Steve, what is your opinion of your Bullet?

OldmatefromOZ wrote:I have some experience with the Bullet. Buying any small air roaster will not help inform your bullet roasts in any shape or form. Rob Hoos recently wrote a great article for roastworld about 150g sample roasts on the bullet. Assuming you have newer model with IBTS, this makes the 150g sample roasts even better at informing on what decisions you will make on larger batch sizes.

I have read that article from Hoos about the 150g sample roasts on the bullet. My thought process, correct me if I am wrong, is to use the KL to test and try to draw out what is possible with a coffee before putting my efforts onto the Bullet. With the bullet, I normally roast 1lb batches, which is quite a lot of beans. I understand that I can decrease the batch size, but there is a limit based on what Hoos wrote in his article. At a certain point (I believe he said 100g or less), the bullet becomes less than ideal for data collection. Since I am not a professional roaster nor do I have a lot of experience roasting, it is difficult for me to see what a coffee can become based on cupping and how to tune that. Hence, the smaller batches from the KL.

Another thing I dislike about roasting with the bullet is the setting up, preheating cycle and cleaning of the bullet. It is not easy because I have to store and clean it after every roast. The set-up and clean-up discourages me from roasting smaller batches.

OldmatefromOZ wrote:For reference my roast preferences align reasonably closely with Luca. TW light roast, Manhattan filter, Passenger and hit and miss with SEY over 1 year subscription with regards to being a bit under for my preference. But I also roast a lot of longer darker roasts for other people and have done so / tested many 1000s of this style, so I can set aside my "filter" style preferences. From memory the KL could do a pretty decent job on darker espresso roasts and there is a lot of happy people using it for this purpose, a major upgrade from stale supermarket roasts.

I too enjoy light roasts. I want to impart as little as I can from the roast onto the taste of the coffee. I want to taste the coffee for what it is, as longer and darker roasts diminish its natural characteristics and taste.

OldmatefromOZ wrote: I also agree that this idea of a "sample" roast quality does not matter, its just to inform green buying etc is not great. Rob touches on this in the roastworld article. When you start making small "profile" adjustments to a sample roast so that its not defective and delivers a light, aromatic and sweet filter brew whats the difference?

This in essence was what I was trying to do with the KL and would happily pay to buy some roasted coffee that would stand up in a blind cupping. I have no problems with being proven wrong and turn out I just didn't know enough about how to use the roaster.

In the end, would you say that the KL was able to produce good coffee or subpar despite all your efforts?

Not keen on making definitive judgements, I may have done this at times in the past and fell into many of the traps in flawed thinking Luca mentions.

Its like most things, you dont know what you dont know and there is likely untapped potential that someone else might discover, especially if the end user is given real time control while able to see the roasting beans...

As for the Bullet (Australian cert V2) appears to be well designed new technology (use of induction) with great potential. I like that the entire dump door is a big window with a flexible LED.

The fully integrated software is really neat and I found it intuitively easy to use. Initial impressions on data from infrared bean temp sensor: seems to be a more reliable way to get consistent results, especially in small batch roasting systems which are subject to huge variations and noise from probes.

luca wrote:Cools, thanks for doing that. That profile looks like the power wasn't flatlining and the target temp and actuals were tracking well, so the machine wasn't lacking for power. You were able to get there sort of at least with a 90g roast, so it sort of suggests that the 110V machine probably isn't underpowered, which would be the main concern for a 110V buyer, I'd have thought.

Found this under the "About this file" tab. During the roast, peak power was 1368.

luca wrote: I have no idea what this boost ... sorry, "BOOST", thing is going to be, beyond the glass roasting chamber, but one would think that, for the 110v KL, at least, there might be power limits that need to be addressed for a 200g roast. Maybe they have some firmware update to remove some internal limit to the power allowed through the element? Or maybe the 200g roasts simply take quite a long time?

