Designing an open source coffee roaster


1. Introduction

2. Small commercial coffee roasters

3. Interesting bits and pieces

4. Draft Design for an open source roaster

1. Introduction

Some day @anu and I want to open a small roastery company in Nepal, and for that we eventually need a roaster. Below, we evaluate suitable models we can buy or modify or build. If we were to buy one in Nepal, it would be the Top Grains Intl. 5 kg LPG fired roaster, imported from the manufacturer in India. If we were to build one, we’d build our own, fully automated, open source coffee roaster with 1-2 kg batch capacity – a draft design for this is included below.

However, we’re not actively building or buying a roaster at the moment – for the start, there are enough roasters in Kathmandu with spare capacity who agreed to roast for us as a service.

Anyway, this is an open source project, and anyone is welcome to contribute to the roaster’s design and to the knowledge collection below. (For that, this very post is a “wiki”, and after a few non-spam contributions a new user automatically gets the permissions to edit it.)

2. Small commercial coffee roasters

Roughly ordered by suitability for a small roastery in south Asia – the most adequate ones first. Contains both models that are ready to use, models that would have to be modified for commervial operation, and models that offer design inspirations for own creations. With a focus on suppliers in India and China, but also including suitable (i.e. relatively cheap) models from around the world.

Model Roasting Capacity Heating Options Price Description
Top Grains International Alfa Coffee Roaster 5 kg gas
185,000 INR
(2400 EUR)
4.5 kW heating. Indian product. Drum roaster, so can be adapted for automatic operation. This is our preferred roaster: the manufacturer is responsive, sends you a proper quotation, can shup in wooden packaging to Nepal. We visited two roasteries in Kathmandu who bought this machine and are happy with it – so this manufacturer is "for real", if you have a genuine contact person of course.
Ace Pack Machines Coffee Bean Roasting Machine 5-10 kg gas
190,000 INR
(2480 EUR)
1.1 kW electrical power and 5 kW heating power. The given 10 kg loading capacity is questionable, as 5 kW heating is more for a 5 kg model. The given image does not apply to the product, it was taken from the Proaster 500 g shop roaster.
China Small Machinery Supermarket Roaster 3 kg
5 kg
120 USD
140 USD
A removable, electrically rotated stainless steel drum with side fill that is used on a regular gas stovetop. Very interesting design. Basically the simplest design possible. Building that by oneself and adding automatic controls via Artisan Roasterscope could do the job. It would be combined with a gas cookstove into one unit. Also, fume extraction has to be solved, probably by connecting something to the centerhole.
Aillio Bullet R1 1 kg electric ~2650 USD Energy efficient, compact 1 kg roaster. 14 kg machine weight. 1500 W electrical heating. Drum roaster, so can be adapted for automatic operation. There is an interesting interview giving some insights about its design. Monthly capacity: 100 kg (equiv. to 40-60 min operation per day). It can roast much more, but will need replacement of the drum motor, belt and fan motor at some point then (source). Also there is a one-hour long "IR window cleaning procedure" to run after each 3-6 kg roasting throughput (source), so it's quite labor intensive for commercial production. However, people do it successfully in parallel to serving coffee (see). Can be computer controlled via USB, using their proprietary Windows and OS X software. Not supported by Artisan so far, but reverse-engineering it for that is not excluded.
Dalian Amazon Mini Gas Coffee Roaster 2 kg gas 2000-4800 USD Price depends on order volume. USB connection available for data logging (seemingly not for control).
Newest machinery Roaster 2.2 kg gas 2500 - 4900 USD Price depends on order volume. Also available as 1 kg model, and as larger models. 130 kg total weight.
Hottop KN-8828KG (Taiwanese copy) 1.25 kg electric 5000 - 6000 USD Also called "KN-8828B KG". Nice compact design, 25 kg machine weight. 1400 W electric heating. Seems to have a USB port at the right front corner (to be confirmed), not sure if it can be controlled by Artisan though. However, uses 110 V electricity (but might be available in other variants as well). Too expensive for a copycat model.
Proaster 1.5 kg Coffee Roaster 1.5 kg gas
9000 USD Perfect form factor. Drum roaster, so can be modified for automatic batch operation. Too expensive.
HQT Coffee glaster 500 g gas 680 USD Used on a regular LPG stove. WIth a separate, external cooler. Interesting design, but too expensive for what it does.
HQT Coffee Mini Coffee Roaster 250 g gas 2500 USD Interesting design, but too expensive for what it does.
Sonofresco 2200 1 kg gas ca. 3700 USD Fluid bed roaster. This type is not very suitable for automation of feeding / draining; however, a fill tube could be permanently added, and also a suction hose for removing the beans.
Ceroffee CRF-800 N/A electric N/A Small electronically controlled desktop roaster by a Chinese manufacturer.
Hottop models (original) 250-300 g N/A ~1600 USD Reviewed with best results among four competitors. However, meant for home roasting and not for commercial operation. Has to cool down 25 min after use (see). The version with USB connection (Hottop KN-8828B-2-K+) can be controlled by Artisan. A 1 kg model has been announced in 2013, but has not been released since.
Small Helmsman 250 g gas
150 USD Hand cranked small coffee roaster to put on a regular stovetop. Interesting design for personal use. Needs electric drive though.
Zjmzym BN90 / Zhoufeng ZF-90 (etc.) 600 g gas
48 USD Interesting design. Could be used / adapted for personal use. Same product but in a funny casing with electronic control and electrical heating here.
Rajarajesjwari Engineering Works Coffee Roaster N/A gas N/A Interesting model because it's a hand-cranked, so supplying only the mechanical hardware while we could add all the electrical parts and software ourselves (based on 24 V motors and fans for safety).
Multi-Type Roaster N/A N/A N/A Interesting design with filling and release from the side of the drum (from top / to bottom), avoiding the precision manufacturing needed for a rotating drum with a fixed front and back wall.

Finding more roasters

3. Interesting bits and pieces

  • Open source roaster control software for desktop computers. With and without sensor input. Note, these sensors can be retrofitted to existing roasters.

  • Roaster control software for Android.

  • WLANThermo. An open hardware, wifi-enabled, 8-channel thermometer. It comes in different variants that are also available for sale as ready-made products for 100 - 300 EUR. Their website is in German, but there is a good article on Hack-A-Day in English. Quite similar to the Arduino-based TC4 shield; in some aspects more elaborate, and in some less so.

  • Using a BBQ thermometers. If you don’t require or prefer an open hardware thermometer to interface with a roaster control software, there is quite a selection of suitable kitchen related thermometers to choose from:

    • kitchen / barbecue (BBQ) / baking thermometers, usually measuring 0-300 °C which is sufficient for roasting; AliExpress has cheap ones like this and its related products
    • Ventus W042 barbecue and baking thermometer. This has a Bluetooth connection, which can probably be hacked to talk to your own Android app. It is available for 41 EUR.
    • cheap dual-probe thermometers from AliExpress, like this
      TFA Dostmann 14151201 barbecue and oven thermometer, available for 39 EUR
  • DIY coffee roaster designs. Lots and lots of them, great for inspiration.

  • Open Source Coffee Roaster. A large drum roaster. So far the 3D designs are ready (but had not yet been released as open source content), and they started to build on the prototype. The original page about the project disappeared, seemingly during a website redesign, so the link goes to the version.

  • coffee roasters based on a Raspberry Pi and popcorn machine parts (multiple models)

  • very nice DIY coffee roaster (electrical, 1 kg capacity)

  • Nirvana Machine. A very nice, very simple coffee roaster. Based on an electrically rotating colander and a camping gas stove. Comes with a full set of instructions. Useful for personal consumption, not for commercial operation.

  • nice chart of an industrial installation / production chain for coffee

  • energy consumption decreases with larger batch sizes in industrial roasters

  • coffee bean cooler made from a popcorn maker, running on 12 V DC

  • afterburner for smoke removal; basically route the exhaust smoke through a flame to burn off the smoke particles and have less smoke. See description at bottom of the page here.

