       |
 |
Back in 2003 I started taking planetary images with a TouCam Pro webcam on my Celestron 8 telescope. In this little essay I want to share my experiences on how to achieve the best possible results. This is a living document and will be updated in the future as my experiences grow. You are also invited! Please give me your feedback on this (
).
1.0 Why Philips TouCam Pro?
That's easy to answer. This camera has a "real" CCD-Sensor (SONY ICX098BQ), which allows a high resolution (640x480 Pixels non interpolated), high color reproductively and sensitivity of darkness. It is therefore ideally suitable for imaging of bright objects such as planets, the moon or our sun. It is also possible to use this camera for deep sky-imaging (integration time > 1/25 sec.). In order to do that, you have to modify the cam (which is rather complicated in my opinion) simply search the web, there are many sites with instructions to modify your cam. For planetary imaging this is not necessary.
2.0 Equipment
1) You need a PC with USB 1.0, the parallel port is not supported. Minimum system requirements are 32 MB RAM, Win 98, ME or 2000. With XP you need an additional program to run the included software (could be downloaded on the Philips homepage), the driver is integrated in the OS. To achieve a fluent dataflow I would recommend a minimum of 128 MB RAM and a processor with a minimum of 600 MHz. (You are lucky if you have a laptop).
2) Adequate disc space! (2 Gigabyte minimum!).
3) A telescope with motor drive in right ascension and (optional) in declination. You can try without motor, but you’ll loose patience soon because the object passes through the viewing field really fast.
4) An adapter to connect your cam to the scope. I use two adapters: The first is 1 ¼" in diameter and fits on the eyepiece holder, so I can use it with a 1 ¼" barlow lens (see picture below).
The second suits to my t-adapter, so I can connect it to the tele-extender and use it in eyepiece projection. It is important that the CCD sensor of the cam is in the center of the optical axis of the system, when it is mounted to the telescope. If it is shifted (even slightly), you'll later get problems to find the object and center it on the sensor.
5) An IR/UV-block filter. If you have lenses in the optical path, particularly none-apochromatic lenses (barlow, eyepiece or a Schmidt-Plate for SC-telescopes), not all wavelength will have the same focus. Especially the ultra violet (UV) and infrared (IR) wavelength will be a problem. With the human eye, we would not see this effect, because it is not sensitive in UV and IR, but the CCD is (see specification for the ICX098BQ). To achieve maximum sharpness, we must apply an UV/IR cut filter in front of the sensor (see image above).
If you want to learn more about the transmission-curves of the IR/UV block filters, please visit this web site for more information (Baader).
3.0 The binning modes of the ToUCam Pro
The full resolution is 640x480 or 355x288 pixels. If you observe planets and need high magnification use one of this modes. 640x480 covers every single pixel of the chip whereas the 355x288 mode only covers the central part. I would recommend the 355x288, because you get less dataflow and this reduces the number of artifacts that occur, when the images of an AVI video are compressed during recording.
The 320x240 mode also covers the whole chip size, but 2x2 pixels are averaged to one pixel. This is called 2x2 binning. The result is a better signal to noise (S/N) ratio. The 240x176 and 176x144 modes cover central sections of the 320x240 image. The 160x120 mode is 3x3 binning.
4.0 The angular field of view
For planetary imaging it is important to know, what part of the sky is captured by the CCD. To calculate the angular field of view of one pixel, use this equation:
Alpha = Angular field of view per pixel [arc sec / px]
d(Pixel) = Edge length of one pixel [micron]. This is very important: CCD sensors with Bayer-Matrix have a bigger pixel size as declared by most dealers. It seems that between the effective pixels there is a none sensitive gap, which must be considered for the calculations. For the CCD ICX098BQ the pixel size is approx. 7.3µm and NOT 5.6µm.
