Editing Tandberg Data TDV 5000 Series (section)

==Firmware Differences==
Four firmware revisions of part 968551 have been analyzed, but there are very likely more. Much of the functions remain the same between these versions, but there are some behavioural differences one should know if planning to use low-level features of the keyboard. In short, v1.4 contains a few notable bugs fixed by v1.7. On top of this v1.9 adds better data-collision handling, and v2.1 fixes an edge-case when recovering from overflow.

===Scancode for key C12 in scancode-set 3===
v1.4 bug: This key will return scancode 0x5D in scancode-set 3 instead of the expected 0x53. This a bug, and it comes from the developer assuming this key in scancode-set 3 having the same scancode as in set 2. 0x53 is correct for the European ISO layout the keyboard sports, whereas it should have been 0x5C in the case of an American layout.

===Set Key Mode Commands===
v1.4 bug: In v1.4 you are expected to set the mode of a key one at a time, and the command is automatically terminated after this. In later versions you are instead expected to keep sending the scancodes of several keys after initiating the command. As a result, for v1.7 to v2.1, an "Enable Scanning" command is required in order to terminate any of the commands setting the mode of individual keys (even in the case the mode of only one key is being changed).

===Key-release of NOTIS-keys in scancode-set 1===
v1.4 bug: Key-release scancodes for the NOTIS-keys in scancode-set 1 will be prefixed with 0x80 0xF0 (as is custom in scancode-set 2 and 3) in addition to setting bit 7 of the base scancode (as is the proper way in scancode-set 1). By v1.7 this is fixed, and the unexpected 0xF0 byte has been removed from the prefix.

===Numlock-bypass release for navigation keys in scancode-set 1 and 2===
For 102-keyboard navigation-keys and PrtScr (originally these were only on the numeric pad of the IBM 83/84-key model F), a make/break-pair of an 'ignore num-lock' scancode is added around the base scancode make/break-pair if num-lock is active. This is as described in the IBM technical references, and also the exact behaviour for v1.4 in all cases. However, in v1.7 to v2.1, the break-part of the 'ignore num-lock' scancode is inserted prematurely, before the next make-scancode if another non-navigation key is pressed while the navigation key is still being held. This enables keys on the numpad to be pressed when any navigation key is being held, respecting the currently selected num-lock state.

===Shift keys and queue overflow===
v2.1 will try to keep re-sending missed shift-key break-codes before anything else when recovering from a queue overflow. v1.4 to v1.9 does not have special handling for this, and will not re-send any scancodes missed to a queue overflow. Due to this, if an overflow happens when a shift key is held on the earlier versions, and shift is let go before recovering, the shift-key will stick until it's pressed again.