On Feb 24, 6:17 pm, Jim Logajan wrote:
"RST Engineering" wrote:
I wonder if a pressure sensor placed inside the bottom of the tank
could be made sensitive enough to "weigh" a column of fuel inside the
tank above the sensor. That would only work for regular sized tanks
(no triangles) but could be integrated over a long enough time to take
care of any slosh.

Just a random thought, mindya...and I haven't run the numbers.

Another thought is a string of LEDs separated from photosensors with,
say, 10 or 15 of them inside the tank mounted vertically.

Hmmm...any other thoughts for liquid level measurements? Sonar a la
Polaroid?

Just looked through my copy of "AIP Handbook of Modern Sensors" by Jacob
Fraden for existing ideas and inspiration. Here's what I got:

It shows one fiber-optic sensor that relies on the change in refractive
index between air and the liquid being measured. If you can find a copy,
check under section 5.8.2 "Fiber-optic sensors". But since it mentions
the company Gem Sensors as already having a product built on the
concept, you can go to their web site for technical details:

http://www.gemssensors.com/content.aspx?id=382

Unfortunately it is a "point level" detector. Gen Sensors also has level
sensors that use ultrasonics in a different way: impedance changes (see
for example):

http://www.gemssensors.com/uploadedF...c_Sheets/exose...

Another idea with respect to refractive index change: use a laser diode
at the top of the tank angled down to a mirror at the bottom of the tank
and a line of receiving photodiodes or phototransistors at the top of
the tank, like so:

Laser
Diode P1 P2 P3 P4 P5 (Line of photosensors)
\ /
\ /
------------------------ liquid level
\ /
\/ (Trying to show change in angle)
-----==========--------- Bottom of tank
(mirror)

As the liquid level changes, the photosensors(s) that receive most of
the light also change. Obviously vibration will make the reception dance
around but time averaging should probably work to resolve that.

Ironically I think vibration noise can actually be used to yield a
higher resolution than that given by the number of receptors. That is,
if you had only 10 sensors but did 100 readings and got 70 hits at P7
and 30 hits at P8 you could reasonably say the average is near (70*7 +
30*8)/100 = 7.3 within (I think) an RMS of +/- 0.1 (That is, 1/sqrt(N)
where N is the number of measurements. But my statistics is very rusty.)
By contrast, if there were no noise the resolution would be limited
to +/- 1.

All of them good ideas, but why are they likely to be more accurate
than the float level sensor?