MVP - Product Design


Current Version of the MVP - Go here for complete documentation = $300 Food Computer

Now we must define the goals and requirements of the physical (and virtual) MVP that the community develops. I believe that of the user groups I listed in my previous post, MVP – Community, the two that will have the largest impact on this project in the near future are makers and teachers. It is for this reason that I propose we develop our MVP to be designed by the average hobbyist maker, and is intended to be operated by an elementary school teacher with the help of their class. It is important that we distinguish what user is the builder and what user is the operator, as these impact our design decisions. For example: a maker may not be worried about their system running without human intervention for a week at a time, but a teacher needs this automation to keep the plants alive during vacation or Spring break.

We have defined who will be the user of this product, which leads us to our second objective: determining our target recipe. It will drastically change our requirements if we are growing leafy greens, as opposed to a tomato. I know that everyone is probably getting sick of lettuce, but I do think that it will have the least system requirements and max potential learnings due to availability of data. @webbhm has already started the data model based upon the Cornell guide, which provides great setpoints for each of our variables:

I am making an attempt to define the requirements myself because I am hoping to inspire critical feedback as opposed to just passive awareness. My goal here is not to say that I know the best way to do something, but instead create a process by which all of us can work together to find the best (per our agreed definition) solution.

  • Product goals for PFC
  • STEAM platform for teaching about ecosystems and life sciences
  • Grow food based upon a standardized OpenAg recipe
  • Hackable, and modular system maximize hardware variation & scalability
  • Control as many variables as possible to maximize value of data
  • Product goals for MVP
  • Online remote sensor and webcam monitoring via mobile device
  • Be constructed with a “home-owner” level of skills and tools
  • Price - Cost less than $300, with parts from minimal suppliers
  • Product requirements of MVP
  • Brain - Should we reuse the PFC V2 Brain? Brain_box? Or something else…
  • Camera - Can we get by with just one? @Eddie Are two camera’s going to be necessary to take advantage of computer vision phenotyping software?
  • Enclosure - Aluminum? Simple wood frame? Foam Farm Box?
  • Lighting - No lux sensor, standardized distance and spectral output based on bulb
  • Reservoir - Determines footprint of the enclosure. Bus tub? Smaller? Just 1 plant?
    • Oxygen - None? (Kratky), Air pump from PFC V2?, or $5 Aquarium pump
    • EC - Manual control - Standardize by using same nutrients & weekly reservoir changes.
    • PH - Manual control - Anyone know of cheaper kits than this?
    • Water top off? - Sensor, peristaltic pump. While this isn’t absolutely necessary, I think it allows for this plant to be left unattended for much longer.
  • Environment
    • Temperature - Sensor, have people experienced that the heater is necessary? Maybe the heat from the lights is enough? Then an exhaust PC fan can cool when needed. Perhaps someone has an idea for evaporative cooling?
    • Humidity - Sensor, humidifier - piezo? Any other creative ideas for raising humidity cheaply?
    • Co2 - Does anyone know a good way to pull this data for cities? I’m thinking maybe we could
  • Software - Initially when PFC V2 stack, off the shelf IoT solution
    • Blynk - @campergeek @MrGadget have used this before with success. I also found there to be a dozen other plant growth chambers built using Blynk in their community.
    • Scratch? In the Media Lab they have a looping video of this being a UI for recipe building. @24Karrots is a fan of the simplicity and appeal to children. I also know it was created at the Media Lab.

Here are some existing examples, I will continue to add more as we find them.

Post 1 - MVP - Community Development
Post 2 - MVP - Product Design
Post 3 - MVP - Future of Process

Need help starting all this
MVP - Community Development
MVP - Community Development
$300 Food Computer
$300 Food Computer

Prototyping frames…thoughts?

Same dimensions same as PFC, we will be able to add the electronics to the top panel as well as a fan to regulate heat from the light.

Under $50, everything from a hardware store, no CNC required, we hope to do it all with just a drill, and a saw.


I wonder what an 80/20 frame would cost. I think the PVC would work well though.


