Fiber Optic very accurate water detection

[Kalvin]

The reflected laser beam from the water and the webcam/image sensor should give accurate results down to few micrometers depending of the image sensor's pixel size. The image sensing will give absolute water level as it doesn't have the problems related counting the fringes or nulls. Should there be any ambient vibration, averaging the image sensor results will cancel out the ambient vibration.

Averaging wont help though, if there is a meniscus on the water surface - the tube ID is only 13.5mm

Wouldn't that affect the contact based measurement like micro level sensing shown above as well? The non-contact optical method doesn't suffer from the surface tension problems or contamination of the measuring probe like the micro probing or any other contact-based method will probably do.

[nfmax]

That's an interesting problem you have there, @TimB100! 

[ajb]

For those wanting to see micro level sensing in action there is a video

https://youtu.be/oA-0UgrdPBU

What I do not understand from the video is how it work.

The description for that youtube video says that the fiber has a silver coated tip.  So when not in contact with the water, the silver coating reflects most of the light back up the fiber, then when the water comes up the sides of the fiber the light starts escaping out of the sides of the fiber through the water, and less light is reflected back up the fiber.  The amount of light lost depends on how far up the fiber the water gets, so there's a narrow range where it's possible to detect a level rather than just contact/noncontact, but measuring that reliably will depend on careful control of the amount of light applied to the fiber.  Using a reflector on the tip of the fiber eliminates the problem of droplets/meniscus clinging to the face causing changes in reading.

The "Y splitter" you'd be looking for is an optical circulator.

[TimB100]

Thanks ajb

That optical circulator looks good but to expensive for production use.

 I have a 1mm od  fiber cable and trimmed the cover of when doing initial tests. Looking at the fiber in the video it seems to be a Single Mode cable which might be only 0.009mm od WOW!

So might try and find a cable with e core of say 0.5mm or less to eliminate the issue of drops clinging to the end

I have come away knowing more than I started with so its been a good day

Cheers

Tim

[nfmax]

Those diameters are CORE diameters, not the actual fibre diameter. Single mode fibre usually has a core around 8 or 9 microns across, while multi-mode fibre is generally 50 microns core diameter. In both cases, the standard diameter of the actual glass fibre is 125 microns. To prevent surface damage causing breakage, an additional protective primary cladding is applied, almost always a soft polyacrylate, giving a diameter of 400 microns. If you buy a reel of 'fibre', this is what you get. Fibre patchleads generally have a secondary cladding, strength members, and a soft silicone jacket giving a diameter of 3mm or so (varies).

Fibre with just the primary cladding is quite robust and can be handled with care. The care is mainly care not to stick bits of it into yourself. 

SM fibre is available (at extra cost) in smaller sizes, generally intended for use in very small coils - we have used 80 micron (glass) diameter fibre. Diameter matters because the water 'drop' that is likely to end up hanging on the fibre will be of around this size. Fibre like this will probably have polyimide coating which is much thinner than polyacrylate (just as good as polyacrylate for protecting the glass surface but not as effective at preventing microbend loss)

The fibre in the video had been tapered - presumably using a fusion splicer. I think how the sensor is working is that this taper has the effect of coupling light from the primary mode, guided by the core/cladding interface, into cladding modes, guided by the cladding/air interface (the primary cladding has been stripped). In air, these travel to the cleaved end, which is possibly metallised (?), and are reflected back, still as cladding modes. When they hit the taper, some at least of the cladding modes are coupled back into the core and go to the photodetector. When you immerse the thin part of the fibre in water, the cladding mode is more weakly guided and so escapes from the fibre into the water. This reduces the reflection. The more fibre is immersed, the more light is lost. However it looked as if the function was not monotonic with depth, which is probably a function of the particular cladding modes being excited.

The fibre can only be tapered if the primary cladding is stripped. The taper and the sensing fibre will be phenomenally fragile!

Edit to add:

SMF-28e from Corning https://www.corning.com/media/worldwide/coc/documents/PI1463_07-14_English.pdf
Small diameter SM fibre http://fibercore.com/product/polyimide-coated-sm-fiber

Fibercore are local to us, and for many years I worked together with the guy who is now their CTO!

[TimB100]

Thanks Nfmax

What I'm going to test as soon as my work load work gives me 5 mins is the light transmission of IR through water. The surface water layer should cause internal mirroring and reduce transmission.  As soon as the Fiber pierces the water the IR should be able to enter it with less losses and be detected by the ir transistor.

I need now to find some better cable than the plastic fiber cable I have.

Tim

[SeanB]

How about using a PTFE rod and machining the desired ID through it. That way you will not contaminate the DI water there, which I assume you are making to injectable water quality. Thus no copper or any metal to touch the water, only either borosilicate or Quartz glass and pure PTFE will do, and all has to be either steam sterilisable or withstand aggressive chemical cleaning. Then you can use a bare fibre top and bottom, fed through a microbore through the PTFE walls ( bare fibre without cladding) and bent gently in the tube so there is a curve in the fibre. Then you can use a constant illumination light source ( or a reference fibre of the same length in a cladding) to illuminate the 2 or 3 fibres evenly ( 3 fibres if you use a red LED in a metal housing with the fibres arranged to see the beam reasonably evenly. otherwise you need another photodiode there to measure light output and drive the LED to a constant illumination level) and then a 3 photodiode block ( thermal mass so each has a constant transfer function that will track) to get the drop in light as the water touches each fibre, and total internal reflection is going to change the light level as it changes refractive index on the curve in the fibre.

