AlbertA
Fractal Fanatic
Will this become the new "audiophile" myth in the modeler age? "The soundspace was not as open until I reinstalled 5 times".
Fl7x is right. There are quite a few checks along the way and for bit errors to go undetected on all of them is a low probability event:
1. Checksums in TCP/IP - yeah burst of errors can go undetected here...
2. CRC32 of the firmware sysex file in the zip - assuming random corruption, the undetected error here is less than 1 in a billion. However, CRC32 is 32 bits so there's a chance of collisions, which increases the undetected probability rate.
3. USB transmission of the firmware, which also does CRC checks
4. Firmware is broken into various sysex exclusive messages, each having an 16-bit checksum (unless this has changed, I haven't looked in a while) CRC would be better but it's something... Detects single bit errors, but two bit errors going undetected is 1/16
Yes undetected bit errors are possible they are just a low probability event.
Just speculation as we don't now how the firmware is laid out, but a few things already mentioned:
1. Corruption in flash storage.
2. Firmware bugs that are affected by global state/system params/other things - which is where various units will differ...
One way to verify is there's such a thing actually occuring is to record a dry track + wet track (just in case). Reamp track in the faulty unit with a known patch.
Reamp with another unit (forum volunteers? and examine any differences.
I wouldn't mind volunteering for the re-amp.
Fl7x is right. There are quite a few checks along the way and for bit errors to go undetected on all of them is a low probability event:
1. Checksums in TCP/IP - yeah burst of errors can go undetected here...
2. CRC32 of the firmware sysex file in the zip - assuming random corruption, the undetected error here is less than 1 in a billion. However, CRC32 is 32 bits so there's a chance of collisions, which increases the undetected probability rate.
3. USB transmission of the firmware, which also does CRC checks
4. Firmware is broken into various sysex exclusive messages, each having an 16-bit checksum (unless this has changed, I haven't looked in a while) CRC would be better but it's something... Detects single bit errors, but two bit errors going undetected is 1/16
Yes undetected bit errors are possible they are just a low probability event.
Just speculation as we don't now how the firmware is laid out, but a few things already mentioned:
1. Corruption in flash storage.
2. Firmware bugs that are affected by global state/system params/other things - which is where various units will differ...
One way to verify is there's such a thing actually occuring is to record a dry track + wet track (just in case). Reamp track in the faulty unit with a known patch.
Reamp with another unit (forum volunteers? and examine any differences.
I wouldn't mind volunteering for the re-amp.
Last edited: