LSM6DSO32 FIFO Compression and FSM's

[Bobby123]

We are using the FIFO compression feature of the LSM6DSO32 which is working well for us if we use it standalone at 833 Hz. When we add the 4 FSM's that handle our motion recognition we start seeing duplicate tags in the compressed output implying that data has been dropped.

In AN5473 it says the FIFO compression will work up to 833 Hz in section 8.10. There is no mention of a dependency on FSM's that I have been able to find - Can someone confirm that this is indeed our issue and the only paths forward are to:

1. Decrease the datarate to the compressed FIFO
2. Stop using FSM's and get compressed FIFO data at 833 Hz
3. Handle compression of the sensor data on a host processor

Are there any other options we could pursue?

Thanks

[Eleon BORLINI]

Hi @Bobby123​ ,

The procedure you are using should be right. Did you check if you correctly implemented the FIFO compression, as for example shown in the C example on Github lsm6dso32x_compressed_fifo.c?

-Eleon

[Bobby123]

Hi @Eleon BORLINI​ 

We are fairly certain that the FIFO compression is implemented correctly as laid out in the application note. We have been pulling the compressed FIFO data and parsing it correctly for a few months.

Our issues arose when we went from 2 FSM's to 4 FSM's at the 833Hz. The frequency of the dropped tags appears to be proportional to each additional FSM after 2. We get significantly more dropped tags at 3 FSM's and very frequent dropped tags at 4 FSM's. Dropping to 416Hz significantly decreases the number of dropped tags but does not solve it entirely.

With the tag only being a 2 bit counter we cannot tell how many samples were dropped, only that some were.

Is this a known issue? We can create a UCF register configuration file that replicates the issue - would it be helpful to share this configuration file, or a binary of the compressed fifo data pulled off the LSM6DSO32?

Thanks

[Bobby123]

@Eleon BORLINI​ Can you confirm this is an issue and whether there are any workarounds? We have are coming up on production and have purchased a couple hundred thousand of the LSM6DSO32 and this appears to be an issue that is not described in the datasheet or any of the application notes. Dependency between the compression of the FIFO and the FSM's seems strange if these are truely happening in hardware - our guess is that this is all running on an internal microcontroller that is being overwhelmed - can you confirm that that this is the architecture?