Oh and also by accuracy I mean angle resolution.
I don't believe hobby-grade servos are really that precise, not even digital ones, because they're not designed to be that precise! Hobby servos are supposed to be used in RC-Whatever (cars, planes, helicopters, boats etc) and they're designed to work with an remote control. All the hobby grade remote controls that I've ever seen have trimmers to fine tune servo motion. The smart remote controls that I have seen don't have actual trimmers but a digital way of doing the same thing. What I'm saying is that it would be stupid for servo manufacturers to design really precise servos knowing they'll be used with remote controls that have "trimmers" to tune the "center point" and the gain when moving left and right
In such situations Economics always wins.
Make sure you get a good estimate of the number of distinct positions your servo can stop at. Just because you can have an infinite number of pulse lengths between 1 ms and 2 ms it doesn't mean the servo can stop in an infinite number of positions! For one thing the digital servos are probably going to use pretty ordinary MCU to implement the PID; This MCU will read the POT position using it's ADC. I'm pretty sure the maximum resolution offered by the MCU's ADC is 10 bit (1024 positions) and more likely 8 bit (256 positions). Next you have a problem with the MCU in the servo receiving the servo PWM: there has to be a limit in resolution, can't guess an number right now but it has to have a limit. Next you'll have an other problem with your MCU making up the servo PWM! Yet an other problem with resolution, you can't generate an PWM with infinite variations in the pulse length. If you do the math you have 3 error sources in the same equation: error when the MCU in the servo reads the POT, error when the MCU in the servo reads the servo PWM, error when your MCU sends the servo PWM. Things are not looking good. I think Webbot
provided an estimate of maybe 90 distinct position for an servo, I don't know where he got that but it seems like an good number, maybe a tad optimistic. And oh yes, just because the servo can stop at 90 distinct positions don't expect those positions to be at precise angles! (don't expect to have exactly 0.5 degrees between each stop). Again, it's a rounding problem and it will cause variation.
My purpose is not to steer you away from hobby servos, they're probably the best solution if you're on a budget (and who isn't)? I just want you to have realistic expectations as to how precise those servos are. To get the same info from a different angle take a look on youtube. I tyred finding some precision robots designed using hobby grade stuff:
This is the milling hexapod, really nice stuff:
The face milled is grate! Yet you can easily see the errors in positioning.
This is the Scorbot robotic arm writing it's name:
This robotic arm doesn't use hobby grade servos, it uses something with optical encoding on each axis. Yet it doesn't manage to write perfectly accurate lines! They're grate lines, but not perfect;