Heat Sink Temperature Sensors - Questions for the experts!

[steve_qix]

Hi !

Working with solid state all the time, the issue of temperature management - no matter how efficient the transmitter of modulator may be - is always there part of the process.

Currently, I use a primitive system:  When the transmitter is transmitting, the cooling fan is operated;  when in receive, it is not.

But, this is not really a good system.  What I would prefer to do is have something which will operate when the heat sink reaches a certain temperature.

I am thinking of using a snap action disc thermostat (or maybe more than one) to control the cooling fan.  I don't have any experience with these devices.  Are they reliable?  How well do they sense the operating temperature?

And most important - what temperature range is appropriate.

I was thinking of using a sensor that closes at 140 degrees F and opens at 110 degrees F.  So, when the heat sink reaches 140 degrees, the fan comes on and stays on until the heat sink temp falls below 110 degrees.

There are other temperature ranges available.  The device is question is the Stancor STC-140.  The fan can be connected directly to the device, as it will switch 120 VAC at 15 amperes.

Is this a good range?  Other ranges are available.  Ideas? Thoughts? Experiences?

Here is the device:

[WQ9E]

Steve,

My only experience with these is in a De Longhi espresso/latte machine and reliability has been very poor.  This is one of the higher end espresso makers with a high pressure pump system and it works beautifully but it has 3 of the snap type thermostats and I have replaced all of them at least once.  Hopefully others can report a better experience and my coffee machine may simply be cursed.

I would have more faith in a thermocouple sensor to drive a fan controller.  Maybe you could use a couple of snap thermostats in parallel for some redundancy or a separate thermocouple with warning light in case the snap thermostat fails.  Personally I wouldn't trust these mechanical devices to protect expensive transistors or conduction cooled tubes.

[WV Hoopie]

The issue with a snap action disc thermostat is dead band. The two points, one where the contact are open and the other where the contacts are closed, are far apart in range. This might be a problem, only experience will solve the question.

A more expensive option would be a type K thermocouple with a PID controller. Don't forget to use type K wire between the thermocouple and controller, else there will be reference junction error. Using this method of control will result great contol with little or no dead band, plus other options that only a PID controller offers.

Craig,

[Mike/W8BAC]

Steve, Have a look at this kit. I think it is exactly what your looking for. He sells the board and you download the parts order list. I use one in my Flex and it's great to not listen to the fan.

I used to use the same type of temperature sensors you pictured. They can fail for a number of reasons. This kit rocks!

http://www.w1ghz.org/small_proj/small_proj.htm

[DMOD]

Hi Steve,

In the Feb 2010 issue of QST, page 44, there is circuit and an IC mentioned, MCP9701, that I am using on my ClassE rig (modified for my application) as another input to the Efficiency Meter/Overload Shutdown circuitry.

Here, they are using it to control a fan in a Power Supply.

The Digi-Key part number given is MCP9701-E/TO-ND.

Phil - AC0OB

[W3SLK]

Craig said:

A more expensive option would be a type K thermocouple with a PID controller. Don't forget to use type K wire between the thermocouple and controller, else there will be reference junction error. Using this method of control will result great contol with little or no dead band, plus other options that only a PID controller offers.

I'm with him on this but would opt of a platinum RTD. Thermocouples, unless they come with a welded joint in the tip or in a well with oil suffer gap temperature loss. Platinum RTD's tend to be accurate right on top of the material. However, if you're not splitting hairs, then use the thermocouple. I work with both and just prefer the RTD's accuracy over a thermocouple.

[WA1GFZ]

Steve, a simple biased diode and comparator works fine. A stud mounted diode bolted into the heat sink is a good thermal interface.  Mechanical sensors are fine as long as they don't have much current going through the contacts. I would use one to drive a relay or better yet a transistor. Direct connection to a fan will live a short life. The problem is they are slow unlike a diode or TC. TC voltages are very low so a pita to interface.
I would start the fan when you go into TX but let the fan run until the heat sink cools down after you are in RX. Also can do a dual speed fan so it is quiet in RX.

[Opcom]

These below have proven useful. Diodes are OK and a stud mount is easy but a purpose made sensor that is well defined will make for precise control of the fan, whether on/off or variable. The thermal lag from the heatsink to the clamped sensor body will in almost all cases be less than the rate of temperature rise of the heat sink. Some people do clamp sensors (or diodes) to the body of the semiconductor under measurement. The GND pin of a sensor can be soldered to the GND pin of a device or to the ground plane of the PCB. But maybe there is no need for such accuracy or variable speed in the fan application.

