Constant current/power/anything implies active circuitry which fails to meet my simplicity requirement.
I can't seem to get it across that I have lots of complex ways to do the job. I'm looking for something simple and elegant. Hot glue and duct tape are my tools of choice. If I have to make a circuit board and a box to put it in, I'll just scrap the whole thing and use a commercial manometer.
I'm more interested in the clever design than in the result.
The tiny thermistors I have are negative temperature coefficient which leads to thermal runaway if you don't have something limiting the current.
My initial thought was to use PTC thermistors aka self heated incandescent light bulbs with the glass busted off, but moisture and dirt would make that a short lived sensor.
I'll take another look at self heating. The thermistors are tiny. I didn't think I could get heat transferred from one to the other to infer direction of flow without a very tiny air channel. The hot resistor between them solves the NTC runaway problem and gives me enough heat to make the air channel manageable size. Something to be said for keeping the heat generation separate from the sensing bridge.
The first crude prototype uses 25mW in a 1206 resistor that fills the channel. Using a -20-0-+20uA center scale analog meter, I can get full scale deflection with 5 Pascals of pressure differential. Since the thermistors are a significant part of the total resistance, the gain is rather temperature sensitive. Zero is still Zero, but I'd like to reduce the temperature dependence if I can do it simply. Yes, measuring voltage helps a lot, but takes active stuff. Simple is, "hot glue and duct tape to hold together stuff that's already in the junk box". I'm sure I'll have lost interest in the project long before anything I ordered would arrive in the mail. ;-)
There's a picture of it here: