I am looking at building a fast and very precise current source that is variable from complete 0mA to 500mA.
I came across the application note CN-0099 from Analog Devices: http://www.analog.com/media/en/reference-design-documentation/reference-designs/CN0099.pdf

Figure 1 shows a circuit which I have been simulating in LTSpice, but I dont understand what the point of the op-amp AD8603 is. If I just remove this opamp, the circuit still works as before.

 

The op amp isolates Rg2 and Rf2 in the AD8276 from R1 and Rload, preserving the common-mode rejection ratio of the AD8276. It also prevents Rg2 and Rf2 from loading R1 and Rload. As the application note explains,

The AD8603 op amp is used in the feedback loop of the circuit and was chosen because of its low bias current (maximum 1 pA) and offset voltage (less than 50 µV). The low bias current makes it possible to interface to a high impedance load without introducing significant offset errors.

With the high current levels you're dealing with (and presumably a correspondingly low Rload), the AD8603 op amp might not make much difference and may not be necessary; nevertheless, the difference is real and should be apparent if you look closely enough at the simulation results.

 

Thanks for your reply
Nevertheless, why don't I need to use one for the upper sense arm.
Since here its only 40k+40k directly to ground.

 

Nevertheless, why don't I need to use one for the upper sense arm.
Since here its only 40k+40k directly to ground.

Ummm... you've answered your own question: you don't need an op amp to drive the Rg1 connection because it IS connected directly to ground, with no extra impedance in series with it to degrade the common-mode rejection.