Misbehaving 74HC4051s - leakage from overvoltage on unselected channels

On a sunny day (Wed, 12 Mar 2008 03:38:44 -0700 (PDT)) it happened Al Borowski wrote in :

You need to do the voltage division _before_ the 4051 input. That chip is not specified for 24 V.

Maybe you can solder some extra resistors to ground from the input pins.

One method: Move the capacitor C7 to the input side of the MUX. This means you now will have 50 capacitors. Make sure that the time constant is long enough that the micro can zip around all of the inputs and discharge them before they exceed VCC. You will have to change the code a bit too.

Another method: Run a Schottky diode from each input to VCC so that it can't turn on the substrate diode.

Third method: Find someone who makes a SOS version of the 4051. In the SOS part the protection diode isn't really the EB junction of a parasitic transistor.

Duh. Have you read the datasheet for the 4051? The chip is not the nut here...

Al Borowski a écrit :

That's a perfect illustration of why some of us here repeatedly say to NOT use the clamp diodes as part of the normal circuit, even if it's tempting.

Now for your pb look for fault protected switches. Some can swallow up to 40 volts input.

The esd diodes do clamp the inputs to maybe 0.7 volts past the rails. But that's still enough to turn on the series-pass fets in common-gate mode in some architectures; most 4051's will do this.

There are mux'es that don't do this, but they mostly have a different (DG508-type) pinout. I think somebody makes a better mux in the 4051 pinout, Maxim maybe. But the Maxim part may have latchup issues.

A schottly clamp might help, but keep in mind it's the pin voltage that's turning on the fets, not the conduction of the internal esd diode. A diode clamp to Vcc-0.7 is better, or your 4.7 volt zeners. The datasheets should point this out, as it causes repeated grief.

Clamp or divide down the inputs *before* the mux.

John

Makes no sense whatosever, and especially with C7 at 0.1u....

If you can tolerate so much error, you could probably just put voltage dividers on each input to map the full DC input range into the range of an ADC (and by extension the analog muxes before it).

Then you can do your DC offset removal in software.

Yes, it's noisier this way, but probably not too noisy to meet your loose requirement. You may also be able to take a lot of readings and average them in the time that you would have spent charing and discharging capacitors.

Hi John and everyone, thanks for the constructive replies. The easiest solution seems to be ditch the 4051 and drop in a heap of 74hc4851's instead, and forget having to mod the boards.

Cheers,

Al