yea that thought crossed my mind
You also have to compensate for long term drift. I wouldn't use a bubbler in fermenting beer, unless you can use an inert gas like nitrogen or co2.
I assume drift will be predictable and can be compensated for.
beer bug does use this method.
Life got busy and this got pushed to the back burner.
Hear is what I have learned.
I could not get the differential pressure method to work predictably. Every membrane I tried to cover the end of the tube was too stiff and using the bubble method would require a constant source of CO2 (even it is valved to turn on just before a reading id taken) which I want to avoid.
So I worked on having a weight in the beer and measuring the change.via a load cell. This works!
The weight needs to have a density of roughly 1.3 , One of the struggles is that CO2 bubbles adhere to the sides of the weight changing its density. Smoother is better. It also must be large enough to offer maximum weight change but the bigger it is the more surface area for bubbles to adhere . I am not sure where the sweet spot is.
The results need to be compensated for time, temperature of the load cell and temperature of the beer/wort.
Load cell drift is not linear so you can’t just use a linear formula to calculate it. It drifts alot at first then levels off. It might be good enough to estimate the drift in the first few hours the calculate it linearly after that. This needs to be studied.
Temperature corrections are pretty straight forward and probably linear . I haven’t studied this yet.
All in all if you are looking for accurate SG figures to 4 significant figures … its not going to happen. If you are looking for relative GS trends with ballpark accuracy I think this method will work.