[sdiy] crystal oscillator oscillator seeming to run at wrongfrequency

[Magnus Danielson]

Tim Stinchcombe wrote:
>> sorry if this question is off topic.
>> I've been looking at a video generator board that has two crystal 
>> oscillators on it, one is at 14.318 Mhz and 17.734 Mhz.
>> ( these are for the ntsc and pal standards ).
>> both use a 74lcx04 ic with 1Mohm feedback resistor and 22pf 
>> caps. but the frequency of the 14.318 crystal measures 14.322 
>> , the other is 
>> much closer to the 14.318.
> 
> Assuming: here you really mean the PAL xtal is closer to its value of
> 17.734MHz than the NTSC xtal is to its 14meg value; that both oscillators
> *are* running and giving something close'ish to where they should be; that
> the circuit design is known to work; and ...
> 
>> the chip the clocks drive is the  Epson S1D13806
> 
> ...the technical manual for this being an impressive 730 pages (= whoa,
> complicated!), I would venture a guess and say that almost certainly your
> problem lies elsewhere!
> 
>> I just wondered if it is possible that this might be a design 
>> fault with 
>> the circuit ?
> 
> However... the Pierce oscillator circuits you mention (simple inverter+large
> feedback R+two caps), are pretty cheap and cheerful oscillators that will
> get you close to the nominal xtal frequency, but you'd probably need
> something a bit more involved to get it *exact*. Hence on the basis that the
> circuit is designed to work with these, I again would be willing to bet that
> as long as they are running and appear close to what they should be (as
> measured say on a reasonably capable scope), they are unlikely to be the
> source of your problem. Though having said that, I note the evaluation board
> in the manual uses a considerably more robust clock 'regime': a little
> 'clock module' for the 17meg, and a more complicated programmable clock
> generator/synthesizer for the 14megs, so it is possible the 'slopiness' in
> the frequencies of the Pierce set-ups is just not good enough...

The CMOS inverter based Pierce oscillator is not totally rubish, but it 
could do with tune-ups. A capacitive trimmer would allow tuning. 
Controlling the crystal current is another classic. However, the CMOS 
inverter contribute noticeable 1/f noise appart from the white noise 
creating noticeable 1/f³ noise (unavoidable, but level is important) due 
to the Leeson effect.
The Q of the crystal is another key aspect. Oven stabilization can make 
minor wonders for such a design.

Anyway, the CMOS inverter design is far from perfect, but adding a 
trimmer allows for ability to trim it at least. Using varicaps allows 
for VCXO operation and thus a multi-turn (20 turn) trimmer can be used 
instead.

Ovenization helps to keep purities of the surface and reduces 
temperature dependent frequency as well as stabilizes drift.

The 14,31818 MHz and 17.xxx MHz has nice relationships to 18 MHz and 27 
MHz so a stable common reference for those could be used if PLLed, but 
that would probably be shooting over the heads here.

Cheers,
Magnus