I prefer load cells for precise test equipment and differential force for big industrial things.
Load cells are usually more accurate and have faster response. They are easier to set up because there is just one analog input. Load cells measure the true applied force. The down side is that they are usually mounted out by the tooling or where the work is being done and often get damaged. Load cells are usually pretty safe in a lab environment. I sometime joke about load cells having a half life as short as their time constants when in nasty environments.
Pressure sensors should be mounted at either end of the cylinder and that is usually far enough away where they don't get damage. The pressure sensors on either side of the piston should have the same response rate. The main problem with differential force is that the applied force can't be measured down below the seal and piston friction. The force that is reported is
NetForce=PressureBE*AreaBE-PressureRE*AreaRE. Where RE is the rod end and BE is the blank end. The problem with that formula is that there should also be a seal and piston friction term. When dealing with a press the differential force works just fine because the rod and piston seal friction force is very small relative to the many tons that are being applied.
Another advantage of differential force is that it can be used for damping and a second derivative for the PID but that is rarely a requirement press control because presses are usually very stiff. A load cell doesn't provide any information about accelerating, just applied force when contact is made. If acceleration info is required then an accelerometer is required and that makes two inputs again.
In both cases the response should be fast. Motion controllers can close the loop at 1 millisecond or faster so the force feedback should be at least that fast. I have customers that didn't listen and used pressure sensors or load cell amplifiers that had response times in the 100 millisecond range. They wasted a lot of their time and our time trying to make it work. The time constant for a force feed back devise should be about 200 microsecond. In 5 time constants or 1 millisecond the analog value will be with 1% of the true value. 200 microseconds is 5 KHz. If the pressure sensors are mounted remotely in a manifold some place because the hydraulic guys are lazy then there are additional delays. The speed of sound, the pressure or shock wave, travels about 4 ft/millisecond or 1.2 m/s. Where the pressure sensors are mounted makes a big difference.
James said:
you cannot correctly measure force applied using pressure transducers with this type of system.
I was long winded so you beat me. However you are wrong about being able to measure force using two pressure transducers but you are right about the piston and rod seal friction. I am pretty sure agarb is talking about presses judging by past post and history so seal friction is insignificant to the total pressures. agarb doesn't need to worry about friction changing as a function of position because presses do the forming at the same position each time.
Also, the debug tools are pretty good now.