F = Focal length [mm]
Example: We want to observe Jupiter with TouCam Pro (7.3 microns, see the next chapter why 7.3 microns) and 2030mm focal length (C8 telescope):
Alpha = 206*7.3 / 2030
Alpha = 0.74"
The number of effective pixels of a Philips TouCam Pro web cam is 640x480 px, representing 0.13°x0.099° of the sky at this particular focal length.
We can also apply this formula to calculate the focal length required to achieve a favored angular field of view:
5.0 The recommended focal length (Sampling)
The theoretical angular resolution of a telescope is depending on the aperture of the telescope and the wavelength of light we are observing at:
Alpha = Theoretical angular resolution in [arc sec.]
Lambda = The wavelength of light [mm] (1 nm = 10-6 mm)
D = The diameter of the lens or mirror of your telescope [mm]
For a typical calculation we choose a wavelength of 550nm. This is a good choice for visual observations because our eyes are most sensitive at this wavelength (green light).
If we choose 550 nm (0.00055 mm), this brings us to the famous formula:
(Rayleigh Criterion)In order to achieve the best possible result for high resolution CCD imaging, matching focal ratio in connection to the pixel size is of essential importance, this is called "sampling". For planetary imaging it is recommended to sample the image at a rate close to the ideal Nyquist criterion. The Nyquist criterion describes that the sampling frequency must be greater than twice the band width of the input signal in order to be able to reconstruct the original signal perfectly. For our aim this means that the wavelength of a signal should be covered by two pixels.
To calculate the desired focal length for a given wavelength and equipment we can apply the following formula:
F = Desired focal length [mm]
d (Pixel) = Size of one pixel, incl. the gap between two pixels [micron].This is very important: CCD sensors with Bayer-Matrix have a bigger pixel size as declared by most dealers. It seems that between the effective pixels there is a none sensitive gap, which must be considered for the calculations. For the CCD ICX098BQ the pixel size is approx. 7.3µm and NOT 5.6µm.
D = Aperture of the telescope [mm]
Lambda = Wavelength [nm] (I choose 550nm in accordance to Rayleigh)
Example:
1) Philips ToUCam Pro web cam with 7.3 microns for each pixel
2) Celestron 8 with 203 mm aperture
3) Wavelength = 550 nm
F = 2000*7.3*203 / 550
F = 5390 mm (Focal ratio of F/26.5)
If we strong oversample an image, we need longer integration times without a gain of detail, we "waste" integration time. If we undersample an image, we waste resolution which also is of an disadvantage. So for planetary imaging we want to sample close to the Nyquist criteria or slightly oversample the raw images.
To better understand the term "undersampled" and "oversampled", please review the following explanation:
The white point in the first image represents the smallest visible detail (diffraction disc) that your telescope can resolve. If the angular resolution of the CCD sensor is better then the theoretical angular resolution of the telescope, the smallest visible detail is only represented by one pixel (Image 2). In this case we experience a maximum utilization of the sensitivity of the CCD but we suffer a loss of information, the image is called undersampled:
 |
 |
| Img. 1 | Img. 2 |
But on the other hand we don't want too many pixels representing the smallest visible detail. Images sampled with more than 2 pixels across the diffraction disc (Image 3) are called oversampled. In this case the angular resolution of the telescope is far better then the angular resolution of the CCD and the sensitivity of the CCD is not optimal utilized (longer integration times needed). As an example, the star in image 4 is heavily over-sampled:
 |
 |
| Img. 3 | Img. 4 |
In image 5 the angular resolution of the telescope matches the angular resolution of the CCD, the diffraction disc is two pixels in diameter. As a result (Image 6), we have a good or critical sampling:
 |
 |
| Img. 5 | Img. 6 |
6.0 The effective focal length
In chapter 5.0 we have calculated a focal length of 5390 mm required to fulfill the Nyquist criterion for a C8 telescope. Now, one could say that we choose a 2.5x barlow lens to achieve this (2.5x 2030 mm = 5075 mm). Well, that is true but before you go and buy "any" barlow, you must know that some barlow-lenses have varying extension-factors. On some barlows, the extension factor is depending on the distance between the lens top surface to the surface of the CCD. With increasing distance, the extension factor is also increasing and maybe your 2x barlow becomes a 4x barlow without your knowledge.
