A special mmc3 board

[byemu]

I found a special mmc3 board with a 74hc32!
board name:HVC-TK1ROM-01
game:http://bootgod.dyndns.org:7777/profile.php?id=1515

anyone know this board
here is big pic :http://i.imgur.com/Q8iWVSh.jpg

[lidnariq]

The 74HC32 allows them to use a 28-pin 128KiB CHR ROM.

[Quietust]

There are several other boards like that - the MMC3 also has TL1ROM and the MMC1 has SC1ROM, SH1ROM, and SL1ROM (though, curiously, only NES-SL1ROM is like that - HVC-SL1ROM appears to lack the 7432).

[Bregalad]

What is special is that, apparently, the batters is soldered on the bottom side.

PS : Since there is supposedly only a single OR gate used to OR /CE with /OE into a single /ENABLE signal, two diodes and a resistor could have done the trick too.

[lidnariq]

PS : Since there is supposedly only a single OR gate used to OR /CE with /OE into a single /ENABLE signal, two diodes and a resistor could have done the trick too.

That's actually not true. Diodes turn into capacitors when reverse biased, and at 2.5MHz that effect could be large enough to break things. The SROM PCB uses NPN transistors instead.

[Bregalad]

That's actually not true. Diodes turn into capacitors when reverse biased, and at 2.5MHz that effect could be large enough to break things.

Mmh really ? Pehaps it depends on the diodes ? I suck at analog electronics...

The SROM PCB uses NPN transistors instead.

If you use "CMOS" (or it's bipolar equivalent) logic you'd need 6 transistor for an OR gate - 4 for a standard NOR gate in series with a 2 transistor inverter.

Of course they could ignore the PNP pull-up transistor and replace them by pullup resistors, then it'd be 3 transistors. Was it what they did on the SROM PCB ?

[lidnariq]

Not inverting logic; they're using two NPN transistors as pull-ups and the resistor as a pull-down. It doesn't draw too much current from the PPU /RD or PPU A13 lines because it's limited by the resistor. (1K might be wrong? it seems a little low, but I think the stripe's red, not orange...)

[lidnariq]

I hadn't looked at HVC-SL1ROM before. It's amusing: they implemented A>B in discretes. Unlike HVC-SROM which implements OR as specified above, I think they've got /A13 > /RD here:

[Bregalad]

OK, and what is the advantage of an NPN transistor mounted as a diode over real diode that would be optimized to have a low reverse capacity ? Price maybe ?

[lidnariq]

What they're using on SROM is a common collector amplifier of a sort, and is really not the same as the diode circuit you're positing at all (that connection to vcc actually changes everything)... The biggest difference is that BJTs don't really go into reverse-biased-variable-capacitor mode. Using two schottky diodes there would work, true, but they have much higher reverse leakage current.

I have no idea if they were more expensive 30 years ago, or by how much. Now both options are around 100USD/10k.

I should go see if I can remember enough SPICE to persuade it to illustrate the differences.

[80sFREAK]

PS : Since there is supposedly only a single OR gate used to OR /CE with /OE into a single /ENABLE signal, two diodes and a resistor could have done the trick too.

That's actually not true. Diodes turn into capacitors when reverse biased, and at 2.5MHz that effect could be large enough to break things. The SROM PCB uses NPN transistors instead.

Have you ever heared about P-i-N diodes? Absolutely not trolling, just these days you can find silicon diodes with very low reverse capacitance.

Bregalad, one gate IC maybe?

[lidnariq]

Ok, so I sat down with LTspice and made some graphs. They're all huge numbers of pixels but trivial number of bytes, so I'm providing links rather than uploading them to the forum. (Is there a way to do this less stupid?)

Here's the schematic I used to test it- http://eamp.org/li/schematic.png [883x445]

Here's a graph of ≈10 PPU pixels (starting with two nametable fetches)- http://eamp.org/li/alltime.png [945x920]
The 1N914s (fairly standard signal diodes) have very large transient currents, even when (not shown) an ohm of resistance is put in series with the signal lines.

Here's a zoom of the moment when the PPU switches from nametable fetches to pattern table fetches- http://eamp.org/li/crossover.png [945x920]
I assume a 10ns transition speed for the lines, without any evidence that that's true. It's a fairly standard assumption made with digital logic. If so, the diode OR here would fall below the "input low" voltage, and the output might glitch. That shouldn't matter—this happens during the start of the PPU's ALE signal, so the right value should win on the data bus moments later, but is bad for power consumption.

Here's a zoom of a successful read (A13 low, /RD low) - http://eamp.org/li/bothtrue.png [945x920]
At 5mA, the 1N914s seem to capacitively couple (or is it inductive effects?) when forward biased, so there's undershoot when PPU /RD falls. Also, the BJTs pinch off current from the signal lines, so although current is drawn from /RD and A13 during the change, after the voltage across the resistor has risen, all the sustaining current comes from the voltage rail.

Finally, here's a trace of replacing the 1N914s with 1N5817s, the first Schottky diode that came up in the library provided with LTspice- http://eamp.org/li/schottky.png [945x920]
That doesn't look very good. I have no idea if this is fair, however.

Regarding p-i-n diodes, LTspice doesn't have one in its library, so I was uncertain which one I should test with.

My conclusion: BJTs have clear advantages. It is probably unfair to use a 1kΩ resistor for the diode OR. Re-testing with a 10kΩ resistor provides a much saner graph, but still higher static power dissipation from the signal lines.

[blargg]

This circuit simulation stuff is really useful for avoiding traps you can barely detect. I also wasn't aware of a diode's capacitance, though it makes sense given its geometry.

[Bregalad]

I completely forgot about this thread (in 1 year), but there is several stuff *really* interesting !

Not only this is the only Nintendo-made PCB with a surface mounted 74-series chip I've ever seen, but also, if they hid a chip below the battery on this board, phaps the've hidden other chips under batteries on other boards ?

Thank you very much lidnariq for your spice simulation of diode-logic, it is very useful and informative. I had completely missed it back then, so I should thank you properly for your work.

Finally, the circuit using the transistor instead for /A13 > /RD is very funny and informative. This could be used to remove bus conflicts easily on the PRG side without a chip, by doing R/W > /ROMSEL.