So after reading several posts about various RTCs available, vs. using mains frequency, I've collected a few thoughts.
I don't see much talking about RTCs from SII (Seiko) or Intersil, which sport (in some cases) 1ppm of resolution (presumably, you have to hold your tongue right, and perform other rites to achieve this). I do see a thread on the Intersil ISL12022. I'm not thinking of buying one of these (yet), just curious.
But if you're really interested in extremes, there's the Atomic clock on a chip, which has yet to go to production, AFAIK. Might be a bit of a wait before we see one on a breakout board from Sparkfun, or someone else.
OTOH, maybe this is all a bit much. There are, apparently, people willing to spend $300,000 on a watch that tells only day and night. Heck, I can duplicate that with a Promini and an LDR.
BTW, in case anyone doesn't get the post title: (long intro version)
So I guess we'll all be dumping our RTCs and start using ATCs soon.
Rob
An atomic clock is actually something that should better be called a frequency normal. If you need very accurate time at a budget go to *bay and buy a used Trimble Thunderbolt module.
GPS is essentially a time-keeping piece. Is there any clock based on receiving GPS satellite signals? Must have poor reception indoors though.
Can mount the GPS unit anywhere and feed the output signal where you need it.
Self contained units that look like a mouse, deck of cards size, cable off to your PC.
Google "gps receiver for laptop", brings up many choices.
AWOL:
Quarter to two.
I'm assuming that's Greenwich. So that means it's actually 6:45.
Egadz. Those things are ~£900 new. And that's just the unit itself. I wish I were operating on your definition of "at a budget". 
(Yeah, I know, e-bay used is cheaper.)
@justjed: "on a budget" depends on what you are buying. I bought a used Trimble Thunderbolt for ~USD 100 incl. antenna and power supply. If you look at other choices that deliver similar or poorer performance this is "on a budget". Especially if you want really accurate time. I use it to calibrate my measurement equipment. Before I bought this I bought a used atomic clock that never really worked. Then I looked for DCF77 based frequency normals and found a lot of expensive solutions and some cheap crap. In the end I figured that the used Thunderbolts are extraordinary cheap compared to all other alternatives of similar or even worse performance.
It also allowed me to go for used frequency counters because I can easily verify/recalibrate the time bases. Thus I already saved a lot of money on the two counters that I have here. The point is: accurate equipment is typically pricey. Thus theses modules are a bargain.
On the other hand: most who ask for "accurate timing" seem to have never ever thought about how accurate they really need. If you do not know what you need or want everything is to expensive.
liudr:
GPS is essentially a time-keeping piece. Is there any clock based on receiving GPS satellite signals? Must have poor reception indoors though.
I have one on my wall in my Study (indoors, heh). Sometimes it stops because it loses reception. I should plug a bigger aerial into it. It was built from an article in Silicon Chip magazine (in Australia).
On wall:
Behind it:
The circuit board with GPS:
Keeps good time too!
The chip (a PIC 16F88) has code in it for allowing for daylight saving, which is nice. Based on the date (and you can connect up a serial cable and change the EEPROM settings) it either stops for an hour, or fast-forwards for an hour, so it shows the correct time when you get out of bed the next morning.
Awesome! My wall clock is just a plain old clock but it doesn't make ticking sound. So essentially your circuit creates the ticking signal, right?
Ah yes, essentially. Here is the author's project page:
The wires you see going into the back of the clock are where you pull out the original circuit board, and connect two wires directly to the stepper motor which drives the clock hands. So basically the processor pulses out positive/negative pulses, once per second, which steps the second hand.
The processor upon power-up fires up the GPS module to find the time, and after adjusting for local time by consulting the EEPROM, double-ticks the hands until they are at the right place. Since it doesn't know where the hands are it makes the assumption that you start it up with the hands at 12 o'clock exactly.
Then to save power it turns the GPS module off and just uses the processor clock for a couple of days. Then it connects to the GPS again. This next time it can work out an error differential for the internal (processor) clock - say for example, the GPS reports a time 5 seconds different to what it thought it was. It then includes that error amount for the next couple of days, thus keeping better time in the future.
