The electronic architecture for the COBRAS/SAMBA High Frequency Instrument (HFI: 150 to 850 GHz), depends on the dimensioning of the system i.e. the final sampling frequency, the bias frequency, the rate and the dynamic of the ADC, and the rate of the modulator DACs actuation. The parameters linked to the practical implementation of the detection system, such as the parasitic capacitances, the wiring lengths, and so on are assumed to be similar to those of Diabolo, since both instruments are based on the same cryostat that cools the bolometers down to 0.1 K, i.e. an open loop dilution system (Benoıt 1994).
From the technical specifications of COBRAS/SAMBA (see report on the phase
A study 1996), the optimal choice of orbit is a Lissajous orbit around the L2
Lagrangian point of the Earth-Sun system. At this location, the Sun, the Earth,
and the Moon are all located behind the payload, where their undesirable
effects are at the lowest possible level in terms of flux. The scan angle,
defined as the angle between the spin axis of the satellite and the telescope
line-of-sight, is equal to 
. So, the HFI beam will follow a circle of
a 
perimeter per minute of time, i.e. 340 arcmin per
second.
The beam FWHM of the highest frequency channels is equal to 5 arcmin. To fulfill the Nyquist criterion and some safety margins (3 data points per beam), a sampling rate of 200 samples per second is needed. This is not the sampling rate of the ADC, but the one of the signal delivered to the on board computer, after lock-in detection. The electric bias frequency of the bolometer has to be an integer factor of this frequency. The digital lock-in allows the bias frequency to be equal to the scientific sampling rate.
To allow proper elimination of transients (see the end of Sect.
5 (click here)), the rate of ADC is equal to 12.8 kilo-samples per
second (i.e. 64 
bias frequency). The range of the ADC is
determined by the ratio of the larger signal measured (a brightness temperature
of about 100 mK on bright sources) to the 
/2 bolometer noise
( = 0.080 mK rms noise at 200 Hz sampling), i.e. 1216, which is
fulfilled with 12 bits.
The actuation rate of the DACs in the digital modulator can be chosen as lower than or equal to 0.0167 Hz, synchronous with the satellite spinning rate. Indeed, due to the speed of the bolometer control by the 4 DACs (i.e. 100 ms), the reset of the signal can, if necessary, be performed at each satellite rotation with a loss of only two samples. However tests on the Diabolo experiment show that using a temperature regulation of the 0.1 K stage allows, on clear sky conditions, to freeze the modulation parameters for several hours. The stability of a satellite system is expected to be still better since there is no atmospheric contamination of the background photon flux.
In order to meet these requirements, an architecture of the bolometer readout electronics of the COBRAS/SAMBA HFI (see Fig. 10 (click here)) can be derived from the system developed and tested on the ground-based Diabolo instrument. As shown in Fig. 10 (click here), the elimination of transients, the signal lock-in, the filtering and the feedback control of the bias (through values delivered to the modulator DACs, controlling the shape and amplitude of the bias voltages) are performed by computers. Each computer is able to manage a group of about 12 bolometers, and delivers the processed signals to the on board computer (BEBA: Bolometer Electronic Box Assembly). There is one analog chain per bolometer (modulator with capacitive load, DACs, amplifier, ADC and digital command logic), and five computer units are needed for the 56 bolometers of the HFI. Bolometers belonging to the same wavelength channel will be dispatched to different computing units to avoid that a particular frequency become unavailable because of a computer failure.
Figure 10: Synoptic of the electronic system proposed for the COBRAS/SAMBA
High Frequency Instrument
This system is robust because it uses current space qualified technologies: the analog part is made of classical low noise electronics, the computer units are based on the Thomson Inmos T805 transputer. In addition to the processed signal, a fully sampled period of each bolometer is passed on to the telemetry at a slow rate. This will be used to check that the signal processing is correctly done, and to perform the first balance of the bias at the beginning of the mission. One of the advantages of the fully digital on board signal processing, with respect to a system using analog lock-in, is that it will be easier to handle in flight problems like an unexpected rate of particle induced spikes or an anomalous background power on the detectors.
In addition, in flight measurements of the bolometer V(I) characteristic are possible. This will allow to find the optimum working point of each bolometer which will depend on the background photon flux. This parameter is difficult to predict before the mission, and it may induce high costs if it is specified within a narrow range (cleanness of the optics). Moreover, recording the bolometer current and voltage allows to deduce the absolute optical power on the bolometer, which gives a calibration of the bolometer response.