[time-nuts] Standards for units
cfmd at bredband.net
Tue Apr 3 05:22:58 EDT 2007
From: David Forbes <dforbes at dakotacom.net>
Subject: Re: [time-nuts] Standards for units
Date: Mon, 2 Apr 2007 23:22:24 -0700
> At 9:57 PM -0700 4/2/07, Tom Van Baak wrote:
> >> >1/meter/39.37 inches = 0.025400051 meters/inch ...
> >> The interesting thing is that this result is off by 2.000004000 ppm.
> >> That's a rather precise error!
> >> The calculator I used is here:
> >> http://homepage.ntlworld.com/r.howitt/Calculator.htm
> >There are two ways to compute relative error, depending on
> >which one you call the reference.
> >For extra credit, note your number is actually this cool-looking
> >fraction* whose pattern you will immediately recognize.
> >* calculated with: echo "80k .002 999.998 / 10 6 ^ * p" | dc
> Yes, I had a feeling it was that case, but I didn't bother to chase it too far.
> This precise sort of error reminds me of the error that RCA
> introduced into the frame rate for NTSC color when they discovered
> that the chosen color burst frequency was right on top of a diathermy
> machine or something silly like that. They bumped the frame rate down
> by exactly 0.1%, creating the 29.97 frames/second that
> cinematographers have grown to love to hate.
We *STILL* suffer from this. With CCIR-601 we where doing good, everything
about PAL and NTSC joins in with 27 MHz as a common frequency, but when they
went for HD-SDI they broke things bigtime as they chose to have 11/2 times
higher rate than SD-SDI. So now we have to support 1.485 Gb/s AND 1.485 / 1.001
Gb/s. Sigh! I would have settled for 5 or 6 times instead, that 11 is what
breaks the magic. This 1.001 error is a major headache since you can't use
the same crystal and pull it either way. If you do many channels it is
infeasable to have multiple crystals and choose between them (we have done that
once already) and the resynthesis chips doesn't have godo enought phase noise
to be useable when they do the 1.001 term.
It also shows up in sample-rates, so we see things like 44056 Hz samplings etc.
It is still broken and its a PINA to get fixed. (Sorry for the strong wording)
> There's a volume of the RCA Review that came out in 1953 that is all
> about the NTSC color system development process. It's fascinating
> reading. Available at your local university library.
Somehow I doubt that. :P
Would like to read it thought.
More information about the time-nuts