If you can bend the fibre in the tube diameter then simply drop them down from the top, and then adjust so the droplet on the end is constant, though you probably will simply have to have the drilled microbores at a fixed distance apart and use that as a calibration, polishing the bore to a smooth finish before installing the fibres. Fiddly, but you can thin the tube at the fibre points or use a larger hole and PTFE plugs on the fibre ( 2 different diameters so you can thread them into the tube, tapered to lock in the wall and then held in position with epoxy and a outer collar) to get them in there.

PTFE and glass is not going to contaminate the water, and the fibre will also do the microsensing you want for the fine adjustment. Just use the small linear range where the water is only covering part of the fibre as sense range before it is either immersed or not.

[max_torque]

You want to measure a "volume" of water yeah?  So, assuming your measurement system is measuring a linear distance or angular one, then you need the thinest "resevoir" possible, so that tiny changes in volume lead to large changes in distance.  Using a 1mm dia meter pipe, that is long would allow you to use basic digital sensing techniques to accurately measure volume.  Even better, use  a very thin ID tube, coil it up, use a IR beam break or TIR sensor when the water in it gets to the necessary length.  If you connect a large volume to the end of the pipe, the pressure rise will be small and linear with volume.


Alternatively, how about just pumping in too much volume, and just drill a hole in your current tube so "excessive" water flows out again!  Lots of complex dosing systems do this, because it's simple and works!!

[TimB100]

SeanB
In one of my original posts I said it was clean steam condensate.

max_torque
The trouble of loads of posts is that new commenters cannot read them all. In one I said that the chamber is emptied in a cycling system.

Anything less than say 8mm you get to much surface tension and it will not drain. Also as soon as you get into small dia temperature changes makes a big change in volume/length.
I have to know when I have ~30ml so at 8mm dia you're talking 600mm tall, 1mm would need a pipe around 40mtrs

As per the title I'm looking for info on fiberoptic level detection.

[Kalvin]

Although you are looking for a fiber optic solution, here is a quick test with a cheap 640x480 webcam bought from eBay measuring the water volume in a translucent plastic container. The container's diameter is approx. 45mm, the initial water volume was approx. 10ml, each additional volume was approx. 0.2ml. The water level change is approx. 37 pixels per 0.2ml, thus one vertical pixel in this setup corresponds to 0.2ml/37 = 0.005ml. With better optics and camera/image sensor, and proper lighting conditions, possibly using polarized light the image quality should be much better. With a thinner 13mm tube vs. the 45mm container used in the experiment, the resolution will be much improved**. However, this was just a proof of concept.

** The change of water level height in 13mm tube vs 45mm tube is almost 12:1. As the height of one pixel row represented 0.005ml change in water volume in 45mm tube, there will be 12 pixel change in the 13mm tube for the same 0.005ml ie. one vertical pixel corresponds to 0.005ml / 12 = 0.0004ml in 13mm tube.

[filssavi]

The simplest thing you can do (for me anyway) is use the capacitive sending, so put two electrodes in the water, apply an AC excitation signal of suitable frequency/amplitude, and measure the capacitance between the 2 electrodes, as water's relative permittivity is much higher than air's One (58/80 to 1) you should see the capacitance increasing as level goes up

[C]

 optical circulator
They use power splitters in cheap fiber optic systems. Think in addition to the power split, it is also somewhat of a directional coupler. I am thinking of the fiber trunk where the power tap is just wrapped around trunk some distance and heat fused. Thinking from lower cost angle.


Going a step, Time of Flight.
Think of using fiber length for a fixed TOF sensor. A split and combine of equal length, no amplitude change. Unequal length of one after split should give a step change for an input amplitude change. Could need a reference  to compare measured tof against, like a three way split.
Your TOF change could be to small for cheap.

If you have a frequency that changes some based on level, today you might think of a x___ phase locked increase. Using harmonics is often forgotten. For a 1 Hz change the 1000x harmonic changes 1000 Hz.
You could use two of the harmonics in to a mixer and use same circuit to repeat this.

[Kalvin]

The OP wanted the measurement repeatability to be 0.01ml which corresponds to water level change of 75 micrometers (0.075 mm) when the tube diameter is 13 millimeters. The TOF measurement will require measurement resolution to be less than 0.5 picosecond (assuming that the light travels in a fiber 20cm/ns: 2cm/100ps, 2mm/10ps, 0.2mm/ps, 0.1mm/0.5ps).

[TimB100]

Hi,

Just to say I'm going to write up after work a very detailed description of all the issues involved in this project.

I really appreciate all the suggestions but 95% of them have already been thought of and dismissed.

