Load cell based fuel gauge

Discussion in 'Burners and their construction' started by Mark's castings, Jun 30, 2020.

  1. Having the fuel needle valve some distance away from the fuel tank might make it a bit harder to implement, I guess with a mirror, a stopwatch, a small pair of binoculars on a tripod and some speed practice on the calculator you could approximate the same function :rolleyes:.
  2. OMM

    OMM Silver

    I think if you can sample time and weight twice a minute or more... and a good display. I think it will work for your needs and many other people.

    I personally would love if this was designed into a fishing hook weight design.
  3. There's no reason why the current load cells can't be replaced with a single bridge load cell configured for hanging up. The HX711 interface can sample up to 80 times a second and down to 0.012 grams (200Kg max) for this particular configuration so it may be possible to average several reading to reduce regular vibration induced noise to a certain extent and still update every second or so. The integration time could also be an option during configuration. As Kelly has pointed out vibration could be an issue although I'm hoping resting the unit on sandy soil some distance away will work...worst case I can put a bike tube under it to isolate it from ground vibrations
  4. Jason

    Jason Gold

    Hey Mark, just thinking about your contraption here... I spend hours in the air and can estimate landing fuel weight within 10 gallons or 68lbs per wing. How can I be so accurate after consuming almost 3,000 lbs of fuel? It's because of the length of time. It eliminates error by the average. I fear unless your scale is SUPER accurate and weight samples are taken several times per second with the oddball readings tossed out, you'll need to extend your time to get something truly accurate. No doubt, what you have come up with will work, but it might take 15mins of uninterrupted burning to eliminate errors. And with that, a postal scale would be just as good. Lest we forget taking into account temperature calculations. That will probably have the same affect as vibrations at a minimum.

    I will now return to watching my dumb 50year old entertainment. (Green Acres) Lisa is SO DAMN DUMB!
  5. Hi Jason, there's no reason I can't calculate both rough and ready instantaneous as well as an accurate long term amount based on starting weight minus current weight, converted to litres or gallons...at least this way it keeps me busy and relatively out of trouble.
    Jason likes this.
  6. joe yard

    joe yard Silver

    Hi Mark
    Although what you are wanting to do can be done. You are making some mistakes in the amount of divisions that you can get from a load cell.
    The ½ bridge is inferior to the full bridge cell and only has the advantage of cost.
    The full bridge cell can only be used to 10,000 divisions to be considered legal for trade in the US. You can stretch the divisions to 100,000 under lab or close conditions, 50.000 is a stretch under most conditions. This is with an indicator that is well designed for the task.
    Load cell output is rated in MV .001 volt times voltage applied. Standard cells are 2MV/V or 3MV/V. At capacity. This means that a 2 MV/V cell with 10V excitation will output a signal of .020V at full capacity.
    By your design you are asking 16,666,667 divisions. I have not been on the forum for a while. Too many other irons in the fire.
  7. Hi Joe, that's certainly something to chew on, I'll be using four half bridges which I can connect as either two full bridges or two bridges in series (I think...). I'm hoping this dedicated load cell amplifier/ADC chip can handle all the heavy lifting: it's 24 bit resolution and a gain of up to 128 but I don't need nearly that much resolution so I was planning to lose four or five least significant bits right from the start.

  8. Al2O3

    Al2O3 Administrator Staff Member

    I think the value in the approach would be real time fuel consumption for the purpose of tuning. Average fuel consumption over the duration of the burn is much less demanding and easily measured. As mark mentioned it's an alternative to inline flow meters. The challenge is accuracy and resolution required to approximate real time measurement given the initial mass of the system.

    If you're not careful you may accidently design a seismograph. You might be surprised how much things move around. If you are in a rural area, it may not be much of a problem. In an urban or industrial area the noise is terrible. You can try to isolate it but sometimes harmonics can be a compounding problem. I wouldn't be surprised if you can detect your own footsteps if you can resolve a fraction of a gram in a 200kg mass.

    Mark's castings likes this.
  9. Managed to get the support hardware built for the ESP32 today: a small 3.3V buck regulator as I'm now going with 4 NiMH cells that give 5.5V when fully charged. There's also a USB programming interface board which was a real pain to get working and even after several hours only runs in manual mode...that's enough to program it reliably and once it's finished I can unplug it. The bluetooth library is hard to find info about but I was able to get it to receive text from the smartphone terminal and echo it back intact with very simple code compared to mucking around with running a WIFI server as the alternative. Once the load cells are fitted and wired I'll just have to write software to get weight readings...how hard can it be? o_O.

    ESP32 bluetooth 3.jpg

    ESP32 bluetooth 2.jpg

    ESP32 bluetooth 1.jpg
  10. The weight scale is working after getting the HX711 software library to work with the ESP32 by lowering the clock speed to 80MHz and changing pin assignments. So far it can "Tare" the load cell assembly, "Calibrate" with a known weight and then permanently save to the non volatile memory. I was able to configure the four half bridges into a single full bridge arrangement which makes it easier to implement than two full bridges. I'm having problems with a drifting zero and with about a one gram noise variation in weight when constantly sampled about two times a second. I'll add a power supply noise filter right at the load cell amplifier and bed the load cells and bases with epoxy resin, the top aluminium plate is full of dimples where it contacts the top of the load cells so that won't help things much. I have some stainless laser cut rectangles I plan to screw and epoxy bed to ensure all four load cells are in firm contact at once. Even with the software library performing a rolling average of the samples there's a +-0.5 gram noise variation which will have to be improved a bit to work well.

