Category Archives: Particle Photon

Optimizing the greenhouse cooling system

My current cooling solution is composed of the following:

  • 15″ x 12″ intake water-cooled heat exchanger
  • 12″ exhaust fan
  • 8″ water-cooled heat exchanger with 240mm fan x2
  • Geothermal loop 5-6′ deep
  • Soil bed and aisle loop 3-4″ deep

Since the beginning of spring, I’ve been optimizing the cooling system.  Now that it’s summer, I can test it’s capabilities.  Here are some of the principles I’m trying out.

Variable setpoint

In my area day/night temperatures can swing violently.  7C at night, and 31C during the day isn’t uncommon.  It will become more consistent as summer progresses, but for now, I want a system that will change the setpoint based on the fact that nights are colder.  I want the system to store more heat during the day so it can keep things warmer at night.  So I have a bit of machine learning built into the system that takes into consideration the night time temperature and allows for a higher setpoint during the day to store more heat.  I have to keep the setpoint within livable conditions for my plants.  Right now, I’ve got 9C as my lowest setpoint, and 34C as my highest and the “ideal” setpoint at 24C.  For every night the temperature drops below 9C, I raise the running setpoint by one degree.  In the winter, that will probably mean that the temperature could reach 34C and maintain there for a while.  In the summer, if the nights are warm, the setpoint will decrease as long as it’s above the 24C.

With this type of “learning”, I shouldn’t have to ever manually set my setpoint.

Oh, and one more thing.  If the setpoint is kinda high -too high for humans to comfortably work in the greenhouse, I have a rule where if the human lights (a white LED strip) are on, change the setpoint to something that a human can tolerate.  For me, that’s 23C.

CO2 On-demand

The rule for this is simple.  If the temperature is below the variable setpoint and it’s daytime, burn some hydrocarbons and make some CO2.  If it’s night, only turn on if the temperature drops below the minimum threshold (9C).  Plants only need CO2 when there’s light, so trying to maintain a certain CO2 level at night will become expensive.

Exhaust only when needed

I have my exhaust set to turn on when the temperature exceeds the maximum threshold of 34C.  Why?  Well, if there’s a certain CO2 level, and it’s 25C (vs the 24C setpoint) we don’t want to exhaust the CO2.  That’d be a waste!

We will, however, exhaust the CO2 if we get a gas alarm from the harmful gas sensor.

Variable speed intake

My intake has 6 200CFM 120mm fans controlled by a photon which can PWM a MOSFET to control how much voltage the fans get.  Using a PID algorithm and the variable setpoint, On the photon, I get 8bits of control resolution (256 steps) so I’m able to suck in just the amount of air needed to cool.  Almost all of the cooling work is being done by this and the other heat exchanger.

Below you can see how well the intake system tracks the temperature (right graph) and also see when the exhaust fan kicks on (left graph).

intake and exhaust

Heat exchanger for geothermal

I redesigned my geothermal system a bit.  Instead of using the same reservoir as the irrigation, it now exclusively uses a 50 gallon drum and a small 5w continuous pump.  This pump pumps water through the geothermal system and back 24hrs a day every day.  In addition, I moved the water cooling pump (the pump that pumps through the intake, LEDs, CO2 generator, etc) to use this dedicated drum as well.

This change allows me to add a bit of chlorine to the water to keep algae from clogging my tubes.  It also frees up my irrigation reservoir for liquid-based feeding (more on that in a future post?).

To compare, here are the designs for both the old system and the new system.

Trip's greenhouse water system

New system:

Trip's greenhouse water system (3)

Attached to the geothermal line is my little 8″ x 8″ heat exchanger with 240mm fan.  The fans don’t push a lot, but it’s the quality that counts.  The air coming from there is cooled from water deep within the earth.


I don’t open my door

With this setup, I have never had the need to open my door so far.  This hopefully keeps pests out and with them, disease.


Over the next several weeks, I’ll be trying to test the potential of this system.  Today’s high was 27C.  My high temperature in the greenhouse was 36C right around the time I increased the aggressiveness of the intake PID algorithm and reset the system (see the missing data on the graph above) .  With that aggressive setting, the intake still only hit 210 pulses out of a possible 255.  My next goal is to optimize the PID settings and tune it better.  This will further give me an idea of the systems capabilities.

LED Stair Lights

Stairs are pretty boring, but they don’t have to be.  I convinced my loving wife that she needed stair lights.  I put the project off for over 6 months while I’ve been greenhousing, but now that that project is more completed, I have time to get back to house projects.  Let’s get building.


First, we will use the APA102 lights.  These are individually controllable.  They are the same lights we’ve used in our ambilight project with the minnowboard max.

I came up with a simple light protocol that supports “instructions” rather than just raw pixel data so it’s fast and light.  I’ve published the library here on github (be sure to use it with my forked Adafruit_dotstar library which has the “driver” for the LightProtocol).  I loaded that onto a particle photon, combined with an level shifter, and powered the thing with a Drok DC-DC power supply and a 24V 5A AC-DC adapter.

To install the LED strips on the stairs, I picked up several aluminum channels from and the corresponding “frosted” covers.  To stick the channels on the stairs, I used 3M automotive double-sided tape.  The aluminum can be drilled and screws can be used to mount, but I didn’t do that.  I used a simple dremel to cut the channels where I needed to.

I’m hiding the photon and the power supply in the closet which is adjacent to the stairs.  I cut a small hole in the wall on the closet side and put a 4 wire, 14AWG cable through the wall.  This is low-voltage (5V), so you don’t need an electrician or a expensive permit to install inside the wall… at least in my area.  On the other side, I combined a 2 socket “keystone” faceplate with a couple two wire speaker jacks.  This doesn’t look half bad.


Using the same python library as the minnowboard max ambilight project, and adding a “driver” that can speak our “LightProtocol” that we’ve installed on the particle photon, we are able to to complex effects and themes on the desktop and change the lights over wifi.

I have three effects coded up: “Chase”, “Random Rainbow Transforms”, and “Rainbow”.  Check out the video for how these effects look on our lights.  What other cool effects can we do?


Building an Internet connected irrigation system using the particle photon

The particle photon is a cool little wifi device.  It’s relatively inexpensive, it connects to the internet, and I’ve been using it for several components in my smart greenhouse project.  I needed to add a second irrigation system so I recorded how I built the controller and wrote the basic code to make it “web enabled”.

This video guide is incomplete, but it gives the basics on how to get up and running with your own web enabled internet device.


Cost breakdown:

Particle Photon – $20
2-port Relay – $7
Electric valve – $30
Case – $9.30
12V power supply – $5
12V -> 5V regulator – $1
Level shifter – $1.50

Total: $73

As a comparison, the Cyber Rain Residential Series 8 zone is $500.  It does control 8 zones, and ours can only control 2.  It also does not come with the electric valves like ours does.  To make a better apples to apples, we’d have to remove the valve from our list and add a bigger relay with more channels.  This relay supports 8 channels and is $9.  We’ll also need an additional level shifter adding another $1.50.  Here’s the new breakdown:

Particle Photon – $20
8-port Relay – $9
Case – $9.30
12V 2A power supply – $5
12V -> 5V regulator – $1
2 x Level shifter – $3

[Apples to “Apples”] Total: $47

Granted, the software of the Cyber Rain is superior.  It supports advanced timers and monitoring.  We don’t have that, but we can add all that by writing more code.  Let me know if you want me to do a walk through on how to add more smarts to the irrigation controller.  I’ve already done it for my setup.

Making Smart Grow LEDs Even Smarter

We know that full sun plants need 6+ hours of direct sunlight per day.  We also know that direct sunlight is about 30,000 to 100,000 lux.  We should be able to say that full sun plants need at least 30k * 6 lux/hrs (180 kilo-lux/hrs) of light energy per day.

I spent the last week automating the redhouse (the geothermal smart greenhouse).  Part of the automation includes giving the grow lights some smarts.  As pointed out in my last post, the grow light controller has a visible light sensor.  At the moment however, it seems my readings are inaccurate.  Full daylight on the sensor reads 1400 while it reads 919 during the night.  This is not ideal, but we can add some crude corrections to see if we are meeting the energy requirements that our plants need.

What we want to do is to make sure that our plants reach the 180klux/hrs (a total of 108megalux) per day using natural light if possible, but using the artificial grow lights if not.  Since we can control the brightness of our grow lights, we can tune the brightness to compensate for any lack of natural light.  If a cloud rolls over, we’ll ramp up our grow lights until we are producing the 30,000lux that we need.

I measured the output of the grow lights using my smart phone’s light sensor.  I don’t know how accurate it is, but at about 12″ away from the lights, it reads about 30,000lux at full brightness.  That’s perfect.

Doing a little math, we can now figure out the difference in natural light to the light we are seeing on the sensor.  We know darkness on the sensor is 919 and full sun is 1400, so we’ll compensate.  All code from this point all will be python, but should be easy enough to convert to any language you want:

directSunlight = 1400 – 919 = 481
growLightMax = 481

correctedCurrentSensorReading = currentSensorReading – 919
energyDelta = directSunlight – correctedCurrentSensorReading
growLightRate = (energyDelta / growLightMax)

growLightRate will give us the percentage of brightness we need to achieve “full sun”.

Another thing we want to do is to not use the artificial light once we reach 100% of our daily needs.  We will save energy (and cost) by turning off the light when we don’t need it.  Here’s how I did it:

totalEnergyNeeds = directSunlight * 3600 * 6

Lux is measured per second, so we will need the equivalent of 108mega-lux per day of energy.  The 3600 * 6 converts 6 hours (full day minimum requirement) to seconds.  We can adjust the number of hours according to the needs of our plants.  We can move it up to 8 for plants that need more light, or down to 4 for part-day plants.

Now that we have computed our daily needs, we need to keep a running total of how much we’ve produced:

totalEnergyProduced = correctedCurrentSensorReading + (growLightMax * growLightRate)

if totalEnergyProduced > totalEnergyNeeds:
growLightRate = 0 #turn off the lights

We should run this code every 1 second.

Using the grow light as a heat source

We will also want to turn off the lights at night time, unless we are below our total energy needs.  But there are reasons to keep the light on at night.

Certain plants not only have different light needs, but also have darkness needs.  This requirement is called “photoperodism“.  Some plants need periods of darkness to “sleep”.  During “sleep”, these plants may initiate flowering among other things.  Tomatoes, cucumbers, roses and melons do not have any darkness requirements.  We could in theory provide light to these plants 24/7 which may help them to grow faster.

I don’t need the plants to grow faster, but the LEDs from the grow light do generate some heat.  This additional heat will help regulate the temperature at night and keep the air temperature warmer than it otherwise might be.  Because these lights are over our tomatoes, we don’t need to worry about causing harm.  I don’t recommend this for plants that require periods of darkness (short-day and long-day plants).  In my greenhouse I know the soil as well as the ambient air temperature.  I’ll use the temperature to determine if it’s cool enough that we will need the extra heat from the lights.  Using a library called “astral” we can also determine if it is night time or not.  Astral doesn’t tell us if it’s night, but it does tell us when solar sunset and sunrise are.

from datetime import datetime
from astral import Location

location = Location()
location.latitude = 45.34342
location.longitude = -122.343426
location.timezone = “US/Pacific”

t =

isNightTime = (t > location.sunset().time() and t < time(23)) or (t > time(0) and t < location.sunrise().time())

if (isNightTime and totalEnergyProduced < totalEnergyNeeds) or airTemp < airSetpoint or soilTemp < soilSetpoint:
growLightRate = 1.0 #full brightness

elif totalEnergyProduced >=totalEnergyNeeds:
growLightRate = 0.0 #completely off


I’ve been fine tuning this logic for about a week.  Everything seems to be working.  I’ve only got one light at the moment over 3 of 9 tomato plants.  The plants under the light have much more growth than plants of a similar age and look healthier.  I’ve also added some variables to track how much natural vs artificial light I’m using and how close I am to my plant’s requirements.  This is my light production for today (from about 10:30am to 7:44pm)

“artificialProduction”: “58%”,
“energyGenerated”: 143407.67,
“nightTime”: true,
“productionCompleted”: “148%”,
“totalEnergyGenerated”: 15344621.17

Today has been overcast for most of the day.  Because of the clouds, 58% of my light has come from the grow lights.  Additionally, I’m 148% of my daily total right now.  My numbers reset at sunrise and the lights will be at 100% brightness for most of the night (it’s already cold enough to have the lights on).  Right before sunrise, I wouldn’t be surprised if I was over 200%.

Smart Grow light

The first law of plant growth is light.  Typically this light comes from the sun at an incredible intensity of up to 100,000 lux (lm per square meter).  Different plants have different sunlight requirements.  Typically these are categorized as “full sun” or “partial sun” plants.  Full sun plants require at least 6 hours of direct sunlight per day (30,000 – 100,000lux).  Partial sun plants need about 3 – 6 hours.

Winters in the Oregon Portland area are dark.  So dark that humans suffer from the lack of light.  This condition is known as Seasonal Affective Disorder.  Most of this period is dominated by overcast clouds which reduces the light to about 1000 lux.  That’s 3% of the minimum  light required for full sun plants.  To grow food all year round, we are going to have to compensate for this lack of light.

There are many different types of artifical lighting.  The most power efficient of which are LEDs.  The problem with LEDs is that they typically have a very narrow range of light.  To make white, LED’s combine 3 diodes with some red, green and blue.  The human eye sees these three as white.  It cannot perceive the gaps in the spectrum that the LEDs do not transmit.
White LED from Cree (TM)

White LED from Cree (TM)

Other light sources cover a lot more of the light spectrum.

Example of a LED spectra

Given that lights light incandescent lights have a wider spectrum, why use anything else?  The answer is, plants don’t need all that light.  Plants use light in the blue and red spectrum and reflect the green and yellow spectra.  That means that any grow light that includes these wavelengths are wasting energy.

LEDs for growing

LEDs for growing do not have the green diodes.  This makes LEDs, which are already the most efficient artificial light source available even more efficient for growing.  While you can purchase may LED-based artificial light solutions, I set out to build my own in such a way that I can control the amount of light my plants get based on sunlight.  If there is full sun, I don’t need to activate the lights.  If there is less than full sun, I can adjust the brightness of the LEDs to compensate.

I found some grow LEDs on ebay for a decent price from the seller sungrowled.  These are 30W but I have also used the 50W variants.


To keep the LEDs cool, I picked up a 36 inch aluminum strip and mounted 6 LEDs evenly spaced on the surface.  I used thermal adhesive to mount them.


On the ends of the strip, I drilled a 3/8″ hole and attached something I could attach to some rope to hang.


I then wired up the LEDs in parallel.


The aluminum strip helps spread the heat, but probably will still get to hot.  To increase the surface area for cooling, I used the thermal adhesive to attach some aluminum heat sinks to the top of the strip opposite of the LEDs.  This unit will be passively cooled.



Six 30 Watt LEDs run at a total of 180W.  Amazon has a bunch of adjustable current/voltage power supplies from Drok.  Oddly, the power supply ratings seem to go from a few watts to 100W  and then jump to 300W and then to 600W.  Really?  No 200W?  Sigh.  I grabbed the 600W power supply version.  To supply the AC power, I picked up a 300W 24V AC to DC converter.

What’s nice about the Drok power supply, is that I can supply constant voltage AND constant current.  LEDs have an upward sloping current draw relative to the forward voltage.  This means if you oversupply voltage, it’ll draw enough current to burn out.

Typical ways to drive LEDs include supplying a constant current, so that you can safely over drive the voltage, a resistor which will also limit the current, or constant voltage.  If you never go over the volts, the current draw will be just fine.  With this Drok boost converter, I have POTs that I can dial both.  I started at 24V and slowly adjusted both the current and volt POTs until I read about 170 Watts on the kill-o-watt.  Waddayaknow!  It worked!  It’s really… really bright too!


To finish off the light power, I put both the Drok buck boost converter, the AC-DC power supply, a couple fans, and  Drok buck 12 power supply in a nice case and mounted it in the redhouse.



Light Controller

At the moment, our lights are dumb.  They turn on and off manually and do not care about sun.  We need a smart controller.  I used the Particle Photon, which is a cheap wifi-enabled MCU that’s only $20.  I’ve been using the photon a lot lately.  This is actually the third unit inside the Redhouse and I plan on using at least one more.  We also need a sensor to measure how much light we are getting from the sun.  I had an SI1145 sensor breakout from adafruit laying around so I used that.

To control the LED brightness, I use a MOSFET.  By modifying the signal’s pulse width on the MOSFET gate, I can control the voltage allowed to pass through the MOSFET.  I wired everything up, wrote a few lines of code using the Particle builder and this is what came out:


The light sensor won’t work without light, so I need a case with a clear lid.  I found a waterproof project case on amazon that had a clear lid.  Perfect.  I installed the board in the case, wired it up, and installed it in the Redhouse.



How much did this cost me?  I try not to think about it, because I’m not building to necessarily save (although, I believe I am).  Here’s a list of components and their costs:

LED Light:

  • 6 x 30W LEDs –  ebay seller sungrowled – $77
  • Aluminum strip and screws – home depot – $15
  • 24V AC-DC 350W converter – – $35
  • Drok 600W buck boost converter – – $21
  • 5 x aluminum heatsinks – – $25
  • Power enclosure – – $30
  • Grow light hanger – – $10


  • Particle Photon Wifi MCU – – $20
  • SI1145 visible light sensor – – $10
  • MOSFET n-channel – – $2

Total: $245

For comparison, you can get a nice (but dumb) 160W grow LED for about $350.  No wifi/internet control.  No smarts.

I think I saved some bucks and get more features.  Here’s the final product growing some nice Roma variety tomatoes:


Do they work?  Early indications suggest they do.  The plants under the light look more lively.  One plant NOT under the light has a livelier branch under the light where the other branches are not as lively.  So, pending future observation, I conclude, it works.  I only have 2 more of these to make :).