nesdev.com

I have read many times on here, as well as in other places, that using a Low Voltage chip (3.3v) in a 5v environment won't work properly. However, I haven't found anywhere that actually says what happens. I have wired a 29L3211 chip (using a 3.3v regulator) to a hirom snes cart, and the game I tested seemed to work fine without incident...so that leaves me wondering, what should I be expecting?

So, that being said, is it just a matter of "it isnt the proper way to do it" or will a game stall or memory fry?

There is also mention on here about level shifters, but even after googling it I haven't found any specific examples of what I should use for this application. What type of shifting system or chip could I use in a cart to make it reliable / have the proper logic between chips? (that part I'm still really green on).

The chip just dies eventually. The ESD protection diodes are sucking up a lot of power turning the 5V into 3.3V and eventually die and then you'll get a catastrophic failure in the chip. The die will pretty much burn up. You can power a LED from a 5V source directly without a current limiting resistor but it won't be lit for a long time...

_________________
http://www.tmeeco.eu

That and you could have the chip malfunctioning at odd times. It's like overclocking a computer, where you trade reliability for speed, which raises the question of what value it is if it's not reliable. Read the spec sheet and in particular the absolute maximum ratings. Beyond these and you damage the chip. Beyond the normal specs and you get unreliable operation.

Even with a voltage regulator? Seeing as the regulator is only putting out the required 3.3v to the chip, wouldn't that stop the chip from ever being exposed to the higher voltage, thus saving the chip? I don't see how, with a 3.3v regulator, that the memory chip would fry from over-voltage...is there something else thats a factor here?

The address and data pins are connected to the SNES bus which uses 5V signals. The chip has input protection diodes which divert signals outside the GND to +3.3V range (actually -0.6V to +3.9V, because the diodes have a voltage drop) to GND and the +3.3V supply. Thus, putting a 5V signal basically connects the power pin to 4.4V, thus powering the chip higher than spec. You can even have a logic chip powered without any connection to the power pin, and have it work fine until all of the signals go low, then have the chip get powered down and lose all its state.

The voltages into the chip on the address lines, if the chip isn't 5V tolerant, exceed what it's been tested with. The exact failure mode depends on the input stages on the part. If it's anything like the PIC's overvoltage protection, you've got a diode from the address line to +3V, and you're stressing the outputs of the SNES's 65816. This could eventually eat the SNES as well as the EEPROM.

If there's no overvoltage diode, you're probably driving a the n-type MOSFET in the buffers at the pins "too" far on, possibly causing gradual erosion of the MOS gate.

Ahh, that makes a lot more sense, thanks guys!

So what sort of level shifter could be used here to make this work? Is there a chip or schematic you guys would recommend?

Last time I was looking into this (FPGAs), the funny GTL FET things looked like the best deal (bits per dollar) — e.g. http://www.nxp.com/documents/data_sheet/GTL2000.pdf .

I recall an old Genesis Flash cart schematics that used 3.3v Flash, and instead of level shifter ICs it used a pile of resistors, one per every address and data line.

So though people that have been putting this 29L3211 into SA-1 and SDD1 cartridge will eventually see the chip fail?

And could this really damage the SNES itself?

I think level shifters are common, aren't they ?

They are but when people hack up carts they are pretty lazy. If you do direct replacement with a 29LS3211 for SA-1 that's part of the hope to avoid having to use an adapter or a ton of wires. But if all the signal lines need level shifters on them then you need wires or an adapter board anyway.

I think that's the example HardwareMan used when I asked him why his infinitenes flash cart used 'limiting resistors' vice level shifters. 'input/output limiting resistors' don't actually solve the problem. You can use a voltage divider on inputs, but for bidirectional data lines you can't use level shifters. Putting a single resistor in series doesn't actually change the voltage present at the pin. While you are effectively creating a voltage divider with the ESD diodes and the input resistor, you're still exceeding maximum chip ratings...

Level shifters are crazy cheap and common. I understand why 'hackers' don't bother with them because you can't put them on donor carts very easily. But for new designs/PCBs we're only talking about 30-50cents. Don't be lazy/skimpy, Just use em!!! Plus they are easier for assembly compared to using a TON of resistors that don't even fully resolve the issue.



I wouldn't expect it to damage the SNES/console. But it's hard to say exactly what will happen when things fail. Before failure you'd be drawing some extra current on the address and data lines, but the console can *probably* handle that okay, I've never measured what that current is, nor do I know the ratings of the CPU's.

But once the ESD diode fails, all bets are off. My guess is they'll shoot-trough and short to Vcc/Gnd but that's just speculation. Because it'll stop working you'll probably stop continually putting stress on the console, and keep from causing permanent damage but it's still uncertain what will happen to these 20+ year old line drivers inside your CPU...

_________________
If you're gonna play the Game Boy, you gotta learn to play it right. -Kenny Rogers

If the EEPROM has overvoltage diodes, it will definitely cause the chip to slowly die by overvoltage, and is loading the address lines of the SNES, and could—since you're basically bypassing the 3V regulator, and the address lines aren't rated for the ≈50mA sustained current—eventually kill it.

I just checked out Macronix's IBIS models: at least all the newer parallel NOR flash they make they do have overvoltage diodes.

Ok, so I looked around and came across this chip 74ALVC164245DL.

http://www.ti.com/lit/ds/symlink/sn74alvc164245.pdf

Looking at the data sheet this looks like exactly what would work...however, it is really confusing (data sheets are not my forte). How would I hook this up between the 29L3211 chip and the address lines? I will be designing a custom PCB for this project, to run upwards of 64Mbits. Does anyone feel up to drawing out a diagram?