CCD vs dSLR

codemonkey (Lee) · #1 12-06-2014, 04:51 PM

Now that I've established that my newly-acquired mono CCD is still broken, albeit in a seemingly different way, I'm taking a step back and revisiting the dSLR vs CCD decision.

I'm interested in hearing why purpose-built CCD systems are considered better than consumer dSLR cameras for the purpose of astrophotography. Previously I had thought the following factors supportive of CCD:

1) Lower noise (due to cooling)
2) Greater dynamic range
3) Sensitivity, in particular to H. alpha

I'm now questioning points 1 and 2 after noticing something. I had a quick look around for some information on the OSC QHY 8 camera, and discovered that it actually has the same sensor as a Nikon D50. This has undermined my confidence in CCD, particularly point 2.

My question is now, why would I buy a OSC QHY 8 over a Nikon D50? The answer seems to be only because it's cooled and because of the greater spectral sensitivity. Of course I could have a dSLR modified to increase sensitivity around 650nm so that nebula are more easily captured, so that's a reasonable solution to point #3.

This leads me to the following question: What would I be losing out by using a modified Nikon D5100? It has greater dynamic range than the D50, I'd gain back the "missing" spectral sensitivity, I'd have a sensor that generates less noise and could conceivably out-perform it if I built a little cooler box for it.

I suppose this is why most people don't bother with OSC CCD cameras and instead use mono CCD cameras. But then what is the advantage of mono?

I've seen the argument that mono CCD are more sensitive due to the lack of a bayer CFA, but I don't understand that; maybe I'm missing something. The reason that I find that confusing is because it seems to me that's actually a problem with reduced resolution rather than sensitivity.

If all the bayer CFA does is put a colour filter over four pixels so that it's 1 red, 2 green, 1 blue, then how does that affect sensitivity? Surely that just means that instead of capturing the full resolution in (for example) red, you're actually capturing less and the rest is "recovered" (guessed) by software interpolation? This would mean that if you were prepared to sacrifice some resolution (and lets face it, these new dSLR have plenty to spare in comparison to available CCD) you wouldn't be giving up much, if anything... hell, without doing the math you might even still be ahead.

So, what am I missing?

Shiraz (Ray) · #2 12-06-2014, 05:18 PM

the cooled OSC gives essentially the same result as a modified cooled DSLR - you get consistent results and much lower noise.

Mono is used in LRGB mode, where you get a high-res low noise broadband image (without filters) and then add in colour taken through filters. In mono mode, each pixel sees roughly 3x as many photons as it would if there was a Bayer filter in front of it - eg the red filter absorbs green and blue photons, so roughly 2/3 of the photons never even get to the sensor. This means that, even taking into account the need to add exposures for colour, the mono system will get to a given picture quality more quickly than an OSC.

In addition, mono subs can be a lot shorter, so you will be less affected by tracking issues. OSC also suffers a bit in resolution because much data is obtained from interpolation - but that is probably not a huge problem for most targets.

RickS (Rick) · #3 12-06-2014, 05:22 PM

Quote:

Originally Posted by codemonkey

But then what is the advantage of mono?

I've seen the argument that mono CCD are more sensitive due to the lack of a bayer CFA, but I don't understand that; maybe I'm missing something. The reason that I find that confusing is because it seems to me that's actually a problem with reduced resolution rather than sensitivity.

If all the bayer CFA does is put a colour filter over four pixels so that it's 1 red, 2 green, 1 blue, then how does that affect sensitivity? Surely that just means that instead of capturing the full resolution in (for example) red, you're actually capturing less and the rest is "recovered" (guessed) by software interpolation? This would mean that if you were prepared to sacrifice some resolution (and lets face it, these new dSLR have plenty to spare in comparison to available CCD) you wouldn't be giving up much, if anything... hell, without doing the math you might even still be ahead.

So, what am I missing?

Your argument is correct as far as it goes, although some of us want every skerrick of resolution we can get

The advantage of a mono camera for RGB imaging is the ability to collect luminance data which includes R, G and B. The theory is that the eye is more sensitive to luminance information than chrominance so combining good quality L data (which you can collect relatively quickly) with lower quality RGB data (often binned) gives a comparable result in less time than RGB imaging.

Mono cameras work better for narrowband imaging too, of course.

Cheers,
Rick.

uwahl (Ulrich) · #4 12-06-2014, 05:37 PM

In addition to the advantages Ray has mentioned many astro CCD cameras have the ability to do 2x2 etc binning to suit different focal lengths with the addition of being able to adjust the gain and therefore dynamic range to match the binning.

Nevertheless modern SLR cameras can take brilliant astro photos that approach the quality of astro CCD cameras.

