Almost Accurrate Clock

[Rob Warnock]

Get a voice-recognition card for your computer, and let it call the
local dial-a-time number... ;-}    No, seriously folks,...

+---------------
| If you ever want a really accurate, cheap frequency standard, try your
| color TV.  For the color to come out right, it has to stay quite close to
| the color burst frequency at the transmitter, so is resyncronized every
| 60 microseconds to keep it within a tenth of a cycle, or so.  The major
| networks maintain theirs against NBS.  (it's 315/88 MHz).
+---------------

True, but there are some things you have to watch out for. (This comes from
memory of an old Scientific American article in the Amateur Experimenter
column, something on frequency standards.) First and foremost, you must be
watching a DIRECT network telecast (that is LIVE from the network, whether
or not the program material itself is live or recorded). "Network" programs
which are taped by local stations and re-broadcast later use the crystal
oscillator in the local station, not the secondary rubidium standard used
to sync up the network. The accuracy required (and maintained) by the local
station is MUCH less. (Instead of 315/88 Mhz  = 3579545.5 Hz, the usual
standards, RS-170A and NTSC, call for 3579545 Hz +/- 10 Hz.)

Next, 3579545 Hz is not a convenient multiple or submultiple of anything
you might want to use for time (such as a 1 or 10 Mhz crystal), so what
you have to measure is the DRIFT of phase of the color burst w.r.t. some
local oscillator. (The Scientific American article went into some extensive
explanations of how to do this.)

(I believe the networks do not "correct" their rubidium clocks, but only
check periodically what the offsets and drifts are from NBS. In fact, the
nets might just use a WWV ==> Heath clock to get the daytime... ;-} )

Then, as a result of the preceding points, you will actually want
to use an oven-controlled crystal oscillator as your local tertiary
(NBS = primary, network = secondary) reference, the sort of free-standing
1 or 10 Mhz units amateur radio operators use for frequency calibration.
(This is good, because such crystals can hold fairly good short-term
stability across "Network Trouble On The Cable".)

Lastly, this still only gives you a frequency (delta-time) reference. You
must obtain an "absolute" reference somehow. ("I know what a second is, but
what is TODAY?") So it's back to the Heath clock, or some other radio, or,
you can carry an atomic (cesium or rubidium) clock to Boulder and get NBS
to calibrate it for you. ;-}

Misc. points:

On black-and-white sets, the color sub-carrier frequency is in the passband
of the normal video, and strongly colored area would therefore have strong
stripes in them, since the color hue/saturation is sent with double-sideband
modulation. But the horizontal frequency DURING COLOR TRANSMISSIONS is not
the usual 15750 Hz, but EXACTLY color-burst * 2 / 455 or ~15734.264... This
means that alternate horizontal lines will have the stripes in alternate
dot positions AND the dots/stripes will not wander over the screen. This
causes strongly colored areas to show up as a sort of halftone stipple
(45-degree "stripes") on a black-and-white set, which is not supposed to
be noticable.  Since there are still 525 / 2 lines in an interlaced frame,
the vertical rate is not 60 Hz but ~59.94... Hz. This can cause slowly
rolling dark bars on black-and-white sets with poor power supplies (the
power grid IS exactly 60.00... Hz ).

(Hmmm... let's move this to net.tv.hardware...)

Rob Warnock

UUCP:	{ihnp4,ucbvax!amd}!fortune!redwood!rpw3
DDD:	(415)572-2607	[ until 10/5/84: (415)369-7437 ]
Envoy:	rob.warnock/kingfisher
USPS:	510 Trinidad Ln, Foster City, CA  94404