AD7891-1: First conversion is incorrect

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

we would like to know your opinion concerning a problem we are facing with
component "AD7891-1".
We use this component in bipolar way, with input voltages between +10V and
-10V. Data reading is effected in parallel.
During our testing we faced the following problem: sometimes the first reading
after PCB starting is wrong (it doesn't matter which channel we read);
outcoming bits do not correspond to the input voltage. All following readings
are then correct. We can state that the first reading is effected when all
voltages are surely stable.


- Reference hasn't had time to power up fully yet. Are you using the internal
reference or an external reference? The internal reference takes 2ms to power
up to 12-bit accuracy.(AD7891 datasheet pg 14 left hand side). The external
reference may take longer, especially if you have a large decoupling capacitor
on the output. Check this by measuring ref out after power up or using an
external reference which you know powers up immediately.Be careful to respect
the absolute maximum rating
Reference Input Voltage to AGND = –0.3 V to VDD + 0.3V

- Not enough time between writing to the part to set the channel and starting
the first conversion by bringing CONVST\ high. (If a conversion is started by
writing to the SWCONV bit, the acquisition time is automatically controlled and
you don't have to worry about it.) The track hold acquisition time is 600ns for
the AD7891-1.

- The T/H goes into Track mode at the end of a conversion. If the T/H powers up
in HOLD mode (no reset pin on the IC, therefore internal logic will be
undefined on power up), the first conversion may not be operating on an
acquired signal but on a spurious value . If this is the case then you will
simply have to perform a "dummy" conversion, ignore the result of conversion
number 1 and start acquiring results from conversion number 2.