Not much I can share about one versus two load cells. I have tried initially 4 load cells in parallel and have quickly reached the conclusion that the tweaking (balance between the four) was way too complicated. I have then switched to Bosche H40 single load cell in a double H frame. To make the frame I found a very convenient site (motedis.com) where you can find all what you need (I can share my list if needed).
BTW, you can see on the right side a small peristaltic pump that I am using to feed the hives with sirup when need be (my set-up can control 6 hives, which is the number of hives I own and therefore this avoids me to go to the apiary too often to feed the bees)
Concerning the H40 load cell, here is what I have learnt: according to Bosche, this kind of load cells is supposed to be “temperature compensated”. In reality it means that the difference between HX711 reading when a load is applied, and HX711 reading when no load is applied (this is often called “Offset”) is independent from temperature. But the Offset IS temperature dependent. In a normal scale, this is not an issue because you first turn the scale on, take a reading before the load is applied (Offset) and then apply the load and take a second reading. In our case, it’s not possible because the load is always applied.
The good news is: Offset seems to be a linear function of the temperature.
The bad news is: the slope of the linear function and the intercept are changing completely upon the load cell unit you are using. You need therefore to calibrate each one separately. If you are lucky you can even find a load cell where the linear function is parallel to X axis (Offset doesn’t change with T).
In order to calibrate, I take a reading of the Offset at 3 different different temperatures because even if I know I am expecting a straight line, this allows me to ensure there is no experiment error. This is actually not as simple as it seems because temperature has to be stable for a long time for the measure to be correct (big inertia). I am using a program that waits until the reading stabilises during a sufficiently long time.
In normal operation, given the temperature inertia of the load cell, the temperature is never perfectly compensated. It may be possible to model the inertia and compensate better but I don’t believe this is really needed.
Hope the helps.