Jump to content

Banner.jpg.b89429c566825f6ab32bcafbada449c9.jpg

Choosing a CCD Camera


narrowbandpaul

Recommended Posts

Choosing a CCD Camera

Buying a CCD camera is not a particularly easy job. Several reasons exist for this. When browsing different manufacturers websites, a vast array of number and graphs can be thrown at you, which can be very confusing. Also, spending several thousand pounds, or the equivalent, is a task you don’t want to make a mistake on. In this we will go through some important factors to consider.

The most important quantity is the Signal to Noise ratio, SNR or S/N. You want lots of Signal and low Noise. So we can break down our analysis in to two parts. What quantities maximise S and what minimise N.

Signal

Quantum Efficiency: Or QE for short. This number is the ratio of detected photons (light particles) to the total number. An ideal detector will detect all photons that hit it, and thus 1 (100%) is the ultimate limit. A high QE will help maximise S. However a high QE sensor can be expensive. Most amateur CCD’s are whats known as Frontside Illuminated. This is the standard technology, whereby light passes through silicon electrodes before interacting with the photosensitive region. The downside of this is a poor UV/Blue response (<400nm), however, most imaging is not conducted in this region, so a poor UV response is not a major deal. A typical front illumination CCD will have a peak QE of between 0.4-0.8 depending on the exact technology used to make the CCD. The range of response will be useful between 400nm-900nm typically.

If one is imaging emission nebula, then a good QE (0.5 is a good guide) is important at the following wavelengths.....H_beta=486nm O[iII]=500.7nm Ha=656nm S[iI] 673nm. This range spans blue/green to deep red. If one is imaging galaxies or stars, then a wide range of response is also important

A typical CCD will have a useful response between 400-900nm.

There are two main types of CCD architecture employed today. Interline Transfer and Full Frame CCDs

Interline transfer CCD’s have a vertical shift register between pixels, and this structure reduces the fill factor (the percentage area of the pixel useful for detecting photons). This reduces the QE, and the peak is around 0.5-0.6. The response at 656nm is around 30%. For this reason Interline Transfer CCD’s are cheaper, so are definitely worth considering. An example of a popular Interline CCD is the KAI 11002 or 4022

Full Frame CCD’s have a 100% fill factor, and offer the highest QE of any frontside device. The QE for the full frame KAF3200ME is 0.85.

To summarise: A high QE will help maximise S. The range of which the CCD is responsive is a good thing for galaxies. A good QE at 656nm (Ha) is important. Interline CCD’s have a lower QE, and cost than a Full Frame sensor.

Thats pretty much the only way to maximise S.

Noise

We want to keep the noise to a minimum.

There is one noise source we cannot minimise through our choice of CCD or camera manufacturer.....source shot noise. This is a noise caused by the statistics of photon counting.

However, the ideal noise performance of a CCD image, is shot noise limited. If your CCD gives you shot noise limited images thats a good thing.

Read Noise This is the noise introduced when reading the CCD out. Camera manufacturers have some control over its value, by carefully optimising the CCD and readout electronics. Therefore, different manufacturers should be looked at when purchasing a CCD. Read noise is the lowest signal you can detect. In some sense, Read Noise is a good measure of low light performance. A value of 4 - 9e- is about typical. If you want to generate a good S/N with low signals, a low read noise is the way to do it.

Summary: Low Read Noise is good

Shot Noise and Fixed Pattern Noise shot noise from the source is unavoidable, but dark current is an avoidable form of shot noise. More later. Fixed pattern noise is not usually specified by manufacturers, and can be removed through the process of flat fielding. A value of around 0.5% is typical. Fixed pattern noise limits S/N, so it definitely needs removed. It is a problem encountered with high signal levels like planetary imaging or deep exposure of bright objects like M42 or M31.

Dark Current.

Even when in the dark, the silicon lattice will emit electrons randomly. This is called dark current. It is a quantity that is very temperature dependent. The colder a CCD operates the lower the dark current. Typically dark current (expressed usually in e-/pix/sec) halves for each 6C drop in temp.

Most CCDs are offered with at least a ΔT of 30C...ie, a cooling of 30C from ambient. Several manufacturers offer cooling of ΔT=70C. This certainly inhibits the dark current. Dark current adds shot noise to an image, as well as dark fixed pattern noise. The more dark electrons are released during the exposure, the greater these quantities will be be.

A value of <0.01e/pix/sec is a good number for imaging. 3e- of noise would be produced for a 15min exposure.

An important feature also is set point cooling. This is the ability to operate the CCD at a fixed temperature irrespective of changing ambient temp. Dark subtraction using dark frames works best when the temp of the darks matches the temp of the lights. With set point, you can take darks during the day while at work etc, as long as the same temp can be reached as the night-time images.

Summary: A good amount of cooling is advisable. Most CCD’s

feature low dark currents when operated in the region of -30C

Residual Bulk Image

Some CCD types are prone to Residual Bulk Image, RBI. Red and Near IR photons interact near the substrate-epitaxial interface, where the generated electrons are prone to being trapped. These traps can become filled or partially filled during an exposure. Then when taking dark frames, these electrons become untrapped via their thermal energy, and may be collected by the pixels, and registered as additional, unwanted dark noise. These traps are also temperature sensitive, and so cooling the sensor will limit the amount that get released per exposure.

The method used to control RBI is to flood the sensor with NIR light, flush the CCD before exposure, then take an image as normal. The NIR flood fills these traps, and stops them collecting signal electrons during the integration. With the sensor run cold, these electrons have a typical ‘trap lifetime’ of several hours (maybe more), so very few leak out during the acquisition of new light frames or dark frames.

You know you have RBI if there is a pattern of signal in your darks that closely matches your light frames....eg, the moon can be seen in your darks!

Some sensors exhibit RBI. These include the KAF full frame range, and those featuring LAB (lateral antiblooming).... for example the KAF3200, KAF 6303, KAF16803. Some manufacturers don’t currently support this flood flush integrate technique, so thats worth checking.

Summary: If a sensor is liable to RBI, deep cooling and flood flush integrate techniques are very advantageous

We have discussed aspects to maximise signal generation (QE) and minimise noise contributions (low Read noise, deep cooling, RBI management). A camera incorporating all these features will deliver a high S/N. However, there are other quantities specified that are important

Other Factors

Full well:

The number of electrons a pixel can hold before blooming or non-linearity sets in is called Full Well Capacity. Full frame sensors, the KAF line from Kodak, feature higher full well than a similar sized pixel in an interline sensor KAI line. The size of a pixel primarily affects the full well capacity. Smaller pixels hold less. A 5μm pixel can hold around 25,000e- (KAF8300) whereas a 12μm hold around 110,000e- (KAF9000).

When the sensor is limited only by shot noise, the signal to noise ratio, S/N is given by the square root of signal. So a higher full well camera is capable of deliver higher S/N. However, the typical astroimage of a deep sky object will not be anywhere near full well.

A large full well would be important when imaging very bright objects, like the moon or planets, but is of limited use in deepsky imaging. What is very important however, is the dynamic range.

Dynamic Range:

This is the ratio of maximum resolvable signal to minimum resolvable signal. This is equal to Full Well/Read Noise. The dynamic range is the CCD’s ability to record both bright and faint signals. If two parts of an object have an intensity ratio of greater than the dynamic range, then the bright part will saturate the pixel before the faint area has been detected. Many objects exhibit a wide dynamic range, for example the core of the Andromeda galaxy is many times brighter than that of the outer spiral arms.

Given a typical well depth of 40,000e- and a typical 8e- read noise, a value of 5000 levels is typical. CCD’s can be found with a DR of 2000-10,000 however. It is conventional to express the DR in decibel, dB, units. Use the formula DR(dB)= 20*log(FW/read noise)

Summary: A wide dynamic range is advantageous for astroimaging, and it is probably more important than full well capacity. A good dynamic range can be achieved through a decent full well and lower read noise. A lower read noise also gives good low light performance.

Antiblooming:

When a CCD pixel collects more charge than it can hold, the electrons spill out and create a blooming spike, which can cover important parts off an object and don’t look particularly nice. CCD sensor manufacturers can incorporate structures to channel excess charge to ground, thus avoiding charge spill over. However the incorporation of these structures reduces QE, fill factor and full well capacity. However an antiblooming CCD can take very deep images without blooming, so this feature can be an advantage. The spikes can be repaired in photoshop, but the image under the spike is lost forever, and the repairing of spikes can take time.

Using narrowband filters, star light is heavily attenuated, whilst signal from the nebula is not. Blooming is far less likely to occur and so a non-antiblooming camera may be superior for this application. However, blooming may still occur so individual preferences will be the overriding factor on choice.

Colour vs Mono:

CCD’s are inherently monochrome devices. They don’t record colour. To make a colour image with a mono camera requires filters to be placed in front of the sensor, and the image then assigned a particular colour in post processing. For a typical RGB image 3 exposures are required per colour image. This can be time consuming. So CCD sensor manufacturers can lay an array of RGB filters over the pixels themselves. In a grid of 4 pixels, the typical layout is RGGB. One pixel corresponds to one colour, so a colour array will 25% of the full resolution for B and R, and 50% resolution for G. A colour CCD will produce a colour picture on every exposure .

However the colour CCD has a lower QE than its mono equivalent, and is not as versatile, in that for narrowband imaging, and general deepsky imaging, a mono CCD is a better choice.

The choice is between having to take 3 sets of images to generate a colour image and QE loss with a colour CCD.

Pixel Size:

Pixel size affects the resolution of the image. At a given focal length, a smaller pixel with produce a higher resolution.

The resolution in arcseconds per pixel is given by:

resolution=206*(p(um))/(f(mm))

P is the pixel size in microns, and f is the focal length in mm. Small pixels, whilst delivering high resolution, will typically have a lower well depth, reducing dynamic range. They will also be more subject to atmospheric seeing. For most locations, a value of 1”/pix will be about the highest useable resolution. The focal length to provide this is f=206*p(um). If your imaging telescope has a higher focal length, you may want to consider a larger pixel.

A large pixel, will have better dynamic range, and full well, but a reduced resolution. It could be the case that the image looks quite pixellated. A value of 1.5-2”/pix is quite good, and the best of both large and small pixels can be achieved.

There is definitely a trade-off with regards to pixel size.

Sensor Size:

The size of the CCD only determines how much sky you image at one time. A large CCD might be required when working at very long focal lengths. A small CCD may be all thats required when imaging say with camera lens and small telescopes. The more area a CCD takes up on the Silicon wafer, the higher the cost. However, technology employed means that some small CCD’s are more expensive than large ones. An example would be the very sensitive KAF3200ME. It is much smaller than the KAI11002, yet the KAF3200 is considerably more expensive. Small isn’t always cheap!

Conclusion

There is always a trade off with picking a CCD for your needs. You need to figure out what you want. Some things may be important to you, others not so much. I would look at quantities like QE, read noise, level of cooling, RBI management, set point cooling, dynamic range.

I advise you scout around the various camera manufacturers, look at the parameters, and see if any vary from manufacturer to manufacturer. Read noise and cooling typically vary a bit between manufacturers.

Thanks for taking the time to read this.

Paul Kent

24/10/09

Choosing a CCD Camera.pdf

  • Like 41
  • Thanks 2
Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.