Hacker News new | ask | show | jobs
by jacquesm 3271 days ago
That's worthy of the term 'hack'. Very neat :)

There should be some silly way of doing this with just a single photo diode and a speaker to provide one axis of motion and to let the Earth's rotation provide the other axis.

3 comments

You can do this even without any motion and possibly even with no focusing optics, just with a single sensor, diffraction grating and a lot of math. This is called single-pixel imaging and is an ongoing research topic in computational imaging.
Could you explain in laymans terms how that would work? If you're only obtaining a single value for a single pixel and there's no movement, how does the pixel value change even if there's a grating in front? Or do you mean the grating can still move?
The pattern still needs to be different at each sampling, yes. Moving the grating is one way to do it, or it can be done with a micromirror array, for example.
Gotcha cheers! A while ago I saw a single mirror MEMS device, could be useful for something like this, I believe you could control the horizontal/vertical angle.
And noise? I used to do long exposure photos of interiors of buildings on film (~2h) but digital cameras seem challenged to me.
That's very interesting re. film. I'd be curious why that is. I've only ever done exposures less up to around a couple of minutes with film & digital. Do you have any examples that you shot for 2hr out of curiosity. Was that with ISO 50 or less, or did you use a darkening filter too?

Edit: I wonder if it is due to reciprocity failure with film - https://en.wikipedia.org/wiki/Reciprocity_(photography) ?

125 asa FP4 roll film (all monochrome and hand developed needless to say, I used a Rollieflex camera on a tripod when I was playing with this stuff decades ago).

I found the correction chart below on t'web [1] just now.

I recollect using 2.5x per stop over 1 second which gives slightly longer exposures than [1], but then I did tend to 'pull' the development a bit (contrast goes up), so a 2 minute metered exposure would be 2.5^7 seconds (10 minutes) instead of 2^7 seconds roughly. A 10 minute exposure on the meter corresponds to two hours and a very significant contrast hike - I recollect printing on grade 1 and grade half filter settings (multigrade paper). I only used extreme exposure times in old dark interiors e.g. churches in the UK in November or something. Inside so no need for any filters - not astro.

I'll dig a few prints out and scan them over the weekend. Nowt astounding. See if you can find a book about Edwin Smith if you are near a library with a very good photography collection.

I have a fantasy of using 10x8 film with a pinhole camera and just doing contact prints... but then I remember the darkroom days and the amount of water that got used up...

[1] http://home.earthlink.net/~kitathome/LunarLight/moonlight_ga...

That chart looks very interesting, thanks!

Cheers, I'd be very curious to see your prints.

I've just had a little look on google images at some of Edwin Smith's photos, they look stunning! The light in the photos look amazing.

I still have an old Pentax K1000 film camera I think it is, I'll have to dig it out again some time :) I'd really like to get a medium format camera, but they're still not super cheap, I was looking at the Mamiya RZ67 (I'd actually like to try shooting landscape with it, which could be impractical due to the weight, but would be fun nonetheless ;).

10x8 film would be awesome to play with!

That's why most actual astrophotography is done with peltier cooled CCDs - firstly a lower temperature means less noise, but more importantly more consistent noise. You can then shoot dark, bias and flat frames, respectively average those, then subtract them from your stacked (averaged using any number of algorithms) light frames.

No reason it couldn't be applied to regular long exposure photography using a DSLR, either.

This is incredibly cool! I wonder what the simplest possible non-mechanical modulator is. (LCD?)
Heh that'd be very cool! Do it!
Can you explain what you mean?
To make an image you need two axis, one horizontal, one vertical. Because the earth is already moving if you have a single photodiode looking at the stars you will scan a line if you wait a while. By adding a second axis of motion such as a speaker which you modulate with a saw-tooth curve you can add a second axis. If you then scan over time you can re-constitute the image by plotting the dots with the intensity captured by the diode taking into account the driving voltage of the speaker and the time that has passed.
I get that. I guess I was just confused as to why your first thought was to use a speaker to move the photo diode as opposed to an Arduino with a stepper motor.
It's an exercise in simplicity. Adding an Arduino is a little against the spirit.
Can you explain how a speaker with a square wave would move the diode? I'm so confused.
Sawtooth, not a square wave. Hook the speaker up to some kind of frame, glue a small hinge to the cone, hinge to an arm that's anchored relatively close to the speaker, diode at the far end of the arm in a tube so the light is directional, instant scanner :)
Now you're thinking with superresolution!
Oh, I just had an even nicer idea that I really will have to try. I won't say what it is yet to not spoil the surprise but I'm sure that it will work :)