MOIRCS Channel-1 Detector Information Page

Current Status and the Update

2008-06-09 update: About the Replacement Plan
We will be able to change the channel-1 detector to the scientific grade chip on coming July.

There was a difficuly in finding new detector for replacement, because the Hawaii-2 array we use for MOIRCS is basically out of production now. The detector available in the company was almost unique, and that was unfortunately not so ideal as we first hoped. There is a big "hole" of low-sensitivity pixels near the center of a quadrant. The area of "the hole" amounts to ~0.8%, and we will need a big dither size (~40 arcsec) to fill the bad area by the dither for imaging observation. For spectroscopy a more careful design to avoid the hole or another bad regions will be required. The readout noise measured by the company is ~15 e-/sec, a bit worse compared the value for the current channel-2 detector. The Quantum effeciency is also a bit (7.5 %) worse in K, but for J it is better by ~18 %. The homogeneity of the QEs across the detector is much better (rms is ~half) than the channel-2 detector.

Considering the serious scientific loss by keeping operation under the current situation (i.e. the operation by the engineering-grade detector), we have decided to accept the new detector. The engineering work will start in July 1st and it will take at least one month for the completion. The image by new detector will be taken in mid August. The information is to be updated on the website.






2007-12-14 update: Gain and Readout Noise Estimate
For the estimate of Gain in chip 1, we use the standard photon-transfer curve method. This method assumes that the rms noise of the image is purely from the photon-statistics and the read-out noise. We take a series of K-continuum dome-off images with various exposure times, as shown in the linearity estimate for less-luminous condition (the same dataset is used). In each exposure we took 12 exposures and combined all images but the first 2 frames, under 3-sigma clipping. Sigma images are simultaneously generated by the IRAF imstatistics task. We measure the mid-point from the combined images, as well as their rms noise from the sigma images. The region around [1025:1225,1025:1225] is used for statistics of chip 1 in order to prevent the affect of the "ring" pattern, while for chip 2 statistics [700:1300,700:1300] region is used.

We plot the variation(=sigma^2) with the mid point values. It should be linar relation. Suppose that the average image count is N (ADU) which is actually shown as g (e-/ADU) * F (e-), and the image rms noise is E (ADU) with the readout noise of R (e-). Then there is a relation as follows.



In Figures 8 and 9 we show the result of the fit for chip 1 and 2, respectively.

Figure 8.

Figure 9.

The result of the gain and read-out noise estimate is summarized below.

Table 3: Gain and Read-out noise Estimate

	Gain (e-/ADU)	Readout Noise (e-)
CHIP 1	2.95	27.7
CHIP 2	3.24	29.5

The values for chip 2 is roughly consistent by the same estimate by Ichikawa et al.(2007, SPIE), who esimated the gain of 2.86 e-/ADU. Small discrepancy may arise from the difference of the way of estimating the image statistics. The readout noise also matches well (30 e- ).

Readout noise for each detector is also independently estimated using the sigma images by 17 21-sec dark images, assuming that the rms noise in each pixel through the 17 set of 21-sec dark is dominated by readout noise. The result is 32.2+-6.2(e-) and 33.4+-7.4(e-) for chip 1 and 2, respectively. The result also well matches to others within errors.




We have shown here that the basic characteristics of chip 1 is not as bad as we first thought. However, we again note that the estimate of linearity, gain, and readout noise for chip 1 described here are all from relatively "clean" region. The data are taken under the controled situation. The characteristics should be much worse around the ring, and any unexpected behavior may appear in real situation. Enough caution should be paid if you try to use the chip-1 data for science.





2007-12-13 update:

1. Linearity of chip 1
The linarity of the Channel-1 chip is measured by taking a series of dome images using the thermal emission in Ks band during a cloudy night in late November. Temperature of the dome were stable with the difference of < 0.3K level. We used partial-readout mode (512 x 512) for the test because the K-band dome flux was relatively bright. The data was taken from short-exposure to longest until saturation with NDUMMYREAD=2. In each exposure we took three contiguous shots and used the last exposure data for statistics. Darks with the same setting (exposure and partial read setting) were subtracted from the image. The median value of the central 512x512(pixel^2) region were used for statistics using the IRAF task imstatistics with nclip=3 option. Figure 4b shows the degree of residual from the linear fit. Linearity is very good, with much less than 1% until 20000ADU level.

Figure 4a.




Figure 4b.

For reference, the result of the linearity measurement in channel-2 chip is also shown in Figure 5. Only the residual from the best fit is plotted. Again, the linearity level is much better than 1 % level till 20000ADU level. Linearity persists 1% level until 35000ADU.

Figure 5.

We also measured the linearity under less luminous dome condition. For this aim, we used the K-continuum narrowband filter. Again, the thermal emission from the wall of dome during cloudy night on early December was used for the light source. Temperature of the dome was 5.0 degree and very stable during the experiment. As the count rate is low, we used the whole read. We took several exposure with NDUMMYREAD=2. Each exposure we took three contiguous data, and only the last image is used. Darks under the same exposure are subtracted.

