Fig. 8.2
Each log stage has a current which is temperature independent, so the maximum output of a log
stage is constant over temperature; thus, the overall slope of the log transfer curve in the logging
region is constant. Finally, at maximum input the last log stage limits earlier at lower temperatures
(because of the higher gain), so the curves reunite at very high power, as also shown in Fig. 8.2.
The simplest way to overcome this problem is to apply a temperature- dependent DC offset shift at
the output, so as to merge the curves in the logging range. However, this will result in an offset of
about 1 dB for the baseline with zero power input. For most applications this does not matter,
since the temperature conformity is only specified in the logging range.
If a DLVA requirement does have a tight spec on the baseline below the logging range, a possible
solution is to make the gain of A1 temperature-dependent by placing a sensistor in the feedback
path. Unfortunately, while this approach does solve the zero-power-baseline temperature-shift
problem, a temperature-dependent gain on A1 can obviously play havoc with recovery nets on A1,
should they be needed.
A second possible approach is to compromise by moving the hot and cold curves only to within 1/2
dB of each other in the logging range; then the hot and cold baselines will be displaced by 1/2 dB
in opposite directions. It is not clear to us why anyone would wish to specify the detailed behavior
of the baseline outside the logging range of interest. However, we have seen such specs, and it is
fair to say that building a DLVA which is active from -40 dBm to +5 dBm, within
½ dB tolerance
over the full temperature range and at both zero input power and throughout the logging range, is
very difficult if something like 200 nanosecond recovery is also needed.
Aside from the problem mentioned above, note that we can come to the same conclusion with
respect to temperature drift at low powers as we did for recovery times - namely, the units are far
more sensitive when the onset of the logging range is set very low. This is again because the
overall gain is higher at low input voltages.
-26-
 |
 |
 |
 |
 |