As an example I found a chart on the TeleVue web site. On this chart we can see, that the 2x and 2.5x Powermates don't change their extension factor significantly, but the 5x Powermate can easily turn into a 7.5x barlow.
So the true extension factor of a barlow lens must be calculated by yourself. In order to calculate the true focal length, it is necessary to image an object which apparent diameter is known from the ephemerides (e.g. Saturn's planetary disc at the equator) and then carefully measure the diameter of the planets disc on the CCD sensor. If you know, how many pixels the object is in diameter, then you can re-calculate the true focal length of your imaging system.
We can re-calculate the effective focal length with the following procedure:
I imaged Saturn on March 13, 2009 with a Celestron C11 (focal length = 2800mm) and a TeleVue 2.5x barlow lens, using a monochromatic DMK 21AF04.AS of "The Imaging Source" (here the size of one pixel indeed is 5.6 microns). The amplification factor of the Televue barlow is 2.5x, but if you have checked the link above (TeleVue web site), you know that the amplification factor is slightly decreasing with growing distance between the surface of the CCD sensor and the top lens surface of the barlow lens. So what is the "effective focal length" of my setup?
At first we must check the ephemerides for the apparent angular size of Saturn's disc (at the equator) for this particular date. Result: On March 13, 2009 Saturn's disc was 19.7 arc sec in diameter.
Now I want to measure the diameter of Saturn's disc on my CCD image, see the following picture:
With Photoshop, I measured the coordinates of the mark 1 and mark 2. Photoshop measured for mark 1: x1=81, y1=112 and for mark 2: x2=176 and y2=61. Now we must calculate the length of a and b using the following formula:
When I apply these formulas I get the result for length a = 95 and length b = 51. In order to calculate the length of "c", we can see this as a right angled triangle and calculate the length of "c" with Pythagoras:
When I apply this formula, I get the result that the length of c = 109 pixels.
Now we know that the planets disc has an apparent size of 19.7 arc sec and this is represented by 109 pixels on the CCD sensor. Now I'm using the "rule of three" to calculate how many arc sec are covered by one single pixel.
= 19.7 arc sec / 109px = 0.1807 arc sec per px.
Now I'm ready to calculate the effective focal length by applying the formula previously shown (see chapter 4):
F = 206*5.6/0.1807
F = 6384 mm
So the true extension factor of the 2.5x TeleVue Powermate in fact was 2.28x for my setup, resulting in a focal ratio of F/22.8! (This is a good result, check chapter 5.0).
7.0 The planets rotation
This is very important, when you are observing planets with fast rotation period such as Jupiter, Saturn or Mars. If you are recording a sequence, you have to consider that the planet is "still in rotation while you are recording" and there is a risk, that fine details can get blurred after stacking if the duration of the recording was exceeding a critical time span. In this chapter we want to calculate the maximum permitted time span of a video record without noticeable motion of the planet.
When we observe a planet through the eyepiece, we see a round disc, but we know that in reality it is a globe with a linear rotation. When we observe details on the planets surface or in the atmosphere, we see that the motion of these details appears to us relatively slowly at the limb and it is of the fastest motion when it passes the central meridian. In particular, a surface-detail is seen in its fastest motion when it is passing the area, where the equator and the central meridian interfere.
The following calculations are considering this "worst case" scenario for a detail that is exactly in the area where the equator and central meridian interfere.
We can express the linear rotation period of a planet with the formula for the angular speed:
Omega = Angular speed
Phi = Angle of 360° for one complete revolution
T = Time to complete a rotation of 360°
For further calculations, we must convert the angle Phi = 360° into radians (RAD), that's easy because one revolution (360°) of the unit circle = 2Pi radians.
