ESD protection for microcontroller pin that is to be touched

[Rene]

Hi!

I am designing a touch sensor on an AVR micocontroller. I have been
working on the Qtouch library for days now but with no luck, there is
nothing to be seen on the pins with the scope after flashing a sample
program into the mcu. It has been a big PITA to even get the program to
compile because the samples are made for AVR studio and Qtouch studio
which are windows-only and I use Linux with avr-gcc.

Now I have tried something else (this qtouch stuff is really starting to
be annoying, and when browsing the web it seems I am not the only one
who feels this way) and that is showing the value the ADC measures on 8
leds that are connected to a port when touching the pin with my finger
(so physical touch detection instead of capacitive) and it shows that
simply the 50 Hz that is in my body is enough to drive it to easily
detectable signals (like in my old home-build touch dimmer but that uses
a special ic). I intend to use a big pull down resistor, I guess 1 Mohm
or so (also as a bleeding resistor, the internal resistance of the ADC
is VERY high, it takes quite a while before charge disappears). The
internal pull up is too strong, then the ADC value gets influenced by
the touch too little. The AVR by the way has internal diodes on each pin
to Vcc and from ground, I believe these should give some protection
against discharges as well.

But now my question is: what series resistor should I use towards the
sensor? I am concerned about the transfer of static from my body onto
the mcu (which, at the same time, I want to measure). To make matters
worse, I'm building this device for a boy in a wheelchair on rubber
tires and filled with plastic sheets that rub against eachother. So I
really see a risk of the mcu not having a long life expectancy without
some precautions. So what value would be safe? If I use 1 MOhm for pull
down and 1 MOhm as series resistor, I get usable results (though ).
Would this be safe in the long term for the mcu?

An alternative would be to do it with a darlington transistor. But my
question still remains: what to do to prevent charge from the body
damaging my circuit?

Thank you very much in advance for your help!

Yours sincerely,
Rene

P.S. I ordered some ic's on Ebay in Hongkong. Today they were delivered
in a small box with their feet stuck into a piece of *styrofoam* (no, it
wasn't something antistatic, it was plain, cheap styrofoam). I guess the
people running the shop are traders and not electronic engineers...

 

[Adrian Jansen]

Hi!

I am designing a touch sensor on an AVR micocontroller. I have been
working on the Qtouch library for days now but with no luck, there is
nothing to be seen on the pins with the scope after flashing a sample
program into the mcu. It has been a big PITA to even get the program to
compile because the samples are made for AVR studio and Qtouch studio
which are windows-only and I use Linux with avr-gcc.

Now I have tried something else (this qtouch stuff is really starting to
be annoying, and when browsing the web it seems I am not the only one
who feels this way) and that is showing the value the ADC measures on 8
leds that are connected to a port when touching the pin with my finger
(so physical touch detection instead of capacitive) and it shows that
simply the 50 Hz that is in my body is enough to drive it to easily
detectable signals (like in my old home-build touch dimmer but that uses
a special ic). I intend to use a big pull down resistor, I guess 1 Mohm
or so (also as a bleeding resistor, the internal resistance of the ADC
is VERY high, it takes quite a while before charge disappears). The
internal pull up is too strong, then the ADC value gets influenced by
the touch too little. The AVR by the way has internal diodes on each pin
to Vcc and from ground, I believe these should give some protection
against discharges as well.

But now my question is: what series resistor should I use towards the
sensor? I am concerned about the transfer of static from my body onto
the mcu (which, at the same time, I want to measure). To make matters
worse, I'm building this device for a boy in a wheelchair on rubber
tires and filled with plastic sheets that rub against eachother. So I
really see a risk of the mcu not having a long life expectancy without
some precautions. So what value would be safe? If I use 1 MOhm for pull
down and 1 MOhm as series resistor, I get usable results (though ).
Would this be safe in the long term for the mcu?

An alternative would be to do it with a darlington transistor. But my
question still remains: what to do to prevent charge from the body
damaging my circuit?

Thank you very much in advance for your help!

Yours sincerely,
Rene

P.S. I ordered some ic's on Ebay in Hongkong. Today they were delivered
in a small box with their feet stuck into a piece of *styrofoam* (no, it
wasn't something antistatic, it was plain, cheap styrofoam). I guess the
people running the shop are traders and not electronic engineers...

I would put 1M from the sensor plate to a junction of a couple of
Schottky diodes, one to ground, one to processor Vcc,then another
resistor, maybe 10k or so, to processor pin. That way the external
diodes carry the static discharges away, and the processor pin only sees
the residual.

The PIC CVD method as suggested could also be adapted to AVR.

Static discharge is still an issue with capacitive sensors, the coupling
can still transfer high voltages.

 

[Joerg]

I would put 1M from the sensor plate to a junction of a couple of
Schottky diodes, one to ground, one to processor Vcc,then another
resistor, maybe 10k or so, to processor pin. That way the external
diodes carry the static discharges away, and the processor pin only sees
the residual.

I'll second that. This is the way it's done right. The pull-down if
needed can also be from that node. If this touch-stuff goes through the
ADC then one can employ filtering to notch out 50, 100, 150Hz and so on.
or in America 60, 120, 180Hz. If the uC runs out of MIPS sometimes the
clever selection of the sampling intervals suffices.

The PIC CVD method as suggested could also be adapted to AVR.

Static discharge is still an issue with capacitive sensors, the coupling
can still transfer high voltages.

Especially on a dry winter day ... bzzt ... pop.

 

[Robert Macy]

Hi!

I am designing a touch sensor on an AVR micocontroller. I have been
working on the Qtouch library for days now but with no luck, there is
nothing to be seen on the pins with the scope after flashing a sample
program into the mcu. It has been a big PITA to even get the program to
compile because the samples are made for AVR studio and Qtouch studio
which are windows-only and I use Linux with avr-gcc.

Now I have tried something else (this qtouch stuff is really starting to
be annoying, and when browsing the web it seems I am not the only one
who feels this way) and that is showing the value the ADC measures on 8
leds that are connected to a port when touching the pin with my finger
(so physical touch detection instead of capacitive) and it shows that
simply the 50 Hz that is in my body is enough to drive it to easily
detectable signals (like in my old home-build touch dimmer but that uses
a special ic). I intend to use a big pull down resistor, I guess 1 Mohm
or so (also as a bleeding resistor, the internal resistance of the ADC
is VERY high, it takes quite a while before charge disappears). The
internal pull up is too strong, then the ADC value gets influenced by
the touch too little. The AVR by the way has internal diodes on each pin
to Vcc and from ground, I believe these should give some protection
against discharges as well.

But now my question is: what series resistor should I use towards the
sensor? I am concerned about the transfer of static from my body onto
the mcu (which, at the same time, I want to measure). To make matters
worse, I'm building this device for a boy in a wheelchair on rubber
tires and filled with plastic sheets that rub against eachother. So I
really see a risk of the mcu not having a long life expectancy without
some precautions. So what value would be safe? If I use 1 MOhm for pull
down and 1 MOhm as series resistor, I get usable results (though ).
Would this be safe in the long term for the mcu?

An alternative would be to do it with a darlington transistor. But my
question still remains: what to do to prevent charge from the body
damaging my circuit?

Thank you very much in advance for your help!

Yours sincerely,
Rene

P.S. I ordered some ic's on Ebay in Hongkong. Today they were delivered
in a small box with their feet stuck into a piece of *styrofoam* (no, it
wasn't something antistatic, it was plain, cheap styrofoam). I guess the
people running the shop are traders and not electronic engineers...

As a 'rule of thumb' I always encourage my staff to use buffers to the
outside world. Several reasons, buffers tend to be more robust, when
hit, they tend to not do much because they're not exercising a
program; and being somewhat sacrificial, don't mind changing them out
if they die.

Next, barring lightning, you can pretty much keep ESD out and do what
you want. BUT you have to think in terms of bandwidths up into the
GHz. just design the protection filter to function over that bandwidth
and you have some hope.

*IF* you insist on direct connection to the uC; I'd put a nice cap
adjacent to the uC pin right to its GND, put a series resistor as
large as you can stand; which takes you over to the connection point
and then put in the three fast tranzorbs, or components like that out
at the pin to be touched between Vcc and GND, pin and GND, and pin and
Vcc. AND right between that pin and the person add that series
resistor to limit current. If components are SMT; you may wish to put
several R's in series so their body capacitance doesn't feed thru/arc
over too much and definitely connect right at the edge of the board,
through a separate connection. There's a lot more to do to keep the
onslaught of planar fields across the PCB and keep out any ESD induced
magnetic field injections, too, but those are a start.

