thermal stabitity for npn transister comparetor

[ataul]

hi friends;
Recently i design a 3 transistor 12 volt battery load cut circuit for emergency light. it works very well. but it little bit thermally unstable.my circuit have facility 1. power fail auto load on and off 2. power on reset both mains on/ff and battery switch on/off 3.Hysteresis 4. load cut at 11.5 volt. 5 battery charging. 6.micro amps self consumption. but i cant fix the load cut at a point when the circuit warm up it load cut point goes to some hundreds Milli volts down.i could'nt solve it as simple way. please any body can solve this i will be very pleased? see my schematics (Attachments)
thanks....

 

[MrAl]

Hi,


From what i can tell about your schematic and your notes it seems
that your bias point is changing with temperature, and that it is
going down as temperature goes up.

The traditional way of fixing this provided you dont need super super
accuracy is to use an NTC thermistor. An NTC thermistors resistance
goes down as the temperature goes up.
Since your bias point drifts down, you would want to connect the
thermistor from the positive supply to the bias point possibly with
another fixed resistor in series with it. As the temperature rises and
the natural bias point starts to drift down, the NTC thermistor resistance
starts to drift down which increases the bias again. The circuit is
temperature compensated when the thermistor exactly compensates
for the natural tendency of the bias to drift down and so the bias
point stays close to the same value over temperature.

To select the thermistor and possibly a fixed resistor in series, you
need to calculate what value resistor you need at a given temperature
where the bias drifts down in order to keep it stable. Since the
thermistor value changes the bias even at room temperature, you
probably also need to change some of the resistor values to compensate
for the addition of the thermistor, then recalculate to make sure
the thermistor will still work at the selected drift temperature. If not,
the thermistor value may have to change or perhaps the fixed resistor
or both.
One way to test your calculations (or just to wing it) is to calculate
the value of the thermistor at two temperatures such as room temperature
(20 deg C) and say ten degrees above that (30 deg C). Now that you
have these two resistance values (we'll call them R20 and R30) you
would force the circuit to run at 20 deg C and insert the R20 resistor
(and possibly another fixed resistor) and readjust the other resistors to
get the bias point right again. Then, force the circuit up to 30 deg C
and insert the R30 resistor and check that the circuit is again biased
correctly. In other words, the two resistors R20 and R30 take the
place of the thermistor at two distinct temperatures of 20 and 30 deg C.
Once you get this working you can then go out and purchase a
thermistor that works the way the two resistors did.

I dont know if you have ever worked with thermistors before and if not
you will need to look into the calculations for these devices before
you start or you could be trying this for the next 10 years
Just in case you havent there is data on the web about these things.
Let us know what you need to do this.

Alternately, if your design can accommodate an op amp or two, you
can build in a little bias compensation circuit that works with quite a bit
of accuracy, better than a thermistor. Let us know about this too.

 

[ataul]

Hi MrAl

Thanks for good suggestion.no i need not absolute accuracy,
thermistor may solve this problem. yes i work with thermistor.my challenge is how can i builds this very low cost and simple way.
using an op amp is far better then this i know. but that's are more complicated and high cost.tlc series op amp costly and another type like 741 is high self consumption. a thermistor is more costly then a diode and transistor in our country but just now thermistor is best way fore solve this.

 

[Chaerl]

I agree you can use NTC for this one, much cheaper. But note, you need to consider that its the ambient temperature you're gonna get with the thermistor, not the actual temperature of junction of the transistor which you should need to know and protect (unless you put the thermistor on top of the transistor). Consider also the self-heating. But anyhow, if you're into more on the accuracy of the thermistor, better use a PTC ones. Normally, PTC are switching type (like a resetable fuse). Certain temperature threshold can rapidly increase the resistance at that point.

For design reference:

[muRata]MMCSV Download Service

 

[ataul]

Thermistor is ok! because of my temperature region is only 20 to 50 degree Centigrade.but i trying to use 1N4148 type diode for temp compensate. is it possible some how?

 

[flat5]

Can you add some emitter resistors to the circuit for stabilization?

 

[MrAl]

Thermistor is ok! because of my temperature region is only 20 to 50 degree Centigrade.but i trying to use 1N4148 type diode for temp compensate. is it possible some how?

Hi there,


As the temperature goes up and the Vbe drop goes down this causes
more base current to flow. Using a diode like 1N4148 in the base circuit
can help a bit by connecting it in series with a resistor and placing
this series circuit in parallel with the base emitter. Then as the transistor
heats up and the base emitter voltage drops and tries to draw more
current, the 1N4148 drop also decreases and shunts some of the current
to ground, which reduces the current flowing into the base. A better
diode however is another transistor of the same type, using the base
emitter of that second transistor as this type of diode will be the same
as the transistor BE diode that is already there so temperature tracking will
be a bit better.

As Chaerl was pointing out, sometimes you want to keep an eye on
the transistor alone, and that would require mounting the thermistor
(or whatever) right onto the transistor body with some thermal
epoxy.