Part Number: PGA970
Dear TI Team,
We have used the PGA970 in some of our projects. Following reports from end customers, we have noticed issues with the consistency of the data provided by the PGA970.
In general, we use this component to read LVDT sensors without any further processing. Therefore, the ARM is turned off, and data transfer occurs only via SPI (configured at approximately 900 kHz). Our analysis revealed that the demodulation registers (32 bits wide) read via SPI are corrupted by the PGA970 itself. Specifically, reading the entire register requires two 16-bit SPI requests. Apparently, the PGA lacks a mechanism to prevent these registers from being updated during the read operation. Therefore, when reading a 32-bit register, the first 16 bits (MSB) belong to the first sample, while the second 16 bits belong to the next sample. This issue is most evident when the value straddles the 16-bit boundary (as a raw numerical position). For example (for a generic demodulation register), given two consecutive samples as follows:
0x0000 FFEA
0x0001 0023
sometimes the SPI read result is:
0x0001 FFEA
or
0x0000 0023
As seen in the example, the two acquisitions made by the PGA970 may, at the SPI transfer level, be mixed in certain cases, corrupting the final data.
Since these issues were identified in a safety-critical project (under DO-178C), I need to provide definitive documentation to submit to the certifying authority (EASA). Is it possible to obtain an “errata” document or an official document confirming this behavior (with a tracking number)?
Thanks in advance.
DS
