Well, it looks that openMSX replicates the behavior of the LS07 gates at pins 6,7. When these bits are LOW the reading of the pins is tied to zero,
// pin 6,7 input is ANDed with pin 6,7 output
And the code of the test program writes a Zero to pins 6,7
ld a,15 out (0a0h),a ; reg 15 ld a,80h out (0a1h),a ; pins low <-- LOOK HERE ld a,14 out (0a0h),a ; reg 14 in a,(0a2h)
The same for the check tap routine of DM system 2 driver
CHECK: CALL PORSEL ; INC B ; B = Max player number + 1 AND 0C0H ; A is state of port 15. OUT (0A1H),A ; 678=L for both joystick ports (!) <-- LOOK HERE EX AF,AF' ; save state of Reg15 with all output pins low LD A,14 ; select register 14 OUT (0A0H),A IN A,(0A2H) ; read state of joystick port AND 20H ; 7=H ? return if bit is high (!) (shouldn't be forced high, its being driven low internally ) RET Z ; if return here, A = state of pin 6, and B=B+1 from the function call
On a real MSX the Ninja TAP circuit is forcing a HIGH level on pin 7, that is the reason why the real hardware is behaving differently from the emulation.
That is the reason why the Ninja Tap circuit looks critical in respect to the use of a chip from the ACT series, that can source much more current than the LS gate inside the MSX can sink to maintain the level HIGH and be detected by the driver (and test program)
Mystery solved :)
I see two problems in the original circuit:
1) The diode D1 is inverted. @Jipe, can you please double-check if it was inverted on the circuit, or maybe a mistake when reverse engineering it?
2) The resistor R1 should not be present. The MSX General Purpose ports already have pull-up resistors on all input pins, and placing another resistor in parallel will cause the resistance to be too strong. Many MSX models have 4K7 pull-up resistors, so this will result in a 824 Ohms resulting resistance. Way too strong.
The Shinobi-Tap version fixed the diode direction, but the R1 resistor also should not be present, even if its value is now 4K7 instead of 1K.
