I've built a few Vbe multipliers as cheap temp sensors. I know there are better ways to sense temperature but I am not interested in accurate measurements, and my question is exclusively about a strange behavior (oscillating/irregular voltage readings) I've witnessed the day after one of the worst built probes fell into boiling water.

One of the sensors, a 5x multiplier with resistor values 100k and 470k, was built with very thin wires (cat 5e), the wrong insulation of the transistor's pins and inadequate heat-shrink wrap. (In my application I intend to seal them with resin or even silicone but, again, that's irrelevant). The following picture shows a drawing of how I assembled the sensor and a picture of the sensor taken after it had fallen in boiling water (it was not supposed to fall in its entirety into the water):

(Bad) thermal probe

As you can see the insulation does not look waterproof at all, and I am sure that boiling water did not help in making it better. (My other probes have better insulation with insulating rings on collector and emitter pins, not the base, and thicker wires)

I used an unmarked PNP with an hFE of 370 and a Veb of 0.71V @18°C.
The circuit was connected to a regulated 5V power supply via a 10k resistor, and the voltage was measured between emitter and collector in order to get a positive reading. This is the schematic with inverted voltage Vout and a known transistor:

Veb multiplier

Now, during 'calibration' at ambient temperature in air I saw the voltage allowed for reasonably accurate and fairly reproducible measurements (shown as blue dots in the the next graph)

16.8 °C 3.547 V
16.9 °C 3.545 V
16.3 °C 3.552 V
16.2 °C 3.557 V

But when I dipped the probe in boiling water it fell fully in - so that tap water reached the terminals and funny things started to happen.

This is a long post divided into four parts

Part 1: the original voltage readings with a DMM
Part 2: replicating the irregular voltage variations of the insulated probe under oscilloscope probes
Part 3: controlling the amount of water on a naked probe
Part 4: replicating the voltage variations on a simple voltage divider

Parts two to four were added after reading comments and answers.

Part 1: The original DMM readings

When the probe fell into boiling water, voltage dropped around 2 V, no accurate measure was recorded because I was too busy cursing but most importantly voltage measurements began to vary wildly making it impossible to obtain a reading on my DMM (Uni-T UT139C). Even when I took the probe out I couldn't get a stable voltage (and no, it was not a gradual variation while the sensors adapted to the new temperature condition, the voltage was all over the place).

Twenty minutes later, it was still unstable - I came back from time to time to see if the voltage stabilized and these were the 'measurements':

18.0 °C 2.8V and rising
17.6 °C 3.30 V and rising, reached 3.37 V in a handful seconds
17.7 °C 3.42 V and rising

I figured water has crept in and I had to wait for it to dry completely.

The day after (temperature had certainly settled to its equilibrium value by then) when I first measure it by turning on the power supply and applying the DMM probes, I found the voltage readings were oscillating:

17.4 °C --- 3.442V, 3.441V, 3.440V, 3.441V, 3.442V, 3.441V, 3.440V,...

and so on, with the 'oscillation' taking a couple of seconds (I am not sure now, it has been a few days ago but I could see the numbers change at least once per second, probably at the refresh rate of my UT139C). This is the behavior I am curios about.
I cannot say how long it lasted; it went back and forth at least three times but at the moment I just thought at it as spurious measurements. I thought "oh well, I broke it..." Only later I grew curios about it.
I also find 'suspicious' the fact that I was lucky enough to witness this oscillation the day after because it was a sporadic event that stopped a few minutes later. Was this a transient due to the application of the voltage supply (it never happened before, but then the day before I had the supply on most of the time and off for at most one hour between measures), or could I have started it with my DMM? Unfortunately I did not record when I turned on and off the supply.

After a few minutes the voltage started to rise again albeit more slowly (let's say once every second and a half, two seconds.

17.4 °C --- 3.465V, 3.466V, 3.467V, 3.468V,...

Fifteen minutes later it was still rising, but ever more slowly - it even appeared to be stable at first.

17.4 °C --- 3.519V, ...., 3.524V, ...., 3,526V

Later on it (I did not write down the time but maybe a couple hours later) it seemed to have stabilized at pre-cooking conditions

16.8 °C 3.546 V
17.9 °C 3.535 V

The following picture shows in blue the measures taken before it fell in boiling water and in red the measures taken several days after it had stabilized after falling in, getting wet and going crazy. It appears the probe is now working as expected and I interpret this as a sign that it was the water that had gotten between the terminals to cause the erratic behavior. Blue (red) line is a linear fit for the blue (red) points, purple line is a linear fit for all points.

datapoints and fits for dry probe

But I would like to know what is the most likely explanation for the oscillation I recorded the day after the bath in boiling water. It appeared briefly at the first measurement for that day (most likely a transient due to the freshly applied voltage), and then voltage started to rise again till it stabilized at normal levels for the given temperature. Tap water is moderately conductive and it is also possible that electrolysis has played a role but I am not sure how to model its presence to explain that first oscillation.

