Capacitor Discharging with CMOS Gate

[Joerg]

OK, I admit I've had too much beer, Maredsous and then Stella, both on
draft, but I can't see how a diode helps either problem.

I had a Spaten Optimator (lotsa volts ...) and then a Huebsch Maerzen.

The 2nd diode goes from cap to VCC so it will never allow VCC to be more
than a diode drop below the cap voltage. Whether that's enough needs to
be investigated but it should take the brunt of the current. You
normally need to go one diode drop above VCC and then some (to generate
a current) to trigger the parasitic SCR and since there is the other
half of the double-diode between cap and output that can't really happen
with any serious current anymore. Now one would have to know what the
SCR trigger current is since the family guide only talks about yanking
outputs the opposite direction, not beyond a rail.

 

[Joerg]

I still recall your 1st response.
That latch up stuff is wayyy beyond my experience.

This may be the time to study it some more. You can do quite bizarre
things with CMOS logic if you fully understand what's going on in there.
Like using them for analog functions. That always generates scoffing
during design reviews. Until the BOM cost totals are presented.

 

[Jim Thompson]

Is not bad practice if the mfr guarantees the latchup current, which
is typically 50 mA when they do. I haven't seen a standard cmos logic
product (HC/HCT, even the HC405x analog multiplexers) latch, ever.

Wonder who it was that reworked the 'HC405x device designs back in
2000 ?

Most of the HC muxes have another shoot-through issue if you overdrive
the inputs, but that doesn't latch.

John

Most data sheets lie ;-)

Go ahead and dump backwards thru an output pin into VDD... at your own
peril... just don't come crying to me ;-)

...Jim Thompson

 

[John Larkin]

I had a Spaten Optimator (lotsa volts ...) and then a Huebsch Maerzen.

The 2nd diode goes from cap to VCC so it will never allow VCC to be more
than a diode drop below the cap voltage. Whether that's enough needs to
be investigated but it should take the brunt of the current. You
normally need to go one diode drop above VCC and then some (to generate
a current) to trigger the parasitic SCR and since there is the other
half of the double-diode between cap and output that can't really happen
with any serious current anymore. Now one would have to know what the
SCR trigger current is since the family guide only talks about yanking
outputs the opposite direction, not beyond a rail.

Oh. Coffee works better than beer in these situations. And schottkies
may work better than PN diodes.

The thing is, we were driving down Valencia street and we
*saw*a*parking*space*!! Of course, we nabbed it instantly, and then
looked around to see what was nearby. It was Frjtz, a Belgian Fries
place, with pretty good twice-cooked fries and some interesting beers
on tap.

That's how you pick restaurants in San Francisco.

John

 

[John Larkin]

Wonder who it was that reworked the 'HC405x device designs back in
2000 ?

You? Whose? What was their latch threshold current before?

Most data sheets lie ;-)

Go ahead and dump backwards thru an output pin into VDD... at your own
peril... just don't come crying to me ;-)

As I said, I've never seen it happen in HC. I rarely use the esd
diodes for clamping, and never allow more than a mA or so on parts
spec'd for 50. No smoke so far.

Sometimes this just happens, as when two boxes are interconnected and
one power cycles.

Recently, one of my better customers started blowing up Tiny Logic
tristate buffers that can gate a clock out to a connector on one of
our VME modules. They never enable the output function, and only apply
+3 dBm RF to the connector as an input, so we can't imagine how the
chips fry. You can tell the dead ones by the tiny bubble of charred
conformal coating on top the SOT-23 package. We tried every sort of
abuse we can imagine and can't zap them here. After a zillion meetings
and conference calls, we decided to remove the chip.

John

 

[Jim Thompson]

You? Whose? What was their latch threshold current before?

Moto/OnSemi. I don't know the latch current numbers... that's a
_measured_ thing. I just added some output features which make
back-flow hi-Z.

As I said, I've never seen it happen in HC. I rarely use the esd
diodes for clamping, and never allow more than a mA or so on parts
spec'd for 50. No smoke so far.

Sometimes this just happens, as when two boxes are interconnected and
one power cycles.

