I want to use the the AD8221 amplifier as a conditioner for my load cell. I saw that to reduce noise I can use an "active guard shield driver" between the "REF" of my amplifier and the shield.
If someone has already worked with something similar or have knowledge on this subject it will help me if you can explain me why and how it is working? And if I really need to use it?
Is it not enough to just connect the "ref" to my GND and the same for my shield?
[This is a more theoretical addendum to the answer by @VoltageSpike. ]
Active driven guard is used when the signal source (the sensor) has a high output impedance. The amplifier input has a (very) high impedance, so when the signal source also has a high impedance, then leakages to and from the signal lines can introduce a significant and uncontrolled offset. This offset would be indistinguishable from the sensor signal.
The key feature of the active driven guard is that the voltage of the guard is close to the voltage of the signal which it guards. The leakages are driven by voltage gradients. If a sensitive signals are surrounded by a voltage that's close to the voltage of the signal, then the voltage gradient will be small and the leakage will be small.
Consider an example. GND, Signal-, Signal+, +5V are traces on a PCB. They run next to each-other, in that order. They are evenly spaced 0.25mm [10mil]. Signal- and Signal+ are the differential signal from a high impedance source. Their common mode voltage is at 2.5V, and their differential component is much much smaller. There will be leakages from the signal lines, and the leakages will be driven by voltage gradients. There will be a leakage current from Signal- to GND, and a leakage from +5V to Signal+.
How could we fix this example? We could add a guard around the sensitive signals: GND, Guard, Signal-, Signal+, Guard, +5V . We would drive the guard with a voltage equal to the common mode voltage of the Signal- and Signal+. Now the traces nearest to the sensitive signals are at nearly the same voltage as the signals themselves. Much smaller voltage gradient, much smaller leakages. We would create a circuit which tracks the common voltage (because it can vary), and adjusts the guard voltage.
If the objective is to shield from RF, then its only necessary to connect the shield on one end to the ground of the PCB or chassis ground (and not on the sensor end which would make a ground loop). It's not necessary to drive the shield. Driving shields is usually used for very high impedance measurements, when the sensor current is in the nA range or below as even static fields can affect the measurement.
For most load cells the sensor current out of the cell is much higher than nA's, error from leakage (from the resistance) of the cable would be minimal (an easier thing would be to use high resistance cables from materials like teflon instead of using a driven shield, make sure the cable resistance is higher than 1MegΩ).
This should work in most cases where you use a shielded cable on all four conductors if using a wheatsone load cell.
Source: https://hackaday.com/2015/03/16/instrumentation-amplifiers-and-how-to-measure-miniscule-change/
In any case if you do use a driven shield (if the load cell resistance was higher than 1kΩ you might want to worry about it) then use the digram in figure 16-7 (Figure 23 won't work because both sensor wires need to be shielded and the ground should not be tied to the shield, the ground should return back to the amplifier).
I don't know if this option is available to you but have you looked into using a load cell with builtin signal conditioning? Instead of running mV signals in the high EMC environment you could have 0-10 or ±10V signals with low driven impedance. This, of course depends of the load cell size and other possible mechanical constraints on the design. Also there are small inline conditioner amps that you could use with your existing load cell: https://sps.honeywell.com/us/en/products/sensing-and-iot/test-and-measurement-products/instrumentation/inline-amplifiers IMHO running low level strain measurements in high EMC (RF is the worst) is tough since shields can pick up all sorts of signals that are out of the bandwidth of your IA. Driven shields should be a last resort when you have run out of all other options.
You can even get load cells with USB or CAN bus output: https://aesensors.nl/assets/uploads/2016/02/dcell_product_sheet.pdf The advantage of these all digital solutions is the calibration is inside the load cells instead of requiring a separate calibration of the sensor and DAS. There is also a niche DAS for crash testing (DTI) that embeds the convertor electronics inside the load cells and uses a serial data stream to multiplex several high speed channels.
You don't need to be particularly concerned about the excitation but note that the leads will "drop" the excitation voltage slightly (10ft shouldn't be an issue). Also on the remote amp method the excitation is also regulated at the sensor so it help niose rejection. Since this is only 10ft you might be fine with a simple LC filter on the signal inputs and Vexc from the DAS. BTW in wind tunnels you might find the leads running 250ft from the force balance to the signal conditioner. Depending on the bandwidth of the signal you are trying to measure some input decoupling might be fine. Instead of putting solutions before requirements could you elaborate on the application: Are you constrained to use a particular load cell? What load range are you measuring? Is the application battery powered or power limited? What is the EMC frequency makeup? What is the signal you are trying to measure (static load vs dynamic)? Is this a cost sensitive application so you need to design as much as possible? How accurate do you need your measurements? Only one channel?
