sunnycoastgreg: i was chasing accuracy in fuel tank measurement (I have 2 large tanks that are coupled together) with only one sender... and a few % is a LOT of litres in my case :) however i quickly realised that the resistive sender unit appears to have quantisation (it doesnt appear to be a continuous resistance?) .. so given all the errors (inc those mentioned by timo) and also the effect of boat movement etc ... i realised that a 20 step readout (ie showing fuel level in 5% increments was the best i could expect ... and realistically the overall accuracy is probably worse than 5%)
remember that the additional voltage divider you add (is not across the 650ohm ... its across the voltage output from the sender unit (the lower leg of the voltage divider formed by sender unit( 0-180)
It depends on what type of fuel level sens it is. The "old good" once where linear 100% resistive sensors. The newer type today is some kind of a pipe with metal or magnetic sensitive switches in series inside. In my case, I have 8 switching points where resistance is added or subtracted depending on position.
With 650 ohms the hole circuit is meant. If I go out from the 180 ohms of the sensor now the additional voltage divider in parallel do not affect the shown value of the gauge at all (180 ohms sensor to 33.1 KOhms voltage divider). If I calculated right this is an error in microAmp. Sensor and gauge together while the sensor have in 180-ohm max. This is values from the datasheet of the gauges/fabricant (Faria Beede).
The movement of the boat has to be compensated by smoothing the measurements. But there are also other sensors for fuel available, "radar" based once. They are very expensive but deliver a more linear result. But I do not think this is worth the money.
I guess using the digital solution now is much more accurate than the gauges ever have been.