view doc/FT232R-notes @ 128:95c2a67e1219

doc, linux-patch: update for failure to mainline DUART28C support
author Mychaela Falconia <falcon@freecalypso.org>
date Tue, 02 Feb 2021 04:32:42 +0000
parents 026dd69e4ebb
children df4bf4e06221
line wrap: on
line source

Unlike FT2232x devices with external EEPROMs, an FT232R device is not expected
to ever have a blank EEPROM in normal usage: these chips have their EEPROM
built in, and FTDI probably ships them with this internal EEPROM already
programmed.  I said "probably" because I have not yet had an occasion to build
my own FT232R-containing board where I would be getting completely pristine
"bare" chips from Digi-Key, thus I have no first-hand verified knowledge.

As an experiment, I have programmed "blank" (0xFFFF in every word) images into
the two FT232R devices I have available for play at the moment (specifically
devices which I could afford to brick if things went badly), and FT232R behaves
the same way as FTDI's earlier chips with external EEPROMs: it runs with a fixed
default config when the EEPROM is invalid.  But this configuration is NOT
recommended for production use - you should always have a valid EEPROM config
in your FT232R chip.

When our FreeCalypso fteeprom tools were first put together in 2019-04, I was
getting erratic behaviour: when I tried to program my own EEPROM config
generated with ftee-gen232r, the resulting EEPROM content became a bitwise AND
between the previous image and the new one, as if the "EEPROM" is not really an
erasable memory, but one of OTP kind where ones can be turned into zeros, but
not the other way around.  I was doing this experiment on a no-name FT232RL
adapter from ebay, thus my first thought was that the FT232RL chip was bad, a
less-than-perfect clone rather than genuine FTDI.  But then I bought a UB232R
module from Digi-Key (presumably containing a genuine FT232RQ chip), and it
behaved the same way.

Further investigation revealed that FT232R EEPROM write operations work
correctly only if they are preceded by this magic sequence:

	ftdi_usb_reset(&ftdi);
	ftdi_poll_modem_status(&ftdi, &modem_status);
	ftdi_set_latency_timer(&ftdi, 0x77);

I can see how FTDI could have reasonably implemented a sort of safety lock on
their EEPROM write operations, allowing them only if a special unlock sequence
has been given - but it completely baffles me why they are doing some sort of
OTP emulation in the absence of the right magic sequence, as opposed to
disabling EEPROM writes altogether.  It is worth noting that this magic sequence
is NOT needed for programming external EEPROMs behind FT2232x chips - were FTDI
folks being deliberately malicious in designing their FT232R chip to simulate
appearance of being bricked when it is programmed with older (or third-party)
software tools that don't know the new magic sequence?  Who knows...