| Bruker Nonius KappaCCD & KappaCCD2000 |
|
 |
|
X-rays hit the detector through a beryllium window and are transformed into visible light by a phosphor. A fibre optic taper reduces the image to the size of the CCD, which converts the light into electrons. This signal is read out, amplified in the detector, digitized in the controller, and send via a special PC board directly into the PC memory. The CCD is cooled by a Peltier element stack, which is powered by a power unit. The 'warm' side of the peltier stack is cooled by a water-glycol mixture of 0°C, supplied by a closed circuit cooler.
The 0.5 mm Beryllium window keeps the detector on vacuum and at the same time isolates for light and for the cold front of the taper. It absorbs part of the X-rays. In the worst case some 15% of the Cu photons are absorbed and some 3% of the Mo photons. These absorption ratios are based on the maximum permitted contamination of other elements. The typical absorption will be less.
Gd2O2S (Terbium doped) is the common used phosphor for this purpose, because it is the most efficient in light photon generation. It is also highly absorbing, so it must be kept thin. On the external side of the phosphor is a thin reflector which reflects the light back to the taper which would have otherwise been lost.
A 25 mg/cm² Gd2O2S is used for Mo radiation. It absorbs 75% of the radiation. This thickness is optimized considering the increased absorption of a thicker layer, which results in a higher point spread and the increased light absorption inside a thicker phosphor. For Cu radiation a 10 mg/cm² Gd2O2S layer is used . This absorbs 90% of the radiation.
The phosphor does not seem to suffer from radiation damage. We have no figures that indicate this, but experience with a similar phosphor and our previous FAST detectors did not show any degradation in the measuring area.
The fibre optic taper consists of glass fibres which transport light by total reflection against the walls of the fibre. In the taper the diameter of the glass fibre decreases from input to output, resulting in a reduced image, but also in extra light loss, which is approximately proportional to the square of the reduction ratio. The fibres are approx. 15 µm diameter at the input, much smaller than the pixel size.
For the KappaCCD the taper ratio is 2.45 : 1. This is the optimum between input size and detector gain. The taper has EMA (Extra Mural Absorption): extra black glass (fibres or cladding) which absorbs the lost light escaping the fibres, which otherwise would result in a higher point spread.
For the KappaCCD2000 the taper ratio is 3.82 : 1. This may seem large, but taking into account the low detector noise this is optimal for the input size (95mm) and a detector gain greater then 1. This taper also has EMA.
A CCD consists of a raster of silicon capacitors (pixels). The light photons generates electrons, which are trapped in this pixel. After the measurement the contents of the pixels is shifted by clock signals to the output amplifier at one corner of the CCD. First all pixels are shifted upwards, the upper row into a read-out row. This row is shifted to the left, to the amplifier readout pixel. The amplifier converts the electrons in this pixel to a voltage, with is digitized with a 16 bit ADC converter. The read-out speed can be chosen 200.000 (default) or 430.000 Hz (= pixeor 430.000 Hz (= pixels/s) with a small increase of read-out noise.
The CCD of the 90mm camera has 1242 (horizontal) x 1152 (vertical) pixels.
After each horizontal row there are (binned) 2 pixels which are insensitive
for light (for measuring the dark current) and 2 non-existing pixels (for
measuring the ADC zero). These could be used in future for control and compensation.
The CCD of the 135mm camera has 1344 (horizontal) x 1300 (vertical) pixels. After each horizontal row there are 4 non-existing pixels (for measuring the ADC zero). These can be used for control and compensation.
The CCD's used in the Bruker Nonius cameras are manufactured by EEV. They were chosen because of their outstanding quality. The special MPP (Multi Phase Pinned) process results in a low dark current, 10 times lower than some other CCD's. This also prevents the need to cool to extreme low temperatures. The CCD is flat, which makes it possible to bond it directly to the taper. For some other CCD's the taper much be ground to the non-flat surface of the CCD to assure a uniform layer of adhesive.
The type of CCD is a full frame scientific grade 1. In a full frame 100% of the area is used for light collection, in contrary to video CCD's, where some area is used for transportation. Scientific grade means high quality. Above grade 1 there is only grade 0. In this application the need for defect free chips does not warrant the much higher price.
The 90mm CCD is normally read out in binned mode. The contents of 2 x 2 pixels is added into one pixel inside the CCD. This is done by shifting the rows twice in the readout row and shifting the readout row twice into the readout pixel.
The advantage is a four times faster readout speed, smaller file size and half the noise because a binned pixel is read out only once. The disadvantage is the larger pixel size, but this is still smaller than the point spread of the complete detector, determined by the phosphor. The CCD pixel size is 22.5 x 22.5 µm which gives at the input a pixel of 110 x 110 µm (with 2 x 2 binning).
The 135mm CCD is normally read out in unbinned mode. The CCD pixel size is 20 x 20 µm which gives at the input a pixel of 76 x 76 µm (in unbinned mode).
The glass of the taper absorbs all radiation, so no damage is possible to the CCD from illumination by the unshielded direct beam.
The well depth is the maximum number of electrons which can be stored in a pixel, typically greater than 400,000. To prevent overflow while binning the well depth of the readout row is twice and the readout pixel four times that size. In practice however this is not used. A binned pixel has four times the well depth of a single pixel, but this is much larger than needed. The gain should be four times lower, which degrades the measurements of low intensities. That's why the gain is adjusted for the output signal to be maximum with one pixel well depth.
The CCD uses 100% of the CCD area for detection, which leaves no space for anti blooming. An overflow will result in spill over to adjacent pixels hence "blooming". For crystallographic measurements it is better to measure the spilled over electrons, even in a wider spot, than loosing the signal, as anti blooming does. Normally overflows will be remeasured.
The CCD and taper are in vacuum, to prevent condensation and to lower the heat flow to the taper.
The vacuum will degrade in time by degassing of components and small leaks. A bad vacuum does not influence the quality of the measurement nor the detector. It is noticeable by the cooling, which doesn't reach the low temperature anymore and a serviceman can check it by measuring the cooling power. Because of the special seals we expect this will take at least a year. For revacuum the detector is equipped with a valve and vacuum connection. A normal revacuum pump (30 mTorr, 0.04 mbar) is sufficient.
The CCD is cooled with a multi-stage peltier element. The temperature is measured near the CCD and controlled to a stable value of -60±0.05°C. The warm side of the peltier element is cooled with a water-ethylene glycol mixture of 0°C, for which a liquid cooler is supplied. This temperature does not need to be accurate because of the temperature regulation of the CCD. It takes less than an hour to cool down or warm up. Higher speed temperature transitions are not recommended by the CCD manufacturer.
All CCD detectors use a 16 bit ADC converter. The effective dynamic range is in fact higher by:
The disadvantage of a full frame CCD is that it cannot be electronically shut on or off. Also before and after measuring and during readout it collects light. So the exact times must be made with the X-ray shutter. For accurate timing a second light and fast shutter is provided. The main shutter still is in function and provides the X-ray safety. This will also prolong the lifetime of the shutter, because it will remain open between successive integrations. The small shutter is not submitted to heavy radiation, which normally wears a shuttch normally wears a shutter.
The Detector Quantum Efficiency is a number to express the measuring quality of a detector. It is the ratio between the Signal/Noise of the input and output:
DQE = ( S/Nout / S/Nin )²
A practical way to use this figure is: T = Tideal / DQE, for instance with a DQE = 0.5 one must integrate twice as long to get the same accuracy as an ideal detector (with DQE = 1).
As seen from the formula, degradation of the DQE can be caused by adding detector noise or by weakening the signal, for instance with a lower absorbency of the phosphor.
The DQE is the best way to see the quality of the detector, but also the less used and understood, and the most misused and wrongly measured. This makes it almost impossible to compare the figures.
The point spread of the detector is mainly caused by the phosphor. It is an important parameter because it determines the minimum distance between two reflections, and so the minimum detector-sample distance (DX) one can use. A short DX gives a high spherical angle the detector intercepts. It also determines the number of pixels for a reflection and so the amount of noise.
There are several way's to define and measure point spread, but for diffraction the best way is to measure with a pinhole and measure the width. Mostly the FWHM (Full Width Half Maximum) is given: a very nice small value, but not so practical because a much larger area is used by the software to integrate the reflection. That's why since several years the FW1%M is used: Full Width at 1% Maximum. This also takes into account the long tails and experience showed this is roughly the size the integrating software uses.
The dynamic range is the maximum signal divided by the noise. It is limited by well depth and noise. This is a theoretical value, because it is calculated with one pixel while a reflection is integrated over several pixels with lower intensities then the peak pixel. It must also be compared with the dynamic range of the input signal: for instance when the measured signal is low it will never be able to reach the detector maximum.
In spite of all confusing values, it shows that the dynamic gain of CCD detectors is excellent for crystallographic measurements.
The nonuniformity is mainly caused by the taper. A well know type is the chicken wire: the borders between bundles of fibres. Also fibres can be broken or melted. The nonuniformity is fixed and will not change in time. For systems used with Mo-radiation it is measured in the factory with a X-ray flood field from a Yttrium sample. For systems used with Cu-radiation an amorphous Fe-sample is used. From these images the software calculates the calibration data, used to compensate the measured images.
The distortion is caused by the taper and will not change in time. It is measured during production of the detector with a mask with holes. From this image the software calculates the distortion and compensates for it. The distortion is low, and can hardly be seen in the image.
A special type of distortion are the shears: a sudden shift, mostly at the border between two fibre bundles. Because of the high quality of the taper these are smaller than 0.5 a binned pixel.
The dark is lowered to the very low value of max. 0.02 e/s.pix (unbinned pixel) by the combination of cooling and the use of a CCD in MPP (Multi Phase Pinned) mode. The EEV process, called AIMO (Advanced Inverted Mode Operation), makes extremely good use of this mode, compared to other CCD's. Because of the low current the resultant noise is very low. The temperature is stabilized at -60°C (±0.05°C), so the dark current will be stable.
Although the average dark current is very low, there may be some pixels with a higher dark current, which are called 'warm' pixels. This is dependent on the CCD quality and is an effect in CCD's of all suppliers. Sometimes they are arranged in horizontal lines. Warm pixels are compensated for by dark current subtraction.
Zingers are high intensities which appear randomly. Although it sometimes look dramatic in the image, they have only minor influence on the measurement. Nevertheless they are removed by making double images, in which they can be detected by their random nature.
There are two causes:
|
Muons: positive electrons, generated by cosmic protons in the upper layer of the atmosphere. They ionize while they traverse the silicon of the CCD. They can be distinguished by their size (approx. 2000 electrons) and sharpness (only in a few pixels, the adjacent pixels not higher than the background). When a muon traverses the CCD non-perpendicularly, they show up as stripes. It is a common effect in all detectors of this type (also in image plates). The count rate is approximate 0.007 cnt/s.cm². The example shown at the left is an exceptional one. |
The main noise of the detector is added to the signal during read-out of the pixel in the CCD and is independent of integration time. The easiest way is to compare it with the signal, and express it in X-ray photons. Most detectors claim to have the famous 1 photon ('photon limited') noise. Of course this is a theoretical value, because the noise of just one pixel is given, while a reflection is integrated over several pixels. So a 'photon limited' detector in practice will never be able to distinguish one photon and will always add more than one photon noise to the signal. Also when using more pixels (non-binning, larger point spread) the total integrated noise will be higher.
The noise is higher with higher read-out speed, so must be specified with the default read-out speed.
Another source of noise can be the dark current. For instance for a 60 s measurement a (binned) pixel will collect up to 60s x 0.02 e/s.pix x 4(binned) = 5 electrons. The noise is square root: 2.2 e, which can be disregarded when comapring with the read noise.
The noise provided by the detector must be compared with the input noise of the signal. With high signals the photon noise (square root of the number of X-ray photons) dominates. With low signals the X-ray background is important. When measuring small molecules the background is low and the detector noise will dominate. When measuring proteins the X-ray background is high, caused by scattered radiation in air, glass capillary (if present) and water in the protein (using the more absorbing Copper radiation) and long integration times. So for a protein detector the detector noise is less dominating.
To the Nonius home