Any takers?

Part 2: badly insulated probe meets oscilloscope

(added after reading the first comments and answers)
I was able to reproduce the effect after dipping the sensor in boiling water, taking it out and leave it alone for a few hours. But let's follow the chronological order: before its new encounter with hot water after more than two weeks, the measures were always stable and the sensor basically flatlined on the scope. This is what happened the first time I dipped it into hot (not exactly boiling) water: a nice and smooth voltage variation

smooth variation

Evidently water had yet to creep inside. I repeated that a few times and the sensor appeared to work smoothly, until...
This is what happen with the fourth bath (I printed the screen while still scanning):

no longer monotonically adapting

The voltage goes up and down while in the hot water. This is not the effect I was curious about but it shows that something odd is happening that should not be happening with an insulated sensor.

Another 'anomaly' I saw after that, immediately after an hot bath is the voltage keeping going down for a while after exiting the water, before rising again:

Still going down when out of the hot water

The sensor does not have enough thermal inertia to account for that. And then I saw sudden jumps in the voltage while cooling in air

sudden jump in voltage

I also witnessed wild variations when touching the probe and for several seconds after that, probably due to the water moving inside. This is one of the screenshot that shows the variation during touch ending with the dashed line, and then the voltage keeping its downward trend before rising again as expected:

dip after touch

But let's get to the behavior I had seen on my DMM: the last digit of my DMM voltage readings going up and down. I was able to replicate that effect after a few hours since the last hot bath. The DMM measures fluctuated multiple times in different ranges when sitting at ambient temperature several hours after the last hot bath

3.310V, 11, 12, 13, 12, 11, 10, 11, 12, 13,...

with measures varying every 2, 2.5 seconds.
A few minutes later I observed

3.330V, 29, 28, 27, 26, 27, 28, 29,...

with about 3 seconds between each different reading.

Here's another screenshot showing a sudden step I saw on the DMM as

3.325, 24, 23, ---> 3.314, 16, ...

sudden down-step

Part 3: Naked probe meets scope and water drops
I have built another probe, but this time it was 'naked' so that I could drop water on the terminals to observe the effect. This is a picture of the probe:

Naked probe

A wet towel on the EBC terminals will result in a sudden drop of voltage and occasional 'oscillations' as shown here:

wet towel on both EB and CB

A drop of tap water on the EB terminals will cause a sudden raise in voltage and occasional, barely perceptible oscillations. It's not even worth posting the screenshot.
But a drop of tap water between the C and B terminals works miracles. Sudden drop of voltage (I do not have a screenshot because the trace fell off screen and when I adapted the vertical scale it was gone) and wild oscillations on the way back to equilibrium. Here is a screenshot with a few seconds later:

oscillations on the way up

And here is a detail of one of the clearest sequence of ups and downs, it almost looks periodic:

oscillation detail

More irregular oscillation happened on the way up after a few minutes. I have several screenshots but this is one of the clearest ones

rough ride to equilibrium

A good probe just flatlines in the same conditions. And here is what happened after I dried the terminals with a tissue: the probe went back to normal with an interesting overshoot:

when dry goes back to normal

So, the presence of water between C and B terminal is the culprit of the oscillation I observed and, from the answers so far it seems that electrolysis is the most likely explanation.

While I still would like to know if it is possible to explain the above variations (extended inertia, ups and downs and sudden jumps) with a theoretical model, I am going to accept the first answer suggesting electrolysis.

Part 4: Taking the transistor out of the equation

I wanted to see what happens without the transistor. I created a voltage divider with two resistors (220 ohm and 1 megaohm) supplied with 3.5 V regulated. Then I placed the two terminals of the 1 meg resistor into a small water vessel, in order to show how the voltage across it changed on the scope's screen, like this:

voltage divider loaded with liquid resistor

With just tap water the voltage started to decrease very slowly. Adding a bit of salt to the solution made the voltage drop faster, as expected, but no irregularities were observed:

gradual reduction in conductivity

stirring the solution caused sudden jumps and the occasional irregular jump, even several seconds after the stirring had ended. But things went 'Crazy Ivan' when bubbles and electrolysis deposits began to accumulate, as hypothesized by Brian Drummond. Here is the wild variation I observed

Crazy Ivan

when the solution was this kind of mess:

Bubble bubble, toil and trouble

I also observed jumps when moving the bubbles around with a toothpick, and also an interesting sensitivity to vibrations (which I do not document here because this post is already toolong).
I believe I can assume the solution proposed by Brian Drummond was spot on: bubbles develops due to electrolysis and that changes the conductivity of the solution in a vaguely periodic fashion as they are produced, grow, and burst. I wonder if the metallic residues that were dissolved into the solution play a role as well, especially in its interesting susceptibility to vibrations.
But that's another story...