Cheers

Tim

[C]

Think you are only thinking of TOF of Fiber only. One side is fiber, side(s) is fiber, water & or air).
So it the ratio of water & air TOF.
With adjustment would result in level change or no level change.
Probably still to small.
Might also be possible to circulate signal to make change larger.

Edit oscillator
  Some here have mentioned using water to change value of a cap creating and oscillator that changes frequency by water level.
Using a time delay you can also create an oscillator. Ten feet of cable vs 100 feet in feedback path.
 So you have an oscillator with radio, light or sound creating a time delay with the water vs air TOF change creating the frequency change in oscillator. Sound for example should bounce back when it strikes the water and container bottom. These standing waves should add at certain distances.

[LaserSteve]

TI makes some wonderful noise rejecting capacitance sensing chips.    FDC 1004 is a favorite around here at the University, especially with its driven shield and low cost EVM with decent logging software. You can get a hydrophobic or super hydrophobic material for the outside of the probe. Flattening
the "Contact Angle" with the right material on the outside of said probe reduces or modifies meniscus to tolerable levels.

Steve

[max_torque]

max_torque
The trouble of loads of posts is that new commenters cannot read them all. In one I said that the chamber is emptied in a cycling system.

Anything less than say 8mm you get to much surface tension and it will not drain. Also as soon as you get into small dia temperature changes makes a big change in volume/length.
I have to know when I have ~30ml so at 8mm dia you're talking 600mm tall, 1mm would need a pipe around 40mtrs

As per the title I'm looking for info on fiberoptic level detection.

It would seem much easier to me to simply use a cheap, easy to drive pump (say a perastalitic one etc)  to push the water out of a narrow tube than spend thousands of man hrs on complex fibre optic detection of infinitesimally small distances.  KISS is a prinicipal that seems to have been forgotten rather these days.....   

And of course, only the bit with the sensor in needs to be narrow (to give you the high resolution), the rest of the pipe can be any diameter you want, so no need to have a "Long" or "large" system.  Use physics to your advantage is my suggestion!

[Kleinstein]

The surface tension / surface wetting will be a real problem. The surface tension per se is not the problem, the difficult thing is that wetting may not be fully reproducible with some more or less sticky points at the surface. Copper is not that perfectly corrosion resistant and thus might have a surface that changes in wetting and might develop nasty spots.

I personally would choose a thinner diameter (which would ease the measurement) and use a kind of glass tube so that an optical measurement from the outside is possible.

A capacitive measurement can get tricky, as at low frequencies the conductivity gets important as the water will be more like a poor conductor, reducing the spacing and not really sensing the dielectric function of water (which is less sensitive to impurities). The cross over can be rather high in frequency (e.g. 100 MHz range). Also isolators might change there properties with humidity / water contend. So while I like capacitive sensing, i would use it here.

[TimB100]

Here is are the notes I made. As they contain pictures I have it as a PDF

[LaserSteve]

What is the minimum amount of water to be dispensed ? What is the dead volume of water, the part that can remain in the cylinder as a minimum load?
That is a big question as well.
At work we use precision pumps for this, off the shelf and controlled by USB or a serial port.
Steve

[TimB100]

Hello LaserSteve

There is a constant flow of condense, I'm sampling it constantly to get the volume/time
The dead volume is whatever I want it to be as the design is not finalised. But it will probably be around 10ml

As per the first answer I'm not dispensing I'm measuring.

[max_torque]

Hello LaserSteve

There is a constant flow of condense, I'm sampling it constantly to get the volume/time
The dead volume is whatever I want it to be as the design is not finalised. But it will probably be around 10ml

As per the first answer I'm not dispensing I'm measuring.

Could you do a simple sketch of the system?  Might help us get an idea for how it looks and sizes  etc.

Is the flow of condensate constant? at what rate?  Is it being condenses else where and being piped to the measurement cell? Average measurenments or real time?


Perhaps some sort of resonant sensor? Measure a drop or two at a time, through changes in the fundamental frequency or amplitude of a tuning fork with a (small) bucket on the end ?  Once full, a servo quickly tips the condensate out. The tips per minute are your flow rate??

[max_torque]

Sorry, just seen the .pdf you stuck up there^^^^ which answers those questions!

[max_torque]

One technique that worked on a larger scale (it was a 220l oil drum!), but could possibly be scaled down is to use a diferential pressure sensor and a pump to drive water OUT of a vertical tube that runs inside a tall thin measurement cylinder.  The pump runs continuously, i used a tiny vane type pump, and takes air from the top of the main cylinder, and pumps it into the measurement tube.  When water collects in the main tube, the pressure differential between the top of the main cyl and the messurement cyl is the height of water above the bottom of the measurement tube, because the pump keeps any water from entering the measurement tube.  You need an accurate diff pressure sensor, but these days that isn't THAT hard, and need to remove the pressure pulses from the pump with software filtering of the dP signal (again not too hard).  Using a diffPressure sensor means it isn't skewed too much by baro changes, but you may want to correct for changes in air density? (measure temp and baro too)

[max_torque]

Something like this (excuse crappy sketch!)