Comparison of Capacitive Soil Probes

Intro: Comparison of Capacitive Soil Probes

This is not an instructable as you will most find here, it is a comparison between two projects -capacitive soil sensors- found here on instructables and as such it does deserve a place here.

In february I published a capacitive soil moisture probe in which the capacitor consisted of some scrap pcb that effectively formed a double sided PCB.

Then in May I came upon another capacitive soil moisture probing project by acolomitchi. His probe design was a more 'interlocking' pattern etched on a PCB.

We got into a discussion on the best design and he graciously offered to send me one of his probes to test.

He used an oscillator based on a the good old '555', in a simple astable mode, whereas I used a feedback oscillator based on an HCT14 which is basically a Schmitt trigger inverter. There probably isnt much difference between those two oscillators but we were mostly interested in the difference in probe design. For both oscillators T=0.67*RC. (for HCTversion) (and 0.8*RC for the HC version)

When I was 'designing' my probe, I just wanted a cheap one quickly and therefore the intricate design of interlocking pads as used by others was out of the question. I hesitated between two designs: a single sided piecen of pvc with two seperate 'plates' on it, or a double sided piece with the two capacitor plates at opposite sides. As I figured the dielectrum formed by the pertinax or polyester in between the copper would be less than that formed by the surrounding soil if humid, while with a single sided design due to the small gap the capacitor might be more sensitive but have a smaller range, I chose for the dubblesided design.

The picture as shown by acolomitchi tells the opposite with the dubbelsided design with plates of opposite polarity back to back, most of the capacitance would be formed by the pcb material between the copper and a much lesser part by the surrounding soil.

This was what actually sparked our idea to do a comparative test. Colomitchi was kind enough to send me one of his probes as well as a pcb for the circuit. This arrived soon enough.
The probe was well made, rather sleek and measured 19x97mm. Definitely smaller than my probe and I thought it would be unfair to compare those two sizes, so I made new one, same size as the colomitchi probe.
Obviously, mine had more copper surface, but as I was testing two designs, I thought it to be fair to test comparable PCB sizes.

Before my comparative tests could start I first had to find a way to coat my new sensor again. That in itself proved to be an entire undertaking.

1. heat shrink
   I had covered my previous,bigger, probe with heat shrink, but when i did that with a smaller probe, the heat shrink proved too thick to provide for a decent capacity of my probe.
2. Spray on laquer
   Initially that seemed to work fine, but halfway through the test the laquer seemed to start leaking. Not being deterred and after some back and forth with Adrian, I tried 2 coats but with a decent curing by the sun in between. That held a bit longer but also started leaking
3. nailpolish
   Nailpolish seemed promising. After all, i use that too to touch up a PCB before etching. However even after a thorough bake in the sun that didnt hold up when put into water. I tried another nailpolish that specifically required curing under UV light, applied that, baked it in full sun but no avail. Also that gave in to being immersed in water.
4. laquer
   I finally tried a laquer that I had previously used on wood. I think it was a regular Flexa high gloss wood laquer, but it was of the 'old' variety. Meaning that it was not water based, but with a decent solvent. I immersed the probe in the laqcer and hung it out to dry in the sun for a day and gave it another coat. It formed into a rather tough coat that indeed turned out to hold well in water.
5. PCB soldermask/paint
   

   A bit too late for this test, I got my hands on some soldermask to coat PCB's. After proper curing this also held up pretty well.

The testing could begin. As we had exceptionally much rain, testing it outside was not really an option, So I decided to test in a 20 cm diameter (3.7 liters) pot, filled with regular potting soil. But first some open tests to set some baselines.

Fully immersed in water both probes measured around 4 nF. In open air it was verging on the limit of my measurement range, but it might have been around 50pF.

The soil that I used is only slightly moist when it comes out of the bag. In that environment my probe measured 300, His probe 500pF. Then started to add equal amounts of water. 50 ml at the rim of the pot (so at about 10 cm distance. No clear immediate effect. I waited and then something interesting happened. "My" sensor started to react after about 30 secs, but only minimal. It became 400pF, but after about 40 secs Colomitchi's sensor kicked in. Mine had gone up to 450 by then and his went to 630 and then both stayed around that value for a while.

