I'm making a board to program SMD Atmega328P's. My programming board has the TQFP32 Socket mounted on it. As seen in the attached photo, the traces for the crystal go about an inch or so, and through a small DP3T switch. At first I was concerned about the correct capacitance for the crystal caps, but if it's not perfect, what does that mean? --that the PPM might be off a little? If so, then that makes no difference, as the chip is just in the programmer long enough to have the boot loader installed and tested.
The reason for the DP3T switch is so that I can program a SMD chip for either 8Mhz External, 16Mhz External, or just an internal clock, and then test the newly programmed chip with whichever crystal it's set for by switching the crystal switch. I'm guess there's more capacitance in those traces, but reading online it seems impossible to calculate such things with math only.
Am I correct that if I use the normal 22pF caps it should be fine for the application it's in?
ps. The crystal pins are on the bottom center of the socket and the traces go over to the right edge, to the DP3T switch as seen by the silkscreen.
AFAIK about 12pf parallel to the crystal is needed to make them resonate at the right frequency.
2*22pf in series (seen by the crystal) is 11pf, plus a bit of board/pin capacitance...
If longer tracks/switches are used, I would drop fixed cap values to say 15pf.
I think the values can be off quite a bit before the crystal actually stops oscillating.
Leo..
The shorter you can make the crystal traces the better, place your switch better perhaps? Missing out the
caps could easily lead to a failure to oscillate, or overtone oscillation, you should not omit them for
a bare quartz crystal.
If you use ceramic resonators instead, no extra caps are needed, but the frequencies are a bit off (which
can affect some serial baud rates).
Crystal oscillators feeding just into the XTAL1 pin of the DUT would be another possibility - no caps needed
(other than decoupling for the oscillators), and much longer trace is fine as you are pushing out a low-impedance
signal.
Switching the signals with an analog switch chip is more reliable than a mechanical switch BTW
Analog switches look good. And I can have the switch IC much closer to shorten the traces since I won't have to physically reach the switch with my finger.
Also, regarding the cap on the 3.3v FDTI CHIP: I forgot, I'll need to switch the voltage, too, sometimes. If I burn a SMD chip for 3.3V, 8Mhz crystal and want to test it, then I can switch the voltage from 5v to 3.3v
It looks like I can switch the crystal, and the voltage with this chip:
