LSM6DSV burst read of fifo

[Geoffrey1]

It appears that the longest spi operation that could be used to access the LSM6DSV (and related parts) would read a single fifo entry (7 bytes) and it is not possible to read all the available fifo entries in a single dma operation.  Am I missing something?

 

Geoffrey

[Federica Bossi]

Hi @Geoffrey1 ,

The LSM6DSV sensor's FIFO can be accessed using burst SPI or DMA transfers. The device supports automatic address increment, which allows you to perform efficient multiple byte reads or DMA transfers from the FIFO data output registers (addresses 79h to 7Eh). This means you are not limited to reading a single FIFO entry (7 bytes) at a time; instead, you can read multiple entries in one operation, maximizing flexibility and minimizing communication overhead, especially when flushing large amounts of data from the FIFO

The practical limit is set by the FIFO's total capacity and the host's DMA capabilities, not by the sensor itself. For efficient operation, configure the FIFO and interrupts appropriately, and use burst or DMA reads to retrieve all available FIFO entries as needed.

[Geoffrey1]

I'm sorry, but I don't see it. Yes you can read the seven fifo bytes in a burst, but to read the next 7 bytes in the same burst addressing has to be circular - i.e. restarting at the tag register .

This does not appear to be the case. every read of a fifo entry requires an address be transferred and seven bytes read.

This each read of a fifo entry is a new 8 byte transaction . This isn't the way dma works. While it is possible to create circular transactions, dma doesn't handle the chip selection line necessary to enable back to back transactions

[Geoffrey1]

I’m revisiting this so others don’t get confused.  The behavior Frederica Bossi describes is not documented in the datasheet.  There is an obscure sentence in an application note (AN5922) 

 

“FIFO output data must be read with multiple of 7 bytes reads starting from the FIFO_DATA_OUT_TAG register. The rounding function from address FIFO_DATA_OUT_Z_H to FIFO_DATA_OUT_TAG is done automatically in the device”

Rounding is a fairly odd word -- wrapping perhaps.  In any case a whole sentence would be more appropriate such as:

Multiple 7 byte reads of the fifo can be performed in a single read request because the internal counter automatically wraps the address from FIFO_DATA_OUT_Z_H to FIFO_DATA_OUT_TAG once FIFO_DATA_OUT_Z_H is read.

Frankly, I’m stunned by the poor state of documentation since this is a fairly fundamental feature.

Thus, what she says is correct, but it is largely undocumented, and not at all in the natural location -- the datasheet.