Russell McMahon wrote: > > Microchip says... > > > > "The current datasheets specify Vss-0.3 to Vdd + 0.3v" > > > > "If you start curve tracing the devices you will find that the ESD > > protection diodes clamp at 0.6v. The 0.3v specification is intended to > > prevent the input protection diodes from conducting. We have learned > > that as more analog peripherals are placed on the device, including > > low power POR/BOR circuits, the currents created when the ESD diodes > > clamp can cause very bad behavior. In one example, the POR circuit > > asserted POR when the ESD diodes were conducted. We do not want the > > ESD diodes conducting so we have specified the voltage range more > > tightly." > __________ > > At last it's official. > All the things I have kept telling people for years about how not to > (ab)use protection diodes, and why, are now nicely summed up in an > official statement. I can now rest in peace and cease my evangelical > campaign against the use of protection diodes for use as anything > except as protection diodes. > > Yeah. Right! > Betcha it's business as usual tomorrow. > People will just keep on insisting on their divine right to inject > arbitary amounts of current into random places inside the IC oblivious > to the pain and heartbreak it causes the poor old processor and its > proliferating peripherals. > > The fact is, it's always mattered sometimes. And picking which time is > the time when you care, or whether you can get away with it in a given > design has always been a matter of luck rather than good design. > > Would-be top class engineers will keep right on insisting that you can > do it safely, and that they've done it for years, and that they have > 37 million devices in the field which are doing it with no bad > consequences whatsoever. Walk quietly away, making no sudden moves, > with your hands clearly in sight. > > So. > Just remember. > When you're tempted to say "but it's only a really really little > current I want to send to its doom down this convenient arbitrary path > into the heart of my precious program processor" ... Recall > > "Wafer thin!" > > > Russell > > > If that doesn't make sense, Gargoyling "wafer thin mint" might. And > might not. The result you get by abusing protection diodes may not be > as spectacular, but it's liable to be even more horrid. > Fully agree with everything you say. But I'm being stubborn and want to ask the questions that follow from this (not previously unknown) revelation. We could write a book about this but I'll try not to... ESD protection diodes are there to protect against cataclysmic damage, and they operate with or without VDD power to the device. No problems when power off, but inappropriate processor operation can occur when power is on. There is an absolute maximum current these diodes can handle, and we can design to prevent destruction of the chip. But there is no ESD diode current at which processor operation can be guaranteed, except zero! If the voltage on any I/O pin exceed Vss- 0.3V or Vdd+0.3V, or any current passes through these diodes at all then we take the PIC outside of it's normal/reccomended operating conditions and there are no guarantees it will function as intended. Not a good place to be for a designer. So how to properly protect an input against ESD, over voltage, induced EMF, and the like, and keep fully within the operational spec of the PIC? Series resistor and +ve and -ve clamping diodes is typical. Silicon diodes no good as Vf is approx 0.6V or more. Schottky is better as Vf is more like 0.3V but only at low currents. Take a BAT54 for example. Max forward drop at 1mA, 25 deg C, is 320mV. This puts us outside Microchips spec of 0.3V. Other ways? Zener diode can clamp +ve voltages ok but -ve we are still stuck with 0.6V. Need to use two back to back. Getting the precise zener voltage to make this work well is tricky. VDR, transorbs etc, not as precise clamping as Zeners. Buffer transistor(s)? Yes. Power consumption becomes a factor, reduced input impedance, bi-directional I/O no longer possible. External buffer chip. Yes. Take 4000B series CMOS gates for example, add a series resistor and the internal protection diodes do their job - like they are supposed to! :-) Adds a lot of extra real estate though if otherwise not required. My own conclusions/opinions? I think it's really quite serious and shouldn't be brushed away lightly. Microchip should take this seriously and build chips with proper input protection circuits, fully spec'd, that can be used as such. The situation as it stands now is not good enough. What to do for now? Schottky diode clamps are probably the best bet, and what I use. Although as I've stated they may clamp above 0.3V for any realistic input currents, if this stays below 0.6V you are *probably* not going to induce too much ESD diode current and this will *probably* not cause any serious problems. Just a reminder this is outside of the absolute maximum ratings and you do so at your own risk. PS. Thanks for the education on "wafer thin mint" - watched it on YouTube! -- Brent Brown, Electronic Design Solutions 16 English Street, St Andrews, Hamilton 3200, New Zealand Ph: +64 7 849 0069 Fax: +64 7 849 0071 Cell: 027 433 4069 eMail: brent.brown@clear.net.nz -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist