Vacuum gauge control circuit

[helobin]

Hi eevblog peoples - long time lurker and first time poster here.

Soo this one has a bit of a backstory, bear with me
I'm trying to make a pirani vacuum gauge to roughly measure low and medium vacuum (doesn't need to be precise at all, just a go/no go) - the device is basically just a filament, the rate of heat loss is roughly equal to the pressure of the surrounding gas.

I've made the filament, a 20mm piece of 0.05mm thoriated tungsten wire (used mostly because it was there). The theoretical resistance is 0.58 ohm but the whole setup (jimmied onto an old BNC vacuum flange) measured around 0.7ohm-ok. I tested the filament just with a bench supply and seemed to work OK, with the power loss proportional to the pressure.

Next I started on the driving circuit - a wheatstone bridge with resistor chosen to balance when the filament is 1.5 ohms (or ~200-250C). Keeping constant resistance is good because it maintains a known temperature and the gauge is more accurate (fairly stable rates of heat conduction down the supply cables etc). So I fed the output of the bridge into a power opamp (L272M, again chosen because it was easy to get quickly before the weekend) but as soon as I switched the thing on it was start oscillating like mad- it would be stable for only a couple of seconds.

I assumed I had just messed up some connection (seems to be 95% of the problems I have is not my schematic, it's just me connecting something the wrong way...), so I tried again but still had the same problems.

I had bypassed the power and the bottom resistor of the reference side of the bridge, so I stuck another cap across the filament and then started playing with different caps from my box of mixed junk. I eventually got it stable-- but I'm not sure why(!).

I'd be super grateful for any insights anyone might have (have I done something wrong?), or if you've got any other ideas of a better way to setup the control circuit for a sensor like this? I can't think of any other ways to deal with the little tiny voltages needed to keep the thing at the right temp.

Another option would be a much longer (higher resistance) filament but within reason the shorter the better.

Thanks for any help!

[Andy Watson]

If the resistance of your filament is less than 1.5 ohms you have overall positive feedback.

[helobin]

Hi Andy, when the filament is less than 1.5ohms it's too cold- so the circuit should increase the voltage which would heat the filament and increase the resistance- or have I got things the wrong way around?

[Andy Watson]

In a sense the circuit is doing what you intended it to do. I think a problem is that the filament has some thermal inertia - i.e. there is a lag in the negative feedback leg. I think the solution is to put a dominant pole in the frequency response of the amplifier, i.e. .the capacitance between the output of the amplifier and its negative input (presently 0.1uF) should be increased to totally swamp the time-constant of the filament - assuming that the L272 is unity gain stable ?

Edit: Introduce a dominant pole and/or reduce the gain of the amplifier.

[helobin]

Thanks Andy,

I think I got some of the way there by just adding more capacitance in out of fustration!

Would it be worth it trying to calulate the thermal inertia of the filament and working out a capacitance value, or just trial and error?

I'm really confused working out the gain... does this make sense: The positive pin always sees V_out * 15 / (15+10) (straight voltage divider), the voltage across the filament should be V_out * R_filament/(1+R_filament) so the voltage on the negative op amp pin should see (V_out - V_filament) * 10k/(10k + 50).
This would make the gain 200, but I've ignored the effect of the 0.1uF cap, which at short timescale would make the gain higher?

If I reduced the gain, by changing the resistance of the 10k resistor in the feedback leg to say 1k, wouldn't that mean the op amp would see 0 volts accross the input pins when the bridge wasn't balanced?

I thought it was getting clearer there for a bit!

The circuit is at my parents place, I'll try playing with the values on the weekend and see what happens but I'll need to grab some capacitors in advance as there is no electronics supplier within 45 minutes drive.

Another question: is it going to cause problems when the output voltage gets really low? Would I be better off adding another resistor off the bottom of the bridge to push up the voltage? Can't fight ohms law but it makes it hard to measure accurately when the votlage across the thing gets really low

[helobin]

Here's an update on what happened with the vacuum gauge if anyone is interested...

It works~! I learned heaps, even it was mostly just realizing how much still remains to be learnt. I've got the electronics bug good and proper now.

I copped out a bit and just read the bridge voltages directly into a micro which then spits out the bridge voltage, I tried to follow all the directions I could about keeping the noisy parts of the circuit away from the analogue stuff, and everything having its own ground path, but there was still quite a bit of noise at the output of the voltage buffer that powers the bridge. I did just sort of stick capacitors all over the place and hope for the best...

I'd really appreciate any feedback (on any of it!) - - it was my first pcb and attempt at eagle cad so I'm sure I've made some classic noob moves.  no one I know does electronics so it's kind just me going round and round getting crazier and crazier!

I've attached a picture of the gauge head (0.05 mm tungsten wire, 40mm long), the gauge seems to read +/- 20% of real value with a (super basic calibration curve), from 1 milli Torr to 1 Torr, which is fine for what I need it for - it's basically just grown to be more and more elaborate.

The commercial gauges are +/- 2%, but they cost $250 for the head and $750 for the display/controller ... I'd love to get my hands on one to see how they do it!

[helobin]

controller

[helobin]

in operation

[helobin]

the board

[SeanB]

Commercial gauges are the same, just they tend to also sense the temperature of the seal area, and are calibrated with a better curve against a master reference gauge. That way you can get the error down, but as it is yours is pretty good and working well. Thiiner wire or a tighter wind would probably help, or try some double spiral tungsten wire out of a blown light bulb. will give a higher resistance and will be somewhat shorter in length.