  • interesting drum roaster design, with the gas burner fixed to the drum and the drum rotating for filling and pouring; might be good as a template for a DIY design

  • GoSol SOL5 solar coffee roaster. The GoSol SOL optical concentrator with attached roaster is reportedly already an open source design. Currently, the roaster has 5 m² mirror surface, yielding 3 kW thermal power at 60% efficiency. This should be enough to roast 2.5 kg of coffee beans in 20 minutes (the maximum recommendable time), when compared to a 6-7 kW commercial roaster for 5 kg coffee beans. Required changes to make it fully automatic would include predicting a 20 min clear sky “window of opportunity”, filling the machine, roasting with automatic temperature control, and cooling the roasted beans afterwards. See the following additional materials about this roaster:

  • Barefootcollege solar coffee roaster by Barefootcollege

  • Explanation of a chaff collector. The proper title is “cyclonic dust collector”, and they are widely used for all kinds of air cleaning purposes. Note that those attached to coffee roasters normally don’t have the motor shown here, as the coffee roaster itself will have the motor integrated. However, integrating the motor here instead could make sense if routing the air first through a cooler, as it would allow a much simpler fan that does not have to be heat-resistant.

4. Draft Design for an open source roaster

Based on the above research, I came up with the following draft design for a simple, gas powered 2 kg roaster with automatic operation (means, automated filling and pouring).

  • Should be a 350 g - 2 kg roaster that is fully automated, using temperature probes and software to control emptying and filling and roasting. This way, the roaster can be small (350 g - 2000 g capacity). This way it can be cheap, allowing to scale the operation by using multiple ones in parallel. Just as with all other machines, technology can be much smaller, so much cheaper, when it does not need human attention and can run on its own. Because then, wage costs are of no concern. The same applies to roasters. Calculations:

    • If the roaster has 350 g batches (the typical “sample roaster”), it can process in terms of green coffee beans: 350 g / 18 min = 1.17 kg/h or 28 kg/d or 196 kg/week or 9.97 t/year.
    • If the roaster has 700 g batches (“double samples”), it can process in terms of green coffee beans: 700 g / 18 min or 2.34 kg/h or 56 kg/d or 392 kg/week or 19.94 t/year.
    • If the roaster has 1500 g batches, it can process in terms of green coffee beans: 1500 g / 18 min or 5.0 kg/h or 120 kg/d or 840 kg/week or 42.7 t/year.
  • For redundancy, it would make sense to use two smaller automated roasters instead of one twice as large (e.g. two 350 g roasters vs. one 700 g roaster). Given that the service of roasting may be charged at 1 EUR/kg, the small 250 g machine could earn up to 10950 EUR per year (minus energy costs). Which would already be a complete business in some countries.

  • For energy efficiency, the roaster would be very well insulated, and heat in the exhaust gases would be re-used through a heat exchanger that heats up incoming air.
    The roaster should have a packaging machine in addition, and be installed at the sales location. It would “notice” how much is sold (taken away from the output queue) and produce only as much as needed to have around 6 packages in stock. This way, customers would always get the most fresh coffee possible.

  • Should be LPG heated, based on a normal off-the-shelf gas stove, either the size for normal household kitchen use or a bit larger (“Hockerkocher”).

  • Should be made from a stainless steel (or normal steel) drum, with the electric motors and all other electric components operating on 12 V DC input for safety. A normal AC adapter of a desktop PC can be used for power conversion. 12 V electrical motors can be used from car parts, so are widely available.

  • Should be mostly built from simple metal parts (sheet metal, tube, square profile …) and otherwise only 3D printed parts. The advantage here is that 3D printing is available at least in the capitals of all developing countries. CNC milled parts from plastic or aluminium, and molded metal parts from aluminium are also possible.

  • Needs an innovative mechanism for fully automatic loading and unloading.

