Thermally stabilizing a circuit without a thermistor.

[quazar]

I need a voltage that drifts as little as possible with thermal changes, I
am amplifying a voltage change that is very slight.

This circuit works ok <see attachment>, however it drifts just slightly
enough to be annoying.

(The LED is the load, the 15K potentiometer is the fake source)

I know there is a way using three transistors to get a more thermally
stabilized circuit, but I can't seem to remember it.

Is there a way to build a thermally stable version of this circuit with
similar voltage gain / bias characteristics, like what multimeter's use or
something like that?

I would appreciate any help or schematics that would help.

 

[Dan Akers]

quazar wrote;
"I need a voltage that drifts as little as possible with thermal
changes, I am amplifying a voltage change that is very slight.
This circuit works ok <see attachment>, however it drifts just slightly
enough to be annoying.
(The LED is the load, the 15K potentiometer is the fake source)
I know there is a way using three transistors to get a more thermally
stabilized circuit, but I can't seem to remember it.
Is there a way to build a thermally stable version of this circuit with
similar voltage gain / bias characteristics, like what multimeter's use
or something like that?"
____________________________________
Re;
Some more info would help. What is the V-ref for? How "slight" is the
voltage change that you are "amplifying"? What is the purpose of the
LED? By my estimation, you'd only get about 3mA max through it.
Additionally, if one assumes an hFE of about 120 for each of the
transistors, the circuit will saturate at about the forward bias point
of the NPN with Ib=0.208uA. With the info provided, there's no
indication that this circuit is what would normally be used for a
reference voltage. Meaning, that the drift you see may be in the 5V
supply itself since there's nothing in the two-transistor circuit to
provide any stability. If you're operating at the bias point of the NPN
(conducting, but not saturated), then any change in the 5V supply
voltage will be greatly amplified at it's collector. But the relatively
low 560 ohm base R of the PNP will allow it to saturate with almost any
conductance of the NPN. In other words, less than 0.2uA in the base of
the NPN would saturate the PNP. Need more info....

-Dan Akers

 

[quazar]

Some more info would help.  What is the V-ref for?

Sorry, I should have labeled this test point, or out to multimeter, or
something like that.

  How "slight" is the voltage change that you are "amplifying"?  

About 0.009 volts according to the DIMM, this is biased so the LED dimms and
brightens with the voltage change.

What is the purpose of the LED?

Really more of a quick and dirty test device to see whats going on.

By my estimation, you'd only get about 3mA max through it.
Additionally, if one assumes an hFE of about 120 for each of the
transistors, the circuit will saturate at about the forward bias point
of the NPN with Ib=0.208uA.  With the info provided, there's no
indication that this circuit is what would normally be used for a
reference voltage.  Meaning, that the drift you see may be in the 5V
supply itself

I am using a small TO-92 version of the 317T voltage regulator and I shorted
it from cold until it got hot enough to burn my fingers, it was still at 5
volts on the dot.

When I heat the transistor by the potentiometer (with a soldering iron), the
LED gets very bright, and quickly (because it is being amplified by the
other transistor).

I can then blow on it, and it will slowly go out (even when cold).

The other transistor connected to the LED gets bright to, but is not as
affected by temperature.

I am using this circuit to isolate and amplify AFC off a slope detector that
I made with two 10.7 MHz IF transformers. I can make it put out more
voltage + -, but this will result in more IF noise and I would need to
shield it better, and it works so well I don't want to touch it.

The problem really didn't occur until the air conditioning came on, it's not
real bad, but it can be really annoying. The rest of the circuit is
unusually stable, I know this because if I disable this part of the circuit
everything is rock solid.

I did come up with one idea that did work by using another PNP as a thermal
reference so that when it got hot it would lower the current by adjusting
the bias; This does work a lot better than I thought it would, but it's
not perfect (but it is very close).

This is mostly because of the other PNP transistor not being accounted for.

I tried many different brands of PNP's however, they all seem react to the
temperature slightly different, I seem to get a little less or a little
more than I need from most of them.

Is it possible that each transistor has different thermal stability
tolerances?

It would be nice if there was a circuit that could use all NPNs that was
thermally stable; then I could just use the same type of transistor.

But the relatively
low 560 ohm base R of the PNP will allow it to saturate with almost any
conductance of the NPN.  In other words, less than 0.2uA in the base of
the NPN would saturate the PNP.  

You can change it to a 100K and it will have a similar effect, but if you
put a the 100K to positive with the 560 ohm still in place; it will greatly
increase the amplification (much more than I need however).

I have a small pile of 510 - 560 ohm resistors, it's just there to save the
PNP from a possible death, your never going to let me get rid of them, are
you Dan. : )