esd diodes as diodes

Isn't MMBT3904 cheaper than your other low-leakage friend?

Can't imagine CMOS gates are healthy for more than 10mA, even if you don't care about latching (e.g. use a whole chip for positive clamps only, leave Vss open). Huh, latching would cause the chip to short all other inputs to the same rail... not pretty.

Tim

-- Deep Friar: a very philosophical monk. Website:

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

As suggested by someone here before, I am using BAV99W (actually, should be BAV99N, since it is narrower than BAV99). 2pF C and 50uA Ir. ESD diodes are either too high in C or too expensive. BAV99W is less than a 5 cents and smaller than SOT-23.

I presume the MUX is also CMOS? Thus it has its own ESD diodes. If you "use" ESD diodes from another chip it's likely all you will get is current sharing and still inject substrate current into the MUX.

Only Germanium or some Schottky's will give you some margin.

How many channels do you need to protect?

Transistor arrays (bipolar) would allow semi-precise clamping right at rail potential. ...Jim Thompson

Lots of chips have latchup current ratings, often 50 mA or some such. My series resistor could be 5 or 10K, so I wouldn't expect much clamp current. Something like an HC240 makes 16 dual clamps, 32 low-leakage diodes, for 60 cents or some such.

I was just wondering if anybody did this and knew of gotchas. Or has other suggestions for clamping a lot of analog signals without having to pick-and-place a lot of parts. I'm thinking about doing a cheapish

John

't

ve

ts

HC240 give you 8 channels.

-1-ND give you 18 channels for same size and same cents per channel.

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

You could use a series resistor before hitting the "protection" chip, then another resistor before feeding the chip you want to protect. That would prevent current hogging.

Many CMOS chips these days just use an Nfet for ESD. You get a diode clamp to the ground rail but a breakdown (impact ionization snap back) clamp for positive hits.

Stating the obvious, there needs to be some limiting on the positive rail if you put a protection diode to it, otherwise the external event will lift the positive supply rail for the whole board. Most regulators only regulate when sourcing, but not sinking.

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 =A0 |

You can add a zener to the supply rail. TVS diodes are just fast zeners. Regular zeners are too slow for signals but not a problem for the power rail.

I proposed adding 26 TVS diodes to our board, but they saw the cost and said "to hack with it, do a BP" (risk the company on it).

At over a buck a pop? Nah. A 74HC240 costs around 15 cents.

The gotcha is that you have to guestimate the total simultaneous spike energy, when it hits nearly all inputs. For example, at some point the bond wire to GND or VCC is going to go ... plink ... *PHUT*

I remember seeing a S100 dynamic memory board with something like a 74ls30 used as damping diodes...

If I set all the pins to tri-state, it has 16 i/o pins, all hi-z, each with two ESD diodes.

That's interesting. I'd have to test it for actual leakage, since this is an analog application.

John

I had in mind using lower rails for the clamp, so the esd diodes in the analog mux can never get forward biased. In most analog muxes, once you go beyond the rails enough for the diodes to conduct, the series switches start to turn on and one over-range channel wrecks all the rest.

Sure, I can handle that part.

John

don't

leave

nputs

..

Yes, cost is a problem. But this little buggy takes 200W peak power, it can certainly beat the HC240 in protections. It does not need to waste silicon for logics.

Those are OK, or the lower-leakage BAV199s. The '99s typically leak around 5 nA at room temp, and a lot more hot, which would pump a lot of offset into my input resistors, or the customer's signals. An actual 50 uA would be lethal. But I'm doing a 64-channel, differential-input ADC, so I'd need 128 of them. That's a lot of pick-and-place.

John

Yes, but at up to 7.2V at 1mA, it does nothing for John L.'s analog mux issue. The mux doesn't want to see inputs get just over the 5V rail, as odd things happen to other input channels.

Grant.

I'm sure the pick'n'placer wont complain too much ;)

You can run the mux chip at 6V? Clamp to 5V at the front for mux, look after the ADC separately on other side of mux?

Grant.

don't

leave

Yes, that's a little-known glitch in most analog mux's.

John

Or when using a cheap CMOS chip of the 74HC type, run its GND a diode drop above GND and its VCC a diode drop below the mux supply. Bypass both independently and clamp those artificial rails somewhere so they can't be pumped up. That way the mux substrate diodes shall never see any significant currents going into them.

Ideally there should be another resistor from clamp to mux pin but that adds pick&place time unless John would use arrays.

Interesting idea. I'll see if it's twistable into an active, lo-Z, clamp ...Jim Thompson

Hey, I wanted to patent that ... oh, too late ...