The electronic system of Diabolo has been tested with special configurations that allow to estimate the noise of the amplifier (Fig. 6 (click here)), to control the stability of the bias amplitudes (Fig. 7 (click here)), to measure the noise of the digital system with a bolometer simulator (Fig. 8 (click here)), and to measure it with one bolometer of the Diabolo experiment (Fig. 9 (click here)).
To evaluate the direct noise of the BEBO, the input is connected to the ground.
The noise spectrum, after lock-in, is measured with a maximum gain (8 105)
(see Fig. 6 (click here)). The average noise level is about
0.5 nV/
, in good agreement with the expectation (JFET noise
0.3 nV/, resistance noise 0.4 nV/).
To control the stability of the square voltage bias, a bridge was made of two
resistors (200 
) and supplied at its two ends with two opposite square
waves (
peak-peak) delivered by two BEBOs. The signal at the
middle point of the bridge was measured by a third BEBO, and its amplitude
spectral density is displayed in Fig. 7 (click here). From 0.3 to 10 Hz,
the noise level of 1.5 nV/ is as expected from the Johnson
noise of the resistors in parallel and the pre-amplifier noise. Below
0.3 Hz, this level slowly increases to reach 4 nV/ at
0.01 Hz. This level is still less than 5 to 50 nV/, the range
in which the intrinsic bolometer noise is expected to remain. Even if the
noise of the bias sources is not negligible at very low frequencies, it
will not be the main source of noise in the system.
To measure the total noise of the electronic system, a bolometer has been
simulated by a resistor (1 k), a capacitance (47 nF), and a low noise
JFET (1.5 nV/). The values of the capacitance and the resistor
have been calculated in order to produce the same Johnson noise at
as a bolometer thermistor of 3 M at 
(i.e. 4 nV/). It can be seen on the noise spectrum
(Fig. 8 (click here)), that the electronic noise is approximately white,
below 4 nV/ above 0.2 Hz, with a slight increase at lower
frequencies reaching 15 nV/ at 0.01 Hz.
From the noise spectrum (see Fig. 9 (click here)) obtained on a bolometer
(
) of the Diabolo experiment, the noise is
approximately equal to 20 nV/ (a 40 M resistance has
a Johnson noise about 15 nV/) and stable down to 0.1 Hz. The
increase of the noise level below 0.1 Hz has been attributed to temperature
fluctuations of the dilution stage because this noise appears to be very
well correlated between the different bolometers and the thermometers.
Subtraction of this noise using blind bolometers is possible and reduces the
level by a factor of order 10.
Finally, the cross-talk between two channels of the full detection chain (from
bolometer to digital output signal) has been measured on a bright planet
(Saturn) to be less than 0.1
, which includes both optical and electrical
possible effects.
Independently of the Diabolo experiment, the use of the Thomson Inmos T805 to
implement a digital lock-in detection has been successfully tested. Up to
twelve channels have been simultaneously measured. The optimum digitizing rate
needed to correctly analyze the signal has been tested. A rate corresponding to
about 64 data points per bias period is enough to eliminate the transients.
In the Diabolo system, the length of the residual transient is about 4 data
points per half period, which have to be eliminated in the lock-in integration.
The efficiency of this AC readout is therefore close to 88. Finally, the
time needed to control the parameters of the bolometer readout by the 4 DACs
was found to be equal to about 100 ms.
This system, tested and used on the Diabolo ground-based experiment at the focus of the IRAM 30 meter radio telescope in December 1995, allowed to observe the Sunayev-Zeldovitch effect at high angular resolution (a 30 arcsec beam and 3 arcmin beam throw) towards several clusters of galaxies (Désert et al. 1996).
Figure 6: Noise spectrum of the amplifier after lock-in
(
)
Figure 7: Noise spectrum of the bias amplitude
(
)
Figure 8: Noise spectrum obtained with a bolometer simulator at 300 K
()
Figure 9: Noise spectrum obtained with a bolometer (
) of
the Diabolo experiment (
)