As will be shown thereafter, replacing the load resistor by a
capacitance (see Fig. 2 (click here), right) improves the system. In this case, the
capacitance is polarized by a triangular wave voltage, which produces a square
wave current. This triangular voltage is built from the square voltage by an
active integrator using an operational amplifier. The advantages of this
substitution are significant. First of all, the capacitance does not produce any
significant heat and is not a source of Johnson noise. It is not any more
necessary to put this element on the coldest stage of the cryostat, for example
at 
. So, only two wires per bolometer reach the coldest stage
(instead of three wires for a load resistor) which may prove a significant
advantage at very low temperatures.
The capacitance has a large input impedance: for 10 pF, the value of the
impedance at 100 Hz is equal to 200 M
, which is, of course, higher
than the bolometer resi stors (1 to 40 M). Finally, it decreases
the spike in the signal, becau se the parasitic capacitance C3 is not
active any more. The spike due to the bolometer remains. For this new
system and for a half-period of bias, the shape of the signal at point S
when the bridge is balanced reads:
where 
.
The bolometer resistance is, in this system and for a balanced bridge,
determined by the ratio 
. With a capacitive load, the
shape of the bias signals can also be adjusted in order to minimize the
effects of spikes and non linearities by adding a small triangular wave
shape to the square wave voltage. In practice, it is necessary to adjust the
current and the voltage on the bolometer in order to reach the best
sensitivity and the equilibrium of the bridge. With analog electronics, the
ratio between 
and 
can only be tuned by manually
adjusting the voltages. To make the system workable for intensive
astronomical observations, a feedback loop was implemented. The amplitude
of the output signal, 
, is pre-amplified, lock-in amplified,
integrated, and is used as the reference voltage to control the amplitudes
of the square waves and . The system works at
constant so that at equilibrium, 
, a
fixed value of the bolometer resistance is reached (see
Fig. 3 (click here)). Other principles could have been used, giving
different laws for the bolometer impedance control. That one proved to be
relevant for operational observations. It allows a direct power sensitivity
calibration of the bolometer: if the radiation power changes, the
electrical power due to the bolometer bias changes in opposite sense to
compensate and keeps the bridge balanced. So, the measure of the bias
amplitude gives a total power measurement. However, for the measured signal,
the feedback control acts as a high pass filter which eliminates the very
low frequencies, 
.
Figure 3: Working point for a bolometer
Such a system has been used on the Diabolo photometer for the observations at the 2.7 meter MITO telescope in Italy (see Benoıt et al. 1996). During the first campaign in May 1994, it worked with a resistive load and in March 1995, with a capacitive load. The system demonstrated its efficiency to avoid 1/f noise. However, the operations and the data reduction after the observations allowed to identify two problems: 1) the need of manual adjustment of the impedance (ratio voltage/current), 2) the absence of direct record in the data of the total power, since manual tunings were not recorded. To solve this problem, the analog feed back loop has been replaced by a digital control of current and voltage, based on a real time digital analysis of the output signal, which is described in the next section.