Btw: for the Renkforce RF1000 3D printer, there is a 3D model available which includes the load cell construction. It can be downloaded from the German-language forum rf1000.de, here is the direct link (download button only appears after registration+login):
https://www.rf1000.de/dl_ext/?view=detail&df_id=32
You can also find a FreeCAD version of the model with my personal modifications here (no login required):
I can imagine, the design can be miniaturised a bit by using smaller load cells. In the original it will probably not fit for most printers…
It should also be mentioned that I see two fundamental distinct possibilities to use load cells in the toolhead. Either the load cells can measure any force applied to the hotend, including the force applied by the extruder motor, or they cannot “see” the force from the extruder motor. This has the following consequences (both positive and negative):
- With measuring the extruder motor forces one can do interesting things besides probing the print bed. E.g. one can measure the force while printing the first layer, which might increase if the hotend elongates for thermal reasons, to compensate the first layer height (this has been successfully implemented at the RF1000).
- If the extruder force is applied to the load cells, those must be stiff enough so that the force won’t press down the hotend too much which would reduce the layer height depending on the current extrusion force.
The RF1000 design can measure the extrusion force and is stiff enough so that it usually does not create problems, but we have seen such effects already if the required extrusion force gets too big (bad hotend, sticky material, high printing speeds - usually a combination of at least 2 of these). As a conclusion I would recommend to make the design either at least as stiff as done at the RF1000 (with 2 load cells rated for 5kg, TAL220B type) or such that the extrusion force is not seen by the load cells. That can be achieved by either putting the motor also onto the load cells for direct drive printers. For Bowden printers the force won’t be applied to the load cells without artificially separating the sleeve from the hotend somehow.
Personally, I don’t see measuring the extrusion force as a killer feature, but others in the RF1000 community will disagree with this statement 
I am even thinking of changing the construction of my printer such that the motor sits also on the load cells, just to allow easier disassembly/reassembly of the entire toolhead. I am currently not using any feature which makes use of the extrusion force measurement (but I am frequently removing the toolhead, since the printer doubles as a small CNC mill).