The most obvious physical difference is the chip size and pixel size. Comparing each monochrome camera I get the following from the Adirondack website:
Lumenera - SKYnyx2-0 Monochrome CCD Astrophotography Camera
Resolution 640 x 480
Sensor Type 1/3" CCD
Bit Depth 8 or 12
Monochrome
Dark Current <1 e-/s
FPS 60
Lens Mount T
Pixel Array 7.4 um
Readout Noise 10 e-
Well Depth 40,000 e-
Lumenera - SKYnyx2-1 Monochrome CCD Astrophotography Camera.
Resolution 1392 x 1040
Sensor Type Sony 1/2" CCD
Bit Depth 8 or 12
Monochrome
Dynamic Range >66dB
Lens Mount T-Mount
Pixel Array 4.65 um
Readout Noise 12 e-
Lumenera - SKYnyx2-2 Monochrome CCD Astrophotography Camera
Resolution 1616 x1232
Sensor Type 1/1.8" CCD
Bit Depth 8 or 12
Monochrome
Dark Current <2 e-/s
FPS 12
Lens Mount T
Pixel Array 4.4 um
Readout Noise 12 e-
Well Depth 14,000 e-
Its interesting looking at the data. The larger 1 1/8 inch chip has a Well Depth of 14,000 e- vs the Well Depth 40,000 e- of the 640x480 chip, so the smaller chip can hold more charge or photons. Generally, the higher the number of photons, the better your signal to noise ratio so the images look rich and full as opposed to thin and grainy.
Bert explained in an older post that small pixels in consumer ccds are analogous to fine grain film – they require a lot of light to ensure that you have plenty of signal to keep noise (or grain) to a minimum.
The 640x480 chip can operate at 60 FPS due to the relatively small number of pixels to be read and downloaded, whereas I note the 1616 x1232 chip spec is 12 FPS.
Somewhere there is a formula which you can use to calculate the optimum pixel size for your focal length. If you have “too many” pixels then you are over sampling and wasting data. If you have “too few” pixels your stars will look blocky. I’ll post it when I find it.
Here is a useful nomogram. Read your focal length in the middle bar, then look at the respective optimum pixel size on the right hand bar for the type of object shown on the left hand bar.
Okay, let’s have a look. We’re looking at 3 things here.
Sky coverage per pixel in arc secs (resolution based on seeing).
The focal length of our telescope.
The pixel size of our ccd chip in microns.
Starting on the centre line (focal length), we know the C9.25 has an fl of 2350mm, so we place an imaginary ruler, horizontally, between 2000 and 3000, where we estimate 2350 to lie. Let’s drive an imaginary nail through the ruler at this 2350 mm mark so the ruler will pivot on this nail.
Next, we would like to do some deep sky imaging of say, NGC253. So, we look at the left hand curved line (sky coverage per pixel) and see a grey block labelled “Deep Sky”. You will note the grey block starts at 1.5 arc seconds and finishes at 2.5 arc seconds. What this is indicating is that these are the typical seeing conditions we might encounter when imaging e.g. galaxies. 1.5 arcsecs is good seeing whereas 2.5 arcsecs is average seeing, where a typical star on a ccd would bloat out due to the seeing.
Next, we pivot our imaginary ruler around the virtual nail so the left hand side of the ruler dips down until it lands in the middle of this grey bar, on say the number 2 for 2 arc secs.
When we have done this, the right hand side of the ruler will have moved up and will cut the right hand curved line (pixel size) at the 24 tick, just below the grey area labelled TC-241 which is a Texas Instruments ccd chip with pixels of around 26 microns.
So, the nomogram is indicating that the optimum size of a pixel for our C9.25 at 2350mm focal length to give us 2 arc sec is 24 microns, when imaging deep sky objects.
Now, if we were to fit a focal reducer to the C9.25 giving an effective focal length of 1480mm, we would pull out our nail and refit the ruler at the 1500 mark which would indicate an optimum chip size of around 15 microns for deep sky objects with 2 arcsec seeing.
However, if we want to image planets and hi-res parts of the moon, the 2 arcsecs doesn’t cut it and if we look at the left hand “seeing” line we note that for Planetary imaging, we would like to resolve around 0.5 arcsecs per pixel. As our ruler pivots so the left hand edge comes to rest at 0.5, the right hand edge slides down to between 5 and 6 microns, which happens to be the pixel size of the ToUcam (5.6 microns).
I think that this is based on the Niquist sampling theory, which simplistically says that to resolve a single piece of data, to need to sample it at twice the resolution (or frequency for a signal). If your pixels are too small, then you are over sampling and maybe using say, 4 pixels when 2 would give you all the detail that is able to be resolved.
If you under sample by using a ccd with pixels that are too large for the fl, then stars begin to look like squares rather than nice round dots.
I think Zac has that model and he is a very experienced hi res imager so I would agree with you. I think the pixel size nicely matches his optimum focal length for hi res imaging of the Moon, and gives a larger area of coverage i.e. 1392 x 1040 so you are not forever creating mosaics using a smaller 640x480 chip.
So Dennis according to the chart, by increasing FL using barlows or eyepiece projection then the pixel size needs to increase.
Yes – I think that’s correct.
Let’s look at a simple white square of sides 10 cms x 10 cms. There is no detail inside the square; it is a 10 cm white square, that’s it.
Now, if we use a ccd chip that takes say, 9 pixels to represent this square, then mostly those pixels are wasted, as each pixel is only resolving a smaller white square but there is no additional detail at that smaller resolution. This means we are dedicating those 9 pixels to reading “no additional data”, and then we also have to read the pixels and transfer the “empty” data to our PC. This is wasteful of our precious pixels and bandwidth and is termed over sampling.
I think the Niquist theorem says something like x2 pixels of 10x5cms would be the optimum required to correctly gather the data for the 10cm white square and of course, we would then only be using 2 of our precious pixels, not 9, and the data readout and traffic would be more optimal too.
Now if we move in the other direction and say use a pixel 15cm x 15cm, then the 10cm x 10cm white square would easily fit inside this larger pixel but the pixel would display as a 15cmx15cm pixel on our display and so we would effectively “lose” the resolution of our 10cmx10cm square – we are now under sampling. That is, we are not seeing our 10cmx10cm white square, we are viewing a 15cmx15cm pixel with the square “hidden” inside it.
Anyhow, that’s my layman’s understanding. I’m sure the mathematically minded and scientifically trained readers will correct me where I am wrong.
Great thread and info Dennis. This is where many beginners in CCD imaging go wrong. You need to match your lens or telescope to pixel size depending on what you want to image.
Then once you've got that calculated, try to match it to your location i.e. seeing conditions. There are few locations that you will be able to reach .05 to .50 arcsec in seeing. This will limit the planetary image resolution.
Lots of valuable information in this thread - thanks Dennis for your excellent contributions.
I'd love to get a SkyNyx 2-1. Maybe someday in the future. Will certainly make lunar imaging much easier, not having to do mosaics all the time. But I'll need a faster laptop I'd reckon, even to be able to get 15fps it would be pushing it on this dinosaur.
It's worth keeping an eye on the cameras being developed by Paolo Lazarotti, too, as he intends on marketing both 640x480 and larger format versions, to be in direct competition with the DMK and LU cameras.
If anyone is interested in getting a DMK, it might be worth contacting Matt at Telescopes & Astronomy, as I believe his prices may have dropped from his website list price of $695 for the DMK-21AF04.
To calculate pixel resolutions use this formula
(205 x Pixel size in microns) / Telescope focal length in mm.
Ideal resolution is between 2.5 arcseconds / pixel to 5, depending on seeing conditions.
Dennis is right about over sampling. And other disadvantage of over sampling is that exposure time has to increase. The amount of the light coming thru telescope will be spread over more pixels so it will take longer to fill the wells of those pixels. I think that effect will be same as if you increase f ratio of telescope.
![[1ponders]'s Avatar](../vbiis/customavatars/avatar45_9.gif){kind=avatar}