@Poitrast I’d be curious to know what 80/20 would cost too, please cost it out and post it as an alternative.

One of the reasons we chose PVC instead of other material is that 80/20 can’t be cut by your average Joe. I’m a huge fan of CNC & have access to a metal shop but I think if it’s possible to expand our “customer” base, and it’s cheaper, then we might as well choose something that allows more people to build this.

Let me know what you think, I would not have gone through the effort of making this post if I didn’t want critique/feedback.

Another thought: I think one of our goals should be to use parts that are available internationally or at least to some extent. I know right now a big struggle is that people can’t get the parts for weeks for the PFC V2. How can we avoid that? What are materials that are generally available everywhere?


You may find the answer to that is “There aren’t any”, at least if you interpret “materials” to mean a complete BOM. That’s part of why I’m enthusiastic about more of a focus on teaching the knowledge that people would need to design a system from local materials to meet a given set of performance specs.


@Webb.Peter The 80/20 pre-cut through 80/20 would run ~$100-~$120 with the hardware to construct the frame. The advantage is they will cut all the pieces to length ($1.95/cut). 80/20 pre-cut from the factory would allow a user to construct the entirety of the frame with a hex key. I am unsure of their international shipping capabilities/rates. The question which remains is can the rest of the unit be built for under $200?


In case anybody’s wondering what 80/20 is and how it works, their gallery showing stuff people have built with their aluminum struts and connectors is awesome:


I hadn’t seen their quick frame material before. It looks as though it is about 3/4 the price of the t-slotted al.


To evaluate its suitability, we need a few more points on the enclosure requirements.
Eg, low cost, sturdy, easy to assemble, easy to source, thermally insulated

It seems to meet all requirements well. I considered adding, suitable for mounting electronics, but the components could sit beside it.
Also what’s the cost for the mylar?

I’ve used aluminium eq angle in mine, is ~1 USD per metre in bulk. Used 10 metres in my frame, which is much larger than this one (1200x700x500)

#10 the electronics were thought to be housed on the top panel, in an attempt to save space. Can you go into more detail about the additional points on the enclosure requirements?

Your proto sounds interesting! care to share a photo?

Reflective Roll Insulation (16-in W x 25-ft L) - Material used in the pictured proto above


Looks simple, easy and neat. My thoughts on it are

  1. I think one transparent side helps in getting inside view without opening the chamber.
  2. PVC pipes are easily available in India
  3. Mylar isn’t available easily in India from my first search, it is imported from US on amazon and ebay. Once a material is imported it cost becomes double so it will create a barrier for hobbyist. I am looking for alternatives and will update once i find out something. There are many products named Mylar Blankets which i am not sure what it is and if it can be used as a replacement.

EDIT:I found this item on amazon, I hope this is the one used in MVP.


Those are probably thin sheets of mylar that are meant to be used as emergency blankets.

I think the mylar material in @Webb.Peter’s photo is probably Reflectix. Reflectix has an inner layer of air bubbles trapped in plastic with reflective material on both sides–it insulates against both radiative heat exchange and conductive heat exchange. Due to the layer of bubbles, it’s also fairly stiff. A sheet of mylar would only insulate against radiative heat exchange (condensation might be an issue), and it would be floppy like thin fabric. The one time I messed with an emergency blanket, the main thing I noticed about it was that it had a strong, unpleasant chemical smell–that’s probably not the case for all of them though.


It is indeed reflectix. It’s surprisingly nice. I think it’s sturdy enough to support the ventilation fan even, which will save us from having to drill a bunch of holes and cut out a circle of plywood with a drywall saw.


That’s mine. Black ABS panels, PETG windows. 40cm deep aeroponics chamber, under that sheet, electronics will go below that. I still need to cut holes for plants.
Mylar is also not very common here. I’m using a car sun reflector for insulation and light reflection. Covers the stem zone like a blanket

What details do you want about the requirements? Those 5 are the only ones I can think about. Thought they might be helpful from a 1st principles approach, eg, easy to source = wood, therefore can we use more wood…


@wsnook I agree with you that we aren’t going to make an entire kit work for everyone, I think though that some products are going to be more generally available than others. I want to start with a plant (lettuce) then use a guide (cornell, other sources) to write the recipe. I plan to start with the needs of the plant and then based upon those requirements, and the defined criteria we’re talking (non CNC, hardware store parts), provide a product that can execute that recipe. I hope to be as transparent as possible about the reasons for decisions on the system design (hence this post and the discussion before just building).

