Are 5v-tolerant inputs clamped?

Started by December 29, 2010
I need a bunch of inputs, so I'm using an i2c port expander.  The one
I've chosen runs on 3.3v, but tolerates 5v inputs.  I don't see any
documentation on the operation of the inputs.

I'm trying to avoid wasting 32 clamp diodes if I don't need them.

Q: Generally speaking, if I feed a 2mA current-limited inductive
glitch into a +5v-tolerant input, Vdd=3.3v, is the glitch voltage
clamped by the device input?  At what voltage?  Most importantly, is
the device damaged?

The datasheet says inputs can handle +6v max and +/-20mA per input.

If needed I guess I could use a dummy logic device as a clamp-diode
array.  Hmmm.

Thoughts?

TIA,

James Arthur
On Wed, 29 Dec 2010 13:07:31 -0800 (PST), dagmargoodboat@yahoo.com
wrote:

>I need a bunch of inputs, so I'm using an i2c port expander. The one >I've chosen runs on 3.3v, but tolerates 5v inputs. I don't see any >documentation on the operation of the inputs. > >I'm trying to avoid wasting 32 clamp diodes if I don't need them. > >Q: Generally speaking, if I feed a 2mA current-limited inductive >glitch into a +5v-tolerant input, Vdd=3.3v, is the glitch voltage >clamped by the device input? At what voltage? Most importantly, is >the device damaged? > >The datasheet says inputs can handle +6v max and +/-20mA per input. > >If needed I guess I could use a dummy logic device as a clamp-diode >array. Hmmm. > >Thoughts? > >TIA, > >James Arthur
As far as I know, "5V-tolerant" inputs are NOT clamped. What would they be clamped to? There's no 5V VDD on the chip to strap a diode to. All that is there is a so-called "snap-diode" (to ground) for ESD pulses. ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
On Dec 29, 4:10=A0pm, Jim Thompson  wrote:
> On Wed, 29 Dec 2010 13:07:31 -0800 (PST), dagmargoodb...@yahoo.com > wrote: > > > > >I need a bunch of inputs, so I'm using an i2c port expander. =A0The one > >I've chosen runs on 3.3v, but tolerates 5v inputs. =A0I don't see any > >documentation on the operation of the inputs. > > >I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > >Q: Generally speaking, if I feed a 2mA current-limited inductive > >glitch into a +5v-tolerant input, Vdd=3D3.3v, is the glitch voltage > >clamped by the device input? =A0At what voltage? =A0Most importantly, is > >the device damaged? > > >The datasheet says inputs can handle +6v max and +/-20mA per input. > > >If needed I guess I could use a dummy logic device as a clamp-diode > >array. =A0Hmmm. > > >Thoughts? > > >TIA, > > > As far as I know, "5V-tolerant" inputs are NOT clamped. =A0What would > they be clamped to?
Some protected MOSFETs clamp overvoltages by incorporating a zener from drain-to-gate. I didn't expect zeners en masse in a digital part, but possibly some equivalent function.
> =A0There's no 5V VDD on the chip to strap a diode > to. =A0All that is there is a so-called "snap-diode" (to ground) for ESD > pulses.
Yeah, crap. Reading app notes for the device family, I'm not confident the device survives >6v inputs, current-limited or not. I'm thinking I'll use a dummy logic part for the clamps. An 'HC244 (outputs disabled) gives me 18 clamps to Vdd and Vss. Does that horrify you? :-) It's only a mA or two, worst case. -- Cheers, James Arthur
On Dec 29, 4:39=A0pm, dagmargoodb...@yahoo.com wrote:
> On Dec 29, 4:10=A0pm, Jim Thompson wrote:
> > As far as I know, "5V-tolerant" inputs are NOT clamped. =A0What would > > they be clamped to? > > Some protected MOSFETs clamp overvoltages by incorporating a zener > from drain-to-gate. =A0I didn't expect zeners en masse in a digital > part, but possibly some equivalent function. > > > =A0There's no 5V VDD on the chip to strap a diode > > to. =A0All that is there is a so-called "snap-diode" (to ground) for ES=
D
> > pulses. > > Yeah, crap. =A0Reading app notes for the device family, I'm not > confident the device survives >6v inputs, current-limited or not. > > I'm thinking I'll use a dummy logic part for the clamps. =A0An 'HC244 > (outputs disabled) gives me 18 clamps to Vdd and Vss. =A0Does that
Oops---------------------------^^^^^^ Make that 16 clamps--the enables have to be hardwired '1'. -- Cheers, James Arthur
On 29 Dec., 22:10, Jim Thompson  wrote:
> On Wed, 29 Dec 2010 13:07:31 -0800 (PST), dagmargoodb...@yahoo.com > wrote: > > > > >I need a bunch of inputs, so I'm using an i2c port expander. =A0The one > >I've chosen runs on 3.3v, but tolerates 5v inputs. =A0I don't see any > >documentation on the operation of the inputs. > > >I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > >Q: Generally speaking, if I feed a 2mA current-limited inductive > >glitch into a +5v-tolerant input, Vdd=3D3.3v, is the glitch voltage > >clamped by the device input? =A0At what voltage? =A0Most importantly, is > >the device damaged? > > >The datasheet says inputs can handle +6v max and +/-20mA per input. > > >If needed I guess I could use a dummy logic device as a clamp-diode > >array. =A0Hmmm. > > >Thoughts? > > >TIA, > > >James Arthur > > As far as I know, "5V-tolerant" inputs are NOT clamped. =A0What would > they be clamped to? =A0There's no 5V VDD on the chip to strap a diode > to. =A0All that is there is a so-called "snap-diode" (to ground) for ESD > pulses. >
seem to remember Xilinx describing the clamping on their old 5V tolerant fpga/cplds as a ~6V zener like diode to ground but if the input signal are current limited well below the max limits and there's no risk of input current lifting the supply is should be fine, Xilinx specificly spec that inputs should be limited to Vcco + 0.5V OR 10mA -Lasse
On Wed, 29 Dec 2010 13:39:01 -0800 (PST), dagmargoodboat@yahoo.com
wrote:

