Fig.9(a) shows the temperature correction applied to the inverting input, and corrects for the A1 output going down with increasing temperature. For the case of the A1 output going up with increasing temperature, connect Rt to VEE and reverse direction of the diode. Note the resistors labeled Rt and Rc are RT0 and RT1 on the board and in fig 2.

Fig. 9(b) shows the temperature correction applied to the non-inverting input, and corrects for the A1 output going down with increasing temperature. For the case of the A1 output going up with increasing temperature, connect Rt to VCC and reverse the direction of the diode.

After the temperature correction has been applied, the amplifier must be re-zeroed by appropriately changing the resistor that was originally used to zero the amplifier. An input offset of 50 µV at A1 will lead to a drift of roughly 1 dB over the temperature range of -55ºC to +85ºC. Thus the output of A1 must be re-zeroed to an accuracy of better than 50*G µV, where G is the gain of A1. So typically it should be restored to within 250 µV of its original value. Note it is not necessary to get the temperature to be exactly the same as it was when the offset was measured before applying the tempco, because you have a first-order temperature corrector in place when you measure the offset after applying the tempco. It may be easier to restore the voltage at the output of A3, as the values will be easier to measure there.

As we suggested above, it is advisable to install a temperature compensation scheme when assembling the board initially, using fig 9b. For a diode that changes by 2mv/degree (C) you should start with a value of Rc of 7KΩ*1.5/2. It will be necessary to change the value slightly to remove the temperature dependence of the output of A1 completely since there is a slight variation from chip to chip, but this should give a good starting value. In the unusual event that an A2 temperature correction is needed, use the same method, but the tempco circuitry must be applied to the negative input (pin 31), as shown in Fig. 2(b). This is necessary since pin 32, the positive input to A2, is driven directly by a very low impedance source. If this is inconvenient for your layout, you can use the positive input provided that you connect the output of A1 to the input of A2 through a 100 or 200 ohm resistor.

Overcoming the problem of parallel shift of the log curves: See (Fig 8.3b in the L-17C app notes)

As noted in the L-17C app. notes, if one zeroes the temperature drift of the output with no input signal, the log curves will be shifted to give curves that are parallel to the room temperature curve, but shifted in the logging range as the temperatures vary. This is caused by the increased g_m of the transistors at low temperature. If it is necessary to have both the output at zero power and the log curves identical over temperature, the method used for the L-17C will not work here.

There are three other ways of dealing with this problem:

1. The RF gain can be temperature adjusted to drop about 1 dB over the -55ºC to +85ºC temperature range.


2. A sensistor can be put in series with a suitable resistor in parallel to the feedback resistor on A1.


3. A diode in series with a 10K* resistor can be placed from the L1 adjust pin to VEE, and a 170K resistor from the L1 adjust pin to ground added to restore the nominal slope.

* This is an estimate, and will depend on the diode used.