Roasting fewer beans makes the controls much more responsive. In theory, it would be a good idea to try to DIY version with the minimum amount of beans. Less heat capacity means more responsiveness. Maybe the DIY should start with 120g.
At $2200 and $70/kg, it will take 260 roasts to break even without even considering raw bean costs (which seem to be much cheaper). But that thing is a beast and I love the data. I hope it lasts a lot longer than that. And I assume you will enjoy using it.
Not necessarily. There are some exothermic and endothermic reactions going on there that will likely be less pronounced with the same airflow and smaller quantities. The crashes and flicks that this cause seem to be a common issue with near-max batches in most roasters, from what I’ve read.
For the fluid-bed systems the airflow needs to be much higher with larger batches so that likely negates that a fair bit, though.
Yeah, I agree, 120g seems like a pretty good spot, actually. I can see why people land here. It results in ~100-110g of roasted coffee which is 5-10 espressos or 5-6 cups of pourover. That’s a good size to easily be able to cup over a few rounds but it’s only ~$3 of green coffee.
For anything DIY, I think for me the primary goal would be good, reliable, accurate bean mass temp and logging. That’s going to be the easiest way to verify that the design is doing what it should. In theory, as long as you’re adequately mixing the mass, it’s following the same profile and you’ve got enough airflow to carry away smoke, the results should come out pretty similar regardless of the batch size… I guess the ideal goal would be something with a wide range of usable masses so you can sample roast at 100g and scale up to 500g-1kg. Failing that, having that be 2 separate setups with some stuff scaled would work. Or having a smaller one but automating it. I’m already imagining turning this thing into some kind of rube goldberg contraption that rotates through a full circle to dump the roasted beans and then has more dosed back in the top. If you had a powerful enough fan you could potentially just do that by launching them out the top, which would be hilarious.
Yeah, the economics aren’t too bad, really. I’d start with $2.1k because it included $100 of green beans. I’ve paid $145 incl tax for 5kg of decaf green beans so $29/kg. That has been dropping from 120g to 104g during roasting so losing ~15% water weight. That brings it up to ~$33.5/kg post roast. We’ve been paying $13-15 a bag depending on specials etc. for 200g of this particular decaf blend we like so $70/kg. $36.5/kg of upside to roasting the beans ourselves. 57.5kg to break even on the roaster investment. We drink an average of 28 coffees a week at 18g per dose so 500g/week, 26kg a year. ~2 years to break even. There are definitely worse toys to spend money on! Sadly our consumption isn’t quite enough to justify buying coffee by the full sack anymore as it’s split between decaf and caffeinated, but it wouldn’t take that much convincing!
Interestingly, it just occurred to me that I’ll have the energy it used logged pretty well because there’s not much else on that circuit running most evenings.
And yeah, I’m absolutely loving the data out of it. I’ve been sending the logs to my colleague who has a PhD in control systems. He’s getting a real kick out of it. I think I’m something like 4-5kg through it already. 2kg of the decaf espresso blend, 1kg of the decaf ethiopian, 1kg of the decaf brazillian, 1kg of the caffeinated ethiopian.
Consistency is foreign to my setup, I’m getting better though. One of these weekends I’ll get the temp probes working.
A main problem with tasting differences is the wait. You really do often need a week or more after roast for some bean flavors to manifest.
Somehow, I missed all these comments. I am getting 6.5-to-7.5-minute roasts on mine lately with 113.5 grams (4oz) of green coffee beans at a medium roast (outside temperature was about 80F). I am currently keeping a logbook and marking the temperature at every minute and also the starting outside temperature, along with 1st crack and when I pull it. This is with the flour sifter/wobble disc design.
I have not tested my sliding roaster design again since my last changes as I am mostly waiting on a calm day but did come up with an idea for wind break on the sides and it should also help to keep it hotter. Here is a mockup using tin foil which might also work, but since 2 sheets of 12”x8” SS is not that much I ordered a couple of sheets. I am going to get the sheets as close to the side of rotisserie as possible. All the plastic parts are mostly away from the heat now, although the rotary motor mount and idler mount might get a little warm.
I love how those projects get so incredibly out of hand. 
It’s really a joy to watch you work (and I don’t mean that ironically, I really like that!). 
Thanks. It has been fun to work on. Fingers crossed, maybe it will work next time.
Ohh I can’t do that the wait would kill i just spoon the Cristals into the cup and add hot water. 
Full strength 
https://www.reddit.com/r/roasting/s/M1GcoOY3ZS
Several kg fluid bed roaster with a 10kW 3 phase heater. Looks impressive.
Are you going to build one of these? He says the SS was not too expensive, but doesn’t say how much it cost him to build.
