H-alpha all-sky survey. C. Stubbs May 25 2015
As was pointed out to me long ago by Greg Bothun, for resolved objects the surface brightness on the sensor depends only on f-number, not on aperture. So as long as we can Nyquist sample faint galaxies, a fast lens will do as well as a big telescope. Zeiss makes a nice 85mm f/1.4 lens that works out to 1 micron with excellent image quality, and covers a 35mm focal plane with a flat field.
At 85mm focal length one arcsecond spans 0.5 microns, and FOV is 28.6 x 24 degrees. A 10 micron pixel subtends 20 arcseconds, so we need arcminute-size galaxies with decent surface brightness. Assume the goal is to get out to redshift of 0.1, at which 1 kpc is 0.5 arcsec. Actually at 6500 A we don't need the fancy IR version, and 3-4 filters get us from z=0 to z=0.1, at 20 nm each. But the key is getting really small pixels and low dark current. So what's the smallest-pixel back-illuminated 35mm sensor? Say we want 5 micron pixels and 35mm diagonal, so 25mm on a side, which means 5K x 5K, TE cooled.
An LSST sensor would be OK, at 10 microns per pixel and 4 k x 4k.
Assume we need no more than 5 arcsec/pixel at 10 micron pixels. That's 10 kpc/pixel at redshift of 0.1. If we have 10 micron pixels, what's the requisite focal length?
For theta=25 microradians = 10 microns/FL, FL = 10 microns/25 microradians = 500 mm. That's a Takahashi at f/2.8. Or, perhaps the Dragonfly lens choice- 400 mm Canon f/2.8. They also make a 200mm f/2 lens.
What about using a Canon SLR camera? Have asked about conversion of 5K Mark III to B&W version, but the challenge will be dark current. Imager is 5760 x 3840, and 36 x 24 mm so each Bayer pixel is 6.25 microns. We aren't going to get images this sharp however and staying in focus is going to be tricky as well, with a fast optic in front.
| lens | filter threads | plate scale on Canon 5D, color pixel | plate scale at z=0.1 | FOV (sq deg) | speed rel. to f/2.8 | FOM | cost |
|---|---|---|---|---|---|---|---|
| Zeiss 135mm f/2 | 72 mm | 1 arcec | 2 kpc | 15 x 10 = 150 | 2 | 2.7 | $1822 B&H |
| 85mm f/1.8 | 58 mm | 1.6 arcsec | 3.2 kpc | 24 x 16 = 384 | 2.4 | 8.5 | $350 B&H |
85mm f/1.2 Canon EF 85mm f/1.2L II USM | 72 mm | 1.6 arcsec | 3.2 kpc | 24 x 16 = 384 | 5.4 | 19 | $1,999 B&H |
50 mm f/1.2 Canon EF 50mm f/1.2 L | 72 mm | 2.7 arcsec | 5.4 kpc | 40 x 27 = 1080 | 5.4 | 54 | $1,449 B&H |
| 50 mm f/1.4 | 58 mm | 2.7 arcsec | 5.4 kpc | 1080 | 4 | 40 | $320 |
| 50 mm f/1.8 | 52 mm | 2.7 arcsec | 5.4 kpc | 1080 | 2.4 | 24 | $125 |
| Zeiss 35mm f/1.4 | 72mm | 3.8 arcsec | 7.7 kpc | 2204 | 4 | 81 | $1543 B&H |
| 500 mm f/2.8 Takahashi | NA | 0.27 arcsec | 0.54 kpc | 108 | 1 | 1 | $5000 |
Well that's not a hard choice! And note that chromatic effects will be highly suppressed in a narrow band, plus the filter is probably easier to get.
Canon even has an f/1.2 85mm lens! That would be (1.4/1.2)^2=1.4 x faster but at the expense of image quality I suspect. Review of 85mm lenses here:http://www.popphoto.com/how-to/2011/03/battle-85mm-f14-lenses
Canon full lens chart is here: EFLensChart.pdf
Filters: Edmund optics sells 50mm dia 10nm wide interference filters. If we want to go from z=0 to z=0.1, need to span from 656 nm to (1.1)*656=720 nm and that requires (720-656)/BW=64 nm. With 20 nm wide filters that requires about 3, and twice that with 10 nm wide filters. The angular span for the 50mm lens is only 20 degrees off-axis, not so bad. Note this is independent of the f-number.
Seems the 85mm f/1.2 Canon lens is a good choice, with 50 mm f/1.2 as a wider field alternative. Ordered the following parts from Edmund and B&H: