Fuel level sensor design (continued...)

[mickeyholland]

Hello all,

I was reading an old thread (thread240-171250), and noticed some comments from Warpspeed and cbarn24050 about fuel sensor design ideas.

The idea of an ntc matrix seemed fun, so I gave it a go!

However, I have not noticed a very large change in resistance for the ntc in or out of the diesel, and my concern is that the temp change may NOT be more than the practical tolerance of a thermistor.

In that regard (appreciating the thread was 2 years old!!), did any one else every give this technique a spin ?

Maybe I am just using inappropriate ntcs. So that would be interesting to know what sensors are recommended or were actually used by someone else.

Actually, I was thinking about 1n4148's instead of ntcs, as they are 10-20 times cheaper per item, but maybe that is really impossible even with op-amping them as the 2mV per degreeC is (it seems to me) well below the rated tolerance (about 25%? 0.62 to 0.72 voltage drop).

 

[mickeyholland]

Hmmm, I solved the ntc config so that actually works surprisingly well. Seeing was believing when I connected the battery in the right place

But I still have the same thought about using some other sensor, as the ntc's all seem fairly expensive (considering I might want a stack of 1000 to measure 50cm height of fuel!)

I was thinking about the 1n4148's originally as a) they are dirt cheap and b) they are diodes (so would be "easy" to charlieplex onto a microprocessor - 32 io pins would allow up to 992 diodes/temp sensors this way!)

I suspect it's not too easy to solve with diodes, but if anyone has an idea I would really enjoy to hear about it / try it.

Thanks !

 

[Warpspeed]

As I have never actually tried to do this myself, I cannot suggest a known solution.

How about building a simple test circuit with one thermistor only, and then experiment with the power pulsing, and monitor the resistance change with an oscilloscope. See how much things change between air and immersion in liquid diesel.

The glass envelope of a normal sized 1N4148 may create a barrier to fast enough thermal coupling, very small surface mount components may be faster. I do not know.

Another approach may be to just use some extremely fine wire. Hot wire flow meters and anemometers are both fast responding and sensitive.

Try blowing on a naked lamp filament while measuring the resistance. This is surprisingly fast and sensitive.

 

[mickeyholland]

Warpspeed: Thanks for your comments.

The speed of the thing may not be too crucial... I suspect it would take some seconds for the fuel to "drip" off the sensor part anyhow as the level goes down, so that will already create a delay of some seconds. Unless these are teflon coated parts (or whatever that high-tech non-stick stuff is)!

I was thinking about the melf type LL4148 to make construction of the final "big beast" simpler, as they can be almost thrown onto a pcb and baked in the oven, rather than hand soldering 2000+ connections! So I will try pulsing and playing with a few of those to see what the oscilloscope shows.

But on your new idea...!

Your hot-wire idea may be workable also, but would that not be a little dangerous in fuel? Maybe in diesel it would be ok, but not in regular gasoline?

If I figure that there would be 1000 thin traces on a pcb edge(perhaps 0.5mm wide, 2mm long, and spaced 1mm apart), then I cycle through them (to detect which is effectively open circuit and which is effectively closed) by connecting some current limited supply (say 5v 1mA) on one trace and measuring the voltage on the next trace.

In this case, worst case, the power would therefore be cycling up and down the pcb inside the tank... would that not create some sparking or gas buildup or something... or would such a low (1mA) current just not be noticed by the fuel, either short term or long term?

 

[Warpspeed]

I cannot see any danger to this, it is no worse than the crude wire wound potentiometers used in many existing fuel tank sender units. And the temperatures should be quite low.

If they can fit electric fuel pumps into tanks with dc brush motors, I would not worry too much about a circuit board.

The current pulse for heating may need to be fairly respectable, but the thermal coupling into fluid of an exposed flat PCB track should be quite good. It would also be a fairly neat cost effective solution for manufacturing, if it works that is!

The tracks could be zig zags or flat spirals to increase the track length. Strain gauge sensors and platinum RTD devices both use some fairly fine geometry tracks quite successfully.

The trick might be watching the resistance change over time after the heating pulse goes away, and picking a suitable sampling point for temperature measurement.

The software algorithm could also be something that hunts backwards and forwards constantly seeking the fluid air interface. It may not need to scan every single sensor in the whole array sequentially, which would be slow.

Another nutty idea might be using small surface mount transistors or mosfets as pulsed self heaters?

All these ideas could be tested initially with only one sensor.

 

[mickeyholland]

When you say "fairly respectable"...

I could sink 40mA from the microprocessor directly, if that would be enough --- it will try it!

