Tiny cheap fluid bed roaster update: background roast test

Leave a comment


Yesterday we tested the “background roast” feature of the free Artisan software, with Artisan activating and regulating the heater in order to let the BT cruise along the pre-designed profile. Before our test we made a few improvements to the hard- and software configuration.

Closing the air leak
Tije sealed the edge of the funnel to fix an air leak. This change boosted the power of the airflow into the bed of beans. Our first impression is that the maximum batch size increased from 200 grams to at least 225-250 grams.

Fighting spikes
In our last test drive we experienced a lot of spikes in the temperature readings. To fix this we installed two ferrite beads that were supposed to suppress the high frequency noise in the circuit. Marko Luther, also gave us suggestions how to get more stable readings. By increasing the filter values in the Artisan Device Assignment tab, we should get better usable readings. Marko indicated that a value from 80 until 90 percent should give us better results. These changes resulted in much more stable readings.


Improving Device and PID configuration

In our previous test we didn’t manage to let Artisan drive the PID drive the roast along the pre defined roast curve. After showing Marko Luther our configuration files, he quickly saw what caused the issue. The PID settings on the Arduino control tab had to be reconfiged. To make it work the source number needed to be changed into “2”. That’s the same as the BT channel number on the Device Assignment tab. Many thanks to Marko Luther for his great help!


Testing roasting along the pre defined roast curve

During the preparation of our test we saw chaff flying into the air pump. This chaff decreased the power of the airflow. Frans fixed this problem pragmatically by installing a little strainer on the air pump.

After making all of the changes we were ready for the ‘background roast’ test. The test was quite successful. We managed to get our prototype to roast along the pre defined roast curve. The graph shows that the PID follows the background. But in the image below that the precision can be improved. In graph below you can also see a big dip after the start of the first crack. This might have been caused by airflow changes during the roast.

background roast tc4

Next steps

More work needs to be done to optimise the PID configuration. Furthermore we are planning to put a kind of hood on the roasting chamber. This is supposed to improve the efficiency of the heating by restricting the loss of hot air. We are looking forward to do the next test run!


Frans Goddijn made (again) a nice movie of our test!


(movie by Frans Goddijn)

Leave a Reply