We have to distinguish between the typical deviation from the "nominal value" of a oscillator (due to production issues) and the actual behavior in your environment, the latter is heavily influenced by temperature, PCB layout, choice of capacitors, and, and.....
Both effects are rarely separated in non-expert dicussions. To get the best out of a crystal oscillator you definitely need a temperature compensation (often done by a "heating" resistor to keep it warm) and a trim capacitor to tune to the requested frequency....
OCXO - Oven-Controlled Crystal Oscillator, as it sounds, it has a temperature controlled chamber with the crystal in. pretty much as accurate as you can get without going nuclear.
TCXO - Temperature compensated crystal oscillator, uses an analogue temperature compensation circuit.
Yes and no. Any crystal is probably more accurate than a resonator. Getting the best possible accuracy even with a crystal requires more attention to details than most people apply (for example, it's common to just plunk in the usual 18pf or 22pf caps, while in reality you're supposed to carefully match the caps to the exact crystal you're using, the characteristics of the chip you're using, and any inherent capacitance present in the PCB you've designed.) Then you can start talking about temperature dependencies, though I think it's a bit overrated for the average "room temperature" environment that applies to most arduino projects.
Yes and no. Any crystal is probably more accurate than a resonator. Getting the best possible accuracy even with a crystal requires more attention to details than most people apply (for example, it's common to just plunk in the usual 18pf or 22pf caps, while in reality you're supposed to carefully match the caps to the exact crystal you're using, the characteristics of the chip you're using, and any inherent capacitance present in the PCB you've designed.) Then you can start talking about temperature dependencies, though I think it's a bit overrated for the average "room temperature" environment that applies to most arduino projects.
I think I already said that some hours ago, but it is most like worth saying that a lot of times
However, I understand it uses a resonator rather than a crystal, and that may be a problem for what I will use it for.
As you can read so far oscillator accuracy and variation can be quite a complex subject and matching a solution to your requirement can range from easy and cheap to quite expensive.
Have you done any math to find out what your actual timing accuracy needs are? Over kill can be a waste if the problem exists only in your mind.
(so how come no one makes crystals with built-in caps, anyway. Sure, you're supposed to tune the caps to the exact blah, blah. But it looks like these days the oscillators in most micros are SO similar that a crystals with (some value) of built-in caps would still be more accurate than resonators...
so how come no one makes crystals with built-in caps, anyway. Sure, you're supposed to tune the caps to the exact blah, blah.
they do, they are called crystal oscillators, and come in four pin rectangular metal "cans" like this
(not mine, random image I ripped from here NewClassD D-Clock)
I have used 40mhz ones with PICs, since it saves an IO pin.
Packed Oscillators are fine and recommandable for many reasons, and are only a little bit above 1 USD. They also contain the DRIVER circuitry, which is a major problem in low cost MC chip design. Driving a quartz needs a lot of current! Against the believe of most people, quartzes are more coils than capacitors This is also the reason why additional capacitors are needed to form a good oscillator circuit.
So supplying the current and the capacity from inside a chip is generally a compromise and rarely as good as an external solution. (BTW the Parallax Propeller is quit good there..)
Professional designers have no problems with external SMD capacitors, and the quartz industry does not work for us hobbyists -
So final answer is that they're using a Murata CST 3-pin resonator which has a 0.3% error. I know these well, and they're quite decent, so I'll take my chances for now, and determine later if to (somehow) swap to a crystal w/caps, or to use an external oscillator.
Actually 0.3% is not insignificant when bit-banging an async protocol, but I may get away with it. Remember that the other side of the bit stream will have timing errors too. The Murata parts are fairly decent, so I'll determine later if I can get away with it or not. (Knowing me, I'll probably do the swap though).
A statement as "0.3%" is made for a professional designer's theoretical proof-of-work, assuming that he or she works according to strict standards in PC design and has full understanding and control of environmental influences.
BTW This even more applies to capacitors, their specific type and presumed accuracy... I have seen things you wouldn't believe..
This is also the reason why additional capacitors are needed to form a good oscillator circuit.