Recently, one of my better customers started blowing up Tiny Logic
tristate buffers that can gate a clock out to a connector on one of
our VME modules. They never enable the output function, and only apply
+3 dBm RF to the connector as an input, so we can't imagine how the
chips fry. You can tell the dead ones by the tiny bubble of charred
conformal coating on top the SOT-23 package. We tried every sort of
abuse we can imagine and can't zap them here. After a zillion meetings
and conference calls, we decided to remove the chip.

John

That's only 0.9V P-P at 50 Ohms. Is there any DC present?

...Jim Thompson

 

[Joerg]

Oh. Coffee works better than beer in these situations. And schottkies
may work better than PN diodes.

Shouldn't matter much since the drop is the same for both diode halves.

The thing is, we were driving down Valencia street and we
*saw*a*parking*space*!! Of course, we nabbed it instantly, and then
looked around to see what was nearby. It was Frjtz, a Belgian Fries
place, with pretty good twice-cooked fries and some interesting beers
on tap.

Ah, did you try a Chimay? Where smoke wafts out of the bottle after
popping the cork?

That's how you pick restaurants in San Francisco.

Now just imagine the SW consultant working on the current project at a
client here, trying to park his Dually crew cab with horse trailer.

 

[John Larkin]

Ah, did you try a Chimay? Where smoke wafts out of the bottle after
popping the cork?

We love Chimay, the red label in the huge bottle with the champagne
cork. But they had these other interesting things on tap.

Now just imagine the SW consultant working on the current project at a
client here, trying to park his Dually crew cab with horse trailer.

He brings a horse to work? Can it code?

John

 

[Spehro Pefhany]

We love Chimay, the red label in the huge bottle with the champagne
cork. But they had these other interesting things on tap.


He brings a horse to work? Can it code?

John

Why does he consume 186.4 watts?
Best regards,
Spehro Pefhany

 

[Joerg]

We love Chimay, the red label in the huge bottle with the champagne
cork. But they had these other interesting things on tap.


He brings a horse to work? Can it code?

Probably not in C. Out here it would not be unusual if someone shows up
on horseback:

http://www.analogconsultants.com/ng/images/horses.JPG

 

[John Larkin]

Moto/OnSemi. I don't know the latch current numbers... that's a
_measured_ thing. I just added some output features which make
back-flow hi-Z.


That's only 0.9V P-P at 50 Ohms. Is there any DC present?

...Jim Thompson

They claim not. I tried up to 20 volts p-p from a function generator,
power cycled that, everything I can think of, and can't kill it. Our
manual says this is a "ttl" i/o, but still a 3 dBm sinewave shouldn't
hurt anything.

It's a Fairchild NC7SZ125.

They also only fail at the customer's system test step, never in the
field. That's good, since it's a test system for B52 radars, deployed
all over the world.

I was thinking maybe a really charged-up coax, but they say that's not
happening. Another mystery.

John

 

[Joerg]

Why does he consume 186.4 watts?

Some of those watts come back out. But those watts are quite smelly.

 

[Joerg]

They claim not. I tried up to 20 volts p-p from a function generator,
power cycled that, everything I can think of, and can't kill it. Our
manual says this is a "ttl" i/o, but still a 3 dBm sinewave shouldn't
hurt anything.

It's a Fairchild NC7SZ125.

They also only fail at the customer's system test step, never in the
field. That's good, since it's a test system for B52 radars, deployed
all over the world.

I was thinking maybe a really charged-up coax, but they say that's not
happening. Another mystery.

Ask them what kind of tile they have on the floor there and what type of
shoes the guys are usually wearing.

 

[Jim Thompson]

Ask them what kind of tile they have on the floor there and what type of
shoes the guys are usually wearing.

Yep. It's that time of year. Walking thru Sam's Club yesterday. I
kept getting zapped until I maintained a constant grip on the cart :-(

...Jim Thompson

 

[Joerg]

Yep. It's that time of year. Walking thru Sam's Club yesterday. I
kept getting zapped until I maintained a constant grip on the cart :-(

Despite the fact that almost everybody has personal experiences like
that I often find it tough to convince engineers that it poses a
problem. "But there is a plastic bezel in front..." Until I hook a scope
to an input, set it to normal trigger, walk a few steps and swipe my
hand across. "Whoa!" phssst ... poof. This week I was lucky, they
immediately agreed. I had put the scope plot into the report ...

 

[Terry Given]

Shouldn't matter much since the drop is the same for both diode halves.

I'd use an R and a BC847BPN configured as a pair of diodes a-la Joerg,
but thats just me. if the logic family spec'd the acceptable clamp diode
current at > 2.5mA, a 330 and a single PN diode would do fine, leaving
me one spare 7V diode/50V diode/8V zener/60V zener/transistor.

you guys inspired me, so I grabbed a Speights Old Dark.

Ah, did you try a Chimay? Where smoke wafts out of the bottle after
popping the cork?



Now just imagine the SW consultant working on the current project at a
client here, trying to park his Dually crew cab with horse trailer.

Cheers
Terry

 

[Terry Given]

Moto/OnSemi. I don't know the latch current numbers... that's a
_measured_ thing. I just added some output features which make
back-flow hi-Z.

I fixed a problem for a customer in China last year, using an on-semi
HC405x (2 IIRC). they fed 5V thru 100R into the 3V3 powerd HC405x. net
result was Vout = 1.4V for a logic zero. no latchup of course, only 10mA
(thanks Jim), but the rest of the circuitry stopped working.

their head of engineering reckoned the chip was faulty; his solution:
- remove "faulty" chip
- cut track
- replace faulty chip
- tack schmitt trigger ontop of another chip
- jumper wires to "buffer" the not-so-zero to get a "proper" zero

my solution:
- change 100R to 1k
- add 1k5 shunt resistor at HC405x input

voila, everything worked again.

their schematics didnt show the chip Vcc connections at all - hidden
pins and power net names (Grrr). I ended up using a multimeter and a
blank board to figure out what chip ran from what supply voltage. they
had repeated this mistake dozens and dozens of times.

they had a 600MHz coax datalink using an agilent
ser4ialiser/deserialiser chipset. This chip had 21 digital inputs, ALL
of which had this exact problem. oddly enough, the chip ran real hot.

 

[D from BC]

No, thaddaway:

14V
signal _|
|-----| \ BAW56
| | |O--|<-+->|--- 14V
+-----|_/ |
| ---
Com --- 10uF
|
Com


Hope it worked, had to copy your ASCII art and edit with keyboard.

I think it's sinking in now.
Vout >> Vdd = Bad
Charging portion of circuit omitted for clarity I guess.

The diode conducts when Vcap exceeds the Vdd rail.
That way the gate doesn't do anything goofy..


D from BC
British Columbia
Canada.

 

[Winfield Hill]

They claim not. I tried up to 20 volts p-p from a function generator,
power cycled that, everything I can think of, and can't kill it. Our
manual says this is a "ttl" i/o, but still a 3 dBm sinewave shouldn't
hurt anything.

It's a Fairchild NC7SZ125.

They also only fail at the customer's system test step, never in the
field. That's good, since it's a test system for B52 radars, deployed
all over the world.

I was thinking maybe a really charged-up coax, but they say that's
not happening. Another mystery.

John, I imagine your output circuit is something
like this, to source terminate a 50-ohm coax?

7sz125 39
--|>o---/\/\----(o) BNC

What if you did something like this instead,

,--|>|-- +V
7sz125 20 | 20
--|>o---/\/\--+----/\/\--(o)
|
'--|<|-- gnd

 

[Fred Bloggs]

Recently, one of my better customers started blowing up Tiny Logic
tristate buffers that can gate a clock out to a connector on one of
our VME modules. They never enable the output function, and only
apply +3 dBm RF to the connector as an input, so we can't imagine how
the chips fry. You can tell the dead ones by the tiny bubble of
charred conformal coating on top the SOT-23 package. We tried every
sort of abuse we can imagine and can't zap them here. After a zillion
meetings and conference calls, we decided to remove the chip.

Are you saying these buffers are off a bidirectional port on your
module and the customer is applying a 3dBm clock to it? How is this
terminated for input mode drive? Some generators are very particular
about termination and can do all kinds of oscillations and non-linear
things when driven into Hi-Z, to include putting DC on the line as well
as damaging peak voltages. This may be especially true of a video grade
signal generator with lots of programmability in amplitude and DC
offset. This would also explain why the failures are not occurring in
the field. Did the numbskulls even non-invasively scope out their input
waveform, or maybe your circuit blows so fast this is not possible.
Obviously you need to hold them by the hand and walk them through it.
Pathetic!