I agree with Bob Davis, for reasons that I had elaborated in the
posting that Dennis Tomlinson
had forwarded to the list (thanks, Dennis).
I only wish to point out that the problem Bob alludes to, charge
injection into the substrate, should be
watched out for, but has simple remedies. Hence some vendors seem to, or
actually might, have a "bullet
proof" solution. In protection structures like the grounded gate MOSFET's,
the main charge flow very
much is between the pin to be protected, and the Vdd/Vss supplies (strapped
by metal connections)
to the chip. There is a substrate charge that could leak out, but can be
collected by substrate/well taps
of the appropriate polarity. One junction (diodes) and three junction
(thyristors) made in a CMOS
process are probably trickier in this respect than two junction (transistor)
devices.It is good
practice to have such substrate tap rings on the inner periphery of your
pad ring, and have them form a continuous ring that just cordons off any
charge flow from the ESD protection structures, and the
rest of the chip.
In addition, it is also common layout practice (and/or should be, IMHO)
to have all sensitive blocks
surrounded with well taps connected to the Vdd/Vss'es of that block. So,
even if there is substrate
charge being injected from whereever (even switching logic is capable of
dumping some charge if
it is poorly tapped and subjected to some adverse power supply conditions,
that cannot always
be entirely wished away ), it is possible to follow some simple layout
rules of thumb and minimize
soft failures.
Regards
Jay
> -----Original Message-----
> From: Bob Davis [SMTP:[email protected]]
> Sent: Thursday, October 29, 1998 10:37 PM
> To: Mike Monett; si-list
> Subject: RE: [SI-LIST] : Schottky diode termination
>
> List Readers;
>
> I think there is a misuse of protection diodes being propagated here.
> Input
> protection diodes are generally designed to prevent the destruction of the
> input of the device. The input of the IC is usually a transistor base, of
> a
> small geometry, or a gate of a MOSFET. These devices have minimum or no
> ability to handle even static discharges from handling. You may remember
> FET's that had all leads shorted together with an extra wire that was cut
> away after the device was soldered into a board. Gate punch through was,
> and
> is, a problem. The protection diodes are there to protect the gate and
> not
> as a line termination device.
>
> Protection diodes are not intended for cycle by cycle usage. These
> devices
> may conduct to the substrate and have a long path back to ground. This is
> injecting charge into the die and can produce unintentional current flow
> in
> the device. This effect is well known in memory systems where there is
> usually specific warning about not conducting significant and repeated
> current into the silicon substrate. Many memory system failures can be
> traced to this problem, which is very difficult to isolate, can be very
> pattern sensitive and usually fails at a customer site. Memory vendors,
> in
> general, guarantee the products with input voltages that are just short of
> conducting current in the protection diodes. There are some vendors that
> seem to have a bullet proof scheme and don't care on the data sheet (a
> source of worry).
>
> The performance of the protection diode is not guaranteed and may only be
> spot tested due to the cost of tester time. Also remember that different
> vendors may implement the protection diodes differently based on there own
> design parameters and process rules as, again, this is not a guaranteed
> function of the device. It is much wiser not to depend on input protection
> devices except to protect the device during required handling.
>
> Back to thinking about the energy and where it goes. Diodes do not absorb
> the excess energy except for some consumed accidentally in unavoidable
> internal resistance. The current goes to ground and must be distributed
> back
> via the holey ground plane causing additional ground plane shift and input
> reference shifting in the target device due to ground lead inductance and
> resistance. This conduction still causes a high energy reflection back
> along
> the transmission line and this energy still needs to be absorbed somewhere
> along the line.
>
> Resistors still provide the best general solution to line termination. As
> with all design issues there is the exception that needs to be handled
> with
> a different approach such as diodes. Diodes, MOV's and similar devices are
> still needed to handle unknown and extreme excursions of voltage from long
> lines subject to acts of nature.
>
> When running Quad I often see the characteristic snubbing of the negative
> excursion caused by the shunting characteristics of the input protection
> diode in the models. This is always a concern for the long term viability
> of
> the circuit and council repeatedly for the addition of some resistance to
> keep that energy from being conducted through the input protection diodes
> in
> any critical or high reliability applications.
> Resistors are cheap compared to just one field repair trip, not including
> the ones that need a rocket ride to fix.
>
> That's my $0.02.
>
> Bob Davis
> Summit Computer Systems, Inc
> Signal Integrity Specialists - High Speed, Critical PCB Design
> 408-353-2706
> [email protected] , www.scsi.com
>
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Mike Monett
> Sent: Thursday, October 29, 1998 3:17 AM
> To: si-list
> Subject: Re: [SI-LIST] : Schottky diode termination
>
> Robert Muir wrote:
> >
> > Some years ago a colleague of used the input protection diodes of
> > logic devices as part of the normal operation of a circuit and
> > experienced an interesting problem.
> >
> > The circuit was an 8-channel digital input module of an industrial
> > controller taking input voltages from 0 to 60V. If I recall
> > correctly the input was taken through a 200K current limiting
> > resistor to the input of a 74HC244 buffer, with a 2M resistor also
> > on the 74HC244 buffer to ground. Even with a 60V input the maximum
> > current through the protection diode is only 0.3mA, well below the
> > limit of 20mA. The circuit worked nicely and was put into
> > production.
>
> [...]
>
> > Regards,
> > Robert Muir
>
> Sounds like the 0-60V signals were logic levels.
>
> Why not change the 2Meg resistor to 18k? This would give 0-5V at the
> input to the 74HC244, which should allow it to detect if the signal level
> was high or low.
>
> This is a simple change to the bill of materials. Then, any 74HC244
> should work. No need for special part selection and procurement.
>
> Best Regards,
>
> Michael R. Monett,
> Automated Production Test
> mailto:[email protected]
>
> **** To unsubscribe from si-list: send e-mail to
> [email protected]. In the BODY of message put: UNSUBSCRIBE
> si-list, for more help, put HELP. si-list archives are accessible at
> http://www.qsl.net/wb6tpu/si-list ****
>
>
>
> **** To unsubscribe from si-list: send e-mail to
> [email protected]. In the BODY of message put: UNSUBSCRIBE
> si-list, for more help, put HELP. si-list archives are accessible at
> http://www.qsl.net/wb6tpu/si-list ****
**** To unsubscribe from si-list: send e-mail to [email protected]. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****