LM135 - analog, -55 to 150 deg C 3%, TO92 is the old standby.

STLM20 analog, -55 to 130 deg C, +/- 2 deg,  SOT323
STLM75 I2C interface, 9 bits, -55 to 125 deg C, +/-2 deg, SO8

[N4LTA]

Steve,

I have used the snap action switches on a big 13.8 volt supply for several years and had no problems - The deadband is needed or the fan will constantly cycle on and off.

It depends on what you need but it sounds like you are not looking for anything more than on/off control to keep the semiconductor in a safe band.

You can use proportional or even go to PID control to keep it with a degree or so - but do you need that type control?

Pat
n4LTA

[WA1GFZ]

Yes a thermsistor or diode mounted directly to the case of a device is a quick way to get fan reaction. I have a TC super glued to one of the FET cases in my 160 meter rig. The case temp is not much warmer than the heat sink fins. So if you have a good thermal design the sensor position is not critical.
A sensor that tracks temperature to adjust bias on a linear is much more critical.
Steve my only point with a mechanical switch is the sparking contact. The higher the current through the contact the shorter it lives.
I'm planning a pair of mechanical switches for the solid state linear, one on each module. In TX the blower will come on. If the heat sink gets hot enough to snap the sensor it will latch the relay on until it cools down even if you unkey the rig. Very easy to do with a 2 pole relay. The relay will have 28 VDC on the coil so one pole will have 28VDC  for the latch circuit and the other 120 vAC for the blower. KISS

[K3ZS]

I would avoid thermocouples in the presence of RF.    They are low level signals and require a lot of DC amplification.   The simplest circuits use thermistors.    Since you don't need linearity, just a temperature setpoint, and they require the least amount of amplification.  Years ago, I built a 2M repeater using some cheap kit built transmitters and receivers.    I brewed up a cheap crystal oven for each crystal in the system.   It used only a few parts, mainly a thermistor and a voltage comparitor and a 2W resistor as the crystal heater.   I would have to dig out the circuit from some old files in the garage, but I think a simple temperature controlled fan could be built with a thermistor and a very few other parts.

[N4LTA]

Steve,

KISS as stated before.

The Stancor snap switches are designed to do what you want and rated at 15 Amps AC. They work. You are not going to have contact problems with a 1/4 amp fan with 15 amp contacts.

Why use op/amps and low level sensor devices for such a simple application?

Pat
N4LTA

[WBear2GCR]

Imo, the snap switches, made in USA work fine...

but a good way to go is with a DC fan and a circuit that uses a 3pin Vreg or a transistor, a thermal sensor (you pick it - thermistor is fine) and a transistor, couple of resistors. You set the minimum speed via a bias set (the fans need a certain amount to start), then the range is from that minimum spin up to max...

I can likely find that circuit if you want Steve.

                 _-_-bear

[N4LTA]

Yep - If I was going to use electronics - I'd use a proportional control system as above - It will give you very good control if required.

[KD6VXI]

I can likely find that circuit if you want Steve.

                 _-_-bear

Share with us all, inquiring minds would all like to know


--Shane

[WD8BIL]

I'll throw a wrench in it....

I use the 140 degree version of that switch in all my amps. They're placed in the airflow within @ an inch of the tubes. The one in the 8874 amp has been in there since 1978 and still operates FB ol man. Blower speed is increased during 'ol buzzards.

I used a 165 degree switch at the cabin as a safety monitor switch for the propane heater. That was 1992. Still working!!

Mounted to the heatsink with the appropriate compound it should be fine, Steve.

As my bestest buddy WB8ZWY says; "There's only one way ta find out!"

[WA1GFZ]

Bear,
I did one once with a 723 but a LM317 would be even easier. Thermistor, a resistor and a cap or two should do it. But, Steve isn't hatching eggs so it does not need to hold 1 degree accuracy.

[WBear2GCR]

this one should be easy enough to modify by putting the thermistor where the pot is... a resistance could still be used to set minimum/trim etc... one could add another transistor if more current is needed for drive, etc... I can't find the nice one that used an LM317 and the thermistor on this machine... i'm on  one of a few laptops.

In case it isn't obvious, the upper cap tries to charge when power is first applied, making the gate look like it is connected to the PS rail, so the fan starts, then as it charges enough it drops down to the "set" level... sweet! DC fan required.

the circuit is courtesy of Nelson Pass.

                 _-_-bear