I've heard its hardware and software but don't know if the hardware is anything other than the new see-through chaff collector. The other improvement is supposed to be an app integration with a wired connection. Not sure if it will allow during roast controls other than end roast.

From KL instagram, new chaff collector https://www.instagram.com/kaffelogic_roasting/

luca wrote:I'm kind of trying not to veer off topic, though - this is about the new KL, not the roest;

Yes this thread wants a larger scope. I've thought about splitting it into its own.

luca wrote:I don't really know what distinction people think there is between cupping roasts and filter roasts.

There might not be much difference. Cupping roasts establish a baseline of bean characteristics to differentiate different lots of the same farm, different farms of same region, etc. Profile roasts try to maximize the bean characteristics towards a personal taste preference. Obviously both want no roast defects and that can go a long way towards making them palatable. The minutia of it is probably beyond my roasting skills and palate, but I'd like to get there.

luca wrote:You can do a pretty easy thought experiment in the KL, or even an actual experiment (Steve and I did this) - imagine you took the KL probe and just bent the sheath down to position the probe lower in the bean mass, or even positioned it in one of the hot air vents. The probe would be positioned in the bean mass, but the readings would be a lot hotter. In fact, from memory I think Steve even had a go at moving the probe to an inlet to try to get inlet temp control, but the problem was a software/firmware limit on temperature for safety wouldn't allow him to set high enough temperatures (since the high temp ceiling is set assuming the higher probe position).

I've thought about checking my stash of probes to see if I there's one that I can hook up to Artisan while it's positioned at KL inlet. This will bypass any heat warnings. But it would have to be a non-invasive installation, don't want to disassemble any of it. A probe at the top of the chamber/chaff collector should be easier. Maybe I can do both and integrate all the readings into Artisan to plot on the same profile

luca wrote:I'm not quite sure what this question is, but the Roest calculates ROR over a 30 second span. I don't know about software smoothing for the individual readings.

What I was talking about was curve smoothing with a delta span of 30 seconds, not ROR in units of Degrees/Seconds. Do you know if ROEST probes are k-type or rtd? My old rig has rtd, skinnier probe thickness can't recall exact number, and I could do 8 second delta span and that would produce smoother data than the KL at 30 seconds delta span. From the source file of a log I see the KL is reading every second and in the log file it's called "mean temp" so that sounds like some sort of averaging is going on probably like Artisan's "oversampling" feature.

For prettier looking software interface, ROEST has the edge. KL Studio does a zoom in feature

Last 30 seconds of roast. The green area represents a boost I had incorporated in the profile but it's a negative boost, told the algorithm to do -1.5F for RoR. Still got actual temp to go above profile. Still tinkering.

hairyco wrote:In the end, would you say that the KL was able to produce good coffee or subpar despite all your efforts?

Might be a shot for you

OldmatefromOZ wrote:Im Steve who had the KL and went down rabbit holes with using different probes and positions / tearing the machine apart.

Nice of you to chime in. I've read your posts over on the KL forum. Do you still use KL? It's not on your equipment list.

ducats wrote:Found this under the "About this file" tab. During the roast, peak power was 1368.
image

I've thought about checking my stash of probes to see if I there's one that I can hook up to Artisan while it's positioned at KL inlet. This will bypass any heat warnings. But it would have to be a non-invasive installation, don't want to disassemble any of it. A probe at the top of the chamber/chaff collector should be easier. Maybe I can do both and integrate all the readings into Artisan to plot on the same profile .

Looks like you have good line voltage, someone on 110V might flatline the 5 min profile?

I actually did similar to what you are suggesting just did not publicly document it all. Getting a bare K type thermocouple to stay on an Inlet is tricky, even going at it from the inside with the machine apart.

Important caveat, it could be that the newer elements and or 110V elements are quite different, so my observations could be completely invalid here!

