The iNES format implies 8 KiB of PRG RAM at $6000-$7FFF, which may or may not be battery backed, even for discrete boards such as NROM and UxROM that never actually had RAM there. This inspired some people on the nesdev.com BBS to come up with circuits to add PRG RAM to the original boards, so that games relying on it can run on an NES. The primary problem is in producing the enable signals for a 62256 or 6264 static RAM or compatible PSRAM.
On the forum, kyuusaku and Bregalad discussed PRG RAM decoder circuits built from 7400 series parts to approximate this behavior in an NES cartridge board. The first tries took two chips[1] or had possible timing problems.[2][3] They settled on the following circuits:
kyuusaku suggested a circuit based on a 74HC10 (triple three-input NAND) stick a pulldown on CE2 to take advantage of Phi2 going high-impedance during reset in order to "offer some write protection".[4]
,-------------- ROM /CE | ____ /ROMSEL --+--| `-. | \ A14 ---------| )o-- RAM /CE | / A13 ---------|____,-' ____ +5V ------+--| `-. | | \ `--| )o-- ROM /OE | / R/W ------+--|____,-' | `--------------- RAM /WE Phi2 ---------+----------- RAM CE2 | < < "big R" < | GND ----------+----------- RAM /OE
He also suggested a circuit based on a 74HC20 (double 4-input NAND), which appears to be the same one in Family BASIC:
The pinout:
Kevin Horton suggested the same circuit.
You could also use the other gate to invert R//W for /OE on the ROM to prevent bus conflicts.
If you don't need bus conflict prevention, you can use a 74HC139 (double 2-to-4 decoder), which may be cheaper or have better timing than a 74HC20. This circuit resembles the decoder in Jaleco's discrete mappers (87 and 140), which uses a 74139 to decode a single mapper register to $6000-$7FFF.
Proof:
1A0 | 1A1 | 1/Y3 | 2A0 | 2A1 | 2/Y3 |
---|---|---|---|---|---|
0 | x | 1 | x | x | 1 |
1 | 0 | 1 | x | x | 1 |
1 | 1 | 0 | 1 | x | 1 |
1 | 1 | 0 | 0 | 1 | 1 |
1 | 1 | 0 | 1 | 1 | 0 |
See further suggestions from kyuusaku.
The PlayChoice version of Mike Tyson's Punch-Out!! uses an extra IC to add battery-backed RAM. The digits in existing photos are hard to read, but it is believed to be a 74HC139. Its wiring has not been traced.
One thing that can complicate adding PRG RAM to a board is the fact that /ROMSEL and M2, used together to decode $6000-$7FFF, do not change at the same time. /ROMSEL is the logical NAND of M2 and PRG A15. This is accomplished by sending M2 and PRG A15 into a 74LS139 two-to-four line decoder on the NES main board. This introduces a small delay of up to 33 ns between the time M2 rises and the time /ROMSEL falls.
If this delay is too long it can cause unintentional writes to PRG RAM when writing to mapper registers $E000-$FFFF.
This is not a problem for the original cartridge hardware because the RAM chips used require a /WE (Write Enable) pulse of at least 50ns to 70ns depending on the chip. This means that the spurious /WE signal generated by this delay (MAX 33ns) will not be sufficient to trigger a write on the RAM chip. The circuits above give even more head-room as they tie PRG RAM /OE to ground and decode to /CE. The /CE to end of write timing is typically longer than the minimum /WE pulse width.
If your RAMs are faster than these timing specifications, your decoding logic must delay M2 by about 33 ns to match the /ROMSEL delay, as in the 74139-based circuit shown above. In this post, lidnariq suggested adding a resistor and capacitor:
card edge M2 --- 1k --- + --- 7420 | 33pF | GND
This delay is obtained by decreasing the voltage rise according to R·C. However, digital chips have minimum permissible rising speed, so the rise speed cannot be too slow because it might produce oscillation inside the digital input. Another option might be not to delay M2 but instead filter quick pulses on RAM /CE down by using capacitor connected to ground or a power supply:
______________________________________________________ card edge M2 --------| combinatory circuit that outputs |----+-----|/CE card edge /ROMSEL ---| 0 when M2 = '1', /ROMSEL = '0', A14 = '1', A13 = '1' | | | card edge A14 ---| 1 otherwise | 1nF | RAM card edge A13 -------|______________________________________________________| | |_________ GND or +5V
If you latch data/address on the falling edge of M2 of writes to $4020-$7FFF, you don't need to worry about this delay because /ROMSEL still has the correct logic level at this point.
Adding battery backup to RAM may be desired for maintaining data while the console is off. Most RAMs have a special low-power data retention mode, which decreases current consumption to few microamps, but the following need to be fulfilled:
D1 ___________ 5V -------------|>|--+ | RAM |---+--| VCC 3.3V battery----|>|--+ | | D2 R1 | | | RAM /CE decoding logic----+--| /CE |___________
D1 ensures that the 3.3 V battery only powers the RAM, not the whole cartridge when in standby mode, D2 keeps the 5 V from charging the battery (permissible if the battery is rechargeable), R1 (100 k should be enough) ties /CE at high level when power is cut off.
If your RAM /CE decoding logic does not become high impedance in power off mode, the voltage at RAM /CE might drop after power down (for example, the AX5202P DIL40 pirate MMC3 chip seems to have its RAM /CE output at very low resistance with respect to ground when not powered, and the above circuit after power loss makes RAM /CE voltage drop to 1 V which leads to data corruption). To protect against that situation, RAM /CE decoding logic path must become open circuit after powering off. The following transistor is essentially a switch that does the job:
D1 ___________ 5V --------------------------|>|--+ | RAM |---+--| VCC 3.3V battery-----------------|>|--+ | | D2 R1 | | | RAM /CE decoding logic--- E C ------+--| /CE \____/ |___________ | B NPN 5V -----------------1k------+
When no 5V is present or 5V is present but RAM /CE decoding circuit outputs high level, transistor is open. When 5V is present and RAM /CE decoding circuit outputs low level, RAM /CE becomes low.
See also: Battery circuit schematic