  • 1
    \$\begingroup\$ What causes transistors to oscillate? There are two things needed. Feedback, and gain. Transistors have gain. So all that is needed is a resistive path from the collector to the base to create negative feedback and the potential for oscillation is there. \$\endgroup\$
    – user57037
    Commented Nov 8, 2020 at 2:30
  • \$\begingroup\$ Also, I think your transistor is upside down, based on your wiring diagram. You put the emitter where I would expect the collector to be and vice-versa. \$\endgroup\$
    – user57037
    Commented Nov 8, 2020 at 2:32
  • \$\begingroup\$ @mkeith it's a pnp: simulation and measurement give compatible results in sign and magnitude. I just put the positive probe of the DMM where the ground is and the negative probe where Vout is in the LTspice schematic, to have a positive value. As for the resistive path from collector to base, well every Vbe multiplier has it, it's R1 in the schematic. But it did not oscillate when dry, and it did not oscillate long when wet. \$\endgroup\$ Commented Nov 8, 2020 at 3:49
  • 4
    \$\begingroup\$ This is a reasonable question, but EESE would be a way more entertaining site if there were more questions of this form: Why does <my circuit> do <funny things> when <wet/on fire/cursed>? \$\endgroup\$
    – Reinderien
    Commented Nov 11, 2020 at 21:26
  • 4
    \$\begingroup\$ @Reinderien "When you are taking data, if you see something that looks funny, record the amount of "funny"" - W2AEW, quoting Bob Pease quoting Tom Mulligan \$\endgroup\$ Commented Nov 11, 2020 at 21:31

2 Answers 2


Good experiment with detailed measurements!

We have no evidence that it's oscillating (in the sense of an L-C or R-C oscillator) though an oscilloscope may reveal something a DMM can't.

It was clearly still drying out. One source of periodicity may be the formation and release of bubbles of H2 and O2 under electrolysis, probably across the C-B junction thanks to its higher voltage.

If you made both 5:1 and 1:1 Vbe multipliers, and operated them both wet with the terminals visible, you may be able to see bubbles on the higher voltage one, and that would allow you to test this hypothesis. You may see a correlation between bubbles and voltage reading.

  • \$\begingroup\$ Yes, I believe I was wrong in assuming it was oscillating. I was able to replicate the behaviour and is more like temporary irregualities in an otherwise monotonic trend. Scope images are not resolutive, but I have at least something to document it. I will edit the question shortly. \$\endgroup\$ Commented Nov 9, 2020 at 1:47
  • 2
    \$\begingroup\$ Just to be clear; I am suggesting that electrolysis (like, formation and bursting of bubbles) is a plausible way of generating periodic variations : the liquid may simply be conductive (causing erroneous readings) but bubbles can disturb the liquid, changing that conductive path, in a vaguely periodic fashion. \$\endgroup\$
    – user16324
    Commented Nov 11, 2020 at 21:24
  • \$\begingroup\$ I plan to do some more experimenting taking the transistor out of the equation. I realized I know next to nothing about conductivity in liquids. \$\endgroup\$ Commented Nov 11, 2020 at 21:28

You have inductance in the collector lead, from the probe wiring.

You have capacitance between collector and base: expect 10 or 20 pF of Cob. Being wet brings the dielectric constant of water into that wire_wire capacitor of base_collector leads, boosting that Cob

At high frequencies, the BETA is rolling off, adding an extra pole.

I suggest you insert 1uH in the collector lead. Add 10 or 20 pF from collector to base.



Nov 9

Given the variations, initially labeled as "oscillations" are extremely slow, I have to agree with Brian Drummond --- electrolysis.

  • \$\begingroup\$ Interesting. And yes, one thing I forgot to mention is that I have some half a meter (almost two feet) of twisted pair between probe and measuring points. Tomorrow I will simulate it and update my question accordingly. +1 in the meantime \$\endgroup\$ Commented Nov 8, 2020 at 4:10
  • \$\begingroup\$ +1 This is a very credible possibility. In a capacitive-multiplier circuit (similar in some respects to this Vbe multiplier) VHF, or UHF oscillations affected DC output voltage measured with a multimeter. A ferrite bead on base or emitter might reduce gain enough at high frequency to quell oscillation. A series resistor on base might be a cheaper alternative (100 ohm?). That twisted-pair may or may-not be involved (in my capacitor multiplier oscillator, it was not required). \$\endgroup\$
    – glen_geek
    Commented Nov 8, 2020 at 14:09
  • \$\begingroup\$ I tried simulating with inductance and capacitors in multiple locations but I was not able to create an oscillation, just a shortlived transient. Besides, the variations I have been able to see on the scope do not look like oscillations. I am uploading now one or two of the more than fifty screenshot I've got now... \$\endgroup\$ Commented Nov 9, 2020 at 1:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.