DIY incubator thermostat anyone?


In the Brooder
7 Years
Nov 2, 2012
I've just completed the design of a stupidly cheap thermostat that is capable of regulating the temperature of DIY incubators to within a few 10ths of a degree.

I did this after having no end of trouble using a traditional "oven" thermostat from China. :( which had a range of 32-40 C on the same setting!

The design, while strictly for those who are moderately competent with a soldering iron, is very easy to build on strip board and uses only commonly available parts that you can get from a local electronics store or eBay at a pinch.

I'm happy to post details here with the strict proviso that it does switch mains (110-240) voltages for the heater - such as a 60w bulb and it should only be attempted by experienced builders who are capable of such wiring without endangering themselves or others.

The controller only employs 12v but has to activate a relay (just like a conventional thermostat) so I want to make that clear.

If everyone is happy, I'll post the circuit and parts list here later.
I can assure you that it does hold to 0.2 of a degree or better.

There is a price for this - it kicks the heck out of the relay. In my experimental unit which I have on at the moment, the only snag is that it looses heat too quickly so I'm making a more robust one now,

The commercial units probably add some hysteresis to the level comparator which widens the range - presumably in firmware.

Others have noted that this project, since it uses mains electric (albeit switched by a relay) could be dangerous for inexperience builders.

So how does it do 2/10ths of a degree?

Without showing the diagram - which I'm happy to share if the Admin agree - it uses a diode sensor.

A silicone diode has a fairly linear response NTC of about 2mv/C. My design amplifies that by a factor of 10 (with a jellybean op-amp) giving the comparator (also a jellybean op-amp) something like 20mV/C to (2mV/10th) to compare which, provided the voltage reference is stable, it a piece of cake. I used LM358 for my unit since I had some in. They are single supply, cheap and easily available. A better quality amp might be preferable since the 358s do drift a little.

Sure, you have to calibrate it manually using a stable heat source and a reliable thermometer but it really does maintain a very stable temp.
Here is my schematic - it's deliberately simple because I don't want to encourage people who might be less able with a soldering iron and could therefore, do themselves (or others) some damage.

The circuit forms three parts - split as a DC amplifier, DC comparator and emitter follower (relay) driver.

Power comes from a small 12V/1A switched supply I had lying around although this design will operate at voltages above about 9V up to perhaps as high as 24V - but that will make trimming difficult.

The diode sensor should be constructed first as it's going to be mounted remotely and ANY cable will affect the measurement. Screened cable is probably better but I just used a twisted pair.

The sensor diode is driven by the 10K current limiting resistor biasing into forward conduction at about 0.6 - 0.7v; I say about as the actual drop varies by diode type (germanium ones I used as child dropped only 0.2v) and crucially, temperature.

This is amplified by the first amp (an LM358 in my working model) by a factor of about 10 [Gf = 1 + 100/11] giving a Vout of 5 -> 7 volts or thereabouts depending on the diode - any small signal diode should do the trick.

The second 1/2 of the LM358 is configured as a simple comparator - swinging from 0v to 10x the diode drop - whenever the the voltage at the non-inverting pin goes below the voltage on the inverting pin. This is set by the cermet preset pot - a procedure which MUST be done when the unit is installed. I recommend inexperienced builders use an LED and current limiting diode in place of the relay to make sure the circuit works as advertised before hanging a relay on there! It's quite possible to turn the circuit on and off just by breathing on the sensor or pinching it between thumb and finger. (See notes below.)

The choice of transistor for the relay drive is up to you, I used a BC378 small signal NPN transistor because it can handle enough current (with lots to spare) to drive a small relay such as a those made by Songle with a 6V coil. It's important to get the right coil voltage as the circuit produces 5-6V to drive the coil regardless of the supply voltage. This is governed by the switchover voltage of the LM358 minus a diode drop for the B-E. junction.

Note 1: the diode sensor should be covered in insulation. I recommend heat shrinkable tubing to fit. Done correctly this protects the diode sensor from water and other hazards which may trigger heating when not required.

Note 2: As tempting as it might be to use a silent (solid state) relay in place of the noisy magnetic type, it should be noted that should the relay fail (the most likely place for failure if correctly built) mechanical relays tend to fail OPEN which means your heat source will shut off. If the relay fails closed, there is a very real possibility of a complete meltdown and, depending on the heater and materials you used, even a fire.

Note 3: For testing purposes, a 200R resistor in series with a general purpose Red LED may be used in place of the relay. This can also be used as part of a over or under temperature alarm when suitably calibrated.
Hello Everyone. I do make several thermostat similar to the previous post. I recomend the using of a triac/optocoupler in place of the relay. The reason is: a relay has a guaranted standard life of 10 000 cycles. In the ones I made the relay stay clicking almost all the time so the 10K happens in days. May be I will need to replace 20 installed boards. I made some others including roller timer using a microcontroller. in them I create a prehysteresis adjust and works fine, still I will migrate the design to triac/optocoupler.

Sorry for my english. I'm actualy mexican.
Since I developed this version, I've done another (simpler) one with a bead thermistor which performs nearly as well. I find the diode versions are more sensitive to temperature change - but your mileage may vary.

What I have seen from experience is that on a properly insulated cabinet (such as a cooler or a styrofoam box) the temperature should not be changing sufficiently fast as to rattle the relay like a set of chattering teeth - so that 10,000 -> 100,000 MTBF will not be reached any time soon - certainly, we're talking months or years not a few days.

This is, of course, a choice which should be left to the person building it.

What I have noted though (and this is a separate issue) is that by wiring the two bulbs in series (to reduce the power) has the effect of increasing bulb life dramatically. Studies have shown that halving the voltage to a given (swan-type) bulb can increase its working life up to 10x! More if you're prepared or able to reduce it further.

Naturally, if a bulb does fail in this configuration, the other bulb(s) also drop off - so that's something you have to be aware of.

A redundant method might be to fit two series pairs fitted in parallel; although that's clearly a lot of extra work and real-estate in your cardboard 'bator.

New posts New threads Active threads

Top Bottom