RTDs and Artisan settings

#1: Post by **LBIespresso** » December 4th, 2021, 1:02 pm

So I set up my Cormorant with RTD's and coming from my 2mm TC's it is definitely different. A touch slower but less noise so I THINK I like it better but I am not calling it yet.

I am just curious what others use for their settings in Artisan. I am using Delta span of 10 and smoothing of 0 and smooth curves of 1. I only came to these settings with trial and error.

What do you guys use?

December 6th, 2021, 12:42 pm

December 7th, 2021, 12:14 am

Thanks Steve.

I was just playing around a bit more. Is there consensus on whether it is better to use a smaller delta span with smoothing or the lowest delta span you can use without smoothing? I thought it was the latter but have no recollection as to why I thought that.

Also, I am using a sample interval of 2 seconds. Do you have any opinion about what to use there? Again, I have no recollection as to how or why I settled on 2 seconds.

December 7th, 2021, 6:35 am

Delta Span affects how far back in time Artisan looks when calculating a delta curve. You can enter different values for the ET and BT curves. Increasing this setting should smooth live-recording delta curves. Increasing this setting will not affect the standard temperature curves (ET/BT/etc.), only deltas. Delta Span is the period in seconds used to calculate basic/core RoR values by dividing the delta-temp/delta-time (so delta-time is the delta-span used for this). A Delta Span smaller than twice the sampling interval has no effect, larger Delta Spans lead to time shifts. - from the QSG.

#5: Post by **Brewzologist** » December 7th, 2021, 7:30 am

I sample at 1 second which works with my Phidgets. I prefer higher sampling frequency for more responsiveness. As for my Delta span and smoothing settings it was more trial and error, again favoring responsiveness over beauty.

December 7th, 2021, 10:40 am

CarefreeBuzzBuzz wrote:Delta Span affects how far back in time Artisan looks when calculating a delta curve. You can enter different values for the ET and BT curves. Increasing this setting should smooth live-recording delta curves. Increasing this setting will not affect the standard temperature curves (ET/BT/etc.), only deltas. Delta Span is the period in seconds used to calculate basic/core RoR values by dividing the delta-temp/delta-time (so delta-time is the delta-span used for this). A Delta Span smaller than twice the sampling interval has no effect, larger Delta Spans lead to time shifts. - from the QSG.

By delta curves, do you mean ROR?

Brewzologist wrote:I sample at 1 second which works with my Phidgets. I prefer higher sampling frequency for more responsiveness. As for my Delta span and smoothing settings it was more trial and error, again favoring responsiveness over beauty.

Also, can you guys check my logic in thinking I would be better off having a smaller sample interval and a larger delta span rather than the other way around to achieve the best signal to noise ratio? Again, this is with 3mm RTD's.

December 7th, 2021, 11:19 am

The following is my understanding which may be incorrect. I also use 3mm RTDs. I look to the Artisan team to make any corrections.

1) Yes, delta curve = RoR curve

2) Sampling interval is independent of any smoothing. IMO, you should set it as low, or close to as low, as your hardware will allow. One way to see the effect of too high a sampling interval is to set it to 3-4 seconds and run the Artisan Simulator; I prefer updated data on a 1 sec interval vs 3-4 sec so I can make more timely gas corrections during a roast.

3) Both delta span and delta smoothing cause time lags, which then delays data appearing in Artisan and thus slows responsiveness and your ability to make timely corrections during a roast. Ideally you'd use none of this, but the problem is the signal can then be too jumpy, which may lead you to make errant corrections during a roast when they aren't needed. For me, a smaller delta span plus a little delta smoothing was the best balance so I could see trends quickly without chasing intermittent spikes which came back to baseline. This is hardware dependent so YMMV. Again, running the simulator in Artisan is a way to see what works best for you. These settings are also nicely documented here: https://artisan-scope.org/docs/curves/

EDIT: You'll also notice I don't use any post-roast smoothing to make my curves prettier. I do this on purpose because I like to use background profiles for new roasts and I want them to appear as they were recorded. This gives me additional confidence in not chasing intermittent jumps in my new roast when the background profile showed similar behavior.

December 7th, 2021, 11:24 am

Thanks to both of you! That is all very helpful. I actually can't believe how useful Artisan has been in helping me understand things and that is with just using settings that someone else sent me without further investigation!

I am thankful to have the help of you guys in my long overdue deeper dive. I will be sure to report back on what I learn.

#9: Post by **MaKoMo** » December 7th, 2021, 1:39 pm

Brewzologist wrote:The following is my understanding which may be incorrect. I also use 3mm RTDs. I look to the Artisan team to make any corrections.

1) Yes, delta curve = RoR curve

2) Sampling interval is independent of any smoothing. IMO, you should set it as low, or close to as low, as your hardware will allow. One way to see the effect of too high a sampling interval is to set it to 3-4 seconds and run the Artisan Simulator; I prefer updated data on a 1 sec interval vs 3-4 sec so I can make more timely gas corrections during a roast.

I don't have a good explanation why, but for me (3mm Phidgets RTDs) I get the lowest noise level (Idle Noise test) at larger sampling intervals (I settled on 2sec as a compromise; 3sec felt too large). I tested from 0.25 (only available if running Artisan from source) to 3sec. The differences are not large but repeatable. I've also tested sync vs async configuration and the async configuration produces a bit lower noise at the same sampling rates. Note that the Idle Noise setup is with all smoothing turned off.

Brewzologist wrote: 3) Both delta span and delta smoothing cause time lags, which then delays data appearing in Artisan and thus slows responsiveness and your ability to make timely corrections during a roast. Ideally you'd use none of this, but the problem is the signal can then be too jumpy, which may lead you to make errant corrections during a roast when they aren't needed. For me, a smaller delta span plus a little delta smoothing was the best balance so I could see trends quickly without chasing intermittent spikes which came back to baseline. This is hardware dependent so YMMV. Again, running the simulator in Artisan is a way to see what works best for you. These settings are also nicely documented here: https://artisan-scope.org/docs/curves/

All fully correct. You do not need to run the simulator. For me the polyfit variant does usually not perform better than the standard algorithm, but for some profiles I saw the polyfit variant was performing better. The difference is minimal.

Brewzologist wrote: EDIT: You'll also notice I don't use any post-roast smoothing to make my curves prettier. I do this on purpose because I like to use background profiles for new roasts and I want them to appear as they were recorded. This gives me additional confidence in not chasing intermittent jumps in my new roast when the background profile showed similar behavior.