Robust Microcontroller Input Protection

[Baliszoft]

I know i may be boring :-).
But i still would like to ask you to do the math.

100v/33000ohms = ~3ma
100v/100000ohms = 1ma

Look up the specs of your processor for maximum input clamping current.
I bet it will be min 10ma.

.... And you could even use a 1Mohms current limiting resistor if you are worried

[Rufus]

100v/33000ohms = ~3ma
100v/100000ohms = 1ma

Look up the specs of your processor for maximum input clamping current.
I bet it will be min 10ma.

.... And you could even use a 1Mohms current limiting resistor if you are worried

If you are only concerned about the processor not sustaining permanent damage look at the clamp diode ratings. Most processors are not specified to operate correctly with any clamp diode current flowing. You can't use 1M either because you would loose a volt of noise margin with typical processor pin leakage specifications.

Using maybe 100k is practical and there is not much to screw up with clamp diode current in a 74HC gate for example.

[Baliszoft]

This is a common practice, not theoretical discussion.

Many automotive ECUs are running with the clamping diodes flowing current (intentionally).
Btw i would not use 1M either, but one could do it and because we are talking about slow digital signals (f.eg on off switch), noise would not be a big problem. BOSCH for example uses a low value cap (max 10nf) for esd protection, a 1-2k pulldown resistor followed by a 33k current limiting resistor. And that is all. Millions of ecus are running this way.

Please do not over engineer, just because you can.

[TerminalJack505]

I know i may be boring :-).
But i still would like to ask you to do the math.

100v/33000ohms = ~3ma
100v/100000ohms = 1ma

Look up the specs of your processor for maximum input clamping current.
I bet it will be min 10ma.

.... And you could even use a 1Mohms current limiting resistor if you are worried

Ah!  This is good info.  I'm going to use an AVR.  Probably the ATtiny 261 Automotive.  I found an app. note where they did basically what you mention.  This might be the app. note free_electron mentioned earlier in the thread.  They use a 1M resistor to implement a zero-crossing detector on mains (110 - 240 VAC) voltage. 

They recommend no more than 1mA through the clamping diodes.  They say this will work for spikes but not surges.  By surges they probably mean something  like IEC-61000-4-5 specifies (mains AC lightning strike.)  This is probably unlikely for an automotive application.

So, in theory, I should be able to use a 100K resistor and a pull-down and maybe add a cap for filtering.

[TerminalJack505]

This is a common practice, not theoretical discussion.

Many automotive ECUs are running with the clamping diodes flowing current (intentionally).
Btw i would not use 1M either, but one could do it and because we are talking about slow digital signals (f.eg on off switch), noise would not be a big problem. BOSCH for example uses a low value cap (max 10nf) for esd protection, a 1-2k pulldown resistor followed by a 33k current limiting resistor. And that is all. Millions of ecus are running this way.

Please do not over engineer, just because you can.

Something like the attached, then?  What kind of voltage rating should the cap have?

[free_electron]

correct, and if you remeber what isaid in that paragrpah it says :
SOME processrs and IC's have actrually factory specified that you are allowed to use the clamping diodes for that purpose.
SOME <- note the usage of the word 'SOME' and the usage of the word 'FACTORY SPECIFIED'. that means they tested it , and it was designed for that ...
Automotive qiualified devices have more robust protection circuitry. it's not just extended temp range... there's other differences too.

Do not try this with a simply cmos gate like a 4011.  there the protection structures are for ESD only... they protect the chip when powered off. send any kind of constant current in them when powered on an there is NO guarantee that you will not turn on some parasitic element that could fry the chip.

the way protection structures work is by deliberately turning on a parasite to shunt the energy. as the energy disperses the shunt turns off all by itself ( just like a thyristor self-extinguishes on zero cross )

if the chip is powered up this becomes a whole other can of worms.. the protection structure kicks in but current keeps flowing after the disruptive pulse has disappeard. current coming from the power supply.. the protection shunt is made for pulse operation and pulse operation only. it can handle a massive overload for a few microseconds. keep feeding it energy for longer than that and it will auto-destruct.