So the formula for the angular speed of a planets rotation period looks like this:
Now we have evaluated a formula for the angular speed. With this formula we can express how long it takes for the planet to rotate one complete revolution. For example we know that the planet Jupiter's rotation period is 9h50'30'' (System I), meaning the planet will turn 360° or 2Pi (RAD) in this time period.
For a linear rotation period, the angular rotation is constant. For our aim this doesn't help us much because a planet always appears to us at a different angular size and so the apparent velocity of a surface detail always appears to us differently (if a planet is close to us, we see more detail and so it seems to us to rotate faster). What we need to know is the tangential velocity.
Imagine you are right above the north pole of a planet and you see an object on the equator which is constantly moving with the rotation of the planet. The tangential velocity describes the path that has been covered by this object in a certain time frame. This is depending on the radius that this object has to the center of the circle. The bigger the radius, the longer is the path that has been covered (whereas the rotation period is constant).
In order to calculate the tangential velocity of an object which is passing the "critical area" (where the equator and the central meridian interfere), we must add the radius to that formula.
The formula for the tangential velocity is:
R = Radius of the apparent diameter of the planet [arc sec]
Now we must take into account the angular resolution of our our equipment (see chapter 4.0 for details). The angular resolution per pixel can be calculated with this formula:
If we want to know, how long it takes for the smallest visible detail to move about the value of one pixel on the CCD sensor (otherwise we will not detect a motion with our equipment), we just have to divide Alpha by v:
or in full detail:
t max = Maximum permitted time span of a video [s].
d(Pixel) = Edge length of one pixel [mm]
T = Rotation period of the planet [s]
F = Focal length [mm]
R = Radius of the apparent diameter of the planet [arc sec]
206265 = Factor to convert radians into arc sec
Following there is an example on how to calculate this effect on Jupiter:
Given information:
1) Date: 16. Feb. 2003
2) Diameter of Jupiter at this time is 45.1 arc sec (R = 22.55 arc sec)
3) Rotation period of System I (area between NEB and SEB): 9h 50' 30'' = 35430s
4) Philips ToUCam Pro with 7.3 microns = 0.0073 mm
5) Focal length: 5390 mm
Calculation:
t max = 206265*0.0073mm*35430s / 2Pi*5390mm*22.55arc sec
t max = 69.9s
As I mentioned earlier, this is a "worst case" calculation. So I think that in real live you can exceed this calculated "max. duration time" by the factor 2 and still have a good final image.
8.0 Choosing a red (R) or infrared (IR-pass) image as luminance layer
In order to achieve a better contrast, most amateur astronomers add to the Red, Green, Blue data a fourth channel, the "Luminance Layer". In the deep-sky business, this is mostly a high resolution 1x1 bin, long exposure clear filter image (sum or mean of multiple images). This adds to the picture the sharpness and contrast, whereas the R, G, B images deliver the color information.
When we want to image the planets, we use this technique to add a high resolution luminance image to the R, G, B data but the "L" image is a R, or IR image. Since most planets show interesting albedo features in the R or IR wavelength which we want to enhance.
But which filter is the one to choose? It depends on the conditions:
IR pass filter
Pro: (1) Bad Seeing has less impact to this wavelength.
Con's: (1) The CCD sensor is less sensitive in IR compared to the RGB light (60% and decreasing, see SONY ICX098BQ, page 8). It's also worth to consider that for a CCD sensor with Bayer-Matrix (as for the ToUCam), the "G" and "B" pixels are not sensitive for IR, only the "R" pixels are sensitive for this wavelength (the IR-pass filter turns off the G and B pixels). So the picture is getting "dark" and needs longer integration times. For the use of an IR-pass filter, the monochrome cameras are more appropriate. (2) The theoretical angular resolution of your telescope is getting lower with increasing wavelength. Example: An 8'' SCT has an angular resolution of 0.68'' at 550nm and 0.97'' at 780nm. So very fine details will get lost in IR compared to R.
