows to control the air flow) by reading a variable resistor driven by some simple mechanics. Finally, most Interface Kits provide also a number of digital outputs that can drive mechanical relays or SSRs, like the Phidgets 1014, to switch on and off various electrical components around a roaster, such as the motor turning the arms in the cooling bin.

Finally, one can go wireless by connecting the Phidgets to a Phidgets Single Board Computer (SBC) (that contains a full-fledged Phidgets Interface Kit 1018 among others) equipped with a Phidgets Wi-Fi USB Adapter. For further details on setup and operation of the SBC see How To Get Your SBC Up And Working and How to Set up Wireless Communication on an SBC.


The use of logging software in coffee roasting goes beyond just observing the measured temperatures, as would be possible by just connecting a simple temperature meter. Minimally, a roast logger allows to automatically generate time temperature protocols, often in an easy to understand graphical format. Software specially designed for roast masters, often allows additionally to register important roast milestones, like the start of the first crack, and gives hints on when and how to change controls (like the heater or damper) and when to take actions (like CHARGE or DROP) by providing additional data computed from the basic sensor inputs. Finally, some roast software allows to send control signals to the connected roasting hardware to modify heater power, drum or fan speeds, triggered by  on-screen software input elements like buttons and sliders, or fully automatically via PID control loops or programmatically via time/temperature triggers.

This blog is about the widely used open-source roast logging software Artisan that works especially well with Phidgets hardware. Artisan supports multiple platforms (Mac OS X, Windows and Linux), its UI is translated into more than 20 languages, it is extremely configurable supporting more than 30 different temperature meters and PIDs from various vendors and plays with even more hardware by implementing the standard MODBUS protocol and an open programming interface. The Overview page list all posts to the various aspects of Artisan.

While Artisan is free to use and targeting the experimental roast development, the authors ask for a donation to support its further development.