In attempting to cross the Atlantic with a zero pressure the key is to travel fast by sticking in the jetstream but also to maintain your altitude at night by dropping ballast. There have been a couple of previous attempts ata trans-atlantic balloon and have use relatively crude methods of controlling how much ballast dropped. The 2 components are a suitable valve or pump and a method of sensing how much ballast you have left. Creating the ballast sensor has been relatively hard - the main problem has been finding a simple yet reliable method that will work in an extreme environments.
At present we are using the IR LED and Photodiode method though I'll document our other attempts to measure the volume of ballast left.
We first looked at capacitance, capacitance is potentially a highly accurate method of measuring the volume of liquid in a tank. It requires a very sensitive ADC (we used the AD7746), as the volume in changes so does the capacitance. The problem is that this very sensitive ADC can be overwhelmed by external noise requiring careful shielding and setup. While perhaps the most accurate method of measuring the volume it was really very hard to setup even on the workbench - getting a reliable system that would work at altitude was too difficult.
Another method is to have a float within the tank with a magnet and then a sliding potentiometer on the outside of the tank, this method keeps the electronics away from the liquid and is relatively simple. Sparkfun stock Magnetopot which is linear potentiometer with an internal magnet, using an external magnet you can move this internal magnet up and down. By attaching this to the outside of your tank and having a strong magnet inside as the level changes so does the potentiometer magnet, the change in the resistance therefore relates to the ballast level. This setup is really simple however we ran into one big problem, if for an instance the external magnet move too far away from the internal magnet then the internal magnet would fall down to the bottom of its run. It would therefore be impossible for the sensor to reset itself until the tank was completely drained. This could be fixed by getting the right strength of magnet, if it wasn't strong enough the internal magnet would fall down, if to strong then the float wouldn't move up and down with the level changes.
In the end we moved away from this method due to the issue of not being resetable, there is quite a lot of shaking during a flight which could easily result in a loss of contact and then the sensor would be useless.
Building upon what we had experimented with we then looked at really simple setups, a quick experiment using a photocell at the top of the tank and an LED at the bottom showed that liquid within the tank attenuated the intensity and as you reduced the volume of ballast the light intensity increased proportionally. As photocells are dependent on temperature we instead used a photodiode and an IR LED, the IR LED is embedded into the bottom of the tank and the photodiode at the top, the photodiode is interfaced to an opamp setup to convert current to voltage. By pulsing the LED and comparing the readings for on and off we are able to measure how much IR light is getting through the volume. There is a rough exponential relationship between IR light intensity and volume of ballast. The setup was improved considerably by averaging out 100 rapid readings.