In my head, a “recipe” is a set of requirements by a plant, whether or not an automated “Food Computer” carry’s out that recipe is another story. For the MVP we can design it so that many of the tasks that are automated with the V2 PFC are done manually, not only does this save us money but it provides an opportunity to learn about what the plant needs to succeed. I would love for the UI to incorporate curriculum and be more of a walkthrough that reminds you to check PH, gives you information about what’s happening to the plant (germination, when to harvest, etc.).

@Poitrast I agree with you that 80/20 is an incredible material and standard from which to prototype. I appreciate you taking the time to figure out the cost. In my personal opinion, I think that it’s a bit overkill for what we need and is pretty expensive compared to alternatives (PVC).

It seems like you have experience with LED design based upon your other forum post. I would love to get your input designing the lights for this MVP. We have been talking about trying to do off-the-shelf lights that wouldn’t require a separate driver. I know the V2 lights are adjustable by intensity and spectrum, I don’t think the MVP needs that flexibility. Instead what we need is 200 μmol m-2 s-1 of light over a 24" area. I have been considering and plan to test using 3-4 10W equivalent LED bulbs at 3000-4000k (this MVP is leafy green only). We could install them directly onto the plywood using regular lighting fixtures. Your prototype looks very slick! I’m curious how you sourced the ABS/PETG materials and had them cut. I appreciate your feedback on the Mylar, from what I understand car sun reflectors are the exact same material and can definitely be used as a substitute. For clarity, are you suggesting that we use wood because it’s readily available? I agree that you are correct, my only concern is again with reflectivity. Lastly, would you be willing to share pictures of your aeroponic pump system/chamber? I think it would be very cool to build a High Pressure Aeroponic reservoir that can be used with a Bus Tub, do you think that is realistic or would it be too shallow?


  1. I agree that a “window” is a great feature for showing off the PFC and getting a peak at the inside. That being said, one of the things the MVP will have is a camera, so you could also view plants from a tablet/phone easily. When considering something like an acrylic panel on the front I always get stuck on a few issues. The first is that of cost, that this will add another raw material purchase to the BOM. Do you have another material or idea for a transparent panel?
  2. Good to know, please continue to let me know if you can/cannot source parts we’re talking about.
  3. That “Reflectix” looks like it would work great. Would this count as an “imported” item, or would it be affordable?

@jimbell @drewthomas89 Thanks for answering questions! I really appreciate everyone’s feedback, the more of us that contribute, the better the end product.


I took a trip to Lowe’s this morning, intending to get materials to build a PVC box/frame; but came away questioning the MVP solution. I thought I would get other’s input before moving forward.
The nice thing about the foil is it is easy to cut, provided good insulation and has a nice reflective surface. The down side is that it lacks rigidity and requires another structure to support fans, lights and anything else.
Lowe’s wants $23.16 for a 24" 25 ft roll of foil, of which you need about 14 ft. The frame is 4 10 ft lengths of 3/4 pipe @ $1.80 and 8 connectors at $1.60. Plus there is a 2x4 ft piece of plywood. The total is about $50.
I was comparing that to two4x8 ft 1" pieces of foam which are $15.79 ($32 total). This doesn’t have the reflective surface, but is easy to cut with a drywall saw (or knife). The down side is hauling around a 4x8 sheet (or cut it in the parking lot).
I did not include the cost of tape, etc.; as I figure these will likely be about the same for both.
Do we want to stay committed to the foil, reconsider this, or go in several directions and try out alternatives? I am wiling to do a foam alternative, keeping it to the same dimensions. Besides price, we are likely to find some little surprises, like cutting holes for fans, as we go along (with all options).


I think my first prototype will be more suitable for an MVP model

It was built on a shoestring budget with common parts. The walls are masonite. The frame is a steel cornice bead(quite a common part). It was designed to match the size of the plastic container it sits on.

I covered the masonite with a layer of corrugated cardboard, for insulation, followed by a layer of regular kitchen foil, and then an adhesive plastic wrap, meant for covering books.Not pretty but functional


Thanks. I Googled thermoplastics manufacturer and found one in my vicinity. They did the cutting for me
The wood is easy to source, cheap and easy to use. See my hydroponics prototype in this thread.

That’s the tub I was testing the aeroponics in. The bus tubs are definitely too shallow. They need to be around 280mm deep, from my research. The depth of the tank for the old food server would be a good reference. The depth needed for aeroponics is the only reason my prototype above is not a desktop model like the PFCs


Now that hardware is firmed up (like Jello), I want to look at the software needs. My concern is defining the requirements, and less with particulars of language; though I will use some specifics as illustration. I am looking at how small of a ‘footprint’ we need, and trying to keep it as simple as possible. This architecture has some growth potential, but will not scale to the level of the PFC. I wish there were some good options (like Blynk or Scratch), but these seem to hit a wall at some point (explained in another post).
For the MVP, it seems to boil down to:

  • Reading a temperature sensor
  • Controlling a ventilation fan
  • Logging measurments (temperature, humidity, fan on/off)
  • Capturing pictures

These seem to break down into three categories of code:

  • Execution time control
  • Applications
  • Services

The Raspberry Pi with Linux comes with Cron - a time based job schedules. Nothing fancy, but it is reliable and works. This is my suggestion for controlling the logging events, thermostat and taking pictures.
There are two ‘services’, applications that need to be called from other applictions and should always be available. These would be the temp/humidity sensor, and a logger. At the moment I am thinking of an HTTP service, possibly using Python Flask.
The applications which Cron would control would are a picture taker, temperature logger, humidity logger and the thermostat. The thermostat would control the fan, and since Raspberry GIPO pins can persist their state after an application terminates, there is no real ‘fan’ code needed - just set the pin voltage.
Whether the logger persists to a file or database is to be determined, as long as there is a persistence interface, the storage can be changed ‘behind the scenes’.

This is still missing some needed features:

  • Manual data input (phenotypic data - use a spreadsheet?)
  • Image manipulation (currently just dumps to a directory)
  • Remote access (where Blynk was nice)

I think this covers the MVP needs in a minimalist fashion. If there are no significant changes, we can try a build.



Blynk is great IoT software, in a few minutes I can install it on a Raspberry and connect things up with my phone so that a LED will turn on and off with a press of the button on my phone; all with no coding. There is a great graphic interface on the phone for dragging and dropping widgets.
I installed Blynk on my Raspberry (actually twice). It comes in two flavors, either as a Java application that runs Java Script, or as a C++ application that runs C++ code.
Here is where things get fun: the JS has more pre-built code, but a lot of the sensors come with pre-written C++ or Python code (the later is currently not an option for Blynk).
The real problem is that the MVP is not an IoT device, and the widgets we need, need to be on the Raspberry (or other device). It does me little good to have a thermostat software widget on my phone when it only runs while the phone is connected to Blynk. It appears that if I need a thermostat on the Pi, I have to write it, and manage the integration and timing of all my devices. This is a major commitment to programming (C++ or JS), and avoiding the coding was the initial goal of Blynk.
I am going to step back from putting much more work into this, and think through the larger software architecture picture (see this post for some issues: Considering alternatives to the Raspberry Pi and Arduino). We need to define the core software needs:
Sensor reading (Si7021 using I2C bus)
Fan controld (analog pin control with thermostat software)
Camera image capture
This is the core, and it needs separate threads for these functions as well as scheduling.
I think that Blynk may be a great option for Spring Break water control - push a button and a peristolic pump starts adding water running; but I am currently not seeing it for the core heavy lifting. I wish it was, as I really like the phone interface and where it is going. The other problem is that it is not open source, and there are (or will be) fees associated with its use.