>I keep hammering the "magic" aspect as I have seen over many many
>years that crystal oscillator design is not amenable to simply
>scrupulous application of the data sheets and that many people,
>both amateur and professional, AND manufacturers continue to have
>problems with it. It is important to apply proper design principles
>and follow specifications

This I suspect is where the biggest problem lies. The limitations of a CMOS
inverter as a linear amplifier to be used in an oscillator does not lead to
an ideal set of parameters for this use, but rather a set which work most of
the time. However the other problem that crystal manufacturers have is
dealing with this broad set of parameters for the oscillator, that come
about from die shrinks and process variations, and then endeavouring to make
crystals that will work with any microprocessor oscillator. The problem
would probably go away if the market was large enough that the manufacturer
could make "PIC crystals" and "805x Philips crystals" and "805x Maxim
crystals" ..... etc as separate lines, so they could tailor the
characteristics of the product line to the manufacturer. The fact that next
year they then need to make "PIC Die Shrink Crystals" is another bother. To
me it is magic that the industry does not have more problems with
oscillators and crystals.

>As RF Jim has suggested before today, a characterisation by measurement
>both of crystal parameters and of the oscillator "amplifier" would have
>the ability to design correctly. To be certain of not being caught out
>this would need to be done on a per component basis (!) and even then
>there are liable to be some parameters which are hidden within the IC
>and interact in a pseudo random fashion so as to confound you.

Well not long after I had finished my apprenticeship, one of the guys in the
development lab where I worked did a paper for internal use in the company,
on the design of crystal oscillators. In the paper he did a reasonable
analysis of the S parameters of the particular IC the company was using as a
building brick for oscillator/RF amplifier design, and came up with some
formulae that allowed one to characterize the required crystal, and the
other component values around the circuit that were required for a
particular frequency of operation. It is these sort of parameters that do
not get published by microprocessor manufacturers, I suspect because they
cannot hold the parameters stable enough with CMOS inverters. I also suspect
it is because of this that the early Intel chips used separate oscillator
chips, and serial resonant crystals, with external tank circuits to make
sure things did not take off at one of the harmonics. It would seem they had
a proper multistage linear rf ampler rather then a single stage cmos
inverter in the chip.

..........

>Serious comment: Many people have found that the human finger is a
>superb promoter or defeater of oscillation. As a rule of thumb
>(pun intended) if placing a finger near oscillator components causes
>oscillation to stop or start then you have a problem which should be
>remedied. Actually touching some parts of some oscillator circuits
>may legitimately affect oscillation (or the applicator!).

Shades of an article in Electronics Australia a good few years back now,
under the by-line "Serviceman Who Tells" (you Americans don't know what you
have missed out on for funny stories) where one of the serviceman's friends
who worked in the outback came to service a radiogram. He found an old
shrivelled up something attached to the aerial terminal of the set. A few
enquiries elicited the information that someone found that when they touched
the aerial terminal the radio reception was much better. A bit of lateral
thinking came to the conclusion that since a finger was just a stick of
meat, then attaching a sausage of meat to the aerial terminal should do the
trick. (Note to Americans, this is not a round patty of meat like you call a
sausage, but a long thin tube of similar meat, I think you may call them
sausage fingers or some similar term).

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads