Water Variable Polarised Capacitor

As part of my monotube boiler I needed a way to detect the water level inside the boiler, so that while water level was low a solenoid value could be opened to let water flow from garden tap. Detecting the water level proved to somewhat more difficult than I’d first thought it would be.

First attempts (failure)…
Attempted a simple conductive probe which would detect water by allowing the water to pass a small current, and having that current switch a transistor. Inside the boiler this simple detector would work for about about 2 minutes and then fail open circuit, and so the boiler would flood completely with water. Seems like some kind of insulating layer was building up on the sensor. Perhaps steam/hot water was corroding the metal, or maybe the limescale was building up on the sensor.
corrosion possibly might have been avoided by less reactive metals. I considered a float valve, but soon decided the physical mechanism was likely to clog up with lime scale. Besides which they were quite bulky, and the only cheep ones I could find were make of plastic and so very likely to melt if the boiler ever overheated.

A solution found (capacitive sensing)…
By chance I came across njhurst’s water sensor . Water levels can be sensed capacitively because water has very different dielectric properties to air and steam. Not at all sure how such sensors would react to lime scale, but such sensors were certainly worth a try. Made up some sensors based on Mr Hurst’s idea and they certainly worked as timing capacitors. Shifted to my favourite 555 oscillator design and got on with work on sensor…

I’d been reading up on how to anodize titanium, and I fairly quickly realised that the oxide layer built up on titanium was very thin, a very good insulator, unlikely to change much in thickness once established and tenacious(stuck well!). All properties which seemed desirable in a sensor based on Mr Hurst’s ideas. Best of all titanium oxide is well known for it’s resistance to heat, Seems like a good property around steam.

For the negative plate…
Sensor needed to be inside pipe work anyway, so used copper brazing rod(melts aprox 800 deg C) to braze a 8mm copper pipe T to some short lengths of pipe and form a conductive outside body for the sensor. To ensure a reliable electrical connection could be made, an electrical push fit spade connector was soldered directly to the pipe T. Despite use of flux the solder kept running around the pipe so it ain’t too pretty but it seems to be a sound joint. Will try a solder resist/mask next time. I’ve heard tipex sometimes works, as does carbon black from a candle.

For the positive plate…
Rectangular strip approximately 5mm wide, 50 mm long cut from titanium sheet with snips. Great care was taken to ensure the strip remained as flat as possible- there wouldn’t be much clearance between strip and the inside diameter of the pipe, and I didn’t want them to touch/make electrical contact. A file was used to “scratch up” the titanium to help key glue to the metal, and a diamond whetstone rounded the cut edges over. A small amount of epoxy resin was mixed up, and left a short time until the epoxy began to heat up a little and began to thicken. Epoxy was then blobbed onto both sides of the strip approximately 20 mm from one end of strip. Epoxy was still somewhat fluid so it would flow under it’s own weight and the strip needed constantly careful manipulation for a few minutes to ensure the strip ended up approximately in the middle of a 10mm spherical bead of epoxy. once epoxy had fully cured small amounts were filed off the bead until the bead would could be slid into a pipe just like a piston. Body was then held vertically and piston pushed down into bore so about 10mm of the Ti strip remained fully outside the pipe and there was about 10mm down inside the pipe to the the top of the piston. Piston was a tight enough friction fit to hold the strip in place. A quick electrical resistance test performed to confirm no accidental electrical contact between body and strip if the strip wasn’t quite straight/parallel to the pipe’s axis. Resistance of >13Mohm suggested there was no real contact, so it was fine to fix it in place. Small amount of epoxy resin made up, and dripped into space above the top of the piston, and bounded by sides of the pipe. A small amount of epoxy was allowed to overspill, so the cut annulus of the pipe would have a thin layer of epoxy to reduce chance of positive push fit connector touching the pipe.
Following shows what this assembly looked like after the epoxy had set.

water variable capacitor
8 mm pipe.

Multimeter mod and test of dry Capacitance…
First step was to confirm the parallel resistance of the dry capacitor hadn’t changed. Rather surprised as it had improved slightly and was now 14Mohm. (is cured  epoxy a better insulator, or could it simply be some measurement error?)
The whole point of the exercise was to produce a capacitor, and as air is an insulator there wasn’t a need to condition the plates. My multimeter happens to have a pair of sockets to test inductance and capacitance but the connectors are a tad award. Rather than implement a permanent mod to WG030 all my mod needed was some small rectangles of copper sheet, with angles bent into them so they’ll act somewhat like a hair spring. Now got tags which I’ve pushed into the capacitance/inductance sockets, and can connect crocodile clips to when I want to test components that just wont fit into those awkward sockets. I’ve just left the tags in the socket and they seem to have a decent enough connections
Leads+tags did add significant capacitance on the smallest range, but it’s simple enough to measure capacitance of leads+tag first, and subtract that from readings.
Dry capacitance measured to be around 22nF

Conditioning Plates So Capacitor Can Operate “wet”…
Positive plate needed conditioned to build up the thin anodized layer on its surface, allowing it to have a low leakage current when filled with water. This process is just anodizing and so long as the capacitor is connected the right way with +ve on the titanium,  it will happen automatically with any leakage current through the capacitor. Once place I’d looked at MUTR, had a nice hint sheet for anodizing Ti, To get the full colour range for anodized Ti you really need a 100V so they suggested clipping PP9 batteries together to produce a pile with output steps of 9V. As an anodizing medium they suggested dipping the titanium in Diet Pepsi. Seems the phosphoric acid in many soft drinks helps to anodize the titanium anodizing, and they claimed experimentation had shown flat Diet Pepsi to work best.