    • Proposal: a drum with only one open end and the gas burner attached near the drum, similar to this roaster. Adding to that a tube as long as the roaster and on the opposite of the pivot. It will position close below the input funnel while allowing the roaster drum to pivot 360°. Three positions are used: vertical for filling, pointing 45° downwards for pouring, horizontal or pointing slightly upwards for roasting.

    • Proposal: the drum should be cylindrical, with one funnel attached at each end. The one at the pivoting end to pour, the other one to fill. The outside diameter of the funnel would be the same as that of the drum in both cases. The drum will turn to vertical around one end automatically (with a counterweight for simple movement) to pour the roasted beans, then close the lower end and pour in new beans from the top. While resting in roasting position, a suction hose will draw in any smoke and chaff from the pouring end, and run it through a chaff collector and exhaust. A great real-life example of this design is this small Chinese shop roaster; also this DIY roaster is similar.

  • The smoke would be pulled through glass wool as a smoke filter and then through 30 parallel copper pipes in a large water bath. This allows to re-use the exhaust heat for water heating, and to use a household vacuum cleaner as suction device.

  • Or even better, we don’t filter the smoke but burn it together with the LPG that powers the roaster, like a rocket stove. This prevents most of the air pollution otherwise created by the roaster. Any chaff contained in the smoke would also be burned, so no need for a chaff collector either. To extract the smoke, it seems better to push in air than to pull it out, as no heat-resistant fan is needed then.

  • There would be no blower mechanism for cooling the beans, instead they would be poured on a piece of sheet metal with many holes in it, in such a way that they distribute over a large area. They can then cool in open air. The reason for active cooling in commercial roasters is usually that they are operated indoors, so the smoke has to be extracted. Also because 5-50 kg of beans per batch will be many layers of beans, needing stirring to be cooled. These are no issues for the roaster here, which will be operated outdoors and only produce 1-2 kg of beans per batch that can be easily cooled in a single layer.

  • Even simpler alternative for cooling: Instead of cooling the beans fast using active cooling or spreading them into a single layer, we will simply calculate the additional heat of slower cooling into the roasting time and stop the roasting earlier. We can simply switch off the gas supply and rotate the drum until the beans are cool. Then pour them directly into a storage container. Since operation is fully automated with this roaster, the longer cooldown period and the inability to roast in parallel to that does not matter at all. With this setup, the drum should be made from light stainless steel sheet metal material, to not waste much heat on the repeated cooling and heating.

  • Should be controlled by a Raspberry Pi. The Pi would run Artisan Roasterscope itself, and would be controlled with a touchscreen.

  • A sheet metal sleeve in 1 cm distance around the rotating drum should be used, and insulation around that, similar to the pot sleeves. It will increase energy efficiency by 30-50%. When pouring the beans out, the sleeve will be lifted, too. It will not rotate with the drum though, and have a large hole at the bottom for the hot air of the stove. There can also be air guides to distribute the hot air from the stove evenly to all parts of the drum.

  • If a chaff filter is needed, it should include an automatically operating one. Not of the vortex type particle filter because these devices are very large, but a perforated stainless steel sheet as a filter, with an automatic brush to clean it from time to time.

  • The whole machine should be small, fitting into a pallet compatible box with a 40×60 cm “Eurobox” footprint. The box should be possible to be shipped as a parcel and as checked baggage during air travel. It should be light enough to carry by one person, allowing to quickly set up the roaster outdoors for operation. By operating it outdoors (for example on a flat rooftop or under a protective roof in a garden area), smoke extraction devices become completely unnecessary.

  • The machine should include a funnel large enough for either 30 kg or 60 kg of green coffee beans (one half or one “bag of coffee”, the usual unit of packaging and accounting for green coffee beans; see The Coffee Exporter’s Guide, p. 16). 60 kg at 1.5 kg per batch would allow for 40 roasts, taking about 13 hours. 30 kg would take 6.5 hours – enough to fill one full shift of a worker when also considering setup time and cleaning. The funnel can for example consist of the upper part of the box (all five sides except the bottom) which is turned around and placed on top of the roaster that was inside the box. The bag that contained the green beans can then be placed below the roaster (assuming it has some kind of extensible legs) so the roasted beans can pour into it.

