Detectors
First Generation Detectors
This section covers information relevant to the first set of detectors used in PEPSI up until July 2023. PEPSI runs one low-noise (LN) STA1600 Charge Coupled Device (CCD) per arm, dubbed PEPSI BLUE and PEPSI RED. Its full frame is organized in two halves each containing an array of 10560 horizontal by 5280 vertical photosites. Horizontal transport registers along the top and bottom permit simultaneous readout of both halves. The pixel spacing is 9μm × 9μm. For dark reference, each readout line is preceded by 8 dark pixels.
Figure 1: Example flat fields and dark exposures. a) 400-nm flat field for the BLUE CCD. b) 10-min dark exposure for BLUE with amplifiers off; average dark read-out is 1.1e−/pix/hour. c) 800-nm flat field for the RED CCD. d) 10-min dark exposure for RED with amplifiers off; average dark read-out is
<1.8e−/pix/hour. This image is rotated -90deg to what is typically seen during use.
PEPSI BLUE
The total system noise (unbinned) and gain ranges between 4.4 e− (gain 0.47 e−/DN) and 5.9 e− (gain 1.3 e−/DN). The median noise in the high-gain mode is 4.6 e−. The total read-out time for optimal S/N is 60 s through all 16 channels (at 125 kpix/s per amplifier). The fast read-out-mode (with the gain set to 1.3) takes 37 s at 200 kpix/s per amplifier. Dark current is measured to be 1.1 e− per pixel per hour at −122 ◦C. Full well is set by the 16-bit ADC limit to about 82 000 electrons.
PEPSI RED
Channel 1 always appears noisiest with 7.0 e−. The other channels range between 4.0 e− (gain 1.475 e−/DN) and 5.9 e− (gain 1.434 e−/DN). Total readout time for optimal S/N is 56 s through all 16 channels (at 125 kpix/s per amplifier). The fast read-out-mode (with the gain set to 1.3) takes 37s at 200 kpix/s per amplifier. Dark current is measured to be 1.8 e− per pixel per hour at −121◦C. Full well is about 85 000 electrons. The device is AR coated for optimal red response and its peak QE was measured 94% between 650–700nm (see Fig. 29) but drops off to 47% at the cutting wavelength of the spectrograph at 907 nm.
Amplifier Glow
There is an amplifier glow in a single amplifier of both CCD’s that began appearing in 2015 after the replacement of the old ARC-48C controller to the ARC-48C. The Blue CCD has a glow rate of about 1ADU/min and coincides with the position of amplifier #1. The Red CCD amplifier glow does not coincide with a particular amplifier position but is adjacent to amplifier #16. The glow rate on the Red CCD is ~ 2ADU/min. The glowing calibrates out and does not affect the spectra. There is a planned detector upgrade Summer 2023 that should resolve this issue.
Second Generation Detectors
After extensive work during Summer Shutdown 2023, two new STA1600 CCDs were integrated into the existing dewars at STA premises in California over August. The new package has much better bonding protection with the preamps closer to the amplifiers, which in addition to other improvements reduces the noise. All went smoothly at STA, with AIP folks participating in the final lab tests and making the focus alignment of the dewars’ windows (which are the last lens groups in both optical cameras of PEPSI). Thanks to their heroic effort of transporting both units by road back to Arizona, the package safely arrived back at LBTO without any damage. Reintegration on the mountain succeeded on the weekend Aug. 26/27 and both dewars were cold on the 29th. Shortly afterwards the first calibration images were taken, which looked as expected from the CCD acceptance protocols earlier this year.
Figure 2: The normalized image of the de-focused flat field in CD35 in ADUs after amplifiers equalization in TAPLINEx of the Archon config file. September 6, 2023
Above is an example de-focused flat field image (taken with CD3 and CD5, i.e. blue and red CCD) with bias subtracted and normalized to a spline fit to remove the blaze function and the spectral orders. This reveals all of the detector defects immediately. Due to the enormous real estate (12k x 10k, including the overscan regions), there are several hot rows on the chips in addition to the usual coating defects. Residual fringing is still visible on the red image (since it has not been corrected yet). The bias level is at 1000 ADUs on both, with full well at 65k. The readout time remains at 80s. The gain and readout noise tables were calibrated and implemented into SDS4PEPSI as part of the automatic data reduction. The gain is on average 2.2 e−/DN, and the readout noise is on average 3 e−/amp. There is no visible fixed-pattern noise structure seen so far, the red-device hot spot is gone.
Figure 3: The CCD gain factor versus ADU in each amplifier. September 6, 2023
Figure 4: The trimmed sum of 72 one hour dark images (to eliminate cosmic spikes). Blue CCD is on the left and red is on the right. There are overall 17 hot/dark columns in amplifiers of both CCDs. A more detailed magnification shows more defects in a smaller scale not visible in this image. Amplifier glow is visible. Several clusters of hot pixels are visible in the red CCD, but lesser in the blue. These are not the cosmic spikes, but rather saturated pixels. September 6, 2023
Figure 5: The trimmed sum of 20 bias images (to eliminate cosmic spikes). September 6, 2023
Figure 6: Several blemishes on the red (left) and blue (right) seen in the normalized image. In the blue the intensity falls by half in the spots, but in the red it reduces almost to zero. These blemishes are present also in the dark image (but not in the bias image) and cannot be attributed to the CCD coating. September 6, 2023