    Also the ESP32 microprocessor board lash up prototype has got to be the most suspicious looking bit of electronics I have built for a long time...I would hate to have it on the passenger seat and be pulled up by the cops :eek::cool:. It has a range of 15 yards with the Bluetooth data link.

    weight scale 2.JPG weight scale half bridge 4.JPG weight scale 5.JPG weight scale 6.JPG weight scale 7.JPG
    Last edited: Jul 7, 2020
    OMM and Jason like this.
  11. Jason

    Jason Gold

    Look at the big brain on Mark! Nice going. Within a GRAM! Now THAT should be good enough for the girls YOU go out with! :p
    Mark's castings likes this.
  12. It's helps keep the boredom at bay in these trying times Jason...how do you know about the girls I go out with?...that was just a house call from my oral hygienist...wait! did I leave that bedroom webcam turned on again?.
    Anyway, if I use 20+ litres per hour, that translates to 333.33 millilitres per minute, 5.5555 mls per second or 4.4444 grams of fuel per second. So if I update every two seconds or so and can cut the noise, I should have a suitable fuel flow meter.
  13. joe yard

    joe yard Silver

    Hi Mark
    I have not posted in quite some time now but your post have drawn me back in.
    It is very commendable to see the progress you have made on this project.
    I will offer advice on this one only because I am quite well trained and was always highly regarded in the weighing field. I worked in electronic scales for the last 25 years of my working life. Factory trained master tech at the Toledo facility in Worthington Ohio.
    You have begun on an endeavor that although looks promising and most certainly can yield good to great results. You are not going to get the results that you think you will.
    With proper weighing equipment it is not hard to get a tank weight down to 1% “100 divisions”. It is not east to get.1% “1000 division” although this is done every day in industry. It takes proper equipment and great attention to detail.
    Over 1000 divisions on a tank containing fluid that will have a line attached to for the purpose of drawing from this vessel will cause complications.
    The suspension under the load cells ai also a source of problems as side torque will effect the cell internal stress causing inaccuracy. This is not a problem when looking at 100 divisions. It starts to show at 1000 divisions if the scale is not designed and leveled properly. At 10,000 divisions it can become a major problem.
    The load cell excitation power supply must be rock stable!!! If it drifts. The displayed weight drifts.
    Temperature will cause problems above 1000 divisions. The HX711 module has no provision for Temperature sensing. I have no experience with this module but it is interesting an I will research it further.
    For the type of cell and the application .1 KG would be realistic
  14. Hi Joe, any advice from an expert in the field is greatly appreciated, from what you say I would have to sample over a minute or so to get a relatively accurate result (which I could also add). This project was always going to be a proof of concept to see how close I could get using cheap bathroom scale sensors and electronics ($7). I'm experiencing up to +150 grams drift off zero after a heavy weight and 1 gram of noise after averaging 10 samples: more or less what you predicted.... I'm wondering if these load cells are the factory rejects!. I'll make some more changes to the power supply filtering and load cell mounting: epoxy bedding and see if that helps with the noise and zeroing. Worst case is I end up with a steampunk bathroom scale I can measure scrap metal on, best case is a roughly calibrated fuel flow meter that is consistent for an hour or two at a time.
  15. joe yard

    joe yard Silver

    Hi Mark
    The zero drift will be the hardest part to control. Often when a cell goes bad. It will drift in one direction until it is out of range. On the cell quality? I an not say. The load cells I have installed over the years with little exception were full bridge. A 50 KG cell would cost the company $100+
    for an S type cell. Shear beam and canister types of cells could go over $500.
    The work you are doing will give you excellent results when averaged over time. I would expect over an averaging time of several minutes you will be within >1% of error.
    You might have mentioned but it is late and I am old. What sample rate are you averaging over time? Does the board you are using have filtering capabilities and if so what type filters?
    There are 2 school of thought on a continuous flow indicator. Old school, slow sample rate and heavy capaciter filtering on the analog side. 2 a fast sample rate digitally processed where the readings must fall within a determined range. All other samples are ignored and then averaged.
    Noise filtration and resinate vibration harmonic electrical noise is handled differently.
    For now there is a few things that you can do to help with noise.
    1 use shielded cable to the load cells.
    2 use ferrite donuts on each load cell at the indicator and on the DC power.
    3 ground all load cells at the indicator to ensure equal potential between the cells and indicator.
    This ground is independent of any further grounding. It is not a safety ground and should be an independent conductor from any other grounds or neutrals.
    I hope this helps. I made a lot of money and lost a lot of hair running down noise.
  16. Thanks Joe, that's a big help. So far I fitted bypass capacitors to the load cell board and let it warm up for half an hour with no improvement after zeroing with a test weight. The load cell integrated circuit has a selectable fixed 10 Hz and 80 Hz sample rates. I'm not even sure how the software library is handling things apart from it taking a rolling average of ten samples and displaying it about second apart. The software library can also reject the highest and lowest readings and average the rest. I suspect from how slow it is to respond at the moment that it just takes ten of the 10 Hz readings at about a second time interval and does the maths. It takes several seconds to respond to settle on a new weight at the moment. The IC can take a separate analogue load cell power supply pin (up to 5.7V), so I'm going to feed that with a 5.5V battery made from four more NiMh cells as an experiment. I have small ferrite suppressor beads I can fit to the load cells too. I'll follow your advice on grounding and I should have a second set of those half load cells arriving soon.

Share This Page