Note that in that appnote from atmel they use a resistor to detect zero cross... so the clamp will extinguish at one point due to the current becoming low enough...
connect that pin to 220 volts DC .... my guess is as good as yours... it could very well be that the parasite remains on and that when powering on or off the system you cause a massive short that fries the thing.

there is indeed no point to over-engineer, but,  if you want to make something automotoive or industrial that has to be reliable a minimum of brain activity and protection is required ... just winging it by slapping on whatever resistor happens to by laying in the junkdrawer on it is 'duct-tape' engineering.

As for common practice... it is common practice for lemmings to commit mass suicide... just because the whole world does it wrong does not mean you should follow the leader.

[TerminalJack505]

Those are good points, free_electron.  That app. note is the only place I've ever seen mention of allowable current through the protection diodes so, who knows just how good of an idea it might be.

How about the attached as a compromise?  The 1k is to limit current through the protection diodes in the case of a reverse battery condition.  (Unlikely, but one of the many cases to consider for automotive circuits.)

[Baliszoft]

There is an app note (cant recall its number) from microchip, which deals with connecting the ac mains to a uC pin thru a 2M resistor (1+1M favourably) directly, and they claim it is safe. Some say it is unsafe, but i do believe the manufacturers recommendations. Where it is possible, internal clamping diodes may be used (not exceeding the specs of course).

[Baliszoft]

there is indeed no point to over-engineer, but,  if you want to make something automotoive or industrial that has to be reliable a minimum of brain activity and protection is required ... just winging it by slapping on whatever resistor happens to by laying in the junkdrawer on it is 'duct-tape' engineering.

As for common practice... it is common practice for lemmings to commit mass suicide... just because the whole world does it wrong does not mean you should follow the leader.

free_electron,

You are getting quite annoying by now, and i bet it is not just with me.
Im not doing any "duct-tape" engineering, the method i just suggested is how they do it at bosch. You called THEM, duct tape engineers.

You'd better go and find yourself a job instead of being the local smart ass here.

[Rufus]

You are getting quite annoying by now, and i bet it is not just with me.

Kinda. He knows a lot but some of what he posts is shit. If you don't already know as much or more about a subject than he does you can't tell which bits are shit. Anything he posts that I didn't already know I have to ignore or independently research and verify - annoying.

[TerminalJack505]

Here's what I'm thinking about doing now.  Use a relatively cheaper bidirectional TVS diode to handle ESD strikes and handle load dumps with the 10k series resistor and zener.  The two 10k resistors should probably be 1W but load dumps are infrequent and relatively short in duration so 0.5W should survive. 

By the way, I should apologize to the original poster for hijacking his thread.  Hopefully you can use the same or a similar circuit.

[free_electron]

is how they do it at bosch. You called THEM, duct tape engineers.

Yes, and what kind of IC's do they use ? Bosh is an automotive/industrial vendor. Those devices ARE characterized for such usage. The input clamps are larger than in commercial stuff.
I happen to have worked on several automotive chips for Bosh ECU's and body control electronics. I did one of the first common rail injection chips for them. I've also done devices for ABS systems of Wabco as well as other automitve stuff for Valeo and Magneti Marelli. The specifications for these devices explicitly call for clamps that can handle automotive environments. Those are NOT the standard ESD clamps you find in a dumb TTL or CMOS chip or off the shelf non-automotive CPU.  Go look it up. And then come back calling my post shit.

I am not saying my solution is the end-all of protections There is others out there that may work better. I do know it protects well. It has passed all kinds of tests and has been in the field for ten years +.

There may aslo be simpler protection scemes that work.
I am also saying that you cannot use the esd diodes in a common chip as protection using a simple series resistance. They are  not designed, tested or qualified for that.

You base your comments on something you have 'seen' on a pcb from people who know what they are doing. But your observations are incomplete as You have no idea how the chips inside are built ( see my comment above. They most certainly are specced for such usage ).

You cannot generalize such an approach. I call such generalizations duct-tape engineering. Slap a piece of tape on it and see if it holds. If not, apply more tape.... There's too many people out there that will blindly clone this approach that are even more clueless.. And pretty soon all stuff ends up being wonky... Because it is a bad copy of a half understood system that is underspecced, underrated and will fail prematurely. But that seems to be the norm now. Cheap crap that dies after a year.

If someone tries to explain the perils and pitfalls he apparently gets wristslapped. You're making it too hard, it's overengineered, you talk shit... That also seems to be the modern way of things. I don't care. You have your opinion, i have mine. I try to do things the right way , no cutting corners. I'm not forcing anyone to use my approach. I am just warning people that using a simple series resistor and 'blindly trusting' that the esd diodes will hold it is not the right way to work. It is a duct-tape approach. Let's wing it and see what happens. If someone loses an eye, so be it ... I don't subscribe to that methodology.

[Baliszoft]

Electron,

Let me apologize, i cant read your whole post right now as it is too long and i dont have the time at the moment.
Yes, bosch is using automotive stuff (even manufacture such things) but in this particular example (i was referring to) they are feeding a simple on off switch to an infineon c167 uC exactly this way. Nobody was talking about measuring analog signals with an adc, nobody asked the protection for anopamp, nobody was talking about an adsl modem, it was just about a plain digital switch. If you download the datasheet of the c167 the clamping current is 10mA, which is not that quite high (being an automotive uC), even a fecking dspic has got twice of this value (max 200ma for vdd/vss). Get yourself an ecu and reverse engineer it. Modern ecus got at least 20 digital inputs (often they are multiplexed with an asic), but they do simply connect them directly to the uC via a current limiting resistor quite often. And not just switches, even freq counters and hall phase sensors (pin is pulled up to vbatt, hall sender pulls it down, excess is clamped by uC).

Dont get me wrong, i do acknowledge your knowledge. I've read many of your posts, which i've found informative, but when you are trying to be smarter than everybody (especially when you are not 100% right), and when you do provoke people (with duct tapes and such), that pisses me off.

[Baliszoft]

You are getting quite annoying by now, and i bet it is not just with me.

Kinda. He knows a lot but some of what he posts is shit. If you don't already know as much or more about a subject than he does you can't tell which bits are shit. Anything he posts that I didn't already know I have to ignore or independently research and verify - annoying.

Lots uf useful info he shares, no doubt. Lots of knowledge, no doubt. Problem with his personality, no doubt. ;-)

[Baliszoft]

An idea: maybe Dave should do some research in this topic for a vblog.

[free_electron]

datasheet of the c167  -snip-

that is an AUTOMOTIVE cpu that has beefed up protections !

This is my whole point. Automotive parts are specced for such usage. There you can use the input protection. But please, DO NOT try this with commercial parts.

I have stated that already like 10 posts ago !

Standard ( commercial, non-automotive) parts use esd protection structures. To be ised only for esd , not parametrized, designed, scaled , not tested , not approved for anything else. Use at own peril
Automotive parts are more expensive for the imple reason that the I/O ring is altered. Its the same internal guts but the bondpads and protection structure is altered. This structure is larger , takes more silicon which makes the chip more expensive. The difference between a commercial and auto otive grade part is not just the temperature... The guts are different. It is also possible they use a different technology. Commercial stuff may be in 90 nanometer whule the automotive may run in 120 nanometer.

trying to be smarter than everybody

I'm the first to acknowledge i'm a pretty dumb guy. I can't type very well, i can't calculate an integral, suck at math , i just happen to know a lot of stuff.

do provoke people (with duct tapes and such), that pisses me off.

i' m not attacking you or cali g you a duct tape engineer. The term duct-tape engineer means : someone who blindly applies somethign without knowing or understanding how it works. If it fails they try the same approach again but slap more un-understood or misunderstood stuff on i. The hope it will hold now.

The reason i am 'growling' here is because

- i know you work on ECU's
- i know they use automotive parts
- those have specified clamp currents becuase they are true clamps and not simple esd protection.
- and you can get away with a simple resistor and cap there.
- i have designed chips for automotive and know what is inside.

But you can not apply that input technique to devices that are not specced for that.. And those form 99% of whatis out there.

Knowing how on the internet 99.99% is beeing duct-tape engineered .. ( blindly copied ) the risk is that everyone starts using a single resistor for their project and design. And before you know it .. Crap design everywhere. The original was workin in spec , designed in spec , but it was misunderstood and blindly copied...

That's what i'm trying to prevent.

[Chasm]

If you do stuff for money you better know WTF you are doing.
It's that simple.

Bosch did not arrive at their minimalist input protection by accident. (Relatively minimalist!) They knew what they were doing. What to expect on the input side, the complete environment, the price point they had to reach and a billion other things that influence the design decisions.
They also did extensive testing! Not only in nice environments and also to destruction.

