Period on telescope	Array Orientation (Degrees)	Plate Scale (arcsec/pixel)
Oct 98	0.65 +/- 0.07	0.09075 +/- 0.00007
July 99 – April 01	0.40 +/- 0.03	0.09088 +/- 0.00005
May 01 – 15 Nov. 02	0.86 +/- 0.03	0.09060 +/- 0.0002
15 Nov. 01 —	0.827 +/- 0.02	0.09085 +/- 0.0002
22 June 08 – Aug. 08	0.506 +/- 0.095	0.09100 +/- 0.00001
03 Dec. 08 – present	0.564 +/- 0.041	0.09123 +/- 0.00013

The UFTI display shows (approximately) North up and East to the left; the array columns are slightly offset from true North-South as shown in the above table (to get columns exactly N-S one would have to rotate an image CLOCKWISE by the angle given above). The 1024×1024 HgCdTe Rockwell array has 18.5micron pixels and the camera operates at f/11. The plate scale is as shown in the above table; with a full array readout the field of view is 92.9 arcseconds. See “Readout Modes” for information on the available subarrays.

The position of a target on the array is set by the “instrument aperture”, a number set by JAC staff that does not change from run to run (it only changes when the instrument or telescope tertiary mirror is taken off/put back on the telescope). At present, a given telescope base position (your target coordinates, for example) should coincide with pixel 533,488 on the full-array, or 257,257 on the 512sqr sub-array. Note, however, that inprecise telescope pointing and inaccurate guide star coordinates can move a target by a few arcseconds, and so your source may not coincide exactly with this pixel-coordinate on the array. If you require very accurate positioning, please discuss this with your Support Scientist.

Port and tilt offsets for the dichroic are specific to each instrument and are entered into the TCS by the Telescope System Specialist (TSS). The TSS also needs to remove and replace the UFTI window cover at the start and end of the night and check that the instrument shutter is open for observing (from his Epics display). If the window appears to be very dusty please report it; in the short term dust particles can be blown off the window with a can of compressed air. If the humidity is high the TSS needs to check the window for condensation; a hair dryer set to low may be used to remove this. (Condensation or ice on the UFTI window will be evident in your images as a diffuse bright patch in the centre of the array.)

Current Filter Set

We have installed J, H, K, 2.166(BrGamma), 2.27(K continuum), 1.57 (H continuum) and 1.644(FeII) filters from the “Mauna Kea Observatory Near-IR (MKO-NIR) Photometric System”. The J in particular is different from the old IRCAM J(Barr). Note also that the I filter bandpass is truncated at the blue end by the dichroic. A K’ filter is available with UFTI; this filter should not be used for general photometry since it results in lower S/N than the K98 filter (remember that UKIRT is optimised for near-IR observations!). However, the K’ may be of use to IRPOL users, because of the increased background from the warm waveplate. Please also note that the H2 2.122 micron filter vignettes the bottom few rows of each frame; see here for an example of this.

See Zeropoints and Photometric Transformations for more calibration information. Transmission profiles of narrow-band filters can be emailed or faxed to users if needed (contact w.varricatt@ukirt.hawaii.edu).

May 2007 –

Wheel 1

Wheel 2

Posn	Filter	50% Cut on	50% Cut off	Posn	Filter	50% Cut on	50% Cut off
1	I	0.785	0.925	1	Prism	–	–
2	Z	0.850	1.055	2	Open	–	–
3	2.122MK	2.103	2.134	3	Blank	–	–
4	2.27cont[98]	2.260	2.280	4	1.69CH4_l	1.617	1.730
5	Open	–	–	5	1.57(cont)[98]	1.560	1.580
6	K’	–	–	6	1.644(FeII)[98]	1.636	1.652
7	J[98]	1.17	1.33	7	2.166(BrG)[98]	2.155	2.177
8	H[98]	1.49	1.78	8	2.173(BrGz)*	2.147	2.199
9	K[98]	2.03	2.37	9	Y_MK	0.966	1.078

Filters previously installed in UFTI are listed here

Sensitivity Tables

Surface Brightness

Measured during March 1999 (March and July for I,Z) engineering. See Zeropoints and Photometric Transformations for more calibration information.