I wasn’t bothered about the colour changes, only the leakage at a given voltage. Wasn’t likely to be running the cap on more than 12V, so anodizing on a normal 12V supply should do the trick if the supply would handle the initial current. Cnnected the cap up on crocodile leads, and put the cap in a cup partially filled with pepsi being careful to get any air out of the pipework, and that the end with the + clip was out of the liquid. Measured the resistance. I forget what it was but as expected it was too low for reasonable use as a capacitor, but I=V/R showed resistance would be high enough that my cheep Maplin “pretend” lab supply could handle the initial anodizing current at 12V. Switched meter to watch current, and connected up the power supply. Current soon halved. Then four times as long to halve again. Exponential decay so resistance clearly going up, but it would take “forever” to top out. To give it a fair time for conditioning, left capacitor connected up to power supply overnight.

Next day took capacitor out of the pepsi, gave it a good rise out, blew it dry with compressed air and tested the dry properties again. Dry capacitance hadn’t changed at 22nF. Dry resistance had remained constant at above 14Mohm. Dipped in a cup full of tap water. Anodizing had indeed added an insulating layer, so now no measurable change between wet and dry resistances. Now measured the capacitance when immersed in water capacitance was now 116nF.

It works and that tingles. (But be careful)…
Having got the water variable capacitor working, I proceeded to test it in a 555 astable circuit. OK the sensor was very difficult to adjust to intermediate frequencies, but that was fine, for the binary type feedback I was needing. What was really puzzling is every once in a while the 555 just died. no magic smoke, but they would cease working.

At first thought it might be the capacitor being too big/small for the 555 I’d used. Timing resistors I had in there, and capacitor size were all within spec. Doubled checked the voltages. There’s no way that 5V was too much. Could it be static. Not hugely likely – seemed to get the same amount of intermittent failures when I had myself grounded. After blowing quite a few 555 (and halves of 556s), I eventually correlated the failure to something happening when the amount of water in the sensor changed. Was it a surge or something?
I tried emulating the capacitance change by using replacing the water variable cap with various combinations of standard caps and switches, and switching the capacitance connected to the 555. No failures, and project was put on hold as until it was figured out how to stop the failures it wasn’t usable

After a break I came back to it, and within 5 minutes I got a major clue. That time I’d set things up so that water from the kitchen tap would enter a funnel, and be directed through the sensor- wet cap. if the tap was closed any water inside the sensor would continue draining out resulting in dry cap. Really easy to change the capacitance without disturbing the circuit, or changing the capacitors. Once again the 555 had been behaving properly, then it stopped oscillating once again, clocked that was just after I’d turned the water off. So were the 555s failing not because of a step change in capacitance, producible by switching capacitors in/out of the circuit, but because of the capacitance of a single capacitor failing at a rate?

Looked up the capacitor equations to see if they would offer any clue. hmm… C=dq/dV . I really needed something containing dC/dV but couldn’t find anything like that, and my differential calculus is too rusty to work it out myself.

Found E=0.5C*V^2 and decided to try working with that. The energy contained in a disconnected (perfect) capacitor should remain constant regardless of what’s changes with it so E1=E2= 0.5*C_1*V_1^2 =0.5*C_2*V_2^2
playing with that lot results in V2= sqrt((C1/C2)*V1^2)
now that applies to a disconnected perfect variable capacitor, If leakage is low, and all devices connected to the cap are slow to react to a change in voltage, it can be used to estimate the maximum voltage spike a real variable capacitor might generate on it’s plates.
V_1=V_supply*2/3= 3.3V
C_1=116 nF

V_2 = 25V
voltage on the cap might have hit 25V, and that’s considerably above the supply rail so it ain’t surprising that killed the 555s. and now that I think about it I’m sure I’d occasional get some “tingles” from cap. Most people know that high voltage caps can be be lethal, but this is a new thing to watch out for in capacitors . I’ve no idea what kind of breakdown voltage these titanium water caps are capable of handling once they have conditioned themselves to handle that kind of voltage. Breakdown voltage must be at least 100V as that’s a voltage than be used to anodize, and I can’t see why a really thick insulation layer couldn’t handle far more. Lets see what a working voltage of 230V (harmonised EU mains voltage ) might produce – 1.82kV. ouch!

If anyone’s interested in obtaining some of these (maybe in different size pipe or whatever) drop me a line, and I’ll consider it.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: