Detectors / Detector Arrays
The detector is a key element in the signal chain, influencing the sensitivity including spectral range and resolution as well as signal quality and measurement speed. It is an intrinsic part of a modern detector array based spectrometer. For operation please see Electronics.
The advantage of an arrayed detector over the single-element detector (SED) is the ability to capture a complete spectrum in one shot and thus collect much more information in almost the same amount of time. By controlling the exposure time (integration time) it also allows to adapt easily to various light levels. However, the signal quality might be reduced over an SED and costs are typically higher.
There are 4 main classes of detector arrays which are used for spectroscopy purposes:
- Si PDA – 190 to 1100 nm, high S/N, low sensitivity
- Si CCD – 190 (400) to 1100 nm, high sensitivity, medium S/N
- Si CMOS – 190 (400) to 1100 nm, high sensitivity, low S/N
- InGaAs PDA – 850 to 1680 (2550) nm, high S/N, low sensitivity
While the InGaAs PDAs are the first choice if it comes to NIR detection, various classes of Si arrays are selectable. For high quality spectroscopy Si PDAs are used when decent illumination signal is available and good S/N is important.
The Si CCDs are used when high speed is required and/or signal levels are low. Back-thinned CCD have a high quantum efficiency across most of the sensitive range, while the alternative front-illuminated show by almost a factor of 2 lower QE and no sensitivity below 400 nm at all. The detector arrays dominate the readout process, the number of pixels (and line binning in case of 2D arrays) as well as the recommended clock frequency limit the readout speed.
Si CMOS arrays have their advantages when it comes to high pixel numbers (high resolution) and fast readout for high speed measurements.
Specialty InGaAs: different from the Si chips InGaAs demands a hybrid solution: InGaAs photosensitive material on a Si chip. This requires a bonding technique limiting the structuration. Lateral pixel sizes below 25 µm are not doable. In many cases, wire bonding leads to the use of two signal processing circuitries for alternating pixels, which again leads to different treatment of odd and even pixels, described as “odd-even effects”. Slightly different gains and dark current behaviors are the outcome.
Standard InGaAs has a cutoff at about 1.7 µm. To push this cutoff further, so called extended InGaAs has to be used. The further out the cutoff wavelength is pushed the higher the dark current rises. This leads to a shrinking dynamic range at longer exposure times thus limiting the useful measurement times to a few 10 ms in case of 2.55 µm quality. In combination with diffuse reflection measurements this is often too short. Back to top▲
For further reading click here
Number of pixels: depends on the type of chip. PDAs are available with up to 1024 pixels, mid to high end CCDs up to 2048 active pixels, CMOS arrays are available with up to about 4000 pixels. Back to top ▲
Pixel size is a structuration parameter. The pixel width (more accurate the pixel pitch) defines the the spectral bandwidth/pixel (pixel dispersion) and thus the resolution, of course in combination with the spectrograph dispersion (grating groove density and setup). It has to be matched to the entrance slit size, i.e. the slit width should not be smaller than at least 2x the pixel pitch, while pixel height and slit height should be the same (1:1 imaging, good imaging quality). If image is larger (higher) than pixel light hits the chip boundaries, is not detected (lost) and may generate false light. A large pixel height is important for a sensitive spectrometer. Back to top ▲
Clock frequency: is the frequency with which the internal multiplexer is operated switching from pixel to pixel during readout. The higher this frequency, the faster the array is read:
tRead = NPixel/fCLK,
Therefore, the higher the clock frequency the fast the sensor can be read. The highest possible frequency is specified by the chip manufacturer, however, it cannot be always reached without sacrificing the signal quality.
Well depth specifies the number of carriers (photoelectrons) potentially involved in forming the signal and is responsible for the signal quality to be achieved. The higher this number is the finer a signal can be rendered. Larger pixel sizes allow a larger well depth. PDAs have well depth up to 150Mio e-, CCDs up to about 4 Mio e-, and CMOS up to 26Mio e-. The achievable S/N is the SQR of the well depth. Back to top ▲
Example PDA: SQRT(150Mio e–) è S/N of 12,250
Dark current: each detector generates a signal even when dark. It increases linearly with the integration time and exponentially with the temperature (doubles about every 7°C). Often a temperature controller is used to reduce the temperature and also to just stabilize the temperature and thus the dark level. The absolute amount depends on chip internal features. Back to top▲
Linearity: is defined as the relationship between the incoming amount of light and the signal generated. Most chips provide values of 98% and better over most of the sensitivity range. The worst part is typically close to the saturation level. By adjusting the Analog-to-Digital Converter to the analog output signal the worst part can be cut off. Fortunately, non-linearities are constant, can be determined and corrected by software.
To determine the amount of non-linearity, an assumption about the “ideal” linear behavior has to be made. We found good results by assuming the detector sensitivity range between 25 and 50% of saturation level to be linear. Extrapolating of this region over the rest of the signal range functions as the reference. The deviation from this linear plot is taken as non-linearity, e.g. in percent of absolute signal value. Back to top▲