I'm using AD8139 as general purpose fully differential amplifier (FDA). The AD8139 is powered via ±4.2V with VOCM = 0V and acts as a fully differential current to voltage converter. Please see the following schematic for details.
The amplifier significantly heats up (drawing more than 70mA, 600mW) under the following conditions:
- When the DAC (AD9744) is under SLEEP mode and BIAS (-1.25V) is not applied (Hi-Z)
- When the DAC is outputting a middle code (each IOUT outputs 10mA), regardless of whether BIAS is applied
In both cases, the voltage on the +IN/-IN pins are close to 0V. The heating is reduced when the DAC is outputting close to lowest / highest code (IOUTA/B = 0mA, 20mA; or 20mA, 0mA, respectively), AND when BIAS is not applied. The +IN/-IN pins reach ~0.7V due to the clamping diodes.
It seems that the amplifier heats up when +IN/-IN terminals approach 0V (even worse when they are around -0.1V), and the power consumption reduces as the common mode input voltage increases (~0.7V reduces net consumption to estimated ~40mA, but still this is higher than the specs on the datasheet).
When the amplifier consumes excessive current, I measured node voltages of all nodes, and they show that the excessive current does not flow through the resistors, i.e. current consumption is due to internal circuit of this amplifier.
Additionally, when shorting +IN/-IN, regardless of what condition, the power consumption reduces to normal, or at least the part is not heating up.

Please also find the LTSpice simulation attached. The simulation did not demonstrate excessive power consumption under ideal or non-ideal conditions (tweaked to simulate resistor mismatch, current source mismatch / non-idealness, etc).
Any advise will be greatly appreciated, and thank you in advance.
