Constant temperature controller

[NaxFM]

Hi everyone, I need to create a circuit as compact and small as possible to maintain a small box at a constant temperature (very constant).
For now i have this circuit which is just a comparator that turns on and off the transistor according to the resistance of the thermistor. By changing the ratio of the resistor R1 and R2,R3 (these two are in parallel) i can set various temperatures. Resistor R5 is there to smooth out the transition between transistor on and off, which would be very abrupt without. With R5 the current in the heating resistors starts to lower when approaching the set temperature.
Do you know of a better circuit for something like this? Maybe more compact or more refined? I was thinking about a microcontroller, but maybe that would be overkill...
My requirements are compactness and constant temperature after it stabilizes for a while.

[Doctorandus_P]

How high is your "+Batt" ?

I would replace R7 and R8 to a shunt resistor, somewhere around 1 Ohm and replace the the BC817 for a darlington in TO220.
With that change you've build a variable current source and the transistor itself becomes your heating element, and the opamp then compares the thermistor voltage with the shunt voltage (= current = power).

This makes your heating element linearly adjustable.

But what do you consider "very constant". is that within 1mK?

[james_s]

I would suggest looking at the circuits used in some of the oven controlled crystal oscillators since they serve exactly the purpose you describe. Most of them use the power transistor itself as the heater, sometimes with additional resistors.

https://syncchannel.blogspot.com/2016/03/schematic-of-cts-1960017-10mhz-ocxo.html

[taste_tester]

I wonder if this is a situation where a tempco resistor would be useful in place of a standard resistor? I understand that they can remain stable in value across a wide range of temperatures.

[BeBuLamar]

You said very constant! You would need a temperature controller with PID loop and if you build one yourself using microcontroller isn't overkill. In fact all of the temperature controller I know has some form of processor. Yet 1/10 of a degree is the best I can hope for. Many times not so accurate.

[Siwastaja]

Define "very constant". +/-1 degC? 0.1? 0.01?

Feedforward + hysteretic (on/off thermostat) is probably easiest to achieve in analog. This means, just make a on/off thermostat with a comparator, make hysteresis really small so that it switches on/off often (possibly many times per second) but just stable enough it doesn't actually oscillate at high frequencies. Then add feedforward: adjust resistor power based on outside temperature. The colder it is, the more you need heating power. This feedforward easily almost completely removes the small remaining outside temperature-dependent offset in the simple hysteretic controller.

[Terry Bites]

You need a well regulated supply for the bridge and low Tc resistors.
The amplifier needs low drift specs. Its hard to know how complicated the servo needs to be without knowing how precise and how stable you need it to be.
Can you specify an operating window? 

[james_s]

You said very constant! You would need a temperature controller with PID loop and if you build one yourself using microcontroller isn't overkill. In fact all of the temperature controller I know has some form of processor. Yet 1/10 of a degree is the best I can hope for. Many times not so accurate.

You don't need a PID loop for something like this, PID makes sense when you have a large amount of power involved but assuming this is a few watts of heat a simple linear circuit is perfectly fine. I've never seen a OCXO with a PID controller in it.

[Wallace Gasiewicz]

I just repaired a circuit kinda like the one you are talking about. It is in a crystal oven.
Here are some suggestions.
Use a mosfet rather than a BJT
Secure the mosfet to a heat sink in the enclosure you are heating. If you use two boxes, one inside the other, you could use the internal metal box as the heat sink. Insulate between boxes.
The mosfet becomes your heater, You can put a small value resistor maybe 10 ohms, (several watts) from the source of your mosfet to ground if you want to spread out the heat, if you think your mosfet will get too hot by itself. Attach the resistor to the heat sink with plumbers' epoxy. Secure the heat sensitive resistor to the heat sink also. Almost any old big enhancement mosfet will do.
Put some sort of buffer circuit between the op amp and the gate of the mosfet. A cap to ground with a resistor to the mosfet gate perhaps might be enough to even things out. Maybe 1 uF?? Dunno. This should keep the fluctuations from the op amp in check.
Another cap to ground on the control input of the op amp would also even things out over the long run.
Suitable caps on the batt  power line would also be suggested to keep other signals out.
You need more volts on the heater than on the op amp controller, Maybe like 15-20 v on the heater and 5 on the op amp. You could use a small volt regulator for the op amp VCC. Op amp here does not use much power at all.
All this stuff goes into the heated box.
Once you are happy that it works:
Insulate the box well with perhaps fire resistant foam in a can (orange stuff). Really two boxes with insulation between them.

This sort of circuit in my OCXO gives me at least 1 HZ stability in 500 MHZ. I am not smart enough to measure more than that.
The xtal is Very heat sensitive and uses the heat control for freq lock with a GPS system.

[Kleinstein]

If the speed of reaction is not critical one often does not need a full PID control. Often simple PI is sufficent and easier.
Pure proportional control has a limited gain and tends to be less constant if the external conditions change a lot. If the system is well behaved with a relatively good coupling from the heater to the sensor this may still be good enough.

The on/of switching type regulators are less suitable to get a very stable temperature.

For a PID/ PI regulator the size and time constant of the system to regulate makes a difference wether a pure analog regulation is easy to build or a digital control is more suitable. A large an thus slow system (e.g. more than a minute of delay) it tricky analog as it would need a rather long time constant. This is better done digital. A fast system can work OK analog, but it a bit more tricky to tune.

For the heater, using a power transistor also as heater is a nice idea, as this gives a linear response. A resistor has heater is nonlinear and thus would complicate at least analog regulation.

