Report on the fringe-free CVFs commissioned in CSHELL

Dear 1997 CSHELL Users,

I have just commissioned the new Circular Variable Filters (CVFs) in the IRTF CSHELL spectrograph and am writing to inform you of the resultant instrument performance and the implications for your upcoming run.

I will start with a bit of history. CSHELL uses CVFs to separate (sort) orders so that only light from the desired wavelength is present in its single-order echelle spectra. Like all other similar IR spectrographs (CGS4, IRSHELL, etc.), CSHELL originally used "stock" CVFs for this purpose. Unfortunately, the substrates of stock CVFs function as cavities which harbor internal reflections, giving rise to constructive and destructive interference which is resolved out by echelle gratings and appears as a wave-like fringing pattern in high-resolution spectra.

Several years ago, several CSHELL users and I developed several narrow-band filters which eliminated this interference fringing problem, but only at individual wavelength settings. The trick in such a design is to never have 2 parallel surfaces in any filter substrate cavity, thereby preventing the conditions of interference. These designs were successful, so I started work on a similar CVF design shortly after the CSHELL upgrade in 1994. The NSF and NASA funded the construction of these filters in 1995, and they were (finally) delivered by the vendor (Optical Coating Laboratories, Inc.) in late 1996. I did some preliminary testing of these new CVFs in CSHELL in November (after which they were removed from CSHELL), but I was not able to adequately test them on the telescope until 11 February this year.

In a nutshell, the new CVFs work very well. The fringing is virtually eliminated; it is below 1% in raw spectra (not clearly detectable to the eye), and point-source spectra are flat-fielded very well by spectra of the continuum lamp (I have not tried this with extended sources yet). Additionally, the new CVFs have greater wavelength coverage (but with identical spectral range per exposure) than the original ones. It is now possible to take spectra over the continuous wavelength range of 0.93 - 5.5 microns. Furthermore, the new filters have better optimized bandpasses for adjacent-order rejection at short wavelengths. Therefore spectra (i.e., equivalent widths) of lines with wavelengths between about 1 and 1.6 microns should now be more accurate. Finally, the improved coatings of the new custom CVFs also provide slightly better transmission at wavelengths above about 1.3 microns, resulting in slightly higher S/N data (more on this later).

However, all is not yet perfect with the new CVFs. One big change that you will see is that the fluxes recorded in flat field exposures of the continuum lamp and in wavelength calibration exposures of the arc lamps are about a factor of 2 lower than before (& also that much lower than the values predicted by the cal_lines program). This is what prevented me from releasing the filters before testing on the telescope. Apparently the optical path of the new filters is different enough from the old ones so that the focus of the lamps onto the slit is changed, reducing the amount of light detected from the lamps. This is of relatively little consequence as it only affects lamp exposures (and not astronomical ones), which are typically less than 10 seconds. Therefore plan on doubling your lamp exposure times for the immediate future.

The next significant problem with the new CVFs is that the middle segment, spanning the wavelengths 1.68 - 3.06 microns, seems to have a small thermal radiation leak near its lower edge (as seen in our VF quick-look data reduction program). Images taken through this segment have backgrounds which are 10 - 20 times higher in the lowest 5" - 8" of the slit than in the center region at short wavelengths (about 2.3 microns and below). The 2.4 - 3.0 micron wavelength region is less affected. The good news is that most or all of this light is diffracted away from the desired light by the grating and does not appear in the resultant spectrum (some light from adjacent orders will make it into the spectrum, but at lower intensity due to the grating blaze). I have measured contamination from this leak at several wavelengths to be 3% or less, and 60 second exposures of celestial objects show that there is no detectable increase in noise along the slit either (CSHELL is usually read noise limited in the near-IR). Thus it is not clear that this small light leak along the bottom 5" - 8" of the 30" long slit has any significant effect on CSHELL spectra. Furthermore, it can be COMPLETELY ELIMINATED by inserting a short-pass blocking filter in series with the CVF (e.g., select CVF/blocker in the filter menu). The filter wheel blocker does reduce throughput somewhat (12% at 1.68 microns; 15% @ 1.80; 19% @ 1.95; 22% @ 2.1; 25% @ 2.2; 29% @ 2.3; and 31% @ 2.4), but you should use it if you need to make very accurate equivalent width or absolute flux measurements. The short-pass filter in the shutter can be used instead, but this one seems to introduce a small amount of fringing (but has somewhat higher throughput).

There are also a couple more minor rough edges which need to be worked out. There are some narrow absorption features associated with the new CVFs which have wavelengths consistent with H2O absorption. These features show up in the 2 shortest wavelength filters only; the 3.06 - 5.5 micron CVF segment has no such absorptions. I recently took spectra with the old stock CVFs at these wavelengths, and they also show these absorptions but are typically only half as deep. The new and old CVFs are made with the same coating technologies, so I anticipate that the absorptions seen in the new CVFs may decrease with time as H2O (and other organics) are removed through continuous exposure to vacuum during pump-downs. In any case, these absorptions are minor compared to atmospheric ones, and they have never previously been noted as a problem in CSHELL spectra. Finally, there may be some measurable spectral contamination from adjacent orders at the shortest wavelengths (below about 1.0 microns), but I have not attempted any measurements of this yet.

I conclude with some performance data which replaces that in the Verifying CSHELL Throughput in the Troubleshooting section of the CSHELL User's Manual (v.2.0.1, 16 Aug. 1994, p. 103). Please use this as a guide in evaluating CSHELL performance or in troubleshooting the instrument. I intend to revise the manual to incorporate all changes related to the new CVFs.

Revised CSHELL performance information from 11 Feb. 97 data with custom CVFs, no blockers (all data values in ADU):
Wavelength  Cont. Spect   Elias Std.      Stellar Image     Stellar Spect
----------  -----------   ----------      -------------     -------------

1.2537      300   / 5 s   HD77281 J=7.11   59,921 / 2 s      3968 / 60 s
1.6593      820   / 5 s   HD77281 H=7.05   51,110 / 2 s      5462 / 60 s
2.1698      1502  / 5 s   HD77281 K=7.03   46,029 / 1 s      4833 / 60 s
3.5260      542   / 5 s   HR4828  L=4.68   19,174 / 0.1s     32,000 / 60 s
4.7010      3150  / 5 s   HR4828  M=4.69   8432   / 0.1s     2500 / 10 s


Wavelengths were chosen to match 63.5 deg blaze angle near JHKLM band centers. Continuum lamp counts are mean of 30 x 30 pixel box near spectrum center.Stellar images were centered in the field and are the star - sky sum in a 30 x 30 pixel box.Stellar spectra were centered midway along a 2" slit and are the star - sky sum in a 10 x 20 (x by y) pixel box.

Throughput as measured by stellar spectra is equal to or better than old CVFs at all above wavelengths.

New CVF position equations (wavelength l in microns; 6 Feb. 97 calibration):
Wheel/CVF        l (microns)             Step                      OCLI part
---------      -----------------       -----------------------     ---------

A CVF1         0.93 <= l < 1.68        (4026.06 * l) +  3981.8      Seg I B

A CVF2         1.68 <= l < 3.06        (-2281.5 * l) + 10999.5      Seg II A

B CVF1         3.06 <= l < 5.50        (1290.3 * l)  -  1906.8      Seg III D

Tom Greene

14 February 1997

LAST UPDATE: February 1, 2010 (format only)