>On Dec 29, 4:10�pm, Jim Thompson
>Web-Site.com> wrote: [snip]
>> >> As far as I know, "5V-tolerant" inputs are NOT clamped. �What would >> they be clamped to? > >Some protected MOSFETs clamp overvoltages by incorporating a zener >from drain-to-gate. I didn't expect zeners en masse in a digital >part, but possibly some equivalent function. > >> �There's no 5V VDD on the chip to strap a diode >> to. �All that is there is a so-called "snap-diode" (to ground) for
ESD
>> pulses. > >Yeah, crap. Reading app notes for the device family, I'm not >confident the device survives >6v inputs, current-limited or not. > >I'm thinking I'll use a dummy logic part for the clamps. An 'HC244 >(outputs disabled) gives me 18 clamps to Vdd and Vss. Does that >horrify you? :-) It's only a mA or two, worst case.
Doesn't horrify me at all... a clever solution to a problem that's been mentioned here before! Only caution... output devices can be big and CAPACITIVE, so you'll take a hit if you're trying for high speed. ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
On Wed, 29 Dec 2010 13:43:03 -0800 (PST), dagmargoodboat@yahoo.com
wrote:

>On Dec 29, 4:39�pm, dagmargoodb...@yahoo.com wrote: >> On Dec 29, 4:10�pm, Jim Thompson wrote: > >> > As far as I know, "5V-tolerant" inputs are NOT clamped. �What
would
>> > they be clamped to? >> >> Some protected MOSFETs clamp overvoltages by incorporating a zener >> from drain-to-gate. �I didn't expect zeners en masse in a digital >> part, but possibly some equivalent function. >> >> > �There's no 5V VDD on the chip to strap a diode >> > to. �All that is there is a so-called "snap-diode" (to ground) for
ESD
>> > pulses. >> >> Yeah, crap. �Reading app notes for the device family, I'm not >> confident the device survives >6v inputs, current-limited or not. >> >> I'm thinking I'll use a dummy logic part for the clamps. �An 'HC244 >> (outputs disabled) gives me 18 clamps to Vdd and Vss. �Does that > >Oops---------------------------^^^^^^ > >Make that 16 clamps--the enables have to be hardwired '1'.
I _was_ going to get the data sheet and count ;-) ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
On Dec 29, 1:39=A0pm, dagmargoodb...@yahoo.com wrote:
> On Dec 29, 4:10=A0pm, Jim Thompson > > > > > > > > > Web-Site.com> wrote: > > On Wed, 29 Dec 2010 13:07:31 -0800 (PST), dagmargoodb...@yahoo.com > > wrote: > > > >I need a bunch of inputs, so I'm using an i2c port expander. =A0The on=
e
> > >I've chosen runs on 3.3v, but tolerates 5v inputs. =A0I don't see any > > >documentation on the operation of the inputs. > > > >I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > > >Q: Generally speaking, if I feed a 2mA current-limited inductive > > >glitch into a +5v-tolerant input, Vdd=3D3.3v, is the glitch voltage > > >clamped by the device input? =A0At what voltage? =A0Most importantly, =
is
> > >the device damaged? > > > >The datasheet says inputs can handle +6v max and +/-20mA per input. > > > >If needed I guess I could use a dummy logic device as a clamp-diode > > >array. =A0Hmmm. > > > >Thoughts? > > > >TIA, > > > As far as I know, "5V-tolerant" inputs are NOT clamped. =A0What would > > they be clamped to? > > Some protected MOSFETs clamp overvoltages by incorporating a zener > from drain-to-gate. =A0I didn't expect zeners en masse in a digital > part, but possibly some equivalent function. > > > =A0There's no 5V VDD on the chip to strap a diode > > to. =A0All that is there is a so-called "snap-diode" (to ground) for ES=
D
> > pulses. > > Yeah, crap. =A0Reading app notes for the device family, I'm not > confident the device survives >6v inputs, current-limited or not. > > I'm thinking I'll use a dummy logic part for the clamps. =A0An 'HC244 > (outputs disabled) gives me 18 clamps to Vdd and Vss. =A0Does that > horrify you? :-) =A0It's only a mA or two, worst case. > > -- > Cheers, > James Arthur
On some designs I was involved in we used to use 74133, 13-input nand gates for the same function. 13 clamp inputs in a 16-pin package. kevin
On Dec 29, 4:51=A0pm, Jim Thompson  wrote:
> On Wed, 29 Dec 2010 13:39:01 -0800 (PST), dagmargoodb...@yahoo.com > wrote: > > > > > > >On Dec 29, 4:10 pm, Jim Thompson >Web-Site.com> wrote: > [snip] > > >> As far as I know, "5V-tolerant" inputs are NOT clamped. What would > >> they be clamped to? > > >Some protected MOSFETs clamp overvoltages by incorporating a zener > >from drain-to-gate. =A0I didn't expect zeners en masse in a digital > >part, but possibly some equivalent function. > > >> There's no 5V VDD on the chip to strap a diode > >> to. All that is there is a so-called "snap-diode" (to ground) for ESD > >> pulses. > > >Yeah, crap. =A0Reading app notes for the device family, I'm not > >confident the device survives >6v inputs, current-limited or not. > > >I'm thinking I'll use a dummy logic part for the clamps. =A0An 'HC244 > >(outputs disabled) gives me 18 clamps to Vdd and Vss. =A0Does that > >horrify you? :-) =A0It's only a mA or two, worst case. > > Doesn't horrify me at all... a clever solution to a problem that's > been mentioned here before! > > Only caution... output devices can be big and CAPACITIVE, so you'll > take a hit if you're trying for high speed.
Naa, they're solenoids and such--slow as dirt. I'm reading back the driver state at its output to make sure it really did what it was supposed to do. "Trust, but verify." Thanks for the feedback. James Arthur
On Dec 29, 4:46=A0pm, "langw...@fonz.dk"  wrote:
> On 29 Dec., 22:10, Jim Thompson > > > Web-Site.com> wrote: > > On Wed, 29 Dec 2010 13:07:31 -0800 (PST), dagmargoodb...@yahoo.com > > wrote: > > > >I need a bunch of inputs, so I'm using an i2c port expander. =A0The on=
e
> > >I've chosen runs on 3.3v, but tolerates 5v inputs. =A0I don't see any > > >documentation on the operation of the inputs. > > > >I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > > >Q: Generally speaking, if I feed a 2mA current-limited inductive > > >glitch into a +5v-tolerant input, Vdd=3D3.3v, is the glitch voltage > > >clamped by the device input? =A0At what voltage? =A0Most importantly, =
is
> > >the device damaged? > > > >The datasheet says inputs can handle +6v max and +/-20mA per input. > > > >If needed I guess I could use a dummy logic device as a clamp-diode > > >array. =A0Hmmm. > > > >Thoughts? > > > >TIA, > > > >James Arthur > > > As far as I know, "5V-tolerant" inputs are NOT clamped. =A0What would > > they be clamped to? =A0There's no 5V VDD on the chip to strap a diode > > to. =A0All that is there is a so-called "snap-diode" (to ground) for ES=
D
> > pulses. > > seem to =A0remember Xilinx describing the clamping on their old 5V > tolerant fpga/cplds as a ~6V zener like diode to ground > > but if the input signal are current limited well below the max limits > and there's no risk of input current lifting the supply is should be > fine, > Xilinx specificly spec that inputs should be limited to Vcco + 0.5V OR > 10mA > > -Lasse
A chip could have one 5v zener and a bunch of internal ESD diodes to clamp positive spikes--that works. If the input isn't clamped at all, ever, that would seem to make it unacceptably fragile--how would it survive minor ESD? FWIW: The most ESD-sensitive part I've ever used is the 2n7002. I can't remember ever killing another chip with static, but I've fried half a dozen 2n7002's just soldering them in, carelessly. -- Cheers, James Arthur
dagmargoodboat@yahoo.com wrote:
> > On Dec 30, 8:39 am, Spehro Pefhany > wrote: > > On Wed, 29 Dec 2010 20:54:00 -0800 (PST), the renowned > > > > dagmargoodb...@yahoo.com wrote: > > > > >16 clamps if I tri-state the drivers and use the outputs as clamps > > >too. I need 13 clamps, so that gets it done with 3 clamps to spare. > > > > We'll have to call you "Jed". > > Well, better Jed than Ellie-May.
Or Ellie-May is still better that Jethro or Granny. ;-) -- For the last time: I am not a mad scientist, I'm just a very ticked off scientist!!!
F. Bertolazzi:

> since 25 mA are way too much for an ESD buildup.
No, I did not write it. ;-) Atmel's AVR MCUs are fairly well ESD protected (never had such a failure), put their clamp diodes cannot take more than 0,5 mA.
> Yes, the SH/LD pin can be tied low.
No, it cannot. But it's not a problem: with an 8 bit expander you can use two lines for clock and LD/SH and four lines for reading in parallel the data from the four HC165s (including shorted drivers and open load detectors).
F. Bertolazzi:

> The input diodes of the HC165 are, in my experience, very robust, also the > datasheet says they can swallow up to 25 mA. A trivial resistor allows you > to protect the inputs very effectively.
Well, let's say 20 mA, but exactly as much as the 244 suggested by krw, solution that Jim Thompson, if I understood well, did not dismiss. By the way, I've seen an 8086 motherboard that used the 244 in exactly that way, just for protecting the inputs of a big MOS chip. With the HC165 I made some 300 boards (all of the firedoors of italian ferry-boats and many of french, dutch and greek ones are monitored by that board) and never had a single failure, despite the fact that ship's service engineers that installed them often are retarded monkeys.
dagmargoodboat@yahoo.com:

> On Dec 30, 10:38�am, "F. Bertolazzi" > wrote: >> dagmargoodb...@yahoo.com: >> >>> I need a bunch of inputs, so I'm using an i2c port expander. �The
one
>>> I've chosen runs on 3.3v, but tolerates 5v inputs. �I don't see any >>> documentation on the operation of the inputs. >> >>> I'm trying to avoid wasting 32 clamp diodes if I don't need them. >> >> If speed is not an issue and you really don't have 3 spare pins on the MCU, >> you coud use an 8 lines I/O expander and, with that, control and read a >> bunch of well known and cheap HC165. [edited, ja] >> >> You can daisy chain them but, if you don't, you have 9 inputs per chip. > > That's a decent idea--it'd take two parts, and eliminates the i/o > expander. It's clean except for using the ESD diodes of the active > device, which annoys our IC designer gurus (Jim Thompson and miso).
Well, actually I meant to use an I/O expander, but an 8-bit or less one, from which you can get the 3 I/Os needed to control the HC165, (that I thought had to be 4 for 32 inputs). The input diodes of the HC165 are, in my experience, very robust, also the datasheet says they can swallow up to 25 mA. A trivial resistor allows you to protect the inputs very effectively. This arrangement does not contradict the gurus, since those diodes are meant, in my opinion, for exactly the purpose you're looking for, since 25 mA are way too much for an ESD buildup.
> It also takes one output pin to clock it, and one input port pin to > read the '165.
Yes, the SH/LD pin can be tied low.
> Alas, I don't have them. Getting them would be > inconvenient. Possible, but inconvenient.
That's why I suggested to have the I/O expander anyway. But an 8 bit one would be cheaper and smaller (and, at least from Farnnel, much more readily available) that a 16 bit one, and, with just three 16 pin chips you would be all set.
> I'm still thinking about the whole thing. My current scheme can't > distinguish between a shorted driver and an open load. A shorted FET > is one of the most likely failures[*], so it'd be nice to detect it.
In this case my solution is even more convenient, because you just need two more HC165 to read that information.
On Dec 30, 10:38=A0am, "F. Bertolazzi" 
wrote:
> dagmargoodb...@yahoo.com: > > > I need a bunch of inputs, so I'm using an i2c port expander. =A0The one > > I've chosen runs on 3.3v, but tolerates 5v inputs. =A0I don't see any > > documentation on the operation of the inputs. > > > I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > If speed is not an issue and you really don't have 3 spare pins on the MC=
U,
> you coud use an 8 lines I/O expander and, with that, control and read a > bunch of well known and cheap HC165. [edited, ja] > > You can daisy chain them but, if you don't, you have 9 inputs per chip.
That's a decent idea--it'd take two parts, and eliminates the i/o expander. It's clean except for using the ESD diodes of the active device, which annoys our IC designer gurus (Jim Thompson and miso). It also takes one output pin to clock it, and one input port pin to read the '165. Alas, I don't have them. Getting them would be inconvenient. Possible, but inconvenient. I'm still thinking about the whole thing. My current scheme can't distinguish between a shorted driver and an open load. A shorted FET is one of the most likely failures[*], so it'd be nice to detect it. [*] or it was, anyhow. I've added short-circuit protection, so those failures should be greatly reduced. -- Cheers, James Arthur
F. Bertolazzi:

> HC161.
Better 165.
dagmargoodboat@yahoo.com:

> I need a bunch of inputs, so I'm using an i2c port expander. The one > I've chosen runs on 3.3v, but tolerates 5v inputs. I don't see any > documentation on the operation of the inputs. > > I'm trying to avoid wasting 32 clamp diodes if I don't need them.
If speed is not an issue and you really don't have 3 spare pins on the MCU, you coud use an 8 lines I/O expander and, with that, control and read a bunch of well known and cheap HC161. You can daisy chain them but, if you don't, you have 9 inputs per chip.
On Thu, 30 Dec 2010 07:01:56 -0800 (PST), dagmargoodboat@yahoo.com
wrote:

>On Dec 30, 9:47�am, Jim Thompson
>Web-Site.com> wrote: >> On Wed, 29 Dec 2010 20:02:54 -0800 (PST), dagmargoodb...@yahoo.com >> wrote: >> >> >> >> >On Dec 29, 6:24 pm, Jim Thompson > >Web-Site.com> wrote: >> >> On Wed, 29 Dec 2010 15:10:51 -0800, Tim Wescott >> >> wrote: >> >> >> >On 12/29/2010 01:07 PM, dagmargoodb...@yahoo.com wrote: >> >> >> I need a bunch of inputs, so I'm using an i2c port expander. The one >> >> >> I've chosen runs on 3.3v, but tolerates 5v inputs. I don't see any >> >> >> documentation on the operation of the inputs. >> >> >> >> I'm trying to avoid wasting 32 clamp diodes if I don't need them. >> >> >> >> Q: Generally speaking, if I feed a 2mA current-limited inductive >> >> >> glitch into a +5v-tolerant input, Vdd=3.3v, is the glitch voltage >> >> >> clamped by the device input? At what voltage? Most importantly, is >> >> >> the device damaged? >> >> >> >> The datasheet says inputs can handle +6v max and +/-20mA per input. >> >> >> >> If needed I guess I could use a dummy logic device as a clamp-diode >> >> >> array. Hmmm. >> >> >> >> Thoughts? >> >> >> >I think different companies are going to do it differently. >> >> >> >I can't remember whether it was 3.3V tolerant inputs on a 2.5V device, >> >> >or 5V tolerant inputs on a 3.3V device, but I do remember that there was >> >> >one Xilinx part that would sorta kinda turn on a body diode (maybe they >> >> >used two in series???) when you gave it the full rated input, leaking a >> >> >bit of current _into_ the chips main supply. >> >> >> >This was no problem, except that in the board we used it at startup when >> >> >there wasn't much current draw on that supply and all the inputs were >> >> >high, and it lifted the FPGA's supply rail. _That_ took us a while to >> >> >figure out. >> >> >> >I'll bet that your data sheet says whats what. >> >> >> Sounds like the LV inputs weren't "tolerant", but had classic ESD >> >> diodes. >> >> >> James Arthur is using the dummy device to +5V, so high outputs can't >> >> lift that supply. >> >> >Actually I figure on tying the dummy device to +3.3v. �That clamps the
>> >'real' inputs to <4-ish volts--totally safe for a 6v max. input. >> >> If your signal is coming from a +5V device, you'll be sorry ;-) > >It's from those inductive loads I mentioned. I'm knocking the 24v >source to 3.3v with a resistive divider, hi-z, which then feeds the i/ >o expander + clamps. > >> Unless you have a series R in there somewhere. > >22k, or maybe 47k.
OK. Much different from passing 5V logic outputs to 3V logic inputs (but 5V tolerant). ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
On Dec 30, 9:47=A0am, Jim Thompson  wrote:
> On Wed, 29 Dec 2010 20:02:54 -0800 (PST), dagmargoodb...@yahoo.com > wrote: > > > > >On Dec 29, 6:24 pm, Jim Thompson >Web-Site.com> wrote: > >> On Wed, 29 Dec 2010 15:10:51 -0800, Tim Wescott > >> wrote: > > >> >On 12/29/2010 01:07 PM, dagmargoodb...@yahoo.com wrote: > >> >> I need a bunch of inputs, so I'm using an i2c port expander. The on=
e
> >> >> I've chosen runs on 3.3v, but tolerates 5v inputs. I don't see any > >> >> documentation on the operation of the inputs. > > >> >> I'm trying to avoid wasting 32 clamp diodes if I don't need them. > > >> >> Q: Generally speaking, if I feed a 2mA current-limited inductive > >> >> glitch into a +5v-tolerant input, Vdd=3D3.3v, is the glitch voltage > >> >> clamped by the device input? At what voltage? Most importantly, is > >> >> the device damaged? > > >> >> The datasheet says inputs can handle +6v max and +/-20mA per input. > > >> >> If needed I guess I could use a dummy logic device as a clamp-diode > >> >> array. Hmmm. > > >> >> Thoughts? > > >> >I think different companies are going to do it differently. > > >> >I can't remember whether it was 3.3V tolerant inputs on a 2.5V device=
,
> >> >or 5V tolerant inputs on a 3.3V device, but I do remember that there =
was
> >> >one Xilinx part that would sorta kinda turn on a body diode (maybe th=
ey
> >> >used two in series???) when you gave it the full rated input, leaking=
a
> >> >bit of current _into_ the chips main supply. > > >> >This was no problem, except that in the board we used it at startup w=
hen
> >> >there wasn't much current draw on that supply and all the inputs were > >> >high, and it lifted the FPGA's supply rail. _That_ took us a while to > >> >figure out. > > >> >I'll bet that your data sheet says whats what. > > >> Sounds like the LV inputs weren't "tolerant", but had classic ESD > >> diodes. > > >> James Arthur is using the dummy device to +5V, so high outputs can't > >> lift that supply. > > >Actually I figure on tying the dummy device to +3.3v. =A0That clamps the > >'real' inputs to <4-ish volts--totally safe for a 6v max. input. > > If your signal is coming from a +5V device, you'll be sorry ;-)
It's from those inductive loads I mentioned. I'm knocking the 24v source to 3.3v with a resistive divider, hi-z, which then feeds the i/ o expander + clamps.
> Unless you have a series R in there somewhere.
22k, or maybe 47k. -- Cheers, James Arthur
On Wed, 29 Dec 2010 20:40:09 -0800 (PST), dagmargoodboat@yahoo.com
wrote:

>On Dec 29, 11:11�pm, Spehro Pefhany wrote: >> On Wed, 29 Dec 2010 19:52:34 -0800 (PST), the renowned "m...@sushi.com" wrote: > >> >No, it is like Jim described. [You wouldn't subject the gate oxide to >> >stress.] You depend on the fet to snap back. That is strictly ESD. If >> >you tied a hard source to the input and forced the protection device >> >to flow current big time, I'm sure the device would be very leaky >> >after such treatment. The absmax in the datasheet is what protects the >> >companies arse. BTW, damaged inputs are very easy to detect in the rel >> >lab via liquid crystal or emission scope. The FA engineer would curve >> >trace the input and immediately sense that you fried it, then do the >> >other tests to prove it. >> >> Do you think this app note is misleading? (page 5) >> >> http://www.freescale.com/files/microcontrollers/doc/app_note/AN2433.pdf > >That gate-to-drain n-FET with a high threshold is kind of like a >sloppy zener. If that were the protection _and_ it clamped to a safe >voltage, then I wouldn't need any external clamps at all. > >That's what I was hoping for. I mean, if it _didn't_ clamp to a safe >voltage, what good is it? > >But, absent info to the contrary, I'll clamp myself to be sure. > >> It's not very clearly written.. I don't see why the clamp voltage of >> the "thick field" transistor would vary with the Vdd, for example. > >Me neither.
I think you need to show us a schematic of this interface. You're about to commit a faux pas :-) ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.