ING Banner
Home > Astronomy > ISIS > Fast Spectroscopy


Fast and faint-object spectroscopy with the QUCAMs

  1. Preliminaries
  2. The L3 CCD
  3. The QUCAMs on ISIS
  4. Observing with the QUCAMs

1. Preliminaries

Two E2V electron-multiplying, low-light-level (L3) CCD cameras are available for use on ISIS red and blue arm to perform fast or faint-target spectroscopy. The QUCAM2 and QUCAM3 cameras are nearly identical, so we will describe only QUCAM2 hereafter. The L3 detector is a 1k×1k pixel, 13 micron/pixel, full frame transfer device, that can be used in two multiplying modes: slow or fast. In the fast mode the readout noise (RON) is nearly zero. Exposure times shorter than one second are possible with almost no dead times (20 milliseconds per full transfer) as the system continuously exposes while reading the previous image. Its technical characteristics are summarized here. The detector has very good cosmetic characteristics, and excellent quantum efficiency. QUCAM3 will be available during semester 2008B so that simultaneous observations with both ISIS arms can be performed.

2. The L3 CCD


The L3 CCD used in QUCAM2 differs from a standard CCD in the sense that it has a frame transfer buffer allowing quick transfer of accumulated charge (the image) into temporary on-chip storage, and it possesses an extended readout register where charge is amplified, allowing very high gains that make the classical RON much less significant. Some important facts have to be taken into account to take full advantage of QUCAM2. A paper with more technical information and analysis of the noise sources can be found here. For some papers describing the use of L3 CCDs please follow this link, and for sample images and observing tips for L3 cameras see this link.


2.1 Frame transfer CCD

With this frame transfer CCD it is possible to drastically reduce the time lost to reading the chip out. Charge is first rapidly shifted into a storage buffer in about 20 milliseconds. While the charge resides in that buffer and is being read out, the exposure can just continue on the active CCD area. In this way observing efficiency is maximized. An exposure time of 0.229 s using a 100-pixel high CCD window over the full spectral length is achievable. Shorter exposure times are not possible as this is the time needed to read the buffer.

2.2 Electron multiplication and very low RON

The electron multiplication takes place in an extra series of stages added to the serial register before the charge is digitized. These stages are clocked with higher than normal voltages, which leads to an avalanche of several hundred or even thousands of electrons for each electron that is collected on the chip. This therefore dwarfs the normal RON produced by the amplifiers. The resulting effective RON is close to zero (0.028 e- in the fast readout mode).

However, the multiplication process introduces an additional noise source called "multiplication noise" (see Tulloch 2007). In the detector noise limited regime (low signal levels) this loss is more than compensated by the negligible RON of the system, but in the photon noise limited regime the multiplication noise reduces the signal-to-noise ratio of the observation by a factor of 21/2, which is equivalent to say that the detector loses a factor 2 in quantum efficiency.


2.3 Two observing modes

The camera has two amplifiers, a conventional and an L3. These are positioned at different corners of the chip so produce mirror image outputs. The L3 amplifier should be used where detector noise would be a limitation. For brighter objects (photon noise limited regime) it is better to select the conventional amplifier, whose behavior is very similar to one of the standard ISIS detectors. Amplifiers are selected using the 'rspeed slow' command for the conventional and 'rspeed fast' for the L3.

2.4 Problems: linearity and clock induced charges.

There are two special characteristics of L3 CCDs that deserve special attention:

1) Non-linear behaviour occurs at high count levels. The system is optimized for faint sources, so it is therefore a good idea to keep the exposure time low to always work in a linear regime (<50000 ADU, i.e. 430 electrons) while observing with the fast mode.

2) The high-gain amplification process generates occasional rogue electrons. These are the so-called "clock induced charges" (CIC). CIC are produced in all CCDs, but are only noticeable in a L3 CCD due to the electron multiplying stage. The effect is that several "bright" pixels, randomly distributed, appear in the image (see Fig. 1). The number of CIC electrons is independent of the exposure time.

specL35s
Figure 1. CIC events in a QUCAM2 spectrum, which show up as bright pixels randomly distributed all over the image.