  • To always use the same batch size, there would be a vibrating hopper or a small conveyor below the funnel, pouring the beans into an intermediate vessel on digital scales. This also automatically serves as the valve to start and stop the flow from the funnel. It might be simpler to connect the digital scales directly to the roasting drum, with a suitable counterweight for the vessel mass if otherwise exceeding the scales capacity.

  • Should be made from just a few types of simple materials. Just stainless steel sheet metal, and steel square tube for the frame. The frame material would allow some options, including T-slot aluminium profiles. Also all electrical parts (motors, microcontroller etc.) have to be of a generic type and easy to obtain. Control electronics should be configurable to be adaptable to slightly different parts (e.g.adjust rpm speed of motors etc.).

  • A 1.5 - 2 kg automatic roaster seems to be adequate for a small roastery. A 2 kg roaster can roast at least 8 kg per hour, so 80 kg in a 10 hour day. It has to be observed for fire safety, but that is easily achievable with even one person (with “work / life integration”). 80 kg per day, sold at 2 USD/kg profit, is certainly enough for 1-2 workers in most countries. Also, the roasting time can easily be extended to 16 hours a day in two shifts, resulting in roasting 128 kg per day, resulting in 100 - 105 kg roasted beans.

  • It should be possible to produce this machine in Kathmandu. Means, if welding stainless steel is not readily possible, it should be bolted or riveted.


Is the need to do this open source as a way to drive to expense of the roaster down? And if so, how cheap could you put one together in Kathmandu compared to these 2-3 kg roasters described above? $1K vs $2-5K?

That, and all the other good reasons: simple repairs, ready spare parts availability, avoiding vendor lock-in for the product or any spare parts. 700 - 1,000 USD seems possible. (And I’m not sure about the prices of the Indian-made models, they are probably more expensive than the advertised ~2500 USD.)

A major reason is also that the roaster would be designed for 16 hours of unattended operation per day. Commercial roasters are only as large and expensive as they are (often 30-300 kg per batch) to get fixed cost degradation for the human worktime involved. That changes when the machine can roast on its own batch after batch – then, the machine can also be smaller and cheaper.

It seems that this new paradigm of small but automated machinery has not yet arrived in the roaster business world. So we need to either retrofit an existing drum-type roaster with automatic feeding and pouring and electronic control tech, or design our own. Not sure yet, but I’m not keen on the work of designing an own roaster just yet.

1 Like

Got it. This is very interesting stuff…

1 Like

There’s only one @matthias :smile:

@matthias Which roaster machine do you recommend for Nepal? :slight_smile:

I recommend a gas-heated 5 kg roaster without electronics. Such as the first two entries in the list above (you could write to the manufacturers in India and ask for the current prices … if the prices are really as in the list, and quality is good, it’s a good option).

These roasters are sturdy and durable and can always be modified later, including with self-made electronic monitoring and control (using the open source Artisan Roasterscope software), and even automatic batch operation. But they are also efficient enough to work with without doing all that, as they roast in 5 kg batches.

The only problem is, they are quite heavy. Around 130 kg. Not too easy to move to a new place.

1 Like

Hi Matthias,

Is there an expected date when you finish with this open source roaster?


Hello and thank you for reaching out. Hmm … at this time I consider the “open source coffee roaster” rather an exploratory exercise. We have more than enough to do with our current project, the open source coffee sorter (with a first machine supposed to be ready by 2018-04-30).

For our (= @anu and me) roastery project in Nepal, we intend to buy a simple mechanical roaster from an Indian manufacturer and will later extend it with open source electronics (Artisan Roasterscope, see the article above). To me, this seems a good compromise for a self-repairable, self-manageable, low-cost small roaster.

A fully open source roaster would still be better of course. You are welcome to take over the idea, it’s open source after all :slight_smile: