Impossible circuit, need someone to figure out how this is possible

[Love_4_old_tech]

So recently I purchased this thermistor called TP2/2, is it a very precise NTC, rated for 2V and 2mA max, starting R is like 12.5kOhm. I want to use it in a precise slow-start circuit, so it increases the signal gradually. After connecting it to a PNP transistor and running 3V through it, I noticed the current gets to about about 1mA and then increasses really fast until the thermistor probably melts, this is fine, I knew it would have a very sharp positive curve, it's NTC and it heats itself, so nothing stange there. After putting a 1.5k resistor I noticed the time it takes to get to at least 1mA is way too long, so I thought of putting a capacitor in parallel with the resistor, so the current bypasses the 1.5k resistor until the capacitor is charged, and then the current would become limited to around 2mA. However, when testing it with the 1uF capacitor, I noticed the current was taking even longer to get higher and max current was even lower... After that I used an extra capacitor in parallel, 10uF, and the current has now got a NEGATIVE curve, and really slowly gets up to only 0.4mA in the last quarter. This might actually be even nicer for me, since now the circuit draws a very low current, meaning the thermistor is not being abused and the curve is nagative, which I prefer (I might use this either for lighting or in some hybrid valve amp for the filaments). But I still really want to know how the heck this is happening. Here's the schematic:

Here's a pic of the thermistor: https://imgur.com/gallery/fVFsh

[Love_4_old_tech]

Ok I finally got it, after speaking to a friend and him mentioning a voltage drop, I remembered that capacitors don't have a constant voltage, but instead the voltage goes up linear to the charge, meaning if it's constantly getting discharged the voltage never reaches 3V, so the 1.5k ohm is actually limiting the resistance greatly.

[Zero999]

What load do you need to soft start?

This seems like a very bad way to do it.

[Love_4_old_tech]

I was just screwing around with this wierd thermistor, trying to use it for different things, I don't actually really need a soft-start yet. Also turns out it can sense IR fairly well while being well-isolated from the outside.

[CopperCone]

well, if you look at precision temperature or resistance measurement, you often want a bipolar measurement to eliminate thermal emf errrors.

One way to do this, is to use a square wave, problem is it rings. SO, bob peace invented a way of doing it with a trapazoidal wave, to eliminate overshoot from ringing. It means heavy slew rate limiting, or soft start.

Considering how well built that sensor is, its not a bad idea to make such a circuit for it.

You can of course use a square wave, but it requires either tuning (to eliminate overshoot, so you have to check when the wiring is changed) or a dead zone where measurements are not taken (as measuring the overshoot will incur error). Using a trapezoidal stimulus is a elegant solution to this problem.

By deadzone I mean a circuit (or MCU program) that will wait x time before recording values after a polarity shift is made and to stop measuring y time before a polarity shift is made. With a trapazoid you can trigger measurements on the correct level being reached in a analog way, I think, for instance feed the signal into a comparator that turns on some logic gates that triggers an oscillator on CONV for a waiting MCU. This way impedance changes dont require programming changes.

I would need to think about this more.

I however doubt there will be mechanical stresses on part of any ringing that are significant, I don't believe you can damage it this way, so its purely for measurement ease/low power circuits/etc. 

[Love_4_old_tech]

Ok thanks for the info, will have to see what I decide to do next.