Have you thought? Using 50/60Hz AC mains pickup works like a charm
UNTIL there's a power outage. Then not so good. If the Operator
doesn't need this to work when there's no AC power, or operate at
remote outside locations, then go ahead and use the technique. If I
were you, I'd make this self-contained -- isolate the pin, drive an
outside ring using the uC's DAC output, or sq wave output will also
work. and sense the feedpath provided as a finger gets in there. All
self contained, requires NO direct connection to the uC and will
operate like a proximity switch. Doing such, you can put the ring and
center pickup BEHIND a plastic membrane and keep a lot of unwanted
stuff out, including dirt.

 

[[email protected]]

I would put 1M from the sensor plate to a junction of a couple of
Schottky diodes, one to ground, one to processor Vcc,then another
resistor, maybe 10k or so, to processor pin. That way the external
diodes carry the static discharges away, and the processor pin only sees
the residual.

I'd use a unipolar TVS diode after a resistor. I don't want the ESD surge
going into my supply, if I can help it, either. A capacitor at that node
helps a lot, too.

 

[Joerg]

I'd use a unipolar TVS diode after a resistor. I don't want the ESD surge
going into my supply, if I can help it, either. A capacitor at that node
helps a lot, too.

But mind that TVS have tons of capacitance. With 1M series resistors I
don't see how an ESD hit could possibly endanger a well-bypassed power
rail. If still concerned make a "waste dump sub-rail" instead.

[...]

 

[Tim Williams]

I would put 1M from the sensor plate to a junction of a couple of
Schottky diodes, one to ground, one to processor Vcc,then another
resistor, maybe 10k or so, to processor pin. That way the external
diodes carry the static discharges away, and the processor pin only sees
the residual.

Agreed, with one caveat:

Using a series 1M resistor against schottky diodes is iffy. Only 2.5uA of
unbalanced leakage and your logic level is wrong.

I'd use 1N4148 instead. They'll still fire before the MCU input
protection diodes, and their greater peak current handling capacity allows
you to use a much smaller series resistance without endangering the MCU.

Tim

 

[Nico Coesel]

Hi!

I am designing a touch sensor on an AVR micocontroller. I have been
working on the Qtouch library for days now but with no luck, there is
nothing to be seen on the pins with the scope after flashing a sample
program into the mcu. It has been a big PITA to even get the program to
compile because the samples are made for AVR studio and Qtouch studio
which are windows-only and I use Linux with avr-gcc.

But now my question is: what series resistor should I use towards the
sensor? I am concerned about the transfer of static from my body onto

You really should read and understand the manual first. The electrode
is not to be touched! Qtouch is a fancy word for capacitive touch
sensing which works because a finger changes the dielectric constant
and thus the measured capacitance. You shouldn't use a scope on the
pins because you won't see anything.

 

[[email protected]]

But mind that TVS have tons of capacitance. With 1M series resistors I
don't see how an ESD hit could possibly endanger a well-bypassed power
rail. If still concerned make a "waste dump sub-rail" instead.

The capacitance, like everything else, depends on the part used. TVS diodes
can be had <1pF. The ones intended for USB work pretty well (1pF, IIRC).
They're small, but that's all that's really needed behind a big resistor.
Where does this "sub-rail" go? You don't want much inductance in it.

 

[Rene]

Wow, I had expected this many replies, thank you all for your efforts. I
think I will go the route described with an extra node with protection
diodes and an extra series resistor. Maybe I'll add an extra resistor.

Thanks again!

Yours sincerely,
Rene

 

[Rene]

You really should read and understand the manual first.

I have read it so many times that my eyes starting hurting ;-).

The electrode is not to be touched!

I know, it is a capacitive system. However, what I wanted to make after
giving up on Qtouch, is a sensor that /is/ to be touched.

Qtouch is a fancy word for capacitive touch
sensing which works because a finger changes the dielectric constant
and thus the measured capacitance. You shouldn't use a scope on the
pins because you won't see anything.

That is not correct. The changed capacitance is detected because there
is a process of charging and discharging of the electrode going on (the
timing in that process is a function of the capacity) and this can be
seen on a scope (which is also mentioned in the manual).

Yours sincerely,
Rene

 

[Rene]

Have you thought? Using 50/60Hz AC mains pickup works like a charm
UNTIL there's a power outage.

To be honest, I had not thought of that. But fortunately that will not
be a problem.