But if I need more, what do you think the upper milli/amps limit would be from a safety standpoint, assuming I stick with 5v supply (which somehow feels more comfortable than pumping in 12v!)?

As for the NPN, some of those are also very cheap, so that is good!.. I originally discounted them as I figured they would act much like the diode and therefore not be so effective given the tolerance/resistance change ratio, but in anycase, now you mention it, I will actually test those too - might do something magic!

But I am thinking this pcb idea might be more like a winner, and certainly easiest to build... I would still need a diode on every "sensor trace" to enable me to matrix the several 100 inputs to 1 or 2 micros, but that is still within reason. .. Actually if I could get really thin traces and leave them coated with pcb green stuff, that thin insulation might also prevent any possible sparking and corrosion of the traces - if it doesn't impair the performance that could be a winner!

Well, this could either work well or explode well, so in any event, I feel compelled to try!

I will definately keep you posted!

 

[Warpspeed]

No need to use the green solder mask over the sensing tracks. The tracks will be tinned anyway, and will not corrode.

As to specifics, I believe it will just need some careful testing to see which idea shows the most promise.

 

[WGJ]

mickeyholland - why would you need 1000 sensor points to measure a depth of 50cm? What's the application?

If it's anything automotive, forget precise resolution - as soon as you start moving, the sloshing of the fuel will mean pinpoint accuracy is virtually worthless.

I tried v.small glass bead thermistors on an instrumentation project on a vehicle fuel delivery module so that I could look at the level of the fuel in the module reservoir versus what was going on in the fuel tank itself.

A vertical row of thermistors was glued (epoxy) onto the inside of a reservoir and connected like this:
img

http://s279.photobucket.com/albums/kk145/bill_gj/sensor/?action=view&current=thermistor_level.jpg

Resistors all the same value, thermistors all the same type and size.
As the self-heating of the thermistors is controlled by the R1 values, the rate of heat dissipation from the thermistor, and hence its apparent temperature is varied by whether it is immersed in fuel or not, so the voltages measured at points A, B and C will change.
As shown, A and B would be same/similar and C would be significantly different. Knowing the position of the thermistors tells you the height of the fuel.

Sorry if I just tried to teach anyone to suck eggs and sorry if this has already been covered in the earlier referenced thread.


Bill

 

[mickeyholland]

Hi Bill, Thank you for your input.

The 1000 / 50cm bit was just an approximation at this stage. As I recall, my calculation of the tank I have worked out that 10mm = 5 liters, so my idea was to have a sensor every 5mm to perhaps achieve 2.5 liter "accuracy".

It is automotive, but the reading would only be interesting at standstill. I do realise that standstill may be on a slope or flat, but this application is for an absolutely square tank, so I wanted to try an idea I had, which was to only be interested in the difference between start/stop level, and not the actual liters. Because the tank is square, and assuming I install the strip of sensors precisely etc. etc., then I was hoping that the difference between 2 sensors would be X liters. And so by determining the start reading and stop reading, I can say the amount of liters removed. It wouldn't mater if I was on a slope or flat, because both start/stop would have that same condition. (Assuming the vehicle does not move from flat to slope mid-operation!).

I would really prefer to use diodes as the temp sensing part, but I cannot seem to get them to heat up (out of oil) to any level that allows me to measure a noticable temperature change. That might just be me though

I also like the idea of using a pcb track, but after calculating the resistance change over a short thin track it seems unlikely that I can do that... but I am still setting up a test rig as we speak!

 

[WGJ]

Sounds interesting stuff -- I used to handle fuel delivery module and fuel level sensor design with a major automotive company and level sensing for a mass-production vehicle was something you could never quite nail down.

Bill

 

[GregLocock]

If you have a simple geometric shape then you should be able to precisely identify the fuel volume at rest using 3 strips of senders.



Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[mickeyholland]

Bill: Sounds like you have plenty of experience with the doubtless infinite params that can play games with this topic then! I cannot imagine how inaccurate those mass-produced milometer computers are we see on modern vehicles! I noticed on mine it can say 100km remaining when I park up, then only 90km when I start it up again! So I guess it makes some averages and assumptions - but anyway if it's around the 10% accuracy mark then it seems that's not a bad achievement!

Greg: May I ask what type of senders those may be and the arrangement? Would that be 3 strips of float senders? I am sorry for being a bit slow this fine Friday, but I just cannot envisage what you meant! Thanks !

 

[GregLocock]

3 points define a plane.

The problem with modern fuel tanks is that the bounding geometry is unknown.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[WGJ]

Sounds like you have a plastic tank.

Bill