Skywalker roaster... | [275] |
Skywalker, the AL... | [168] |
Dereks 1-2Kg drum... | [115] |
Skywalker Roasts | [91] |
SR800/SR540 Fan C... | [82] |
Where to filter?
|
|
shortyjacobs |
Posted on 01/25/2020 5:03 PM
|
Newbie Posts: 44 Joined: September 24, 2014 |
Where should most/all of the signal filtration be applied? I?ve got a TC4+ and artisan hooked up to it. Both offer filtering options on both BT, ET, and ROR. So if I plan on using Artisan for all control (manual or PID), should I just zero out all the filters on the arduino sketch, so Artisan can do its own filtering and be working from the ?Raw data?? Is there a problem with that? (Lag?). Better to use firmware PID and do all the filtering in the sketch? It seems silly to filter twice, kind of a copy of a copy situation. Help me out geniuses. -Keith
|
|
|
mg512 |
Posted on 01/25/2020 5:42 PM
|
1/4 Pounder Posts: 189 Joined: March 04, 2018 |
If I recall correctly, even if you configure filtering in Artisan, it will still actually happen on the Arduino. The configuration in the sketch will set default filtering parameters. The configuration in Artisan will override these at runtime. But there's no filtering happening in Artisan, it merely passies on the configuration parameters to the Arduino. You can check this yourself, either in the Artisan serial log or by enabling echoing received commands to the LCD on the Arduino; it should send a FILT command with the parameters you configured in Artisan. That overrides the defaults configured in the sktech. In my personal experience, less filtering leads to better PID performance. That may be different for different setups though. |
|
|
renatoa |
Posted on 01/26/2020 3:24 AM
|
Administrator Posts: 2965 Joined: September 30, 2016 |
Less filtering, should equate to jumpy input for PID. So the error value will be also jumpy, same for output values, which is not good for a low inertia roaster. Beans don't like sudden temperature change. Maybe you have a low P term set in your PID, lower than half of ultimate gain, and rely PID output action on I term mainly... which is a filtering in disguise, smoothing the input jumpiness. Relating lag, the following values for RISE_FILTER will led to the corresponding lag for temperature at the input of RoR computation code: - 50% => 1 seconds - 67% => 2 seconds - 75% => 3 seconds - 80% => 4 seconds - 83% => 5 seconds - 85% => 6 seconds If the same value will be used for ROR_FILTER, then the lag above will double. The above are math determined, not guesses or practical measurements, and are independent of temperature reading resolution. What is dependent on resolution is the need for the minimal filtering to apply, for a desired RoR precision. As a reference, for a 0.1 C degrees temperature resolution measurement system, at least 50% filter values are needed for 10% precision reading, i.e. 1 digit for the middle of the road 10 C degrees / minute in a roasting process. Other way said, if your temperature measurement system is single digit precision, and NO noise in the system ! you need at least 50% filtering to be sure when you read 10 C RoR, then this is rounded 10 value you can rely, i.e. between 9.5 and 10.49 Once again, the above are valid without any noise in the system, which is out of the roasting world, just for the record. When a noise comparable with half of the system resolution is present, then the filtering needs are increasing, in the 75% ballpark. For more noise increasing filtering will make the lag unacceptable, it should be fixed at the source. Or compute RoR and simultaneously filter using better algorithms than a rolling average, as is Savitsky-Golay first derivative smoothing, providing better results with less lag, but needing more processor resources than an Arduino. |
|
Jump to Forum: |
Thread | Forum | Replies | Last Post |
---|---|---|---|
Locally sourced alternate material for top filter? | HotTop Roaster | 4 | 02/04/2022 11:34 PM |
Camera lens filter as sight glass? | Drum Roasters | 8 | 03/09/2020 1:53 PM |
Thermocouple Response and Filter Setting | Dataloggers/Controllers/Rate of Rise Meters | 11 | 03/15/2019 9:44 AM |
Yet another filter thread... | Preparing Coffee | 5 | 10/06/2018 7:09 AM |
Filter Options | Preparing Coffee | 10 | 03/04/2017 8:14 AM |