Question
We use the AD595AQ as part of a temperature to frequency convertor.
The convertor is housed in a DIN rail mounted component box. A
k-type thermocouple is connected to terminals on the component
box which in turn connect to the pcb and pins 1 and 14 of the
AD595. Pin 14 is also connected to 0V and pin 13. Pins 2,3,5,6,10
& 12 are not connected.Pins 4 & 7 connect to 0V, pin 11 is connected
to 24V and pins 8 & 9 are connected to the input of a AD654JN
V to F convertor via a 10k resistor . On recent boards the output
of the AD595AQ does not appear to be 10mV/degree C. e.g. for a
23 degree input the AD595 output is 350mV. This value varies from
board to board. Also if the AD595 is touched the output increases
in value quite rapidly. Circuits that give the correct output
do not appear to do this. Q1. What is the likely cause of this
behaviour Q2. The layout of tracks on the PCB has been changed
recently (although not the pin connections to the AD595) Could
factors such as pad size and shape or track width be causing this
behaviour.
Answer
First thing to check is that you have provided a path for the input bias
current of the input amplifier. If possible, connect pin1 to ground ( as shown
by the dotted line in figure 1) or if you require the thermo couple to be at a
voltage other than 0V, connect a 10k resistor between pin 1 and ground. Also
check which "ground" you use in your system. if you hare powering the AD595
from a 24V single supply, ground in this case is the 0V line - not the mid
point of the two supplies.
For the cold junction compensation to work, the AD595 must be at the same
temperature as the cold junction; the AD595 and the cold junction must be on
the same thermal gradient. Remember the cold junction is the point at which
Alumel and Chromel wires meet the copper tracks of the PCB. Ensure that the
cold junction is indeed on the PCB at the connectors marked CON2 and not remote
from the PCB. As you increase the ambient temperature, see if you can measure
the actual temperature at the cold junction and at the AD595. Any difference
in the thermal gradient will could cause the drift you are seeing. Factors such
as pad size and shape should not affect accuracy but certainly location of the
AD595 relative to the terminal block will have a significant affect on accuracy
for the above reasons.
You can check that the Cold Junction compensation circuitry on the AD595 is
operating correctly by connecting in Celsius thermometer mode as shown in
figure 13 of the datasheet. To do this, short pins 1 and 14 together with a
short connection and measure the voltage directly at pins 8 and 9. At 0degC,
the output should be 0V (or within the gain and offset specs for the electrical
grade you are using). The output voltage will rise by 10mV/degC. If it does,
you have proved that the ice point compensation is working correctly. If it
does not, please get back to me and we can look further.
Remember that long thermocouple wires make an excellent ariel and will quite
happily pick up any RF (or mains) interference in your lab. It may be necessary
to add a filter between the TC and the AD594. And certainly if the problem is
related to RF interference, touching the AD595 could change the output.