atmegacapacitorcrystaloscillator
I am reading this guide and it states that, in order to use an external 16 MHz crystal for an ATmega328, I need two 22 pF capacitors.
I found this bundle of two 20 pF capacitors and the crystal (sold specifically for this purpose).
I'm wondering if there's any difference between using the 22 pF and the 20 pF capacitors provided with the bundle.
I am looking pretty silly now... the footprint (I created, unfortunately) of the crystal was incorrectly pinned, and so I've rotated the crystal 90 degrees and put the 39pF caps back on and its now correctly oscillating. I should have checked this before bothering you guys!
Checked with 10X probe.
Sorry for wasting anyones time!
If you want to use a DCM to generate the necessary frequencies inside the FPGA, then it is more dependent on the divider ratio in the DCM than the crystal itself - the DCM is going to generate more phase noise than a decent crystal oscillator. Also, above a certain point, you need to use LVDS or LVPECL or similar to connect the oscillator to the FPGA with controlled impedance traces, otherwise you can get reflections. For example, for 10G ethernet, you need a 156.25 MHz crystal, a clock buffer, and controlled impedance differential traces to all of the transciever clock input locations. A 40 or 50 MHz crystal will cause fewer issues if the phase noise is not a major issue - it generally isn't unless you're multiplying it up to a very high frequency, like 10.3125 GHz for 10G ethernet. The PLL phase noise is generally dependent on the lowest frequency that you have to divide down to, so you may get better performance if you select a crystal size that allows smaller PLL coefficients - 50 MHz just needs to be multiplied by 3 to get 150, 40 would need to be divided by 4 and then multiplied by 15. 283 MHz is a fun one, though - you may need to do some screwing around in Excel to get that one figured out. It may be a good idea to pick an oscillator that gives you a small rational fraction for both of the frequencies you need. Alternatively, it might be a good idea to use two different oscillators.
Generally all of the boards I have worked on have used 50 MHz or 100 MHz oscillators for the core and 156.25 MHz oscillators with clock management circuitry for the transceivers, and then we have just ignored the 50 MHz oscillator and connected everything to one of the transceiver clocks.
Best Answer
Pedantically, the difference is 2 pF.
Practically, these kind of crystals (parallel resonant) are "cut" for a specific series capacitance. In the 16 MHz range, typical series capacitances are 18 pF to 30 pF or thereabouts.
In this case, you would hope that the crystal is cut for 20 pF, and hence they are providing you with 20 pF caps. In actuality, due to stray capacitance from the board, the caps should actually be a little less than the value that the crystal is cut for.
However, for this frequency of crystal and level of performance (it's an Arduino, frequency accuracy and stability is not terribly crucial), 20 vs 22 pF is no problem.
Finally, remember that if you can deal with really horrible frequency accuracy (10% tolerance IIRC) and a slower clock, the ATmega328 has an internal 8 MHz oscillator, which eliminates the need for the crystal and its associated caps. For minimal Arduino-like clones, it's a great way to minimize component count. However, this is only feasible when you don't need any timing critical functionality, such as the UART.
Edit: You're in luck! The crystal in the Adafruit bundle has a datasheet. Per the datasheet, the nominal load capacitance is 18 pF. Now, 20 pF will likely work fine, but it's not how you're "supposed to do it". Typically, you would assume 3 pF or so for stray capacitance (if you're working on a breadboard, it's probably much higher), and then size the capacitors as \$C = C_L - C_S\$ where \$C_S\$ is the stray capacitance. Ideally, you'd be using maybe 15 pF caps, but again, it's not too crucial in this situation and 20 pF is fine.