I found that depending on where the thermocouple was placed the temperature variations were quite large. The hottest part of the inlets where there was a build up of creosote, regularly went over 320C.(where the element glowed bright red). As opposed to the opposite quadrant reaching only 50 - 70C. So the system relied on mixing of different temperature air within the chamber to even things out and profiles with low fan RPM can be problematic or dangerous, anyone who has messed around with popcorn popper will know.

ducats wrote:What I was talking about was curve smoothing with a delta span of 30 seconds, not ROR in units of Degrees/Seconds. Do you know if ROEST probes are k-type or rtd? My old rig has rtd, skinnier probe thickness can't recall exact number, and I could do 8 second delta span and that would produce smoother data than the KL at 30 seconds delta span. From the source file of a log I see the KL is reading every second and in the log file it's called "mean temp" so that sounds like some sort of averaging is going on probably like Artisan's "oversampling" feature.

The drum temp and the experimental inlet temp are k type; the sensors in the roast chamber are PT100 RTD I think. The CSV file records temp values every second; the firmware on the PCB calculates ROR over a 30s interval, I gather. But what I don't know is if each of the individual 1 second point temperature values are themselves the result of averaging/smoothing or whatever.

OldmatefromOZ wrote:I found that depending on where the thermocouple was placed the temperature variations were quite large. The hottest part of the inlets where there was a build up of creosote, regularly went over 320C.(where the element glowed bright red). As opposed to the opposite quadrant reaching only 50 - 70C. So the system relied on mixing of different temperature air within the chamber to even things out and profiles with low fan RPM can be problematic or dangerous, anyone who has messed around with popcorn popper will know.

Oh yes, that's a good point, I remember that now. Geez we jumped through a lot of hoops on our journey! I remember I think you built your own sheaths for the probe to see if some sort of hardware smoothing might help, and I built an entire raised false bottom for the roast chamber out of stainless steel mesh so that the coffee wouldn't hit the hottest vents and so that, maybe, the air could mix in the bottom part before shooting up and contacting the green coffee. Steve, did they ever send you any roasted coffee? I asked them to send me some so I could see for myself what they thought a good roast tasted like and said I'd be very happy to pay for it, but they never responded, which is fair enough, but of course didn't really encourage me to keep going.

Ducats, I guess one point worth observing here is that Steve and I tried a lot of stuff for a long time before giving up. So like we're unlikely to answer any questions that you or anyone else may have with some sort of simple answer that gets you great roasts. If we were, we wouldn't have given up! So you guys are sort of on your own to figure it out, but please do report back if you get good results for light, aromatic roasts. I have a few friends with KLs and they'd appreciate knowing how to do it.

Steve, thanks for the early thoughts on the Bullet. Sounds positive. Also thanks for documenting your journey of the KL forum, very helpful.

Ducats, you only have 90 watts spare overhead in your last roast (1460w available, 1368w used), do you think that is enough to handle the proposed two thirds increase in capacity? I am assuming because you have a pre release unit you may have inside info.

No roasted coffee, they were very generous helping with parts to refurb the machine and then I moved it on.

Used it for quite some time with a 2.5mm sheathed probe in the bean mass, was not as noisy / needed less smoothIng and can not remember if this was a good thing or it just made me feel better.

For those interested there is fairly active unofficial group on Discord.

Here are a few images of experiments I've made with using different probes. It's a bit crude, but did the job. I think it would be fairly straight forward to locate a 2/3mm sheathed probe through the top casing bolt/hole and drill into the bean chamber. Will look to possibly try and do this with a PT100 probe. Though my current probe that I use on my gas roaster is particularly sensitive to insertion depth for accurate readings.






Cheers Phil

OldmatefromOZ wrote:Used it for quite some time with a 2.5mm sheathed probe in the bean mass, was not as noisy / needed less smoothIng and can not remember if this was a good thing or it just made me feel better.

Not as noisy would make me feel better as well. Any final thoughts on power profiling? Did you try it after the firmware came out that allowed zone 3 to be functional?