Apple1 Video Terminal - undestend.

Hi! First time poster here.I'm in the middle of synthesizing this part of the circuit on an FPGA. I'm probably still a a few days away from getting it working, but here's what I think is happening....The pattern you see at startup is due to some glitchy behaviour of the uninitialized cursor circuit. Immediately after startup the 2504's that store the screen characters are all clear, so the entire screen is filled with '@'. The 2504 at C11 is used to store the cursor position, with 1 bit per screen character. A '0' indicates that the cursor is at the current position and a '1' indicates otherwise. During ordinary operation only one of these bits should be '0', but at startup this entire chip is also clear. Its output is fed through a D flip-flop in C13 and the negated output drives the "CURS" signal a cycle later. To see what happens next, look at the upper right hand corner of C7; you'll see CURS being fed into one of C6's NAND gates. This is the part of the circuit that signals for a character to be written into memory. In order for this to happen the PIA normally sends the DA signal into C7 to indicate that a character is ready, and then a cycle later it gets fed into the C6 NAND gate which then waits for the CURS signal to go high. Once both conditions are met the WRITE line gets triggered and the character gets written into memory. Where does this character come from? RD1-RD7, which are also being fed in from the PIA. What I believe is happening is that all of these lines (i.e. DA and R1-R7) are set to an initial state of '1' when the PIA is first powered up. Ordinarily this would cause a '?' character to be written to screen memory (i.e. 1111111), but the 555 at D13 controlling the cursor blink is also starting up at the same time and it keeps the MSB low via the C12 "AND" gate that it feeds into, and so a 0111111 underscore character gets written instead.So why do we see an entire page of this? Because the 2504 at C11 is still filled with '0' bits (i.e. cursor position indicators). During startup the circuit sees the first '0' in this chip, thinks the cursor is present, writes an underscore character to the next position and clears that cursor bit by setting it to '1' (via the AND gate directly below C11). What I'm not quite clear on yet (and wont' be until I examine the RTL simulation) is why it's doing this for every second character instead of each one. Either way, something causes the write to be skipped the next cycle, and so we get an alternating pattern of '@' and '_' which the 555 timer then starts blinking half a second later. If we could see the contents of the C11 cursor chip we would see 01010101.... and since both the repeating pattern and the 2504 size are multiples of 2 a stable image is generated until the main CPU and PIA get around to initializing the screen properly.

No, my theory doesn't require any processor. It does, however, require something to keep the DA and R1-R7 lines initially high, if not a PIA then internal pull-up resistors or something.