The linearity becomes a bit poor for both channel. Under the usual assumption with y-intercept of 0, the deviation from the best fit line is ~1% level for both channel. The results of chip-1 and chip-2 were each shown in Figure 6 and 7, respectively.

Figure 6.





Figure 7a.

Figure 7b.





2. Partial Read Mode
The table below is the result of how dark level changes with the size of partial-read (PRD_SIZE). The same dark data we as we used for linearity measurement is used for image statistics. For image statistics the area [1025:1225,1025:1225] is used.

Table 1: Dark Level difference

PRD_SIZE	MIDPT (CHIP1)	STDDEV (CHIP1)	MIDPT (CHIP2)	STDDEV (CHIP2)
512	164.7	8.54	1.092	5.751
1024	94.84	7.329	1.009	5.76
2048	23.59	5.839	1.754	5.726

We then check how the counts of the K-band lamp-off dome image changes with the PRD_SIZE. The data was taken during the same night as we took linearity data. Exposure time is fixed to 21 seconds with three different (512x512, 1024x1024, 2048x2048) read-out size. Three images were taken for each setting and the last image in each setting were used for statistics. Darks with the same setting were subtracted. For statistics we use the IRAF imstatistics with nclip=3.

Table 2: The Difference of K band Dome Level

PRD_SIZE	MIDPT (CHIP1)	MIDPT (CHIP2)
2048 (1st)	14501.41	17715.98
1024	14854.86	18209.54
512	14959.48	18353.05
2048 (2nd)	14568.63	17813.49

There is a systematic discrepancy in mid-point counts between the whole read and the partial read. Though each count shoud be close with each other, partial-readout data always show a higher count by ~424 (chip1) and ~588 ADU (chip2), about 3% level. As the darks are already subtracted, the change of dark level cannot be the cause. Also, the change of the counts in whole-read (2048) mode shows only small rise, the change of the dome temperature cannot explain the result. The cause of this ~3% discrepancy is not yet known.






2007-12-12 update: Characteristics of Dark in chip 1
The dark level on chip 1 varies time to time. The figure below is the sequence of how the median value of the darks changes during several set of contiguous data acquisition. The data were taken with 21-second exposure under NDUMMYREAD=2 option during the observations in late Nov-early Dec. Statistics are measured using the [1025:1225,1025:1225] area avoiding the "ring" region, with the IRAF task imstatistics with the iterative 3-sigma rejection (niter=3). The same data for chip 2 is also shown for reference. In the figure, points connected by lines are the data taken contiguously. These series of data were taken at the start of observation (1st, 2nd, 4th sequences) or flushing after taking high-count data (others: 5th and 7th sequences are after saturation). All sequences show high dark level for 1st data, without regard to the situation. For data after saturation, the count values after second data continuously drops, implying the affection of latent.

Figure 3a.

Next figure shows the same statistics but on the "ring" pattern in Figure 2 (the region [1800:2000,1724:1924]). For comparison, the chip 1 data shown in Figure 3a is also overplotted. Clearly the characteristics of the ring region is different from other clean region. The affection of the reset anomaly as well as the latent seems large for the ring region.
If we ignore the first image in each sequence, the variation of darks are always less than <30ADU for clean region, but may sometimes as large as ~150 ADU for the ring region.

Figure 3b.

Figure 4 shows the behavior of the dark noise with the median counts. The same data in Figure 3a and 3b are used. Dark noise at the clean region is almost the same as the noise in chip 2. But at the ring region the noise level is 8-9 times larger.

Figure 3c.

2007-11-07 update:
There is a very large "reset anomaly" on Engineering-grade chip. Each quadrant showed a prominent gradient along slow-read direction. When we took 5 dark frames contiguously with normal mode (NDUMMYREAD=0), first three images showed negative count (~-800ADU) with strong negative gradient up to -2000ADU level. However, with the use of a special readout option (NDUMMYREAD=2 mode) the situation improved greatly. This mode with additional dummyread, however, requires more overhead. The minimum exposure time for whole read will beas long as 20.5 seconds.

The image below is an example of 600-sec dark frame (14 frames are median-combined). The data were taken with NDUMMYREAD=2 option under 8-times multi-sampling. The image is displayed with the range between 0 to 512 ADU, with 5 by 5 boxcar smoothing. A prominent high-dark region (called the "ring") is seen (fits format available: CHIP 1 / CHIP 2 for reference). _{- updated Dec 13, 2007}

Figure 2.





A cross section view along the rows 1800-1900pixel(upper) and 800-900(lower) are shown below. Except the bad doughnut-like region, the background count level is around 75ADU. For reference, the background level on chip-2 side is ~22 ADU.

Figure 1a.

Figure 1b.

2007-10-19 update:
MOIRCS was transferred to summit from Hilo base and already started vacuuming. No problem was found on the test exposures after transfer. Next week we will start the pre-cooling of the main dewer, which will take nearly whole the week. The first detector test will start on the late weekend. The thorough characterization of the detector may require several weeks.

2007-10-14 update:
Inspection of the channel-1 chipbox has finished and we decided to replace the detector to the existing engineering-grade one. No problem has been found during a preliminary test after installation.

by Ichi Tanaka