Where astro CCD cameras really score is if you want to do any science (eg filter photometry)

pluto (Hugh) · #5 12-06-2014, 05:39 PM

The D50 raw NEFs are 12bit as opposed to 16 bit raws from the QHY8.

alistairsam · #6 12-06-2014, 05:48 PM

hi,

There is also the difference in the image format

DSLR's are usually 12bit or 14 bit, OSC's are 16bit. the difference does allow for a greater range in shades of grey or colour.
I found I could stretch my qhy8 images much more than the dslr images.
just my experience though.

Cooling - the QHY8 has a dual stage peltier that can easily hit -30C.
This is a significant advantage over peltier cooled dslr's that can also be cooled but reaching these temperatures pose significant challenges due to physical constraints and condensation and add a lot of bulk, usually are not regulated but you can add regulated cooling.
the qhy8 has unregulated cooling but I think you can use the DC201 power supply and get regulated cooling.

if its a whole body replacement like the CDS cooled dslr's then results should be similar as they can cool pretty well.

As for mono vs colour and the Bayer matrix, that has been covered several times here, but mono provide greater flexibility with narrowband imaging which is a huge advantage if you want to do narrowband.

Another minor difference - batteries with the dslr's vs dedicated power supplies for ccd's, of course you can get an adaptor for the dslr, but then...

finally, backfocus.
if you want to add a filterwheel, OAG etc, the dslr's are very limited as they already take up 45mm to the ccd, whereas ccd's are usually 15 to 20mm.
this is important if you use a coma corrector, reducer etc where the backfocus is typically 55mm to the CCD and you want to use an OAG and filterwheel.

Again, solutions are to get a zero profile bayonet to T2 and a slim OAG, filterwheel would be pointless.

I'm very tempted to get a QHY8 and remove the CFA but then that's another story!!.

Cheers
Alistair

alistairsam · #7 12-06-2014, 05:59 PM

hi,

There is also the difference in the image format

DSLR's are usually 12bit or 14 bit, OSC's are 16bit. the difference does allow for a greater range in shades of grey or colour.
I found I could stretch my qhy8 images much more than the dslr images.
just my experience though.

Cooling - the QHY8 has a dual stage peltier that can easily hit -30C.
This is a significant advantage over peltier cooled dslr's that can also be cooled but reaching these temperatures pose significant challenges due to physical constraints and condensation and add a lot of bulk, usually are not regulated but you can add regulated cooling.
the qhy8 has unregulated cooling but I think you can use the DC201 power supply and get regulated cooling.

if its a whole body replacement like the CDS cooled dslr's then results should be similar as they can cool pretty well.

As for mono vs colour and the Bayer matrix, that has been covered several times here, but mono provide greater flexibility with narrowband imaging which is a huge advantage if you want to do narrowband.

Another minor difference - batteries with the dslr's vs dedicated power supplies for ccd's, of course you can get an adaptor for the dslr, but then...

finally, backfocus.
if you want to add a filterwheel, OAG etc, the dslr's are very limited as they already take up 45mm to the ccd, whereas ccd's are usually 15 to 20mm.
this is important if you use a coma corrector, reducer etc where the backfocus is typically 55mm to the CCD and you want to use an OAG and filterwheel.

Again, solutions are to get a zero profile bayonet to T2 and a slim OAG, filterwheel would be pointless.

I'm very tempted to get a QHY8 and remove the CFA but then that's another story!!.

QE - although most CCD's are above 55% QE (broadly speaking), I don't know if there would be a difference with QE measurements between the D50 and QHY8 as QE calculations for the dlsr aren't that straightforward.

The D50 is an exception amongst a few others in that it uses a CCD vs CMOS in canons. the QE of most canons are calculated to be between 25 and 40% with the exception of one or two that are quite high.

binning - as someone else mentioned, this is also an advantage with CCD's depending on your sampling and system image scale.

As for your broken CCD, why not get a replacement if its under warranty?

Cheers
Alistair

gregbradley · #8 12-06-2014, 06:14 PM

Generally speaking modern DSLRs have very small pixels compared to most Kodak chip based mono CCDs.

Small pixels generally speaking usually means more noise, less dynamic range and smaller well depth (meaning bright objects fill the pixel too easily and you get bloat and overexposure).

You see this sometimes in DSLR images where the stars are all white because they are all overexposed due to small wells of the small pixels.

Some DSLRs have large wells and pixels. You have to do your research but a 24mp APSc sized sensor is going to have small pixels probably half or less than half the size of a typical CCD (9 microns).

Mono of course as said already makes every pixel count so no 4 pixels to make 1 image pixel. Bayer CFA rob light transmission. This is obviously a big point as Sony recently improved a CFA on one of their cameras and claimed a 20% increase in light transmission from a new set of colour dyes for their CFA. Mind you there would be light loss from RGB filters as well. Perhaps not as much as they are higher quality filters.

DSLRs are usually CMOS sensors. Rarely CCDs are used, older Nikon D50 and D70 are probably a rare camera that used one. CMOS means the circuitry like analogue to digital converter and amplifiers are etched in the area around each pixel ie each pixel has its own circuitry. This typically takes up 40% of the surface area of a sensor so only 60% takes light, some of that is lost through the CFA, there is blur applied by the AA filter and no doubt light loss there as well (it also blocks the Ha and IR response).

So smaller pixels, 40% loss due to circuitry etched in around every pixel as opposed to circuitry being completely separate on a CCD so 100% surface area used for collecting light, CFA light transmission losses, then 4 pixels to create 1 image pixel is a substantial light loss, plus the AA filter creating a blur and more light loss on top of small wells, low QE (often below 25% compared to around 50% for most CCDs).

Add to that most CCDs are 16bit and most DSLRs are 12 bit or later models are 14bit which also translates to lower dynamic range. Read noise may be higher in some cases as well.

So a lot going against them. DSLRs vary a lot in QE and well depth though so some research would be good. The new Sony A7s may be the ultimate "DSLR" (its a mirrorless) with large pixels and super high ISO performance (not to mention 4K video).

Greg.

Bassnut (Fred) · #9 12-06-2014, 06:25 PM

Get the CCD fixed, no brainer, stop "revisiting".

To use a DSLR for a remotely similar result you will need vastly more imaging time, serious extra skill in processing and a whole LOT more time stuffing around generally.

Your choice, time or money. DSLR or CCD.

LightningNZ (Cam) · #10 12-06-2014, 07:04 PM

CCDs generally have bigger pixels for the same area - that's an advantage for the obvious light gathering potential but it also makes processing much quicker. Mono images are also likely easier to process.

For the same money you tend to lose a lot of area though. Why don't you just get a mono'd D5100 with a cold finger cooler? So you lose a little on fill (which is normally compensated by the micro-lens array) and you definitely lose on QE, but that's still a big, high-quality sensor right there.
-Cam

Peter Ward · #11 12-06-2014, 08:46 PM

Ah...this old Chestnut...

CMOS (read DSLR's) sensors have quite bit more "architecture" over and around their pixels.

This causes diffraction spikes and scattering (image bloat) that are common in even the best DSLR images. (OK, there are now a few Sony back-illuminated exceptions ).

Once you know what to look for these artifacts are a bit like a dunny in a desert.

QE and noise are however are the big killers. When you push the envelope...ie Narrow-band sub-frames of 30-40 minutes..DSLR's simply don't cut it....and it is here, that cooled CCD's become objects of wonder.

LightningNZ (Cam) · #12 13-06-2014, 07:32 PM

I know you shoot some great stuff Peter, but this image was produced with a Nikon D7000 (same chip as my D5100): http://www.eprisephoto.com/nebula/h42f1922#h42f1922

I would be over the moon (so to speak) if I could take an image as good as that with ANY sensor money can buy.
-Cam

Quote:

Originally Posted by Peter Ward

Ah...this old Chestnut...

CMOS (read DSLR's) sensors have quite bit more "architecture" over and around their pixels.

This causes diffraction spikes and scattering (image bloat) that are common in even the best DSLR images. (OK, there are now a few Sony back-illuminated exceptions ).

Once you know what to look for these artifacts are a bit like a dunny in a desert.

QE and noise are however are the big killers. When you push the envelope...ie Narrow-band sub-frames of 30-40 minutes..DSLR's simply don't cut it....and it is here, that cooled CCD's become objects of wonder.

Ken · #13 13-06-2014, 07:49 PM

As Greg mentioned the Sony A7s looks like a camera built for low light work, full frame 12 mp with large 8+ micron pixels. Low noise can be pushed to amazing iso of 409k

Maybe the big camera manufactures have seen the light and reverse the trend for increasing the mp on each new model.
This new cmos sensor also has larger lenses overlapping the grid structure between the pixels giving it higher than normal sensitivity.
Clear skies Ken.

Spookyer (Brett) · #14 13-06-2014, 08:33 PM

I have used both a DSLR and CCD camera now.

I think as far as the chips go the modern DSLRs are way ahead in terms of noise. My Nikon D4 at 1600ISO produced almost noiseless images at ambient temperature. By comparison my new SBIG STT8300 produces massively more noise even when cooled to -30 degrees. There is simply no comparison. It uses of the "old" KAF chip, they cannot match the modern full frame DSLR chips. It is not surprising really when you consider the sales volume driving chip development in the DSLR area compared to the very modest market for specialist CCD cameras.

