1 Introduction

The continuum emission of various astrophysical objects in the millimetre domain has long been proposed as one important clue to many physical processes in the Universe: such emission includes dust, free-free, synchrotron emissions, but also fluctuations of the Cosmic Microwave Background (CMB), either primordialy (Smoot et al. 1992) or due to intervening matter (Sunyaev & Zel'dovich 1972). In the past 15 years, the field of millimetre and far infrared measurements has tremendously grown. The advances in instrument technology have allowed many discoveries, with ground-based observations of our Galaxy and of extragalactic sources, with the many successful ground-based and balloon-borne CMB anisotropy experiments, and with the instruments onboard the COBE satellite. Following the experience acquired with the submillimetre balloon-borne PRONAOS-SPM experiment (Lamarre et al. 1994), we have devised a millimetre photometer called Diabolo, with two channels matching the relatively transparent atmospheric spectral windows around 1.2 and 2.1 mm. This instrument is designed to be used for ground-based observations, taking advantage of the large area provided by millimetre antennas such as the 30 m telescope of IRAM, and of long integration times that can be obtained on a small dedicated telescope. Such observations are complementary to those that can be made with highly-performing but costly and resolution-limited space-borne instruments or short duration balloon-borne experiments.

There are two main disadvantages to ground-based measurements, which are:

• a larger background, which not only produces a larger photon noise but also limits the sensitivity of bolometers because of their power load, especially when one tries to obtain broad-band measurements with a throughput ($A\Omega$) much larger than the diffraction limit;
• additional sky noise, mainly due to the fluctuating water vapour content in the atmosphere, and which is usually the main limitation of ground-based instruments unless properly subtracted (see e.g. Matthews 1980; Church 1995; Melchiorri et al. 1996).

There are two usual methods for the subtraction of sky noise, either spatial or spectral ones. The spatial subtraction method uses several detectors in the focal plane of the instrument and takes advantage of the spatial correlation of the atmospheric noise. For this technique to work, the source size must be smaller than the array size. It is especially suited for big telescopes for which the beams from the different detectors have not diverged much when crossing the 2-3kilometre high water vapour layer. Kreysa et al. (1990), Wilbanks et al. (1990), and Gear et al. (1995) have used this technique with bolometer arrays (MPIfR bolometer arrays, SuZie photometer and SCUBA arrays respectively).

The spectral subtraction method takes advantage of the correlation of the atmospheric signal at different wavelengths. If the source signal has a continuum spectrum different from the water vapour emission, one can form a linear combination of the source fluxes at different wavelengths which should be quite insensitive to sky noise. This technique has been used in various photometers. For extended sources like clusters of galaxies (see below), it has been used by Meyer et al. (1983), Chase et al. (1987) and Andreani et al. (1996, see also Pizzo et al. 1995). In particular the spectra of the 3 K CMB distortions (either primordial or secondary) are quite different from the water vapour emission as can be seen in Fig. 1. This technique implies that the smaller wavelength channels do not work at the diffraction limit, so that the beams at the different wavelengths are co-extensive. Hence, for broad continuum measurements, the detectors can only be large-throughput bolometers.

Figure 1: The distortions of the 3 K Cosmic Microwave Background expected in the millimetre domain are shown in the upper panel. The dotted curve corresponds to the Sunyaev-Zeldovich effect with a comptonisation parameter y=10-4. The dashed curve corresponds to a Doppler effect of ${\Delta T/ T}=10^{-4}$. The thin curve shows the spectrum that comes from a typical fluctuating part of the atmospheric spectrum, normalised to only 0.15 $\,\mu {\rm m}\,$of water vapour (notice the very different colours between 1 and 2 mm). In the lower panel, the curves show the normalised transmission of the 2 Diabolo channels centered at 1.2 and 2.1 mm and the atmospheric transmission for 3 mm of water vapour (with y axis between 0 and 1). Atmospheric curves were deduced from the ATM atmospheric model kindly made available by Pardo (1996)

Figure 2: Cold optical plate of the photometer (diameter of 250 mm): PFR is the cold pupil entrance with a lens and some filters, F1, F2 are some submillimetre cutoff filters, LC1 is the focal plane lens with its diaphragm DC1. A dichroic DI1 splits the radiation between channel 1 in reflection and channel 2 in transmission. BP1 and BP2 are bandpass filters. MR2 and MV1, MV2 are plane mirrors to fold the two beams. S1, S2 and S3 are three pillars holding the optical plate to the cryostat

Figure 3: Vertical cut of the cryostat showing the cryogenic plate of the photometer and the optical plate. For each channel, a lens LV reimages the beam onto the entrance of the Winston cone CL, through a bandpass filter FI. The dilution fridge provides cooling of the bolometers to 0.1 K. It is shielded with a dedicated 1.8 K screen. Each bolometer can also be fed via an optical fibre in the back of the bolometer, with near infrared ligth provided by a diode which acts as an internal relative calibrator

Once and if the sky noise can be subtracted, the need for sensitive large-throughput bolometers implies the lowest possible working temperature (see Sect. 3 & Subsect. 4.3). Diabolo has been built following this line of thought. It is a simple dual-channel photometer, with two bolometers cooled to 0.1 Kelvin for atmospheric noise subtraction using the spectral subtraction method adapted to small telescopes. Its design and performance are described in the rest of this paper, which is organised as follows. Section 2 describes the optical layout of the photometers and the filters we use for the proper selection of wavelengths. Section 3 describes the dilution cryostat that is used to cool the bolometers. Section 4 deals with the design and testing of the 2 bolometers. Section 5 gives details on the new bolometer AC readout electronic circuit which is used for the measurements. Section 6 gives the characterisation of the instrument that was possible with the first observations at the new 2.6 metre telescope at Testa Grigia (Italy). Finally, we discuss in Sect. 7 the recent improvements that have been made over the original design.