5V Tolerant Ideas

[MrAureliusR]

I'm in a bit of a jam. I really would strongly prefer to use a CPLD or FPGA for a design I'm working on, but (some) of the inputs need to be 5V tolerant. However, looking at Digi-key, the only 5V tolerant CPLDs (that aren't Microchip -- their programmable logic tools are absolutely terrible and have zero support, or so I've been told by someone at MC) are Lattice. I actually really like Lattice parts, but all those ispMACH CPLDs are marked as Not For New Designs or Obsolete. MicroSemi lists a bunch of FPGAs as 5v tolerant but they are super pricey and way out of the BOM.

I'm curious if anyone has any ideas. I need probably 50-75 I/O pins, most of which would be 5V. This chip will be replacing a LOT of 74 series chips and a couple other small things, so it doesn't need a huge amount of logic. My current solution ideas are:

• Use diode clamping and resistors to limit the current
• Use 74LVC or some other level-shifting buffers (I would need to get them in a really small package, and ideally with the same pin pitch as the FPGA so they can literally sit right in front of the pins and level-shift)
• Try and run the 5V logic at 3.3V (which may or may not work, and probably not *reliably*

This board could potentially be ramped up to the tens of thousands produced, so I need chips that are active and will be available for a while. I've been searching for an easy solution, and it seems like there aren't any, but I would be really curious to hear what others have to say, and how people are dealing with 5V programmable logic interfaces now that the Lattice parts are going obsolete.

[Yansi]

Diode clamping = what you will do with the current injected in the supply rail? That's just bad idea.  Also the resistors will trade off speed.

If it was only a bunch of 74 logic, wha not using some older CPLDs like Altera MAX7000, which are fully 5V? These can still be sourced.  Good enough it is only a design for a run of only a couple of boards. I think that Xilinx still makes XC95 series CPLDs that are/should be 5V tolerant (but not sure, never used Xilinx parts myself).

And what about using a LVC logic (bus drivers) to convert signals from 5V to 3V3?

Also, there are bus xceivers that do just this, 16 lines at a time: https://assets.nexperia.com/documents/data-sheet/74LVC_LVCH16245A.pdf (which is also in a SSOP48 package with a 0.5mm pin pitch)

Try and run the 5V logic at 3.3V (which may or may not work, and probably not *reliably*

Why to "try"? Logic levels are pretty darn good defined for all logic families. 74HCT (and any TTL) will work absolutely fine with 3V3 as logic H. Remember that TTL H is anything above 2.4V.

[phil from seattle]

Lots of buffers/line drivers have 5V tolerant inputs.  For example 74LCX760.  Or the venerable 74LS07 using 3.3V pullups. For 3.3 to 5V output, HCT logic?

[MrAureliusR]

Diode clamping = what you will do with the current injected in the supply rail? That's just bad idea.  Also the resistors will trade off speed.

If it was only a bunch of 74 logic, wha not using some older CPLDs like Altera MAX7000, which are fully 5V? These can still be sourced.  Good enough it is only a design for a run of only a couple of boards. I think that Xilinx still makes XC95 series CPLDs that are/should be 5V tolerant (but not sure, never used Xilinx parts myself).

And what about using a LVC logic (bus drivers) to convert signals from 5V to 3V3?

Also, there are bus xceivers that do just this, 16 lines at a time: https://assets.nexperia.com/documents/data-sheet/74LVC_LVCH16245A.pdf (which is also in a SSOP48 package with a 0.5mm pin pitch)

Try and run the 5V logic at 3.3V (which may or may not work, and probably not *reliably*

Why to "try"? Logic levels are pretty darn good defined for all logic families. 74HCT (and any TTL) will work absolutely fine with 3V3 as logic H. Remember that TTL H is anything above 2.4V.

I'm talking about the other part of the circuit we are interfacing to. They are older microcontrollers and such, and they are all 5V spec'd. In my experience they can *sometimes* run at 3.3V, but it's not really reliable enough for production.

I have been looking at bus transceivers, but hadn't come across that chip. Thanks for pointing it out, I will take a look at it!

[nctnico]

Definitely Xilinx XC95XL series. If you power them from 3.6V then the outputs are even 74xxHC compatible.

[Yansi]

There is no sometimes. Look what the voltage level spec of the MCU IOs is, and design for it. Violating the spec would be a mistake, even if it seem to work. The specs are there for a reason.

So are the IOs you need to interface exactly a 5V cmos levels, ie. 0.3Vdd ViLmax and 0.7Vdd ViHmin?  Then you really need a bus level translator ICs.

For example exactly this will convert between 3 and 5V logic, both directions: SN74ALVC164245
http://www.ti.com/lit/ds/scas416q/scas416q.pdf

[nctnico]

0.7*5=3.5V so at 3.6V the minimum is satisfied.

[NorthGuy]

TI has sn74lvc16T245 in TVSOP (0.4mm pitch) and BGA (0.65mm pitch) level shifters which can use any voltage between 1.65 and 5.5V on either side. You'll probably need only 3 or 4 of these.

[Yansi]

0.7*5=3.5V so at 3.6V the minimum is satisfied.

I wasn't replying to you, but yes, will be satisfied. Although I find using such non standard supply voltages a bit awkward. 

I think using a proper level translator won't hurt anyone. SN74ALVC164245, SN74LVC16T245, etc. Especially Texas Instruments has loads of these. I have used also such as TXS0108E, these are bidirectional and do not need direction steering signal (but they are obviously not of the fastest).

[maginnovision]

The atf1508 are 5v tolerant, still active, Enough IO for your app and the tools are fine. They have WinCUPL and prochip which allows you to use vhdl or verilog. I've only used Wincupl for them because It's more familiar to me but as far as I know prochip works just fine.

[SiliconWizard]

I second the use of buffers in the 74LVC245 series. I've found the NXP line (74LVC16245A has 16 I/Os) to be very good performance- and protection-wise. In the TSSOP48 package, the pitch is 0.5mm which would most likely fit your CPLD/FPGA pitch if you use a QFP package.

As for the CPLD/FPGA part, since you're thinking of using Lattice parts, I'd suggest switching to the MachXO2 line.

[technix]

I have been investigating TXB0108 and TXB0104's. Maybe they cna work for your case too?

[jmelson]

Xilinx 93xxXL CPLDs are 5V tolerant.  I'm using them in a couple places.
I use the SN74ALVC164245 translators where the CPLD or FPGA connects to the outside world, they are VASTLY more robust against ESD than the Xilinx 3.3 V FPGAs and CPLDs.  I've had some boards where the ALVC164245 had holes blown out of the package, but the FPGA survived.  So, the only place I trust the 5V tolerance is where the 3.3 V - 5V interface is wholly within one board.

Jon

[FrankBuss]

For my Kerberos project I've used 74LVC8T245 level shifters to output a nice 5V signal, and a XC95144XL CPLD, which has 5V tolerant inputs and runs at 3.3V, controlling a SRAM and flash chip, which run at 3.3V as well. Not thousands of boards, but hundreds of boards work without problems.