Perio said:
jrista said:
raptor3x said:
jrista said:
I definitely want the IQ improvements. The 1D X is the noisiest sensor Canon has. Compared to the noise levels I'm used to working with these days, it's INSANELY noisy.
Just curious where you're getting this from. I've owned pretty much every Canon FF body at this point and this has been the exact opposite of my experience. DxO and Sensorgen seem to disagree with you as well.
To quote sensorgen:
"Model EOS-1DX 100 80 38.5 90101 11.2"
Find me a camera this generation or later with base read noise that high anywhere else. 38.5e-! That is insane!
I use my 5D III occasionally for landscapes, but I can't stand it's read noise either. It's 33e-. Still insanely high. High ISO is certainly a different story, but I can get low read noise (4e- or less) at high ISO with just about any camera on the market, I don't need a 1D X for that.
I also intend to use any future Canon DSLR I get for astro as well. Buying a full frame CCD camera is a minimum of $10k, and a better large sensor CCD is as much as $45,000. The best way to get a full frame field with astro is with full frame DSLRs. I use the 5D III as low as ISO 400, but the poor low ISO DR stops me there. The 5D III has so little DR at ISO 400 and up that I never get away with unclipped stars. Even 12 stops of DR on a 5D IV would be huge for my astro work.
I don't do astrophotography, but why you do not consider Pentax 645D if you need a CCD sensor? It can be bought for relatively inexpensive prices these days.
A proper CCD camera is purpose-built for what we as astrophotographers do. For one, they are monochrome, thus delivering a 100% fill factor (vs. the 50% green and 25% red/blue fill factors with a bayer CFA matrix.) CCD cameras usually have extremely high quantum efficiency these days, 78% EQE for modern Sony ICX sensors. Even a high IQE CMOS sensor has at most 40% EQE in any given color channel, and often as little as 30% or less. They use multi-stage peltier (TEC) cooling, and the cameras I am interested in can regulate temperature to within 0.1°C at -45dT from ambient. That means throughout the year, I can maintain sub-freezing sensor temperatures, which is a critical factor for minimizing noise from dark current. A proper CCD camera is either compatible with, or in the case of the cameras I am interested in, has built in, a filter wheel that can hold LRGB and narrow band filters. With filters, you get that 100% high sensitivity fill factor for every color channel. With narrow band, you can image under heavily light polluted skies for specific emission bands like hydrogen alpha, hydrogen beta, oxygen III, sulfur II, nitrogen II, and others if your interested (although they get significantly more difficult to pick up for other emission bands). CCD cameras are also usually built with much cleaner readout noise. For a brand like QSI or FLI, there is no banding, no pattern...just pure, random read noise that follows an ideal gaussian distribution.
Having purely random noise is key to getting the most out of the stacking process, where multiple individual sub frames are combined to produce a single high signal strength/low noise "integration". Random noise averages out and is suppressed by the stacking process. Non-random noise, such as hot pixels, banding, etc. will CORRELATE through the stacking process, and only become reinforced, like any other signal.
CCD cameras often have higher read noise than DSLRs/Mirrorless. The common KAF sensors have around 7e- RN, and the Sony ICX sensors have ~5e-. In practice, sometimes they are slightly less. CMOS sensors in DSLRs often have as low as 2-3e- read noise at higher ISO. Another BIG difference between CCD and CMOS sensors, though, is the dynamic range. A CCD camera will usually be 16-bit, and will use a gain value that is specifically chosen to maximize the potential use of the full potential well. Dynamic range of a CCD camera can be significantly higher than a CMOS camera at high ISO. This gap in dynamic range becomes even more true when you account for the additional dark current noise that a DSLR has, vs. the minimal dark current that a temperature regulated CCD has. The additional noise from dark current reduced dynamic range of the DSLRs even further.
As an example, my 5D III during the summer has as much as 3-5e- dark current PER SECOND per pixel. For a 300 second sub exposure length, that is as much as SQRT(5*300) additional noise, or 38.7e-! Combined with read noise at say ISO 800 (one of my commonly used ISO settings, with a rather high read noise level of 6.1e-!), it would be SQRT((5*300) + 6.1^2), or a total of 39.2e- noise. Dark current can and will utterly decimate dynamic range in a camera like that. In contrast, a Sony Exmor may have around 0.8e-/s/px dark current at the same temperature, and 3e- read noise. That leaves us with SQRT(0.8*300 + 3^2), or 15.7e- noise. However a CCD is still significanly better than either of those. A KAF-8300 sensor (very popular sensor) has about 0.02e-/s/px dark current at a regulated temperature of -15C. For a 300s exposure, dark current is a mere 6e-, and read noise is 7e-. That gives us SQRT(6+7^2) or 7.4e- total noise. A Sony ICX sensor has as little as 0.003e-/s/px dark current at only -10C, and 4.5e- read noise, so SQRT(3 + 4.5^2) or 4.8e- total noise. The Sony cameras have around 20ke- FWC and the KAF has 25.5ke- FWC. That gives us dynamic ranges of 11.8 and 12.1 stops each. Compared to my 5D III DR at ISO 800 of 7.9 stops and say a D800 at ISO 800 of 8.8 stops.
Canon cameras have the highest pattern noise, particularly banding, of any camera I've ever used. By a SIGNIFICANT margin. CCD cameras are effectively devoid of banding and have low pattern (i.e. hot pixel, stuck pixel, dead pixel) noise. Anything that uses a Sony Exmor has almost ideal gaussian read noise characteristics, although usually not quite as good as a proper CCD. However all DSLR and mirrorless cameras except maybe one of the Leica lines are bayer CFA. The use of a color filter array decimates overall sensitivity, and the quality of the filters embedded into each pixel is low compared to the quality of say a set of Astrodon LRGB filters.
Noise is everything when it comes to astrophotography. We often gather as little as 0.12 photons per
minute per pixel for faint targets, so our final signal strengths are utterly minuscule, even after sub exposure lengths as long as an hour or so. Noise is EVERYTHING to us astrophotographers.
That's probably why I am so obsessed with it.
;D
