PIA TTL: Falling Edge Detection Issue (800XL PAL)

[8bit-Dude]

I am getting ready to launch the Atari 8bit cable for my 8bit-Hub (see: http://8bit-unity.com/?page_id=551)

 

The system connects to Joy Port 2, using PIA for I/O. Rapid toggling of the first pin (falling edge TTL) is required to trigger interrupts on the Arduino.
I have carried out testing with Filippo on various systems, and found the following:

 

  - NTSC 800 (1 unit): OK
  - PAL 800 (1 unit): OK
  - PAL 600XL (1 unit): OK
  - PAL 800XL (4 units, some modded with U1MB, Sophia2...): ALL NOT OK!
  - NTSC 1200 XL (1 unit): OK
  - PAL 130 XE (2 units): OK

As you can see above, the PAL 800 XL is the only model giving an issue with TTL edge detection.
I wonder if there is anything different with the chips used in 800XL PAL region???

Edited January 10, 2022 by 8bit-Dude

[Rybags]

Have you tried swapping PIA with another machine where it works OK?

 

Other than that all I can think of is resistors/components on the port lines - by the schematics it looks like the 800XL ports are done differently to the 600XL.

[TGB1718]

Not all PIA chips are equal, although they may have similar part numbers, they can

differ slightly in how they perform

[ClausB]

See this thread:

 

[8bit-Dude]

Thanks for referring this thread ClausB, some very interesting information in there.

I am going to try and play with switching frequency, to see if that improves the situation. Maybe not enough time is allocated for the 5V signal to rise back to the point where it can trigger a falling edge again...

[_The Doctor__]

it pulls down fast but does not rise fast, you can try to add a slight pull up to help or you can detect the lows or you you can let the Atari pump low and invert the signal or give a longer recharge time as you describe by spreading out the time ie frequency... but it may be best to pick two so to speak

[Rybags]

You'd probably want several cycles worth of high value at least.  Is there a required frequency for this?

Generally the 6502 is sufficiently slow anyway that you should be holding it high for sufficient time.

[8bit-Dude]

After some further investigation, it turns out that the issue is with the Joystick Trigger (used as Acknow flag).

Switching the Joystick state from low to high seems to take an eternity on the 800XL.

I am going to experiment with inverting the default state (low instead of high).

[Rybags]

You can use latches on TRIG though the main purpose is for 1->0 transitions (button pressed)

[8bit-Dude]

1 hour ago, Rybags said:

You can use latches on TRIG though the main purpose is for 1->0 transitions (button pressed)

 

AFAIK, that's what the Arduino does in Digital Output mode.

 

I set the state to HIGH / LOW, which should cause the Trigger to transition betwen 1 / 0.

 

This works fine on 800, 600XL, 130XE, but on the 800XL it takes such as long time that a ldx #255 counter decrements all the way to 0 before the trigger transitions from 1 to 0. It is as if the Atari is fighting against being pulled down on that line...

 

Edited January 15, 2022 by 8bit-Dude

[TGB1718]

Don't the trigger's use CTIA/GTIA ?, not PIA ?

[Rybags]

Yes, but they can be subject to the normal array of resistors, caps, inductors or whatever crap Atari saw as flavour of the month at the time.

Also you have some TRIGs potentially shared as a lightpen input to Antic which might affect behaviour as well.

 

Then you have the FGTIA (French) which has to free up pins to do the SECAM output so TRIGs on it are read in serially (per scanline?) and no idea if latches work at all.