Then not so good. If the Operator
doesn't need this to work when there's no AC power, or operate at
remote outside locations, then go ahead and use the technique. If I
were you, I'd make this self-contained -- isolate the pin, drive an
outside ring using the uC's DAC output, or sq wave output will also
work. and sense the feedpath provided as a finger gets in there. All
self contained, requires NO direct connection to the uC and will
operate like a proximity switch. Doing such, you can put the ring and
center pickup BEHIND a plastic membrane and keep a lot of unwanted
stuff out, including dirt.

What you describe is what is called the Qmatrix method in the Qtouch
library, there is also the Qtouch method which uses only one electrode.

I think first I will try to make things work with physical touch and
later I will sink my teeth into capacitive sensing (and that cursed
library ;-)) again.

Thank you!

Yours sincerely,
Rene

 

[Rene]

With 1M series resistors I don't see how an ESD hit could
possibly endanger a well-bypassed power rail.

I would be very gratefull if you could explain to me what a
well-bypassed power rail means, I do not know what it is.

Thank you!

Yours sincerely,
Rene

 

[[email protected]]

I would be very gratefull if you could explain to me what a
well-bypassed power rail means, I do not know what it is.

Lots of distributed capacitance to ground. The capacitors should be of
different sizes (small for high frequency and larger for low).

 

[Robert Macy]

Lots of distributed capacitance to ground.  The capacitors should be of
different sizes (small for high frequency and larger for low).

?? "The capacitors should be of different sizes..." Did the industry
change their thinking here?

I have NOT found that technique effective unless done with EXTREME
care. Using HFSS to analyze.

Rather than risk your electronics, just use the IEEE model for a
'personal' ESD event and model the circuitry using LTspice, or such.

In other words, pretend you've just attached a 100pF capacitor charged
to 18kV, and ask yourself, "I wonder what THAT does?" be sure to
consider all those pesky parasitic paths, too. Oh, and also ask
yourself what that resulting spike of current might induce into your
circuit someplace else.

 

[Joerg]

Lots of distributed capacitance to ground. ...

With a full plane you can live with a few sprinkled 0.1uF plus on 10uF
or so. Because the plane itself is a small and very good distributed
capacitor.

... The capacitors should be of
different sizes (small for high frequency and larger for low).

That was in the last century

Rene, plus one very important thing in addition: If you get a barrage of
hits and your system is extremely low in power consumption you might
want to have a limiter that bleeds excess charge. This is because
regulators can typically only source current but most (except for some
sync bucks) cannot sink. A TLV431 is nice for this, just set it a few
hundred mV above your nominal rail voltage. High enough that tolerances
won't make it come on but low enough so it does come on before chips go
kaputt.

 

[Joerg]

Agreed, with one caveat:

Using a series 1M resistor against schottky diodes is iffy. Only 2.5uA of
unbalanced leakage and your logic level is wrong.

Don't use Schottkys for this.

I'd use 1N4148 instead. They'll still fire before the MCU input
protection diodes, and their greater peak current handling capacity allows
you to use a much smaller series resistance without endangering the MCU.

The BAV199 is very low leakage and "budget-priced".

 

[[email protected]]

With a full plane you can live with a few sprinkled 0.1uF plus on 10uF
or so. Because the plane itself is a small and very good distributed
capacitor.

"Few" is left to the engineer.

That was in the last century

Not so much. Still gotta supply current.
n

Rene, plus one very important thing in addition: If you get a barrage of
hits and your system is extremely low in power consumption you might
want to have a limiter that bleeds excess charge. This is because
regulators can typically only source current but most (except for some
sync bucks) cannot sink. A TLV431 is nice for this, just set it a few
hundred mV above your nominal rail voltage. High enough that tolerances
won't make it come on but low enough so it does come on before chips go
kaputt.

It's easier just to dump through a TVS directly to ground.

 

[Joerg]

"Few" is left to the engineer.

Or guys like Muntz

Not so much. Still gotta supply current.

That's what the power supply is for. Nowadays many 1uF ceramics are good
up into the gigeehoitzes.

It's easier just to dump through a TVS directly to ground.

That no workie. A TVS has huge tolerances and a large spread between
when it starts conducting and then conducting some more. On the usual
logic rail, however, you typically have only 10%-15% margin between
operational at full clock speed and *PHUT*

 

[Joerg]

Wow, I had expected this many replies, thank you all for your efforts. I
think I will go the route described with an extra node with protection
diodes and an extra series resistor. Maybe I'll add an extra resistor.

Make sure to use the BAV199 or something similar. 5nA leakage guaranteed
at 75V reverse (room temp), tough to beat that and you get a dual-diode
for under 5c a pop in reels.