How important is the smoothness of the rate of rise curve? Do you notice any difference on the cupping table when you curve is smoother?
Cropster says it is useful to measuring consistency of the development of your beans and seems to imply that smoother is better (https://www.cropster.com/focus-on-the-rate-of-rise/)

I'm not sure Carl Staub's (http://www.sweetmarias.com/roast.carlstaub.html) recommendation for ET is useful for me because I don't understand how much *heat* is being transferred to the beans at that particular ET (for e.g. my understanding is that would depend a lot on airflow).

My delta BT graph in Artisan is really lumpy. Is this just a data sampling issue/ graph smoothness issue? (I have Smooth Deltas on 15 and I'm using a naked bead thermocouple to measure BT)

Welcome
RoR should be looked at as just another tool. Makes available a single number for the relationship of BT and time that we would be normally considering. Makes post roast analysis easier. The idea of all of this was to help a roaster be able to make slight targeted profile tweaks from roast to roast. And to help make learning easier at the start. Roasting is still really learned the same as it ever was, lots and lots of trials.

The Cropster posts are a bit complicated by the language translation plus the European terminology factors. The smooth in the early post I think was more about just being able to have tighter control.

With a bead tip and RoR noise improvements in all of the available software offerings,you should be able to have a workable limit of noise(sharp zig-zags) in your graph. if that's the lumpy graph issue.

There's quite a number of good threads in the archives on ET and MET, And relationships with batch size and bean types.

faustroll wrote:My delta BT graph in Artisan is really lumpy. Is this just a data sampling issue/ graph smoothness issue? (I have Smooth Deltas on 15 and I'm using a naked bead thermocouple to measure BT)

It's your setup.

http://artisan-roasterscope.blogspot.de ... te-of.html

I'd like to suggest not to use naked thermocouples for roasting. The lumpiness factor also depends the meter you are using. I argued in the blog post above that this cannot be fixed in the logging software completely.

Hi. Long time lurker first time poster here. My writing style is boring so please bear with me.

I've had the same struggle and formed some thoughts going through this problem. From the article and suggestions, naked "fast" probes arent rcommended, and curve smoothing is bad.

My question is, isnt using a slower, more dampened probe the same as curve smoothing? To me it is. One can model a slower probe with a mathematical system (linear time invariant), very similar to a digital filter with a high dampening ratio like you do your curve smoothing using a sliding window (convolution). In FIR like windowing, it's just a longer window, and IIR a higher dampening factor. And yes I would suppose a slower probe would also shift your curve a bit to the right in time, too, just like you would with curve smoothing, because of the probe's larger heat capacitance and therefore the lag.

On the other hand, we want to filter out the noises of the equipment. We want to get to the "true" bean mass temperature. I would question how much of the mentioned equipment noise is really seen on the graph that hasnt been smoothed out already either by the mass of the probe or the software itself. True, there is aliasing and noise enhancement effect when sampling unfiltered noise, but that should be very minimal. Bottom line for me is if you run your roaster with no bean, from room temperature to 500F, you see no lumps on your curves. Is that what everyone sees, too? That to me shows none of that equipment noise is visible, unless the probe reading can incur noises from physical abuses of the beans sloshing around in the drum, or from the cracking noises. At this point I have not investigated that possibility because I'm frowning upon myself, but one cannot discount the possibilities before having proven it.

One source of noise that I actually see on the Artisan curves is digital noise. Now this is different than you may have in mind. Sometimes I see my BT or ET spiking all the way to 600F (or whatever max is on yoru graph) from 300F (example), before falling immediately back to 300F. I suppose this type of noise is digital, but not sure who's at fault. I suppose somewhere between the probe exclusively, and the USB port inclusively. In this case, the type of probe cant help it, and too much curve smoothing would ruin the graph indefinitely for taking into the spike values into the averaging. At that point I would just use a text editor to overwrite the offending values to your liking.

To me it is helpful to think of what type of noises are you seeing on your curves, who is doing the bad things, and what type of noises may be OK to be on your curves and you actually want to see. The above reasoning led me to feel comfortable to continue using my naked probes. Yes I see them lumps, and bumps, and humps, and I ride along. Some of these humps could've been contributed by the energetic cracking of the beans, and how that affects the local temperature where your probe is situated. I want to see those on my curve and not have the slower probe or the software smoothen them out.

To a certain point I also ask, when should I worry about these humps? When should I start to say, hey there's a problem here. Thsi is because, I dont want to just ignore all the humps on the curves. To answer this question I have to go back to ask, why do I see the lumps to begin with? What were the beans doing? Did they catch a fire in the chamber for a moment? That's important to know so I would lower the heat and the ET. If that's not it, what was I doing at that instance? Was I going crazy with the knobs? Is your system PID'ed, and the automation going crazy with the knobs? If that's the case I would expect to see ET humps along with BT bumps. ET for me is air temperature within the drum. To me (correct me if wrong) if I see smooth ET, I would not worry at all. If your roaster control is out of whack then I would expect to see ET bumps as well. That I would personally avoid on my Hottop with a perforated drum. But then your setup may be different, and with a solid drum and enough airflow, even a bumpy ride on the ET curve may be OK.

Last but not least I would ask, if my ET is smooth, why would my BT be bumpy? One can form their own opinion on this. I usually see this in the development stage of a SHG/SHB bourbon. It may have something to do with the physical release of so much energy in each crack. I donno.

Let me know what you all think.

Take care.

A very happy Hottop / Amprobe / Artisan user

ROR should be very, very smooth. Whatever bead tip and data smoothing you need to use to get your readings less than jumpy is an entirely different thing that you need to take care of first.

fu11c17y wrote: Last but not least I would ask, if my ET is smooth, why would my BT be bumpy? One can form their own opinion on this. I usually see this in the development stage of a SHG/SHB bourbon. It may have something to do with the physical release of so much energy in each crack. I donno.

Let me know what you all think.

Take care.

A very happy Hottop / Amprobe / Artisan user

You shouldn't save up 9 months worth of thoughts, stream of consciousness and unload them all at once! What you've written is interesting, but it took me more time than I'd normally dedicate to a reply to take in.

That being said, the ET probes measure a fluid environment of changing air temps. It should be relatively smooth, artifact is a fault somewhere along the line. BT probes on the other hand, are having changing temp'd beans tumbling and rattling against them the entire time. And as those beans are swelling, their volume changes and the way they bounce off your probe changes. Simply put, micro-artifact on a bare wire thermocouple would be expected, more so than on a ET probe.

I witness harmonic noise on my Artisan ROR readings. This, what I imagine, comes from the device's sample interval being at a different frequency than the software's. On the graph I see a blip up, then a bit later, a blip down. No amount of smoothing seems to fix this entirely, so I use very little and live with it. Of course, by BT and ET curves are perfectly smooth unless my thermocouples are acting up... a problem I've entirely taken care of.

I like what fullcity say's about using a fast probe and letting software fix things. My roasting style keeps my eyes glued to the ROR graph, and I know full well that I'm "making my moves" late as a result. Oddly, I can't believe how soon after touching the gas knob I see results on the screen. It's almost immediate... which conversely makes me think it's not such a big deal to have a heavy probe and 1sec sample interval.

fu11c17y wrote:My question is, isnt using a slower, more dampened probe the same as curve smoothing? To me it is. One can model a slower probe with a mathematical system (linear time invariant), very similar to a digital filter with a high dampening ratio like you do your curve smoothing using a sliding window (convolution). In FIR like windowing, it's just a longer window, and IIR a higher dampening factor. And yes I would suppose a slower probe would also shift your curve a bit to the right in time, too, just like you would with curve smoothing, because of the probe's larger heat capacitance and therefore the lag.

You are right with the assumption that the mathematical system "convolution" should have the identical effect as using a slower probe (with more mass). In an ideal world. However, the two systems, smoothing via the probe vs smoothing in software, are different in that the first is applied directly to the analog signal while the second on a very curse sampled digital signal. My experimental experience (by using both shielded and naked probes for roasting) is that the signal quality on "smoothing by probe" is way higher and generates a lower time lag as less smoothing is needed to achieve a good quality (say readable) signal.

fu11c17y wrote: One source of noise that I actually see on the Artisan curves is digital noise. Now this is different than you may have in mind.

Yep. The digital noise you are referring is caused by communication errors, ground loops, electromagnetic inductions and any combination thereof. Those should be eliminated by corresponding counter measures like optical isolation, the use of ungrounded probes and others.

fu11c17y wrote: To a certain point I also ask, when should I worry about these humps?

It depends on how hard you are hit. I saw RoR curves that were all over the place were the underlying signal was kept fully hidden by noise.

fu11c17y wrote: A very happy Hottop / Amprobe / Artisan user

Thanks for this positive feedback.

Thanks Marshall, Tom, and Marko. You guys are awesome.