LM334 + LM336 2.5V reference - What Am I doing wrong?

[gildasd]

Hello,

This is a reference design that can be found on TI's data sheet of the LM336 :

This is what I have done (same thing - I think).

My output is 5V (adjustable from 7.6 to 4.2V) and the trimmer set at 68 Ohms gets hot, thus getting the output out of whack.
Not the 2.5V 3% advertised...

Overall the two 336 I have are acting weird and out of spec of the ref designs. Am I doing something very dumb here?
My overall idea is to get a 2.5V reference that is cleanish at the 5mV
This is fed by a 12V brick supply, fed into a 7809 delivering a slight ripple (but this should not matter, I get the same thing with my bench supply).

Should I just forget it and use a 7805 with a 10kOhms voltage divider?

[daddylonglegs]

Gildas,
  First check your pin-outs, one datasheet gives the TOP view, the other datasheet gives the BOTTOM view for the TO92 packages.
  Check that the potentiometer can handle several mA of current through the wiper. You might be better off using fixed resistors in this application.

  Split this into two circuits and check you have each part working:

If the LM336 is fed from the 8.9V supply through a 5.6 kOhm resistor it should give you 2.5V. The resistor value need not be exact, 1 kOhm to 10kOhm should do. Alternatively connect it to a 5V supply with a resistor in the range 1 kOhm to 5 kOhm.

If the LM334 has a 68 Ohm Rset it should produce 1mA (roughly). Connect this in series with a 2.2 kOhm resistor and you should see 2.2V across that resistor.

  Try these two tests and feedback to the mailing list.

Good luck,
dll

[rdl]

Yeah, I still remember how I destroyed my first LM34 because of the data sheet showing the pin out from the bottom view instead of from the top like most ICs.

[gildasd]

Yeah, I still remember how I destroyed my first LM34 because of the data sheet showing the pin out from the bottom view instead of from the top like most ICs.

That explains why the 7809 was warm too...
I'll post new results when I've finishing digging the ICL7660's dips out of my forehead. Kids, clean your table before slamming your head on it.

[gildasd]

Gildas,
  First check your pin-outs, one datasheet gives the TOP view, the other datasheet gives the BOTTOM view for the TO92 packages.
  Check that the potentiometer can handle several mA of current through the wiper. You might be better off using fixed resistors in this application.

  Split this into two circuits and check you have each part working:

If the LM336 is fed from the 8.9V supply through a 5.6 kOhm resistor it should give you 2.5V. The resistor value need not be exact, 1 kOhm to 10kOhm should do. Alternatively connect it to a 5V supply with a resistor in the range 1 kOhm to 5 kOhm.

If the LM334 has a 68 Ohm Rset it should produce 1mA (roughly). Connect this in series with a 2.2 kOhm resistor and you should see 2.2V across that resistor.

  Try these two tests and feedback to the mailing list.

Good luck,
dll

- 336 test:
At 8.9V, gives a stable 2.465V from 9.8kOhms to 250 Ohms... Replaced pot wit 10kOhms resistor, works perfect AND is stable over time.
Bonus, it keeps it's value for about 8 seconds after pulling the power supply out.

-334 test:
Your values are bang on: 2.222V drop with 86 Ohms Rset and 2.23kOhms series resistor.
No heating of the Rset.

Now doing the 334 with the 336 test, but the previous result begs the question: is there any use going above the basic 336 circuit for a reference?

Edit:
After 30 min, the 336 resistor combo has not budged, while the 334+336 has drifted 10mV.
After 40 min, 16mV, not good.

[daddylonglegs]

Yeah, I still remember how I destroyed my first LM34 because of the data sheet showing the pin out from the bottom view instead of from the top like most ICs.

  Oh yes, I've come to grief that way more than once.
  That National/ TI switched between top and bottom views for related parts in TO-92 is a bit precious though.

[daddylonglegs]

- 336 test:
At 8.9V, gives a stable 2.465V from 9.8kOhms to 250 Ohms... Replaced pot wit 10kOhms resistor, works perfect AND is stable over time.
Bonus, it keeps it's value for about 8 seconds after pulling the power supply out.

--SNIP--

Now doing the 334 with the 336 test, but the previous result begs the question: is there any use going above the basic 336 circuit for a reference?

  You know what you need from the circuit (and what you want to learn) so it's your call.
  The LM336 datasheet uses a 1mA bias current for most of the specification.
  The combination of the LM334 and LM336 is suggested for accomodating for maintaining a roughly constant current [1] over a wide voltage supply range.
  If you already have a regulated voltage rail available in your circuit (you mentioned a 7809) then a resistor from that rail with a value that yields a roughly 1mA current should (hopefully) give you pretty decent stability.

[1] But the current is proportional to the absolute temperature (PTAT) of the LM334?

Edit:
After 30 min, the 336 resistor combo has not budged, while the 334+336 has drifted 10mV.
After 40 min, 16mV, not good.

  The datasheet for the LM336 that I'm looking at quotes a maximum variation of 12mV over a current range of 400uA to 10mA so I'm not sure what is causing the 16mV drift. Perhaps the devices were still warming up or these devices are still settling after being overcurrented earlier this week?

[gildasd]

The 2.5V gets fed into 5 sensors and then to an Opamp... So the power usage is minimal (µA's)...
So I'm wiring the 336's directly into the sensors for now. If that does not work, I've left space for a buffer.

The 334 that is drifting is not the one that got hot. It probably does not like the trimmer... In any case, I'll revisit this circuit later when not in a rush.

In the picture you can see the two 336. In 3 hours they did not drift 1mV, stayed bang on 2.5V.

Thanks for the help!