I got a CCD capability for two reasons
1/ So I could collect and combine LRGB data with Narrow band data
2/ Get more sensitivity on the sensor to certain wavelengths like Ha.

Brett

Peter Ward · #15 13-06-2014, 09:08 PM

Quote:

Originally Posted by LightningNZ

I know you shoot some great stuff Peter, but this image was produced with a Nikon D7000 (same chip as my D5100): http://www.eprisephoto.com/nebula/h42f1922#h42f1922

I would be over the moon (so to speak) if I could take an image as good as that with ANY sensor money can buy.
-Cam

Sorry, with big bright objects (well, like M42) it's a bit of a doddle.

As I mentioned earlier, with NB it's very much a different ball game. Show me a narrow band DSLR image as good as say this one and I might change my mind

Note: the uber-clean SII data, lack of scattering and needle like stars...

MrB (Simon) · #16 13-06-2014, 09:24 PM

Not comparing myself to Martin here

I'm a beginner with modest equipment and no real processing skill.

60Da, 70mm f6 acromat, Ha and OIII:
http://www.astrobin.com/full/98888/0/?real=&mod=

graham.hobart (Graham stevens) · #17 13-06-2014, 09:47 PM

jeez Peter , who took that picture ? It's amazing!
I have waxed and waned about this choice for a few years- where I live it's a royal pain trying to get LRBG data because of poor weather - which is why last year I sold my mono Ccd and concentrated on my dslr. Granted I only aim for happy snap quality but my personal experience tells me I need much more dslr data for the same quality than mono. it can be done - also using my choice of software I find it much harder to process unless I have (in my terms) reams more data . That you would expect given better SNR.
Two things I personally would be interested in doing is trying an LP filter and comparing to a cooled dslr? My cooled 60 da currently is back in Korea getting fixed so I can't compare it directly. That has the cooled advantage plus a greater Ha sensitivity but when it was working I was getting max 10- 15 degrees below ambient in summer which compares ok with the Afore mentioned older style qhy 8 but not with a more expensive Ccd.
which is why I bought another mono Ccd!
I have to say the choice of imaging in my collection is made easier with the flexibility and cheapness of dslrs, I know I can quickly set up my canon with a small lens and get a wide field shot, doing that with a Ccd is expensive and requires more than ok fix that camera here with tape type planning.
I think the qhy 8 is a great Ccd camera and even with the same sensor as the Nikon it holds it's own in intro Ccd class and has the Ccd and software advantages of " made especially for the Astro peeps" about it. ( only reason I am selling mine is I bought the big brother second hand recently!)
In my mind I think they both have pros and cons but the major con for CCDs is cost and more complicated acquisition , but having a the cooling regulated, software coming out of your ears, and narrow band options is enormously advantageous . I think getting amazing dslr pics require more data, really accurate calibration unless cooled, and pitch perfect processing. What Rowland is doing on here wide field with Dslrs is inspiring . Regulated cooling is massive here - my canon 5 d will often run to 30' c during a run, exit temp from APT.
I used to have the cool box from TS express Europe, it worked done to a delta of maybe 10'c but was ungainly, heavy and condensation was an issue. Also in the older canons when live viewing you had to change B and M settings to focus so this opened the box and let the cold air out etc etc.
Sorry for the long answer, watching the footy and home alone with the two girls as SWMBO gone out.
Graz

graham.hobart (Graham stevens) · #18 13-06-2014, 09:59 PM

mr b that is impressive too!
I have a book called objects of the southern sky by Nicola Montechiarri -, it is a gorgeous book full of delightful Astro prints. All of the pics were taken with spectrum modded canon 350 and 450d.
admittedly he had really dark African skies ( Namibia ) bit different to Hobart clouds punctuated by these bloody search lights currently.

alistairsam · #19 14-06-2014, 12:34 AM

Quote:

Originally Posted by Spookyer

My Nikon D4 at 1600ISO produced almost noiseless images at ambient temperature. By comparison my new SBIG STT8300 produces massively more noise even when cooled to -30 degrees.

hi Brett

would you have any 10min dark frames from both to compare?

my experience with a qhy8, qhy8L and stf8300m has been that I've never had to use darks though they're recommended.
so an ambient temp dslr sensor having lower noise than a cooled ccd is a bit odd.

Cheers
Alistair

Amaranthus (Barry) · #20 14-06-2014, 12:41 AM

I find Spookyer's claim extremely hard to believe without some empirical evidence from the darks, especially if taken at exposures of, say, 2-10 min.