A few thoughts... As far as I can tell, smoothing only gets applied after stopping the roast (ie, hitting 'off'). But I'm not an expert, that is just my observation from using it for a year. I use 5s sampling rate and oversampling, and this produces reasonably smooth RoR values during the roast. And after a "smooth delta" value of 4, the finished curve is extremely smooth. ;) Also, regarding Rao's guidance with respect to RoR, he says that a flat RoR "even for just 1 minute" will have a negative effect, but I interpret this to mean that you should not worry too much about sample to sample RoR (4 or 5 seconds) as that is probably impossible to control even with the best equipment. And regarding Rao's "1st crack to drop" percentage, have you found a way to have Artisan display that in real time? ie, if first crack occurs at say, 10:00, how to get Artisan to show a recommended range for ending the roast based on that? Rob On Mon, Sep 15, 2014 at 8:36 AM, Fred B <fbacher1(a)san.rr.com> wrote: -- "If this is coffee, please bring me some tea; but if this is tea, please bring me some coffee." -- Abraham Lincoln "The thing about quotes on the internet is that you cannot confirm their validity." --Abraham Lincoln Sent from my iPalmDroidBerry7

Dear Fred, I think the RoR core algorithm was updated in v0.7.3 (moving from averages to weighted averages), however, around v0.8.0 a second algorithm was added (tick Smooth2 for this in the extras dialog) provided by Nicholas that is based on approximation. To my opinion both algorithms perform similar (but require different parameter settings). Those two algorithms are used during recording. A third algorithm is used after roasting. That third one is the best of all (using the same smoothing settings as the other two) as it does not produce any time lack. The first two produce a considerable time lag (depending on the amount of smoothing configured) that is unavoidable (as they cannot know the future). I wrote somewhere more general about the smoothing issue over on HomeBarista. In summary, - smoothing should be done as early on in the change (probe, meter, firmware of meter, software/Artisan) as possible - different probes and meters deliver different amount of noise to Artisan (Phidgets deliver a rather smooth signal out of the box as the Omega 506RA, which is way to slow thus) - probes for roasting should not be thinner than 3mm (and in general ungrounded) - some "meters" allow to configure the amount of smoothing applied to the signal (eg. Arduino/TC4) or have some other mean to smooth out signals on the way to the software side (Phidgets; not supported in Artisan v0.8.0, but in the next version still to be released) - smoothing of the underlying temperature curve makes the RoR curve smoother - additional direct smoothing of the RoR curve is needed in most situations Oversampling is applied only for sampling intervals greater or equal to 3sec. It essentially takes a second reading in the middle of the sampling interval (so minimally every 1.5sec) that is used for averaging. Oversampling helps somewhat, but does not have a hugh effect. Still 3s+oversampling is better than a 1.5s sampling interval. The reason is that one of the problems with smoothing out the RoR signal is the small resolution on the temperature scale vs the time scale. Technical details suppressed here as they have been discussed somewhere else. So in summary, keep your sampling interval at 4sec or increase it to 5, tick over sampling, apply smoothing to the curve itself ("Smooth Curves" setting under extras) and to the RoR itself ("Smooth Deltas" setting under extras) and expect some time delay. If you are still not happy, replace you probes by thicker ones, move to a different meter (TC4 or Phidgets 1048 with the next Artisan). Best, Marko On 15.09.2014, at 16:36, Fred B wrote: