Transformation tables

[Graynomad]

At present an entire transformation table must be uploaded to the MCU, however if the transform function is linear the table input/output max/min values could be uploaded instead, thus only 8 bytes can define the transform function rather than a 1000-byte (or whatever) table.

[garylcamp]

Not my area of knowledge but I always advise low traffic where possible and this is especially true on a low speed network.

[Graynomad]

Admittedly uploading a transformation table is a one off event, but I still think just sending a few bytes to define the table is a good idea.

The next question is, does the MCU use that info to create a lookup table in RAM (wasteful of space but faster) or will it do the interpolation on the fly when a Point value is received.

[garylcamp]

Typing out loud (as it were)...
If the table is of more than one format (linear end points vs 1000 byte TABLE or ?) then you would need a way to identify it as such. If it is big, you might not want to send it often. If there is local room, add a cheksum for safety and a flag saying it is in RAM and allow it to be overwrote as needed and request a resend. I am leaning to keeping it in RAM, which gets bigger and cheaper every day. In ten years, it will be free. In 20, they will pay you to take more RAM (helps the economy).

[Graynomad]

If the table is of more than one format (linear end points vs 1000 byte TABLE or ?) then you would need a way to identify it as such

True, this needs more work.

In theory a TT should only be sent once, or at least very infrequently. I agree keeping it in RAM is the way to go, with maybe an option to burn into EEPROM as a power fail backup.

Naah, if the power fails spend 5 minutes uploading again. This is supposed to be a system that stays up all the time.