Floating above Gnd?

Current source is a 10 V reference, through 1 Meg, out and into fet opamp to float the reference. GH

Sorry that's not the whole story, there was also a 'big' DC offset when the heater circuit turns on and puts ~100mA through 'local' ground.

GH

The best way to meas temp with a "diode" is with dV/dt for between two currents. I.e., I changes by 10x, dV=58mV at room temp. With a fairly-fast dI measurement, DC offsets are OK.

Why not make R1 zero ohms?

You might put a cap across the diode, so it doesn't rectify RF and spikes and such. 10 uA isn't much current.

Are you measuring cryo temps?

Perhaps a voltage divider between whatever V- and V+ that amp is using, to get a "mid-way" voltage, bypass it, and send it through a voltage follower. Use that as a "virtual ground" for the diode circuit.

Right. I've done that, diode connected transistors are more 'ideal' than just diodes, or either junction.

for dV/dT the diode voltage is better. (not as good as thermistors... ) George H.

That's how it started, and when the heater turned on the 'error' voltage got bigger... yuck. The heater and diode sensor share a DB9 connector... two layer pcb, traces get skinny around the connector.

Anyway I've got a current source and can float it several volts above ground. (There's x10 gain in the first Int amp. I didn't show that.)

Right, I think I have a BFC (1uF film?) across the Int amp inputs. I've never tested caps at low temps. Are them some that work/ survive (say) 77K?* smd ceramics?

Right. Well the sensor craps out at ~130K (or something) so we want to keep it above that.

George H.

Sharing a DE9 connector is not a problem as long as the sensor ground and heater ground are on different pins. Perhaps you can keep the grounds separate until they meet at the power supply. R1 could be zero - unless you actually want the zero point to be 10mV?

piglet

Yeah, I guess that is right. I had both pins 'grounded' at the DB9 connector. I don't really care about 10 mV offset. It could be more. George h.

It'd be preferable (having an instrument amp) to make a three-resistors-and-a-diode Wheatstone bridge, and really USE the instrument amp common mode rejection.

Ground-wire currents won't change the bridge output, unless it changes the excitation current. It does, however, mean you need more wires as far as the bridge resistors' position (though not necessarily to the diode).

huh, I never thought (saw) a diode in a bridge. I was hacking one circuit to do something else. The original had a four point measurement for the diode.

But I found three leads was good enough.. a remote voltage sense of the diode 'Gnd'. And then I posted this question about DC offsets of sensors... and piglet said put in two, one external. And Lasse said, put in two back to back. ~No offset and double the signal. So that's what I did. (Thanks guys, :^) If I've made a mistake about who suggested what.. I'm sorry, my error. But I had to give up my third wire of the diode, for the extra sensor lead. (the exact T is not important, stability first.) So here I am. I'm thinking first; return the (big) heater current right to PS ground. (Which is a lesson I learned years ago, go figure.) And second I'll make a spot for an offset R, (and think harder about local gnd around the input Int amp....)

Thanks again guys! (I think I've been lazy in 'ground everything hard' and that isn't always best at DC- ~1kHz ranges. It's more work to think about ground currents. :^) George H.

Reference the "ground" of the electronics(INCLUDING supply) to that 'local' ground. That should be the only "float".

In other words, the Tektronix way.

Also used in combination bandgap reference/temperature sensors, e.g. the REF01, and other PTAT sensors.

Cheers

Phil Hobbs