[time-nuts] finding time astronomically.

Jim Lux jimlux at earthlink.net
Mon Jan 23 16:08:04 EST 2012

On 1/23/12 12:29 PM, Chris Albertson wrote:
> On Mon, Jan 23, 2012 at 12:02 PM, Jim Lux<jimlux at earthlink.net>  wrote:
>> This chat of zenith cams, etc. is interesting.
>> How well could you do with something like the camera in the iPhone4 facing
>> up. The front camera is VGA resolution.
>> Say you're on another planet?
> You can use a stick pounded into the ground and wait until the shadow
> has minimum length.   But I assume we need better accuracy?

An interesting approach, because it could conceivably get 
"magnification" without using lenses or mirrors.  Imagine the shadow tip 
of a 2 meter long stick, and I have the camera positioned so that I only 
see about 20cmx20cm.  (of course, the shadow isn't that well defined, 
because the angular extent of the sun is huge)

A similar scheme if i use a pinhole to project an image of the sun, and 
image that, instead.

> If you use a camera, accuracy will be limited by your knowledge of
> where you are aiming the camera.  If you are off by one degree then
> the error is about 1/360 times the length of the day on your planet.
>    So finding the time is really about discovering where you have aimed
> the camera.    This is best figured out at night when you can see
> stars.    You can actually aim the camera at random, so long as you
> measure the aim point and don't let it move.
> That said, I think if you were to leave a cell phone in a fixed
> position, un-moved all night you can likely get to 1/10th of a pixel
> angular resolution.     So what is the angle subtended by  one pixel
> on your phone divide that by 10 then multiply by one day.        A
> total guess is "about 1 mSec" if you use a full night's data.  Just be
> warned that reducing the data is not simple there are many steps
> involved just one of then is matching your data to a good star catalog
> and this implies having a good catalog.

iPhone cameras (and most webcams, etc.) seem to have a FOV about 45 
degrees, so one pixel is around 0.1 degree.  At 4 minutes time per 
degree, that's about 24 seconds per pixel.

(It's not a monochrome sensor, either, so although it's NxM pixels, that 
doesn't mean that you could actually resolve a planet to that scale, 
depending on color, and how the image is processed)

> You really can get to 0.1 pixel. You fit a function to the "fuzzy
> blob" image of each star and then maybe 100 pixels contribute to a
> solution.

tricky on a iPhone type camera, since star images are one pixel at best. 
  On the cameras I've seen that were designed to do this, they have a 
cleverly designed optical system that blurs the image.  (and another 
scheme uses a camera with a multi pinhole mask in front, to render the 
image in multiple places across the sensor.

More information about the time-nuts mailing list