How to use CRC32 generator for an effective CRC16?

Question

I am writing in C for an embedded STM32F105 microcontroller. I need to implement a CRC routine to validate a message sent over the air.

The microcontroller has a CRC32 generator built into its hardware. You feed it 4 bytes at a time and it calcs the CRC without additional processor overhead. It's non-configurable and uses the Ethernet CRC32 polynomial.

I want to use this hardware CRC generator, but I only want to add two bytes (not four) to each data packet. The packet will vary in size between 4 and 1022 bytes.

Can I simply use the two high (or low) bytes of the CRC32? Or can I always feed the CRC module 2 bytes at a time, with the high bytes being zero?

Is there some other way to get what I'm looking for?

Answer 1

For most applications, sure, you can just use the low two bytes of the CRC-32 as a 16-bit check value. However that will not be a 16-bit CRC. It will be as good as any other hash value for checking for gross errors in a message.

It will not have certain desirable properties for small numbers of bit errors in short packet lengths that are afforded by CRCs.

There's no point in feeding the CRC generator zeros. Go ahead and give it four bytes of data for each instruction.