Sampling Modes
Gemini has several distinct software readout modes for each array. These include single-sampling readout, Fowler sampling, and 8 converts/pixel mode, and a subarray readout mode for the SBRC array. Each sampling mode is outlined below, and a table of minimum integration times and noise performance is below.
| Mode | Minimum Itime NICMOS (sec) | Minimum Itime SBRC (sec) | Typical Noise1 NICMOS (DN) | Typical Noise1 SBRC (DN) |
|---|---|---|---|---|
| 1 | 0.070 | 0.070 | 5 | 3.8 |
| 2 | 0.083 | 0.072 | 3.0 | 4.2 |
| 3:2 | 0.19 | 0.19 | 2.1 | 2.8 |
| 3:8 | 0.76 | 0.76 | 1.2 | 1.5 |
| 3:32 | 3.04 | 3.04 | 0.92 | 0.99 |
| 6 | 0.4 | 0.3 | 1.7 (splotchy) | 1.7 (splotchy) |
| 7 | --- | 0.004 | --- | ~4 |
| 8 | --- | 0.005 | --- | ~4 |
| 9 | --- | 0.020 | --- | ~4 |
Sampling mode 1: Single-sample mode
Mode 1 is just a straight forward single-read mode, with no correlated double sampling or other effects. The appearance of frames taken in this mode is dominated by the bias structure of the chip, including quadrant-to-quadrant offsets in the NICMOS chip and column-to-column offsets in the SBRC. This mode is recommended for checking saturation levels and for observing in high background situations. The dynamic range of each array in DN in this mode are:
NICMOS: -4000 DN to +4000 DN
SBRC: -7000 DN to -1000 DN
Sampling mode 2: Correlated-double-sampling
Mode 2 is a standard CDS mode. The chip is first reset pixel-by-pixel, then read out once. After the integration time has elapsed, the chip is read out again, and the resulting frame is the difference between the second and first reads. Ideally, there would therefore be no bias - a frame with zero photons should just be uniformly zero. In practice, both chips have a slight bias in this mode (caused by a slow settling of the chip after its been reset), at the tens of DN level. A complication is that for high signal levels, the chip may saturate wholly or partially between the second and the first reads, and will therefore be near zero; detecting saturation in CDS modes can be quite difficult, especially in short exposures. Noise performance in this mode is dominated by a 13 kHz noise pattern, which appears as faint diagonal lines across the frame. Therefore this mode is best employed when the observation is highly background limited.
Sampling mode 3: Multiple-read mode (Fowler Sampling)
This is the recommended readout mode for long exposure times (many seconds or more). As in mode 2, the array is reset pixel-by-pixel. Then in this mode, the array is read out several times (set by the "multiple reads" parameter) at both the start and end of the integration; the results are averages to reduce noise. Each multiple read takes approximately 0.1 seconds. Thus the minimum integration time is (0.1 x # of reads). Also, there will be 0.1 seconds of overhead per multiple read. Thus, this mode is best employed for longer integrations, where the overhead is not a significant portion of the duty cycle. Typically, 8 multiple reads are used for integrations on the order of 10 seconds. More (16 or 32) reads may be used for 3-5 minute integrations. Finally, the NICMOS chip suffers from amplifier glow if many multiple reads are used. This can be as much as 100 DN in the corners of the array. This glow is very stable, however, and should flat-field out.
Sampling mode 4: Movie-mode
Mode 4 is reserved for a future implementation of a movie mode.
Sampling mode 5: Multiple-converts-per-pixel
Mode 5 is a multiple-converts-per-pixel waveform intended to reduce noise. After each pixel has been selected, it is converted 8 times in a few microseconds before the next pixel is selected, and the results are averaged. Mode 5 is only a single-sampled implementation of this function, and is therefore of limited use. Minimum integration time is on the order of 0.3 seconds in both channels.
Sampling mode 6: Multiple-converts-per-pixel CDS
Mode 6 is a CDS implementation of the multiple-converts-per-pixel waveform. The array is reset, then read out with each pixel being converted 8 times, then after the integration time has elapsed, the array is read out again, with 8 converts per pixel. The result is a lower noise CDS-style frame. The sampling rate is automatically set at 80 kHz, and the minimum integration time is about 0.4 seconds. Saturation is again difficult to detect. A drawback to this mode is that the 13 kHz noise which is present in the system appears as a "splotchy" pattern on the chip in this mode. Therefore mode 6 is only recommended for highly background limited situations.
Sampling mode 7: 64x64 subarray (SBRC only)
Mode 7 reads out a 64x64 subarray of the SBRC array. This is a single-sampled mode, so there will be lots of bias structure; it is extremely important that bias frames be taken in this mode, with exactly the same integration time as was used for the data, due to the time dependence of the chip bias. Minimum integration time in this mode is 0.05 seconds. the data is read out as a 256x256 frame with the pixels that weren't read out set to zero. The autoscale function does not distinguish this mode, so it is best used with autoscaling off.
Sampling mode 8: 64x64 subarray CDS (SBRC only)
Mode 8 is a CDS implementation of the 64x64 subarray mode for the SBRC array. It is still recommended that dark frames be taken with the same integration time in order to remove any residual bias structure.
Sampling mode 9: 128x128 subarray CDS (SBRC only)
Mode 9 is a CDS implementation of a 128x128 pixel subarray for the SBRC. The minimum integration time is 0.02 seconds.
Last modified: Mon Apr 21 04:07:27 PDT 2008 by Elinor Gates