1 sigma / 1 hour sensitivites

Filter	magnitudes per pixel	magnitudes per sq.arcsec.	10-20 W/m2/sq. arcsec
I	27.1	24.5	24.6
Z	27.0	24.4	—
J	26.6	24.0	12.3
H	25.9	23.3	15.5
K	25.6	23.0	8.7

With narrow-band filters, for continuum sources the source signal and background signal are both reduced by the reduced passband of the filter, i.e. by about a factor of 10. The Signal-to-Noise ratio for a continuum source will therefore decrease by the square-root of the ratio of filter bandpasses, sqrt{broad/narrow} ~ 3.2, and the magnitudes in the above table will DECREASE by 1.25. For an unresolved emission-line source, only the background signal is reduced by the reduced passband, so the Signal-to-Noise ratio for a line-emission source will increase by the square-root of the ratio of filter bandpasses, sqrt{broad/narrow} ~ 3.2, and the magnitudes in the above table will INCREASE by 1.25.

Point Sources

Measured during March 1999 (March and July for I,Z) engineering.

5 sigma magnitudes; 0.9″ seeing, 2″ aperture

Filter

Exposure Time, seconds

	1	10	60	600	3600
I	17.2	18.4	19.4	20.7	21.6
Z	17.1	18.3	19.3	20.6	21.5
J	16.7	17.9	18.9	20.2	21.1
H	16.0	17.2	18.2	19.5	20.4
K	15.6	16.8	17.8	19.1	20.0

5 sigma magnitudes; 0.4″ seeing, 2″ aperture (estimated)

Filter

Exposure Time, seconds

	1	10	60	600	3600
I	17.5	18.7	19.7	21.0	21.9
Z	17.4	18.6	19.6	20.9	21.8
J	17.0	18.2	19.2	20.5	21.4
H	16.3	17.5	18.5	19.8	20.7
K	15.9	17.2	18.1	19.4	20.3

Notes:

• For point sources the sensitivities assume that the source is jittered on the array and separate blank sky frames are not necessary. If sky frames are necessary, for example for extended sources, then only half the time is spent on source and the sensitivities are lower by 0.4 mag than those given in all the Table.
• For point sources if image quality improves and smaller s/w apertures can b e used the sensitivity goes up, e.g. if a 1″ s/w aperture can be used, the signal stays the same, the noise decreases by a factor of two, and the sensitivities improve by 0.7 mag.
• For surface brightness with other apertures increase the signal by the increase in area, increase the noise by the square root of the increase in area, and hence increase the S/N by the square root of the increase in area.
• To calculate S/N for a different magnitude after the same length of time, multiply the S/N by the ratio of the fluxes.
• To calculate S/N for the same magnitude but different exposure, multiply the S/N by the square root of the ratio of the exposure times.
• For narrowband filters the S/N is reduced by the square root of the flux reduction factor. The narrowband K filters are typically 1.5% wide compared to the broadband width of 15%, hence the S/N for these filters should be reduced by a factor of 3, or the magnitudes made brighter by 1.3mag.

Readout Modes – speed, read types, read areas and overheads

Speed

Two speeds are available – normal at 3.66 microsecond/pixel, and fast at 2.0 microsecond/pixel. Although the readnoise in fast readout is slightly lower (compare the figures given in the next section), hot pixels tend to “bleed” in this mode and more pickup-type structure (horizontal banding) is seen.

Stare and Non-Destructive (NDSTARE) Reads

Because of the “long” minimum exposure times associated with UFTI (see below), it is very unlikely that observers will ever need to use STARE mode; this mode will probably only be used in engineering operations to look at the bias structure of the array.

Mode Name	No. Resets	No. Reads
NDSTARE	20	2
10_NDSTARE	20	10

There are two forms of non-destructive (NDSTARE) readout available (see table above). Both do a number of resets to counteract latency on the array. The counts measured are fitted by a slope; the output image gives the increase in counts for the requested exposure time, normalised to one coadd if more than one coadd is taken. The two modes differ, however, in the number of times the array is read out (and consequently the overheads associated with each exposure). ND_STARE is read out only twice; 10_NDSTARE is read out 10 times, which results in a factor of ~2 improvement in read noise. 10_NDSTARE may therefore be useful when observing with the FP and sometimes with narrow-band filters, where background signal may be low and so background-limited performance difficult to achieve. The additional overheads associated with 10_NDSTARE are tabulated below. In both NDSTARE and 10_NDSTARE modes the bias is subtracted from the raw data frames.

Gain – Normal or “HiGain”?

As well as being able to select between non-destructive readout modes, UFTI users are also able to chose between two different gain settings (discussed further below). The “normal” gain should be used in most circumstances. However, a HiGain mode has also been implemented for low background or faint objects. The background limited regime will again be reached more quickly in the HiGain mode; however, the array will also saturate and/or the non-linear regime be reached more quickly (with fewer photons) so please be aware of this. As with 10_NDSTARE, the HiGain mode may be a better option when observing with the FP or (possibly) with narrow-band filters.

So how do you chose between 10_NDSTARE and HiGain? In a nutshell, the former gives lower read-noise, though at the expense of increased array readout overheads (about 30 seconds per full-frame exposure); the latter doesn’t give lower read-noise, though it does allow you to reach the background limit with less light (though beware of non-linearity and saturation!). As a general rule of thumb, normal gain and normal ND_STARE readout will be best suited to most observing programmes. If you are background limited (and not saturating on-source) then there is no need to use 10_NDSTARE or HiGain modes, though please discuss these options with your Support Scientist if you are not sure.

Subarrays

For many observations the full 1024×1024 area of the array should be used; for compact or point sources (e.g. standard/flux calibration stars) a sub-array is preferable, because of the reduced overheads. There is also no overhead to the choice of subarray, unlike IRCAM where different OCCAM waveforms have to be loaded, although the 512×512 array is usually used. The centre of each subarray is offset from (0,0) on the sky; the telescope and crosshead/guide-camera will be offset accordingly by the new telescope control software. The available readout areas are:

Name	Field	Region	Offset
1024sq	92.9″x92.9″	Full	0 0
512sq	46.5″x46.5″	Top left	23″W 23″S
256sq	23.2″x23.2″	Top left quad inner corner	12″W 12″S
512×1024	46.5″x92.9″	Left half	23″W 0
1024×512	92.9″x46.5″	Top half	0 23″W

Maximum Exposure Time

For very long exposures there are problems with the output image due to an overflow in a buffer which accumulates the sum of time squared statistics for NDR’s. Exposures longer than shown in the table below should be avoided; these times are much longer than time to be background limited and so this should not be an issue.

Mode Name	Max. Exp., seconds
NDSTARE	750
10_NDSTARE	350

System Gain, Noise, Linearity and Persistence

System Gain

The system gain in normal readout mode is either about 5.5e-/DN or about 7.5e-/DN, depending on which array controller we’re using at the time (we have two). In fast readout mode the gain is about the same. We have also implemented a “Higain” mode for low background and faint target conditions, which halves the number of electrons per DN.

Speed/Gain	Readout Speed	System Gain
Control Name	microsec/pixel	e-/DN
Normal	3.67	5.461
Fast	2.00	~6
Higain	3.67	~3

Read Noise

The read noise values for the common readout modes (see previous section on Read Modes ) are:

Speed/Gain	Readout Mode	Noise e-
Normal	STARE	56
	NDSTARE	26
	10_NDSTARE	12
Fast	NDSTARE	24
Higain	NDSTARE	20
	10_NDSTARE	10

Linearity

Measurements in 2000

Currently no linearity correction is applied to UFTI data by the ORAC-DR, since the array response appears to remain essentially linear up to about 4000 counts. Note that a 1% error in the translation of photons received to counts reported by the array is equivalent to a photometric uncertainty of only about 0.01 mags. Other sources of error are likely to dominate. Nevertheless, observers who require high photometric precision should try and keep the calibrator and target at a similar count level.

UFTI Linearity plot

The plot above shows how the array begins to saturate at about 10,000 DN and goes into hard saturation at ~14,000 DN. The rather complex fit to the data could be used for linearity correction for normal ND-Stare readout with normal gain (observations in Hi-Gain readout yield a very similar plot). The polynomial fit above is:

   y = -1.38.10-13x3 + 1.38.10-9x2 - 5.73-6x + 1.0

where y is the ratio of measured to expected counts, and x is the measured counts.

New measurements in 2009

The linearity measurements were re-done in 2009 and a plot is given below. The red line shows the observed counts and the black line shows the polynomial fit to the measured/expected values.

The non-linearity remains essentially the same as in 2000. The fitted polynomial is different. The equations given for the old measurements do not reproduce the non-linearity very well due to rounding off errors and probably due to the lack of higher order terms. The new equation should be used in preference to the old one for correcting for non-linearity of UFTI data.

For correcting for nonlinearity, divide the observed counts (“x”) by “y”. Currently no non-linearity correction is by the ORAC-DR while reducing the UFTI data. Detailed instructions and tools for correcting for this in the ORACDR will be provided shortly.

Persistence

Even with multiple resets before every image, we do see “persistence” or image “latency” at JHK at a level of 0.3-0.4%; at shorter wavelengths the effect is worse – 0.7% at Z and 0.9% at I. This acts like an enhanced dark current so that it gets worse with long exposures. For example, after exposing standard stars to ~4000 counts, even three frames later, with an 80 second blanked off exposure, a latent image of the star could be seen at the standard star jitter positions which had ~2 counts above the background.

This persistence makes flatfielding difficult and, as the UFTI flatfield is very stable, you may want to take a sky flat with a clean array at the start of the night and use that throughout the night (as opposed to creating flats from median filtered jittered sets). Take clean dark frames at the start or end of the night also. If possible put any bright objects in a different part of the array than your faint targets, and keep your counts at the “couple of thousand” level, as opposed to 5000 or greater. Finally, and most importantly, blank off (set dark) between sequences so that while you’re acquiring the next object, or preparing the next observation, you are not illuminating the array. If you do get a bright object onto the chip you will want to take a couple of short, 4 second darks (note: longer isn’t necessary, as the latency acts like an enhanced dark current) before taking more science data. You can use the observation called “Flush Array” in the ORAC-OT UFTI Template Library to do this.

Count Rates, Exposure Times and Sky Brightnesses

Counts on Sky and Dark Current

Filter	Counts/second	Time to be Background	Sky Brightness
	/pixel	Limited, seconds	mag/sqr arcsec
Dark	0.04	—	—
K	30	8	13.4
nbK	~3	80	~16.0
H	40	6	13.6
J	7	36	15.6
Z	1	250	17.6
I	0.5	500	18.6
*FP	—	*100	—

Read-Noise or Background Limited?

When observing in the near-IR, observations will either be read-noise or background-noise limited (there is also a dark-current contribution to the noise, though the low level of this “thermal” signal makes this about 5-times lower than the normal NDSTARE read noise of ~26 electrons). In the read-noise limited regime, which may be encountered with e.g. the FP, the signal-to-noise ratio (S/N) increases linearly with time because the statistical noise associated with reading out the array remains constant. For background-limited exposures, the S/N increases with the square-root of integration time, because of the steadily increasing Poisson or photon noise associated with the background flux on the array. Co-adding background-limited frames will also improve the S/N by root-time. There is therefore no advantage in integrating for much longer than the time needed to reach the background limit UNLESS the object frames are to be used to construct a flat-field. In this case a strong signal is needed in the background sky – of at least a few hundred counts – so that the noise contribution from the normalised flat field is not too great. Observers of bright sources using short integration times should ALWAYS obtain a separate flat-field!

Co-adding short-exposure frames obtained while read-noise limited is also not the best observing strategy. For example, co-adding four short integrations would only increase the S/N by two. By comparison, a single read-noise limited integration with 4-times the integration time would improve the S/N by four.

We therefore arrive at the following conclusion:

When possible, one should aim to observe in the background-limited regime, obtaining enough background signal so that a good flat-field may be obtained from the median average of the frames (if needed).

With a readnoise of about 26e- and a gain of 7.2e-/DN, counts should be >>90 to be background limited, say DN~250. Time to achieve this is given in the table above. You will have to compromise between being strictly background limited and not sitting too long on your target so that sky variations affect your ability to accurately flat-field and/or subtract the sky signal. A good rule of thumb at JHK is not to integrate for longer than one minute per exposure, or two minutes on a good night. Sky stability at I and Z is somewhat better than at longer wavelengths, so longer integration times are acceptable. Time to reach the non-linear regime on sky is in all cases greater than 5 minutes and not worth worrying about.

In “Higain” mode the readnoise is lower and the gain is higher (this mode may be preferable for narrow-band imaging and when using the FP). To be background limited requires about the same number of counts, ~250DN, but this corresponds to about half the number of electrons so your exposure times should be about 60% of those tabulated above. However, you do need to be more careful of non-linearity and saturation effects, since these limits will be reached more rapidly/with fewer photons.

Saturation Magnitudes

Saturation depends of course on the seeing conditions. On a night when the seeing was poor and varying between 0.7″ and 1.0″, saturation on point sources was measured to be: normal gain mode K=8.1, H=8.2, J/Z/I=8.5; HiGain mode J/Z/I=7.8. This is for an exposure time of 1.0 seconds, the minimum exposure for the 512×512 subarray. With the full array (likely to be used with your science programme) the shortest exposure is 4.0 seconds, so saturation magnitudes will be fainter.

Exposures for Standards and Times to Complete Jitter Patterns

Standard star sequences for the UKIRT Faint Standards are available in the ORAC-OT Standards Template Library; these use the 512-sqr sub-array and have integration times set appropriately. As an additional guide, however, the following table gives the exposure times appropriate for a given magnitude, using NDSTARE and Normal Speed/Gain:

K	J	H,K	Time to observe a 5-point jitter in three colours (e.g. J,H,K)	Time to observe a 5-point jitter in three colours (e.g. J,H,K)
magnitude	exp x coadds		1024sq	512sq
	seconds	seconds	minutes	minutes
8-9	2 x 3	1 x 6		6.0 (28%)
10-11	10 x 1	5 x 1	6.5 (30%)	5.5 (35%)
12	15 x 1	10 x 1	7.5 (44%)	6.5 (51%)
13	30 x 1	20 x 1	11.0 (61%)	9.0 (74%)
14	60 x 1	40 x 1	17.5 (77%)	15.5 (87%)

The above table also includes the time taken to execute a 5-point jitter pattern in three colours (including all overheads, e.g. time to jitter telescope, set filters, etc.), using the exposure times given in the second and third columns. The times do NOT include initial source acquisition (i.e. time to slew the telescope across the sky). The sequences each include 2 darks, as well as a final jitter to offset 0,0 and a set dark at the end of the sequence (to guard against latency from bright stars).

The figures given in brackets in the last two columns are the efficiencies, that is, the fraction of time spent actually integrating on source/dark against the total time taken to complete all of the observations. Clearly, observations of brighter sources are less efficient, since they involve shorter integration times, although of course higher signal-to-noise may still be attained.

Below we give the equivalent time and efficiency estimates for 9-point jitter patterns in 3 colours (again including all overheads, darks and jitters).

K	J	H,K	Time to observe a 9-point jitter in three colours (e.g. J,H,K)	Time to observe a 9-point jitter in three colours (e.g. J,H,K)
magnitude	exp x coadds	exp x coadds	1024sq	512sq
	seconds	seconds	minutes	minutes
10-11	10 x 1	5 x 1	9.5 (34%)	9.0 (36%)
12	15 x 1	10 x 1	12.0 (47%)	10.0 (57%)
13	30 x 1	20 x 1	18.0 (63%)	15.0 (76%)
14	60 x 1	40 x 1	29.0 (78%)	26.0 (87%)