3. The QUCAMs on ISIS


3.1 Spectral resolution and wavelength coverage

The spectrum is dispersed along the x-axis. The L3 CCD covers about 1/3-1/4 of the spectral range covered by the standard ISIS CCDs.


ISIS/QUCAM2 wavelength coverage and spectral resolution on the blue arm
Grating
Blaze (Å)
Dispersion (Å/mm)
Dispersion (Å/pix)
Total spectral range (Å)
Unvignetted range (1024 pix) (Å)
Slit-width for 54 mu at detector (arcsec)
Slit-width for 27 mu at detector (arcsec)
R158B
3600
120
1.56
1597
1597
0.8
0.4
R300B
4000
64
0.83
850
850
0.8
0.4
R600B
3900
33
0.43
440
440
0.9
0.45
R1200B
4000
17
0.22
225
225
1.1
0.55
H2400B
Holo
8
0.11
113
113
1.2
0.6


ISIS/QUCAM2 wavelength coverage and spectral resolution on the red arm
Grating
Blaze (Å)
Dispersion (Å/mm)
Dispersion (Å/pix)
Total spectral range (Å)
Unvignetted range (1024 pix) (Å)
Slit-width for 54 mu at detector (arcsec)
Slit-width for 27 mu at detector (arcsec)
R158R
6500
121
1.57
1608
1608
0.84
0.42
R316R
6500
62
0.81
829
829
0.88
0.44
R600R
7000
33
0.43
430
430
0.97
0.48
R1200R
7200
17
0.22
266
266
1.24
0.62


4. Observing with the QUCAMs

Observing with QUCAM2 is very similar to observations with normal CCDs (see the ISIS Cookbook). Nevertheless, some differences related only to the CCD operation are discussed below:


4.1 Settings
The spectral direction with QUCAM2 is parallel to the x-axis. Windowing in the y-axis makes readout faster. In order to have enough sky at both sides of a point-like source spectrum we use a 100-pixel high CCD window covering the whole range of the x-axis. In our tests we used:

SYS@taurus> window qucam2  1  "[1:1072,540:639] "

Windowing in the x-axis is NOT recommended since it will reduce the gain of the system.

Only y-axis binning (spatial direction) has been implemented. On-chip binning is not generally recommended for L3 observations except to reduce readout time. In a conventional CCD, binning offers a noise-less method of adding pixels. Since the RON of an L3 chip is close to zero, any binning can be accomplished post-readout with very little penalty. Post-readout binning can then be done in a flexible manner.

The L3 CCDs have two amplifiers. One is a conventional, low-noise amplifier with very similar characteristics to the EEV and RED+ ISIS cameras. This conventional amplifier can be selected by using

SYS@taurus> rspeed qucam2 slow

prior to image acquisition. This amplifier is preferred if the observations are to be photon noise rather than detector noise limited.

The second amplifier uses avalanche multiplication in an 'L3 register' to give a RON << 1 e-. It can be selected by using

SYS@taurus> rspeed qucam2 fast

prior to doing a run. It is preferred for observations that would otherwise be detector noise limited. The noise of this amplifier is sufficiently low to easily allow the detection of single photons and, as the object brightness falls, the image can be seen to break up into discrete spots, each corresponding to a single photo-electron.

Since the readout amplifiers are positioned on opposite sides of the CCD, when switching between amplifiers the observer will notice that the images become mirrored, i.e. flipped in the spectral direction.

4.2 Taking spectra of the target

Although all the usual UltraDAS commands can be used as with the other ISIS CCDs, some commands specific to QUCAM2 have been introduced in order to allow for better performance in the application of rapid spectroscopy. The new commands (currently available only on the QUCAMs) are as follow:

SYS@taurus> rmode qucam2 spec [n_reads]
SYS@taurus> rmode qucam2 simple
SYS@taurus> run, glance, flat, arc, etc.

To enable the rapid spectroscopy readout mode, the command 'rmode qucam2 spec [n_reads]' is used. For example, to set the camera up to perform 360 readouts:

SYS@taurus> rmode qucam2 spec 360

Once the readout mode is set, the normal run commands (run, glance, etc.) are used to perform exposures. The time specified in the run command is then used as the integration time for each readout in the readout sequence. For example:

SYS@taurus> run qucam2 5

performs a sequence of 5-s exposures (but see Sect. 4.3 below), with the number of exposures determined by the [n_reads] parameter used in the 'rmode qucam2 spec [n_reads]' command.

Please be aware that the first image of a sequence of exposures suffers from a signal drop as it appears to have an exposure time 6 per cent shorter than the rest. It must therefore be discarded.

The time taken to perform a rapid spectroscopy run will be approximately:

t = n_reads * (exposure_time + readout_time)

To disable the rapid spectroscopy readout mode, just set the readout mode back to simple:

SYS@taurus> rmode qucam2 simple

Image sequences obtained using 'rmode qucam2 spec' are stored in a single FITS file with a number of extensions equal to [n_reads] (please remember to discard the first image of every run).
NOTE: zero second exposures are not permitted. 1 ms is the minimum requested exposure time.

4.3 Calculating exposure times

Exposures for the L3 cameras are timed using the processor clock of the SDSU controller. The crystal clock is specced to 100 ppm across its whole temperature range. Since the temperature of the controller is pretty stable, the actual timing accuracy will be a lot better than this.

When doing 'rmode qucam2 simple' exposures there will be a timing error (probably at the level of 10s of ms) associated with the opening and closing of the mechanical shutter and this will dominate. When doing image sequences using the 'rmode qucam2 spec' command timing will be much better.

There is an important feature of frame transfer operation that the observer should be aware of when using the 'rmode qucam2 spec' command to do a sequence of spectra. If for example we request the following:

SYS@taurus> rmode qucam2 spec 100
SYS@taurus> run qucam2 0.1 "Test run"

we will NOT get 100 images with an exposure time of 0.1 s each. The actual exposure time will in fact be 0.1 s + the time it takes for a frame to read out. During a frame transfer sequence there are no clears between frames so the exposure time = cycle time of system. The QUCAM2 web page indicates that reading of a 100-pixel high CCD window will take 600 ms in fast speed. The above command will actually generate 100 frames with an exposure time of 0.1+0.6=0.7 seconds. It is the demanded rather than actual exposure time that will appear in the EXPTIME field of the image header. The true exposure time can be calculated from the UTSTART fields of two consecutive frames.


4.4 Data volume

If image sequences with minimum exposure time of 1 ms are taken, the data rate will be 592 kB per second or approximately 1 GB per 30-min observation. Run lengths greater than 30 min or with data volumes greater than 500 MB are not recommended.


4.5 Photon counting

If the images are not too deeply exposed (<0.1 photo-electrons per pixel on average) then it is possible to do photon counting. A threshold is applied to the images and any pixels over that limit are interpreted as containing 1 photon. This has the benefit of removing multiplication noise. Some care must be taken in the choice of the threshold. If too high then the effective quantum efficiency will suffer as some photo-electron events will fall under the bar. If set too low, then RON will cause false triggers. The RON is about 3-4 ADU, the average height of a photo-electron is about 116 ADU. Setting the threshold to about 20 ADU (above bias mean) would be a good starting point. In the case of a spectrum containing both strong and faint emission lines it should even be possible to apply photon counting analysis to the faint lines and proportional analysis to the stronger ones.


v

4.6 Calibration frames

Arcs and biases can be obtained in the usual way. Flat fields, however, have to be obtained at low signal levels (less than 50000 ADU). Thus, to achieve a good signal-to-noise ratio in the fast (L3) mode many flat field frames are needed.



Top | Back

Contact:  (ISIS/QUCAM Specialist)
Last modified: 06 May 2015