When you select an amplifier IC, whether a basic op amp or a more complex function such as an instrumentation amplifier (in amp), one basic choice is between an uncommitted device with external gain-setting components, or one with internal fixed gains. For the former, you establish gain via one or more external resistors; for the latter, you select the gain via jumper or a digital interface in decimal steps (×1, ×10, ×100), binary steps (×1, ×2, . . . ×128), or even instrumentations steps (×1, ×2, ×5, ×10).
But external variable or internal fixed: which is the “right” choice? As in most engineering decisions, the right one to use depends on the application and the circumstances.
Consider the uncommitted gain-value device. On one side, it lets you pick the specific gain value which maximizes your dynamic range performance, matching the channel span to the A/D converter span. For example, for a sensor with full-scale output of 0.63 V, and an A/D converter with a 2 V input span, setting the channel gain to 3.17 will map the sensor's output to all available bits of the ADC, so you get the resolution and range you paid for.
On the proverbial other hand, there are downsides to having to set the gain yourself via external components. Most obvious is the cost, space, and availability of the precision component(s) needed to set and hold that desired gain value, and that tight temperature coefficients and even matched tempcos are needed.
But more challenging is that the amplifier function can no longer be fully specified by the amplifier vendor, and you have to do some serious error analysis on component tolerances, drifts, and other issues. In contrast, vendors of fixed-gain devices can test, trim, specify, and guarantee the amplifier performance at each gain value.
Finally, and most insidious, is that the amplifier performance is now determined by the vagaries of your layout and parasitics, since the gain-setting components are outside the IC package. This becomes especially challenging at higher frequencies, where the placement of the external devices is no longer a straightforward matter, and issues related to off-package pinouts, PCB traces, passives, and parasitics and strays have significant effect.
As usual in engineering decisions, it's a tradeoff: the flexibility and range-matching you get when you can set the gain value to any value you wish, versus the component cost and layout headaches that such flexibility offers, compared to fixed-value, in-package gain-setting amplifiers. Keeps it interesting, right?
-x x x-