R filter
Pro: (1) Better angular resolution and lower integration times than IR, but
Con: More depending on good seeing.
Solution: If you have good seeing, choose the red color filter instead of the IR, if the seeing is bad and you are using a monochrome camera, you will gain better results with the IR pass filter.
However, if our interest lies in details of the atmosphere, maybe blue or green filters are appropriate. On Mars, a blue filter greatly intensifies ice-clouds or the polar caps and a green filter can intensify dust storms in the atmosphere. Some amateur astronomers also apply a green filter as a luminance layer in order to reveal the smallest details on the Moon.
For more information, please visit the web site of Christophe Pellier and if you have some more time, check out his planetary images - they are really brilliant
.9.0 Now it's time to get started
It is very important that the temperature of your telescope is equal to the surrounding temperature. So let your scope at least 30 minutes to cool down. Orient the mounting to the true north as close as possible, otherwise the object is permanently drifting out of the field of view. The next step is to check collimation, this is a very important step - you won't gain high resolution, if the telescope is misaligned.
Lets assume we want to observe Jupiter. Remove the star diagonal, pop in an eyepiece of high magnification (10–15 mm), center and focus the planet carefully. Now replace the eyepiece with one of lower magnification (I use 25 mm) and assemble the webcam behind this configuration.
Now start the driver of the webcam and disable the automatic functions. Change the settings to maximal brightness, integration time 1/25 sec. and 10 – 15 frames per second. Change the resolution to 320x240. You see nothing on the screen? Simply repeat the procedure if it’s still not working, remove the eyepiece and repeat the procedure again. If you see a bright blob on the screen, congratulations! Now try to center the blob carefully with your motor drive. Make yourself familiar with the controls for the motor drive, because one slight push and the planet could drift off the field of view dramatically.
When turning the focus knob of a Schmidt-Cassegrain-Telescope the object is shifting. So be patient when turning the focus knob and always re-center the object in short intervals. It is recommended that it is the best to turn the focus knob counter-clockwise for focusing (then the mirror is pushed and is more stable in this position). When you focused too far, simply go back by turning the focus knob clockwise and then again counter-clockwise to focus.
When the planet is in the center and you think it is relatively sharp, change the settings in the control box. For the TouCam the best adjustment is:
- Brightness: 50%
- Gamma = 0 (or near to 0)
- Saturation = 80-90 %
- Gain = 50%
Change the settings for the integration time so that details of the planet become visible. (With a C8 and a 25 mm EP 1/25 and 1/33 sec. works fine). Only use integration time and gain for changes. Now again try to focus the image as sharp as possible. (This procedure sometimes takes me 15 to 20 minutes).
Now if you have a sharp image showing some details, we can start the recording. Change the image size to 355x288 pixels (check focus again!). Select 5 or 10 fps - and start! It is recommended to collect as much data as possible, meaning at least 500 frames per sequence, this means a lot of data.
Take several sequences with different settings, always check focus before starting the recording. Seeing sometimes changes within minutes and so the sharpness of the image often changes very quickly.
10.0 Some remarks about image processing
This is all half of the work. Back indoors, the sequences must be processed. The main steps consist in dividing up the sequence into individual frames. Then sorting the best images, registering and stacking them to one final image. If you want to save the file, choose a 16-bit TIF format, don't save as 8-bit JPG or BMP, otherwise you will loose the dynamics in the image.
I use Registax for this operations, all these functions are carried out automatically by this software. Nowadays I heard that Avistack should produce better results on the Moon, but have not yet found the time to proof this.
 |
 |
| Single raw image. | Processed image: 200 frames stacked, unsharp mask |
If you want to know more about image processing techniques, I can highly recommend Paul Haese's and Mike Salway's web site. You can find on both sites a very detailed tutorial how to proceed.
That's all, thanks for reading.
-Frank Brandl-
© 2003-2009 Frank Brandl / Disclaimer