Added another 50 ml and as the path obviously had been wetted by the previous amount, both probes went up. His to 1.2 nF, mine to 900pF. adding another 50 ml sent his up to 3nF and mine to 1.6nF. Till now I was pouring the water all at one side, directly facing the plates. So, as the soil at one side was rather wet, decided to pour 50m at the otherside. That had a dramatic effect on my probe, while his went up to 3.2 nF, mine went up to 3.5. adding another 50 ml at that second site made both probes go up till about 4. adding more water did not have a significant effect.

     HIS  MINE
     500  300
50ml 500  320
     500  400 (after 30 secs)
     630  450 (after 40 secs)
50ml 1200 900
50ml 3000 1600
50ml 3200 3500 (poured at opposite side)
50ml 4000 4000 (poured at opposite side)

Then I decided to repeat (my floor really has become a mess) the test with fresh soil, but this time pour the water facing the edge of the probes. Starting values again were practically the same. Pouring 50 ml again each time.

        HIS  MINE
        500  320
50 ml   500  320  after about 1 min  550 -450
50 ml   600  450
50 ml  1300 1200
50 ml  2500 2600
50 ml  3900 3900 (climbing to 4nF)

That prompted me to try some rather dry and compact sand and pour water at one side: facing the broadside

         HIS  MINE      
         --   --
50 ml    --   --  waited 5 min... same result
50 ml    --   --  waited 5 min  : 400 --
50 ml    590 --
50 ml    800 --
50 ml   1000 --
50 ml   1700 300
50 ml   2000 350  Waited 10 min 1900 350
OK time to pour some water at the opposite side
50 ml   2000- 400
50 ml   2100-1700
50 ml   2250-2100
50 ml   3000-3000
50 ml   4000-3900

Well which probe is better???? :-)

Not completely surprising my probe is very directionally sensitive. When water comes from one side, his probe already has half of its plates being able to 'act', whereas my probe doesn't. Now if the soil is a bit moist, that isnt such a problem as my probe's backplate takes part in the dielectrum as well, but in very dry soil there is just no decent dielectrum between my probe's two plates (or better around the my probe's two plates) even though one side of the soil may already by moist. So... when it rains.... I think both probes go up equal (but then there is no need for irrigation) but that if irrigation comes from one particular spot, my probe may let irrigation on for longer than necessary, but ofcourse that is all a matter of choosing the right level to switch off. However his probe's linearity is better. for my probe the same amount of water can cause a jump from 350-400 or from 400-1700 So I think my probe can definitetly see a good difference between dry and wet, but if the soil is very dry to begin with it may have some trouble differentiating between wet and very wet. Now obviously with an irrigator in place the soil is not intended to get very dry, but still my probe may be a bit better suited to differentiate between, "yeah it is time to irrigate" and "it is really really time to irrigate."

I am still a bit baffled by the first result in which my probe seemed to earlier sense the water, but that could well have been an artefact.

Conclusion
It is hard to say which probe is 'better'. But my guess would be acolomitchi's probe. However, the most important thing is that both probes can distinguish between humidity levels pretty well and they are not saturated in the range they are supposed to measure. My recommendation though would be acolomitchi's sensor rather than mine. Mine may seem cheaper, but colomitchi's probe, coming from a fab house, is already covered with a proper coating, whereas for mine you would need to coat it yourself and you may have a few misses there.

Comments

[JeremieF1]

This is probably one of the best educated instructable I ever read, thank you guys for reviews + modelling + cooperation! Makes my day :)

[diy_bloke]

Jeremie, thanks, it was fun doing and in fact I was curious to see how cheap one could make those probes.
As a bit of update, the cover of my probe suffered a bit from the cold, I presume the frozen soil was hard enough to probe some holes in the protective layer, but that was fixed soon.

Putting it all together, capacitive probes definitely have advantages over resistive probes, but they are slightly more complicated. So what would I now chose?
Well I am very happy with the two capacitive probes I have, but I am also very happy with the resistive probe opamp version, as well as with the resistive one that is being read by an Attiny85.
But all in all i have chosen for resistive probes with addition of some hard/software that will cut off the power to the resistive probe for most of the time and only switch it on for the few ms that I read every 4 hrs. Does wonders for the metal of the probe.

Given the fact that resistive probes actually measure molecule concentration rather than true humidity, if I were a commercial farmer I probably would opt for both, just to be able to get some feedback on the fertilizer I might be adding.

As it is, both types of probes have given me satisfactory results.

Currently I have changed my irrigation set up. Irrigation is now triggered by a combination of soil measurement, temperature and weatherforecast. I measure all my growbeds with 2 pcf8591 ADC's, use the mean to establish if dry or not and use the individual results to reduce or increase irrigation to that specific bed. Used to be something self written but bringing it all into OpenHAB

[JeremieF1]

BTW did you consider tar paint? I am about to try, as this stuff is weatherproof and soft, kind of indestructible in the long term (e.g. I used some to protect the metal flat beams I soldered to protect the underside of my dirtbike engine, and there is no single trace of rust 20 years ago!). It may be thick though, so I consider heating it before dipping the probes in it. Not sure at all about its dielectric properties though :p

My probes are made of milled PCB, with alternating strips of about 3mm each and 1mm gap on a 5x3 zone. I count the pulses from a low power 555 variant in astable configuration, that fire at about 100kHz (I used to go much faster but I can no more use the MCU interrupts -- I think the faster the better against EM noise). For now I used only a single side, but they already give good and reliable results (still to be assessed further). I will check double sided ones when I receives them from seeed studio, I guess I will get much better results.

My issue with resistive probes is they are quite dependent on the soil composition. This may be no issue when you know where your are going to use them and they are calibrated against it, but I have no clue where people might use mines so far (and my garden soil is amazingly heterogeneous anyway).

I use MQTT a lot nowadays, this protocol really rocks imho. I did not try any interface yet (beside some dirty homemade web), and I have to give OpenHAB a few days soon or later.

Thanks again for the update! :)

[acolomitchi]

1. to me, came as a surprise that, in wet soil, both probes have almost the same capacitance. When I'll have time, I think I'll set my hands on an electrostatic modeller to see exactly how the field line go

2. when I experimented with probe pattern designs, I needed to factor the scarcity of water - I simply could not afford to over-water the trees for 1 month only to fail them afterwards.
How's that relevant? If you don't have water restrictions*, going with the cheapest design seems sensible. If a simple piece of double-sided PCB is good enough then it effective and, surface to surface comparison, less expensive than the PCB you'd be using.

(*do consider all factors related to water restrictions. Throwing half a litre more/plant on a veggie patch won't cause the same troubles as a potted plant set above your high quality carpet; your significant half may object the most strenuously to the amount of water you use, be it automated or not).

[ancienthart]

When I modeled the data above in Excel, I found that the derivative (dC/dV) with respect to water volume almost perfectly fit the logistic equation with a centre roughly between 120 to 125mL water, depending on the sensor. http://mathworld.wolfram.com/LogisticEquation.html...
Since the differential equation of the logistic equation is dC'/dV = kV(T-V), (note that this is a differential equation of a 1st order derivative btw.), this suggests to me that the electrostatic field is being 'partitioned' between the soil and the plate in a series connection (hence the multiplication of the two factors, rather than addition).
As V increases, the factor in the brackets becomes weaker, while the kV factor outside the brackets gets stronger.
I'd suspect that as the soil becomes wetter, the additional electrostatic field moves out into the water (factor outside brackets), meaning eventually this completely dominates any (weak) electrostatic field left in the sensor (factor inside the brackets). I also suspect that this is purely due to the water, as it seems to be pretty independent of the soil type as well.

This is all empirical btw, trying to fit a theory to the fact that I got the logistic equation in my data. Maybe someone else can actually explain this from first principles rather than shoe-horn a theory to fit the data.

[acolomitchi]

> trying to fit a theory to the fact that I got the logistic equation in my data.

The hypothesis (your theory) seems reasonable enough.

> Maybe someone else can actually explain this from first principles rather than shoe-horn a theory to fit the data.

Unfortunately, I'm not able to. I reckon plugging in an "electrostatic field solver" (for the Poisson equation) is the least-resistance (pun unintended) way to explore what actually happens.

[diy_bloke]

Correct. It seems that your probe is slightly more linear/ has a better resolution over a range that probably is most relevant to determine levels of adequate moisture. Nevertheless, your soil is different from the soil I tested, but it probably is more accurate than mine for the soiltype I think you have. Whether in practice that is relevant for your trees remains to be seen. As I didnt have double sided PCB, I used 2 pieces of single sided glued together. I wonder if there would be a big difference between them

[acolomitchi]

> As I didnt have double sided PCB, I used 2 pieces of single sided glued together.

That may explain why the electric field lines may "jump to the chance" of closing through moist soil rather than the PBC (once the soil is moist enough) - the most usual PCB is 1.6mm thick, double that for two pieces back to back glued together:3.2 mm, maybe 3.4 adding the dried glue thickness.

Given that the PCB substrate has a Kr (eps-rel) of around 4-5 and water around 80, with a narrow-but-thickish probe one may "convince" the entire area of the electrodes to "send" field-lines through outside of the probe rather than through the PCB substrate (see Principle of least action)

Would be interesting to see the probe capacitance in water varies between a narrow probe (say 10x200 mm) and one with the same area but squarish (say 45x45mm) - if my hypothesis is right, the narrow one should reach a higher capacitance; in the case of the square one, the areas close to the centre would prefer to close their line fields through PCB (their "field line trajectory" becomes too long if going through outside, even if the outside is more ...umm... "electrically permittive" than the inside).

[diy_bloke]

It may, I will see if I still have some scrap left do you mean single sided with a separation over the middle or double sided? (so effectively 2x45x45)?

[acolomitchi]

As you already have 1 probe approx 20x100mm, with 2 single-sided back-to-back, I think that 2 single-sided back-to-back 45x45mm (too keep the same surface, same PCB thickness, but minimize the fringe).

As a third data point, and only if you manage to find scrap, a 2 single sided 10x200mm would be interesting, especially in low moisture content.

[diy_bloke]

i kinda forgot about this one.Will see what i can do still

[diy_bloke]

I will see what I can do :-)

[ancienthart]

I've graphed the second data set in Excel and visually, there isn't that much difference between the two capacitor plates, well within what I'd expect for experimental error. Both stay pretty flat until 100mL has been added, and then climb in an almost linear manner.

Fitting a trend line to the 100mL to 250mL measurements I get the following equations:
"His": y = 22.2x - 1810, R^2 = 0.9798

"Mine": y = 23.5x - 2075, R^2 = 0.9852

Obviously the sand test shows a considerably different result, but I'd argue that this little experiment shows evidence that we don't need to overthink capacitive soil sensors. Just use a bigger plate. My local electrical supplier actually sells blank double sided pcb, so I may give that a go for an even easier approach.

[diy_bloke]

Thanks for yr work. I fully agree. The sand test is indded a bit different but though i dont recall the exact findings right now, sand is a rather different medium as itwill probably compact more as it gets wet.

Though i see a lot of advantages with capacitive sensing, I have now years of experience with resistive sensing and found that he electrodes actually last prety long if i put a voltage on them for only a few mSec, every 3-4 hrs.

I also changed the irrigation protocol. If dry, i irrigate for a set number of minutes, The advantage is that i dont have to keep measuring (and corroding) while the pump is on.

In short..... I am kinda deliberating between going resistive or capacitive...I may just keep a mix.

[satsatt]

in terms pf corrosion - has nobody considered corrosion free electrodes ?

what about grafite or stainless steal or ( platinum - hiiii ) ?

[ancienthart]

Some suppliers do have gold-coated probes, but as diy_bloke commented, with enough electricity, nearly anything will corrode. I haven't tried these, but I'd assume that if the gold coating was scratched or scuffed, the copper metal underneath would just get eaten out through the crack.

[diy_bloke]

I have used graphite works well, but somehow they also do dissolve.
I have used stainless steel, also works well, but the key is to just apply current when you are measuring, say every 3-4 hrs

[Tecwyn Twmffat]

Nice work thanks! I'll be testing one of acolomithci's probes in the winter.

[diy_bloke]

Tnx Tecwyn. In the end, finding the proper coating proved the hardest
and testing proved the messiest. Not all my doing. Acolomitchi had a
frequent back-and-forth about it.
As a matter of fact, for sheer looks I found yr probe with the DS18B20 in it the best :-). My probe
has an NTC as sidekick, just to signal me when it is time to plant
carrots. Hope both perform as well as your dads method