The DMA only supports the widths mentioned in the error. You could expand your signal using a “concat” block and concatenate zeros.
You would ignore the upper bits.
There will some inefficiency, but may be good enough for your use case.
Regards,
Cathal