5

I'm building a small capacitive water level sensor for a reservoir (16cm tall x 10cm x 7cm).

This prototype gives me 10pF on free air and 20~30pF on medium water levels, so its definetively doable (its a piece of wire inside a brass tube).

enter image description here

I'm looking into possibly getting better measurements (as in bigger difference between free air vs with water). So before chopping the rest of the brass tubing I have, I'd like to ask if there are some known recommended geometries.

One that I find online a lot is two tubes, one inside another. I have also considered building a three layer config (wire, tube, tube), with the innermost connected to the outermost (connection not show in this mock-up):

enter image description here

I have tested with one of the armatures not isolated/directly touching the water and I have way bigger difference between free-air vs in water but I need it to be galvanically isolated.

So, again, is there a way to get more contrast between free-air vs on water? Level sensing is nice but detecting an empty reservoir is even more important.

Wesley Lee
  • 7,844
  • 6
  • 35
  • 53
  • Your construction of the sensing element aside, what method of detection/digitalization do you intend to use? Depending on your analog front end, recommendations would be drastically different. For example, using a frequency to digital converter (FDC) it's possible that a single piece of wire could suffice. Then again, reliable empty reservoir detection might depend on the software more than on the hardware. – sekdiy Jun 03 '16 at 21:37
  • I think DC measurement is not good because of progressive damage of the sensor due to electrolisis. AC feeding to the capacitive sensor is preferred. – Claudio Avi Chami Jun 03 '16 at 21:43
  • @sekdiy I'm considering feeding pulses ("AC") through an RC (C being the probe) and measuring the result (which seems to be the FDC you mentioned). This is not fixed though, I only tested this with a meter, no MCU yet. – Wesley Lee Jun 03 '16 at 21:59
  • @ClaudioAviChami I mentioned I plan to have the probe isolated from the water so electrolisis probably won't be an issue. The measurement I made with water contact was mostly out of curiosity. – Wesley Lee Jun 03 '16 at 22:01
  • @WesleyLee That's not quite FDC in the true sense, as your output signal is a (low pass filtered) voltage, which is then sampled by a voltage-ADC. But it would still work. The strength of a true FDC in that application is that it directly detects changes in capacitance. With an RC filter and an ADC you have to consider frequency, salinity, temperature, charge pickup (drift over time), etc. It's easy to get started that way, but more difficult to get reliable (especially if a missed empty container has consequences). – sekdiy Jun 04 '16 at 06:45
  • FDC might be more feasible in your application if you look for an MCU with an FDC built in. – sekdiy Jun 04 '16 at 06:47
  • @sekdiy I have temp sensing and its a closed system so salinity should not change (my worries about empty reservoir is detecting leaks and evaporation). Do you think its feasible to overcome these issues you mentioned with the RC filter I mentioned? – Wesley Lee Jun 04 '16 at 21:30
  • 1
    It's been done before. Unfortunately I don't have the time to form a full answer. In short: for what you want to accomplish, look into other sensing element constructions, too (like parallel instead of coaxial, dedicated bottom/empty detector instead of single element). A quick read on a possible alternative: http://www.ti.com.cn/cn/lit/an/snoa935a/snoa935a.pdf – sekdiy Jun 04 '16 at 23:31
  • @WesleyLee it looks like your sensor has concentric electrodes in a sealed outer tube to prevent water from entering. Is this correct? Or does the water fill the space between the wire and the tube as the sensor is immersed in the liquid? – Colin D Bennett Jul 20 '16 at 18:01
  • @ColinDBennett -- The sensor is sealed. – Wesley Lee Jul 20 '16 at 18:51

1 Answers1

4

The design of your sensor is not very well suited to measuring the liquid level. Since a capacitive water level sensor operates by forming a capacitor for which the dielectric can be either air, water, or a mixture, we need to maximize the role of the water in the capacitor forming the sensor.

Unfortunately, in your sealed coaxial sensor, the dielectric between the electrodes is always air, so it is ineffective:

Sealed concentric level sensor - bad

Since your sensor has only air between the inner electrode and the outer electrode (the tube), its capacitance should actually not change at all when inserted into the water! But in fact there will be some other capacitive effects that may cause a small change in capacitance, but these are not related to the primary capacitance should focus on.

Because the dielectric constant of water is greater than 50, and air is about 1, a perfect capacitive probe would have a capacitance of more than 50 times higher when immersed in water. Now there will be some stray capacitance due to the probe wires etc., so it is desirable to maximize the sensor capacitance as much as possible, by (1) increasing the area, and (2) decreasing the distance between the electrodes of the sensor.

The simplest capacitive level sensor would consist of two parallel, electrically insulated plates. The space between the plates would be filled of air or water, depending on whether or how far the sensor is immersed in the water. However, parallel insulated wires will work as well: it's not as ideal, but it should work.

Your coaxial capacitive sensor would work if the water were allowed to enter between the electrodes:

enter image description here

However, unless the electrodes are electrically insulated from the water, the conductivity of impure water may be a problem.

Take a look at Texas Instruments' capacitive liquid level sensor reference design TIDU736A. (This design goes farther and gets rather complex including an out-of-phase technique to mitigate effects of external influences.) The sensor here forms a capacitor with coplanar electrodes on the inner layers of a four-layer printed circuit board. The sensor basically works like this:

enter image description here

Colin D Bennett
  • 482
  • 3
  • 8