Detector Information
Appearance of the data
There exists one-pixel-width non-data rows/columns between quadrants for both detector. The unreadable regions are [1:1024,1025:1025], [1025:2048,1024:1024], [1024:1024,1:1024], and [1025:1025,1025:2048]. The detector for channel-1 has a large (~100-pixel radius), circular dead pixel island, as well as a few bad-data lines and some bad (high-dark) regions. There are no dead pixel islands with the size larger than 1.5 arcseconds on the detector for channel-2, though several scratches or small holes do exist. Channel-2 detector has a picture-frame-like high-dark region around the edge. The linearity is worse for the region due to the complex behavior of the dark current. Examples of raw images can be downloaded from here (CHIP 1 [gif, fits] / CHIP 2 [gif ,fits]).
The detector for channel 1, which we installed in July 2008, has on average 27% and 5% higher sensitivity in Y and J bands than channel-2 detector, respectively. In H band both channel is almost similar. In Ks-band the channel-2 detector is on average ~5% higher than channel-1 detector. The upper-half of the channel-2 detector has ~20% higher sensitivity than lower half (see the example of flatfield data below as a reference of the sensitivity map).
Linearity
The figures below are the result of the detector linearity measurements executed in Nov 2008 (current set of detectors). The linearity is maintained to well within 0.5% up to over 25000 e- under the low-illumination condition. Note that under the CDS readout the full well will vary depending on the position on the chip and the input flux density (above values were measured near the center of detectors: i.e. deepest well is expected). Therefore we recommend to keep the raw data count below 22000 ADU for science data.

Figure 11:
Figure 11: The result of the detector linearity measurements for the chip 1 (red) and the chip 2. The lower panel shows the residual from the linear fit assuming that the count is propotional to the exposure. The linearity is both very good.
Latent Image
The "latent" or "residual image" is the common problematic characteristics in HAWAII and (pre-RG) HAWAII-2.
It will appear on the position where the strong light is illuminated. For example, the latent from very bright stars will put its "footprints" on the data during the dithered imaging observation. For spectroscopic mode, the latent from the slit images, bright stars during the alignments, or sometimes the spectra from the alignment hole from the previous observation may appear after long exposure. Enough care should be paid for the possibility of the spurious objects by latents on the reduced image.
The typical amplitude of the latent is about 0.05% after 60 sec under unsaturated case. The e-folding time this case is roughly 200 sec. If saturated, the latent could be stronger than this. If the saturation is really severe, it may remain for several hours.
Reset Anomaly (Bias Tilt)
Reset anomaly (Bias tilt) is also another problematic character seen in some HgCdTe FPAs
such as HAWAII and the old HAWAII-2. It can be reduced by operating an array continuously and sampling by the CDS (Correlated Double Sampling) method. The current channel-2 detector shows only a small level (a few %) of the reset anamaly. Currently there is no data for new channel-1 detector. Taking data with dummy-read option (NDUMMYREAD=2) will suppress the residual reset anomaly much.
The instrumental minimum exposure times are listed in the table at the top of the page. In the minimum exposure data we see relatively strong reset anomaly. This can be suppressed when we set a bit (+1 to +1.5sec) longer exposure times than the true minimum. So we usually use the minimum exposure of 13.0 sec for NDUMMYREAD=0 and 20.5 sec for NDUMMYREAD=2 for full readout case.