Solar Observatory to provide time.
I share a trait with Captain Hook - we both like clocks. I've had a fascination with precision timekeeping since forever. At the present time, a bank of homemade 5th overtone OCXOs have adequately addressed short-term stability needs. What seems lacking is a long-term reference. If a government, social, nuclear, or volcanic meltdown disrupted services like GPS, WWV, and GLONASS, where exactly would one obtain the correct time to start with?
So I'm building what could be considered the ultimate sundial. It's really more a solar observatory. It measures the precise timing of local apparent noon (using lens, long tube, and sensor array) and applies the Equation of Time higher accuracy methods described by J. Meeus in Astronomical Algorithms. The effects of small perturbations are minimized by time-weighted filtering of daily measurements. I live in an area that is sunny with annoying regularity, so most days I get a datum.
It is worth noting that errors are not accumulative. Indeed, four times a year the measurement is exact. So, with an already stable oscillator, corrections are infrequent and tiny.
My solar clock is being operated independently of any other time source and, while accurate references remain available, being compared continually. The resulting graph is expected to provide useful information to guide the optimization of the weighted filter.
Elevation adjustment is provided by a lag screw driven by a gear-reduced stepper motor. Observatory longitude was obtained with great precision from Google Earth.
Shortly, I expect to operate the entire observatory independent of mains power.
More to come...
So I'm building what could be considered the ultimate sundial. It's really more a solar observatory. It measures the precise timing of local apparent noon (using lens, long tube, and sensor array) and applies the Equation of Time higher accuracy methods described by J. Meeus in Astronomical Algorithms. The effects of small perturbations are minimized by time-weighted filtering of daily measurements. I live in an area that is sunny with annoying regularity, so most days I get a datum.
It is worth noting that errors are not accumulative. Indeed, four times a year the measurement is exact. So, with an already stable oscillator, corrections are infrequent and tiny.
My solar clock is being operated independently of any other time source and, while accurate references remain available, being compared continually. The resulting graph is expected to provide useful information to guide the optimization of the weighted filter.
Elevation adjustment is provided by a lag screw driven by a gear-reduced stepper motor. Observatory longitude was obtained with great precision from Google Earth.
Shortly, I expect to operate the entire observatory independent of mains power.
More to come...
Comments
But I'm a little curious: why did you use the sun and not a different star?
As your question implies, a good optical telescope and CCD array would likely deliver a more precise measurement. If I could put together such a thing on a shoestring budget, I might give it a try. Alas, spending as little as possible is perhaps 50% of the fun.
BTW, it occurs to me that "ElectricAye" would be a great name for the observatory... Would I have to pay you a nominal consideration for its use?
Interesting. I've always wondered if something like this could be made cheaply for a star like Sirius or Vega, etc. But then the fun would be in how to make a "light bucket" that was both cheap and time-precise, I spoze.
I am honored you would think it an appropriate name for your end-of-the-world scenario time-keeping system. No royalty check required. At the end of the world, your technological monument will no doubt be the only surviving scrap of evidence that I had ever existed.
Perhaps a small memorial plaque will be attached and, behind it, a vault containing a repository of 1000 ridiculously funny photographs and their captions, in hopes that future civilizations retain a sense of humor.
BTW, speaking of Sirius and Vega, I've long thought of building an optically crude telescope (perhaps with a Fresnel objective), the sole purpose of which would be to gather light from a distant source and train it on a high efficiency solar cell, and then use the output of the solar cell to power some small thing, like a radio receiver or a clock. It may not do much, but it would be out of this world.
That's not the Gray Pacific Northwet I'm familiar with.
-Phil
Sounds like some Sirius power technology! I'm guessing there's a way to calculate the theoretical harvest on that, but here's a caveat: your star-tracking system might end up using more power than the starlight can provide (unless, of course, the star just powers your clock device only as the star passes by). On the other hand, general skyglow might provide some steady light, but that doesn't sound as cool as raw star power. Would it be cheating if, in the daytime, you stored solar power only for the star-tracking device, then utilized the star power from a photocell at night to gather power from the star? In other words, the sun provides power for the tracking mechanism, but the star provides power for your radio or clock at night. Some Sirius hybridization, maybe?
Keep us posted on how this goes.
I don't know why, but I've always envisioned UserName as being dressed like Captain Nemo and living in a desert perched atop a mountain overlooking something like the Umatilla Wilderness, living in a house that looks like the front end of the Nautilus, laughing at the stars but scowling down into the valley whenever thoughts on the futility of the human race cross his mind. Which happens a lot.
1 - Mount the camera chip in a case with a tiny pinhole (ideally so small the sunlight falls on one pixel) .
2 - Mount the case in a fixed position so it would face the sun directly at noon on March 21st (or Sep 21st).
3 - Use the row/column of the pixel the sunlight falls on to determine the date/time.
Of course the light spot will probably fall on more than one pixel so the centroid of the lit area would need to be used for the date/time calculation.
@PhiPi: Port Townsend is much closer to the ocean. The bulk of my property is located in forested mountains near the Oregon/Idaho border.
@kwinn: You have exactly the right idea regarding the pinhole and camera chip. That's just what I would like to do, and it would dramatically reduce the size of my installation. The problem I encountered was access to the array. Kinda wish I could go back in time and buy one of the original optical sensors Micron sold, which were just windowed 64k DRAMs. Nowadays, it seems the available sensors are either inexpensive but difficult to query with a bare microprocessor of modest throughput, or else they are too expensive. I mean in concept, I'd just use a $10 Chinese keyfob camera. But realtime access to the images is problematic.
As for the Vernal and Atumnal equinoxes, turns out those are not the dates when apparent noon and mean noon correspond. Those dates are 11 Feb, 14 May, 26 July, and 3 Nov.
IMHO however the best image sensor would be a single chip that has an image sensor (say 1024x1024) and a similar sized memory array with interface circuitry that converts the image pixel analog voltages to integer values in the memory array. Then the processor would access that memory array just like it would any other memory chip.
It started with a spur-of-the-moment decision to fish through my box of table sweepings for a crystal oscillator. I found a 14 MHz Saronix DIP oscillator that looked promising. I cemented four 10 Ohm resistors to the top of the can and four to the bottom, wired in series-parallel for a total of 20 Ohms. To the side of the can I attached an NTC termistor with RTV. The little module then was packed into a small cavity made from suitably shaped blocks of 3/4" pink foam from Home Depot, left over from the packing of my most recent paramotor. Gaps were filled with foam produced from a gorilla glue/water mix. Once it was well-insulated, I applied (separate) power to the oscillator and heater, and hooked up a frequency counter to the oscillator's output. The temperature was slowly ramped up from about 10 C to 80 C, plotting the frequency response every few degrees. The pattern followed the familiar AT-cut 3rd-order polynomial-sort-of curve. The lower deadpoint was at about 15 C and the upper one at 70 C. So a simple controller was quickly whipped up to keep the temp at 70 C.
So far, the output is very stable. Long-term stability remains to be determined. But the results certainly seem to belie the simplicity and low cost.
Very cool. Keep us updated.
Funny, when I first read this, I thought for sure you wrote paramour. I was going to ask if you were packing your paramour off... or in.
Understandable confusion. The difference is that the paramotor's packing can be used for other things since it is now out of the bag.
Use these to power it, they seem like really durable end of the world material.
Plus you could use them to help find the sun based on output, put one on all sides of apparatus, rctime each output to determine suns approx location.
FWIW, the commercial frequency counter I've been using has reached the limits of its utility. This week has largely been devoted to the design and construction of a more accurate and precise counter. The University of Limerick, the Technicon in Haifa, and MIT have nice online documents on metastability that were a big help in this endeavor. Gotta love the internet!
Thanks to the fantastic timing capabilities of the Prop, this was very easy to determine. What I found was that the periodicity of the reporting interval was quite ragged, varying by more than +- 4% in a haphazard way. I imagine a histogram of this variation would be interesting, and I may obtain one later. The salient point is that the start of messaging is not an accurate reference by itself.
This is where the Prop comes in again. It was easy to make a digital PLL by summing each second's error (by subtracting 80_000_000 from each measurement and then accumulating the signed values thus obtained). This accumulated error was then measured against thresholds. When the value was found to be in excess of one threshold or the other, the 80_000_000 value was incremented or decremented by 1.
(The long-term accuracy of the cheap GPS is essentially perfect, since it's ultimately phase locked to the international time standard. So the accumulated error is primarily the result of the crystal I selected and the temperature of the room it's in.
Anyway, with very little effort and at no additional expense, I ended up with a low jitter 1pps reference that is phase-locked to UTC. For more demanding applications, considerable optimization could be done on the control loop to improve (minimize) static phase offset.