Attached is a circuit for a simple analog PI regulator, that uses a transistor as a heater. I think R73 should be smaller, depending on the ref voltage. The time constant is about right for rather close coupling, e.g. with the transistors side by side screwed to a metal piece.

[David Hess]

You need a well regulated supply for the bridge and low Tc resistors.

That would be true if a voltage reference was used, but in the bridge configuration the reference voltage and thermistor output track ratiometrically so supply voltage variation is removed as an error term when only balancing the bridge and not making an absolute measurement.

RC oscillators using a comparator can do the same thing to make oscillation frequency independent of supply voltage.

The amplifier needs low drift specs.

The high sensitivity of the thermistor relaxes the requirements of the operational amplifier.

[james_s]

This sort of circuit in my OCXO gives me at least 1 HZ stability in 500 MHZ. I am not smart enough to measure more than that.
The xtal is Very heat sensitive and uses the heat control for freq lock with a GPS system.

It's amazing how sensitive crystals can be to temperature. I have a GPSDO with an onboard OCXO and I found that even with the OCXO, just placing my hand on top of the can or even blowing on it causes a small but noticeable shift in frequency. I have some other oscillators that utilize a double oven and they are somewhat more stable.

[Wallace Gasiewicz]

Xtals are responsive to temp.
Some xtals are relatively insensitive to temp changes.
Some xtals that are placed in ovens are very stable in a specific temp range.
Some xtals are very sensitive to temp and temp is used to control the freq of the xtal, this is the case I was describing. The temp has to be very stable. I was just using this as an example of how stable the temp control can be with a relatively simple heater arrangement, using a Mosfet as the main heater. And trying to describe the heater control even though I have no schematics (too bad).
When I repaired the surface control board, I found out that the surrounding caps and resistors took the "band bang" out of the response, and I ended up with a pretty stable "control" voltage for the heater, after it got up to operating temp.
I damaged the control board with a bad PS arrangement, this was my fault. So I fixed it. It was somewhat problematic taking the xtal heater can apart for the repair. HA!
The heater (mosfet) is controlled by an Op AMP and the input from a heat sensitive resistor. There is a resistor, actually three 3 ohm  resistors in parallel to ground on the Mosfer output to ground. There is also a big cap there. And other caps on the input and output leads of the op amp.  All surface mount, cannot see all the markings.
The caps tend to even out the heating voltage.

I think it depends on how the xtal is cut from the xtal block. The way they are grown and cut determines the properties of the xtal

Wikipedia has an article, more than I ever wanted to know.

[Wallace Gasiewicz]

BTW, I think Kleinstein's schematic shows the caps on the output of the heater and the resistors that I mentioned.
Thank goodness someone has a schematic.
The difference is that my control used only one Mosfet and the op amp. His schematic uses a heating BJT and another BJT as a driver.
The sensor in my circuit is the temp sensitive resistor and a BJT (I think) The operating VCC of my op amp is 5 V (5 V regulator) and the inputs of the op amp are lower also. The control voltage of my heater (one of the op amp inputs) comes from another circuit that compares the GPS output to the 10 MHz square wave oscillator output (another story).
The oscillator makes nice harmonics that I can see to at least 500 MHz.

[Kleinstein]

In my circuit 1 transistor is the heater
the 2nd transistor is the temperatur sensor (used as the diode, as the TO126 case is easy to mount) to get fast response
the 3rd transistor is used to limit the maximum current and thus the maximum heater power.
Taking the feedback from the resistor and thus including the current limit (and not just what the OP wants) helps to limit integrator windup. It is not full anti-windup, but an easy step in this direction.

I have used such a circuit with a small oven for a voltage reference (JFET based) with dimentions of some 25x25x15 mm³, so a relatively small system.

[wizard69]

Hi everyone, I need to create a circuit as compact and small as possible to maintain a small box at a constant temperature (very constant).

This tells us almost nothing useful.

• "Very constant" doesn't mean much, in some processes that is 10 degrees and other it is 0.01 degrees.
• You can't avoid mechanical / thermal properties of the item being controlled.   Some systems require considering insulation and thermal conductivity to achieve good temp control.   A common problem here is system fans or external air currents (air conditioning).

For now i have this circuit which is just a comparator that turns on and off the transistor according to the resistance of the thermistor.

Bang bang control is almost always a crude approach to temp control

By changing the ratio of the resistor R1 and R2,R3 (these two are in parallel) i can set various temperatures. Resistor R5 is there to smooth out the transition between transistor on and off, which would be very abrupt without. With R5 the current in the heating resistors starts to lower when approaching the set temperature.
Do you know of a better circuit for something like this? Maybe more compact or more refined? I was thinking about a microcontroller, but maybe that would be overkill...

A micro controller may very well be overkill for mass production.   However if this is a limited production item, a micro controller would be ideal, especially if you can support a communication channel to the outside world.   Between that and the ability to modify PID parameters as you zero in on optimal control can save you a lot of development grief.

My requirements are compactness and constant temperature after it stabilizes for a while.

Neither one of those requirements means much.   You have a bunch of guys expressing experience with ovens for crystal for example.   My experience runs more along the lines of temp control for industrial processes and optical metrology equipment.   Here a temp control unit might sit along side a desk with an interferometer to control temperature in an inspection cell.   We don't build our own PID controllers, but rather use purchased hardware as it would be a waste of time to reinvent the wheel.

By the way one thing learned over the years is that not all PID controllers are the same in their ability to work properly.   This includes expensive laboratory equipment that frankly had to be swapped out for a competitive unit from a different manufacture.   I bring this up because a DIY PID controller is not a cake walk if you are after very precise control.   There are however some open source solutions floating about the net that might be a good starting point.   Just don't expect to have a working set of parameters in the first go.