It’s tempting. I’m interested to see how he goes with temperature control because there seems to be quite a lot that goes into the control systems on my one.
Hopefully he posts a breakdown of what he has spent so far because it certainly doesn’t ‘look’ cheap.
Cheaper than the 3 phase power upgrade to my house.
I got it pretty close to working yesterday. I did 2 test roasts. The 1st one I just had the 12x8” SS sheets on the sides and open at the top and got the thermocouple up to 190c which was 20c hotter than before. The 2nd time I wrapped aluminum foil around the top to keep the heat in better, and it did get hot enough, but I never heard 1st crack. It got to 138c (280f) in 6 minutes which is above the temperature I was getting from the other roaster. Even though the thermocouple is a similar location as other roaster with the aluminum foil in place, it makes sense that thermocouple gets hotter. Here are 3 photos. I am not sure I will proceed further from here.
I see a couple of burnt beans here.
I’m going to add a USB connection to my Ryobi portable battery. I thought it was going to make my electrical connection less complicated, but the USB connection I got with the DDPS5005 Digital Control unit only allows either or connection to the USB or the 12v connection I have. It is not that big a deal to have another USB plug into the wall, but I did want to simplify the plugging for each roast. It certainly is more complicated than I first thought to wire it up, but easier setup for each roast as I can keep the USB wire connected all the time. I had to add a switch in order to cut off the 12v to the nema17 motor. I just realized I do have a button on my Arduino controller for turning the motor on/off, so that will give me 2 options to see which I like better.
I got ChatGPT to do the wiring diagram and after 3 tries, it got it pretty close and has the correct images for the parts. I had to erase a couple of extra ground wires it put on coming off the bottom ground wire to 1 & 2 of the 4-lever connector and the ground wire from the DSP5005 to the + line. It also did not have a couple of the + lines in the right place. These errors were easy to spot and correct for my use. At least I think I have the wiring correct. I am waiting on the LM2596 and USB panel from China before putting this together. I also have to redesign the 3d printed case to allow for these extra parts.
I did see one other problem with that wiring diagram. There should be no ground wire coming from the Rocker switch. I was able to put it all together today. I had a 16-gauge wire from before for the 12v lines, so kept most of that. From what I have read 20 - 22-gauge wire should be adequate for 1 or 2 Nema 17 motors on the Arduino. If I convert my other PS, I might go to 18 - 20 gauge where the 16 gauge is to make it less crowded.
I initially set the trim pot of the buck converter to 5.05v using alligator clips off the output +/-. After I connected it to the USB panel, my little USB checker showed it around 4.7v I think. I adjusted it further with that USB display of Voltage to get it back to closer to what I wanted. Looks like it is 5.07v if that display is correct. I also checked the voltage on the banana jack with the 12.03v showing on the display and it was 13v, so I guess that display is not accurate.
[edit] Looks like there might be a calibration menu that was not in the paperwork that came with it. I will have to see if I can calibrate it to a closer value.
I used a little hot glue to secure the USB panel as I could not get a tight fit with that. I only needed 3 port lever connectors for the + and - connectors and mounted them at an angle for clearance so I did not have to increase the size of case.
It turns out the DSP5005 display was correct and my 19-year-old GreenLee DM-20 meter was off. It was way off because the meter’s 9v battery was down to 6v which is my fault for not noticing that low battery on the display. Using the multimeter I bought last fall, that measured exactly what the display had. Replacing the 9v battery it was still off about 0.8% off showing 11.95v instead of the actual 12.05v. Think I was still in the mindset of using the old meter which is why I was not using the newer one.
Yeah, that’s a common issue with cheaper multi-meters. I have a couple of Uni-T ones that I use as ghetto dataloggers and they have a habit of going a bit weird with low batteries. Inaccurate and fluctuating readings. At least in your case it was clear that the batteries were low!
I’ve never had an inaccuracy issue with my Fluke or HP/Agilent/Keysight meters, right down to the meters refusing to turn on. The batteries also last MUCH longer in them, interestingly.
I might be going down the metal 3d printing rabbit hole some more and reprint this wobble disc in titanium. I had not thought of using titanium before, but it does seem to be food safe. It is more expensive, but the weight is about half that of the 316L SS and has some other good qualities. I have the price down to $92 before shipping. I changed my design to use M6 threaded rod and added dimples for the serrated washer to connect better. I also changed the thickness from 2.0mm to 1.9mm. Probably could go to 1.8mm.
I started down this rabbit hole when looking at flat burr coffee grinders. There is one that has an option for SS or titanium coated burrs.
Here is a screen shot of the modified wobble disc. The only visual difference is the bumps for the serrated washer to grab.
Here is what Google AI says about using titanium here.
