Several observers have been finding that the brightness of spectral lines from the heterodyne receivers are not constant. Sometimes they are not the same as a year ago, and sometimes they vary from night to night. In late 1993, the calibration appeared to be much worse than normal. This was traced to specific faults with both RxB3i and RxC2; observers with data taken at this time should be aware of these faults.

In both cases the problems in late 1993 were eventually traced to standing waves within the receiver dewars. With RxB3i, this was a result of a fluorogold covering over the quartz thermal blocking filter becoming detached. The problem with RxC2 was caused by a layer of polythene on a similar quartz filter peeling off. It is believed to be just coincidence (or Murphy's Law) that both faults occurred around the same time. The standing waves resulted in receiver temperatures and sideband ratios which were highly frequency dependent. The overall receiver performances were also worse than normal, and there was some small variation with elevation.

The problem with RxB3i was noticed in October-November and can now be traced back to the warmup at the end of July 1993. At frequencies around 345 GHz there were sometimes variations of approximately ±30% in the final line calibration. These were somewhat less (15%) at higher frequencies (around 360 GHz). The filter was fixed during a period when the receiver was not required in early February 1994.

Again hindsight suggests that the problem with RxC2 started after a cold head failure in mid October 1993. We estimate the calibration uncertainty after this was approximately ±20% at 460 and 490 GHz. The problem was diagnosed and fixed in mid April 1994.

Even if no specific faults exist in a receiver, there are many other possible sources of error in the line intensities. Below I list some of the more significant, with estimates of their values.

Factors affecting line intensities.

1) Sky variations. Mainly caused by changes between the time of the calibration and the sample. This is negligible under good conditions; test measurements on a source with repeated calibrations and integrations with a stable sky gave ~3% variation with RxA2 and ~6% with RxB3i. Under variable sky conditions (not necessarily correlated with poor sky transmission) this effect can introduce factors of 2 variation in the line brightness. Some work will be done in the Autumn to improve this by using the reference position in each sample as the `sky' for the calibraton.

2) Baselines. For linear baselines, the signal:noise ratio in the integrated line intensity is decreased by the ratio of [linewidth(MHz)]/[baseline(MHz)]. For example, a s:n of 5-sigma in Ta*.dv will be reduced to 3.3-sigma if the line covers 60% of the useful IF passband. This will obviously be worse for higher-order polynomial baselines. Larger s:n is better. No baseline subtraction is also better (ie BMSW).

3) Standing waves. These could be off the windows or filters in the receiver (even when working properly, there will always be some reflection; the filters in RxA2 were intact, yet low-level standing waves have been seen). Signals will also reflect off the hot load, mixer and the secondary. The level can be estimated by observing continuum sources, and variations of as much as 5% can be seen.

4) Beam size. These are measured in the case of RxA2 and RxB3i to typically better than 2%, although note that there is some variation with wavelength. This would result in ~4% error in the line brightness for a point source, although an extended source would show no change.

5) Beam efficiency. For RxB3i, this is known to an accuracy of ±7% at present, although with more measurements we should reduce this below 5%. The RxA2 value is more accurately known, and RxC2 is more uncertain (±10%).

6) Sidelobes. The integrated emission over the inner sidelobes becomes significant at higher frequencies when the accuracy of the dish surface is low. For extended sources, this provides a significant uncertainty. At 690 GHz, this may contribute 50%, and even at 490 GHz, there may be ~10% error . This can be corrected by careful measurement of the coupling to compact and extended planets. It causes an excess in Ta*.

7) Receiver sideband ratio. This is around 1.05 for RxB3i, requiring a correction of ±2.5% to either sideband. RxC2 and RxA2 have values of ~1.1; check with the recent measurements at JACH.

8) Receiver tuning reliability. I repeatedly retuned and integrated on the same line under good conditions using the prescribed tuning technique. Variations of less than 5% with RxB3i and 3% with RxA2 were seen, although these variations may actually have been caused by (1) above.

9) Pointing. This can be significant; eg for a compact source, a 16% decrease in line intensity can be introduced if you're only 1/4-beam off source (that's only 3 arcsec with RxC2).

10) Focussing. This is usually most important around sunset or sunrise when the telescope temperature is changing. An error of 0.5 mm in z will give a ~10% error in the intensity of a compact source at 345 GHz.

11) Errors in the atmospheric model. This is used to determine the sky transmission in the two sidebands, so errors will be largest when the ratio of transmission (a) is largest. It is difficult to quantify the possible error, but tests with RxC2 at 460 GHz (worst case) over a range of a show less than 10% error. It is probably under 3% at say 345 or 230 GHz, where a is close to unity.

12) Telluric Ozone or CO lines. These are not in the atmospheric model, but are taken out, to first order, with the DAS channel-by-channel calibration. The lines are not bright, so this effect is negligible.

13) Variation in physical temperature of hot load. This could change by up to approximately ±5K, resulting in less than 2% error in Tsys. RxC2 and all future receivers will monitor the load physical temperatures.

14) DAS sampler non-linearity. Improvements to the DAS have reduced this down to a level of ±1%.

15) Emission in reference beam.

16) Not enough signal:noise in the line (!).

With the above list, it might appear that you stand no chance of ever getting calibrated data. But it's not as bad as all that. Under good conditions, and assuming you're pointed and focussed etc, the errors combine to a total of ±10% with RxA2 and RxB3i, and ±15% with RxC2 (these are 1-sigma assuming the errors add in quadrature). For a point source, the predicted uncertainties are ~3% larger.

We have been collecting standard spectra from known sources over the last few months, using data from PATT observers, and from E &C time. These data therefore represent a normal range of observing conditions. Analysis of the results so far indicate rms deviations of ~6-12% with RxA2, 6-11% with RxB3i and 10-15% with RxC2.

Standard Spectra.

All spectral line standards are kept as hardcopies and are available on-line on the JCMT summit computer. Summaries of the spectra (both hardcopies and the ones on-line) are found in:-

DISK$USER:[JCMTUSER.REF_SPECTRA.**]

README_**.TXT

and listings of the on-line spectra are found in:-

DISK$USER:[JCMTUSER.REF_SPECTRA.**]

DIR.LIST

where ** is either A2, B3i or C2.

The data themselves are available in specx files within separate receiver subdirectories, and the filenames are of the form: SOURCE_TRANSITION.DAT

I would recommend that whenever one of the JCMT standard line sources is used for pointing (all of them are in the pointing catalogue) a standard spectrum is taken and compared with the on-line data. Once the spectrum has been observed, it can be logged using the command LOGSPEC from ICL. All logged data will be reduced at a later date by local staff. This serves not only as a check of the system operation before the start of observing, but also adds to our database of standard spectra.

Bill Dent, JAC