If you can't test as much or don't need to squeeze out the last cents because it is low volume or high price equipment it may be easier to use too much input protection (including layout!) - as long as you know that it meets your requirements.

[Baliszoft]

Im not arguing with you, because you are always smarter (in any topic).

I only believe to the facts.

If my datasheet says a processor can contiuously clamp 20mA on a pin (more than the 16+ years old automotive uC) without any damage, i will believe it, then test it, and use the hundredth of that value (without any worries).

F.eg in my ecu in which i am using a dspic, 14V is fed to the controller via a 1206 43k resistor. Most of the times (excluding load dumps and transients), the resistor is dropping 8.5 volts. That is roughly 0.19 mA going into the pin. Never had any problems in the past 2 years with this setup. Take a load dump situation, nearly 100v goes into the resistor which drops 94.5v and results in 2.2mA. Still, nine times below maximum clamping current. Cannot see where the problem is. And it still works and will be working even if you cannot believe it.

[Baliszoft]

Btw, dspic33 hi temp versions are getting aec qualified.

[Rufus]

If my datasheet says a processor can contiuously clamp 20mA on a pin (more than the 16+ years old automotive uC) without any damage, i will believe it, then test it, and use the hundredth of that value (without any worries).

It will also say

Stresses above those listed under “Absolute Maximum Ratings” can cause permanent damage to the device. This is a stress rating only, and functional operation of the device at those or any other conditions above those indicated in the operation listings of this specification is not implied. Exposure to maximum rating conditions for extended periods can affect device reliability.

I have yet to see a Microchip data sheet with with clamp diode current ratings where functional operation of the device is implied - so you are winging it. I remember reading somewhere they were going to start characterising and adding specifications for some parts.

If you look at an Infineon C166 series datasheet (the one I looked at today anyway and it wasn't an 'automotive part') you will find lots of specification of allowed clamp diode current where functional operation is guaranteed which they call overload. Limits of both polarities, different limits for analog and digital pins, multiple pin combined overload current limits, and coupling factors indicating what proportion of overload current might leak into adjacent pins.

An ATMega datasheet I looked at didn't even specify an abs max clamp diode current rating.

From abs max ratings Infineon parts would appear to be less robust than Microchip parts but as far as tolerance of clamp diode current with Microchip I have no idea. I know for sure that clamp diode current does affect the operation of some Microchip parts, starting with analog functions such as ADC, comparators and RC oscillators, PLLs are probably sensitive. How much current on which pins has how much affect on which functions is unknown and a 20mA abs max rating tells you nothing, you are left to guess or test and hope the next batch of parts isn't any worse - in other words wing it.

[free_electron]

Never had any problems in the past 2 years with this setup. Take a load dump situation, nearly 100v goes into the resistor which drops 94.5v and results in 2.2mA. Still, nine times below maximum clamping current. Cannot see where the problem is. And it still works and will be working even if you cannot believe it.

And it may still work for another 5 years.. and then all of a sudden something will fry it.
One sample , 2 years is not a representative.. Take 1 million cars 5 year spread and we'll talk.

One of (many) things i did in the past is fault analysis. I have seen many an esd protection shot to pieces because of incorrect usage. Customer returns with abnormally high failures. Cut the thing open and investigate . Chip dead, esd diodes are shot and the entire input structure is fried.

What did you do ? Do you guys use transils externally ? Customer : (sheepishly) eh... no , we trust the esd diode to hold it...  Manufacturer : Sorry, no refund as this is abnormal usage. Here is the catalog with transient suppressors, we'll happily sell ...

There's catalogs full of transils ,tranzorbs ,tvs and other devices. There is a reason these things exist... they don't make these for fun to store them on a shelf , unsold because they serve no purpose.

Yes, you may get away with it ,yes it may work for a long time , it may even never fail on bench. Put it in production and Murphy will lash out like a wild animal...

There are governing standard like ISO7637 and others specifically for automotive. ISO7637 outlines the test methods , what the impulses look like and how you can perform the test.

So, here's the million dollar question ... did you go through those certifications ?

[Chasm]

An idea: maybe Dave should do some research in this topic for a vblog.

No, this one is definitely for Mike. Much more high voltage induced flames that way.
I suppose that Mike, in the name of science, can use something a bit less energetic than the Destruct-o-tron. 

Really, it would be interesting to see how different protection circuits react to (repeated) misuse. Also which potential problems like delay or bandwidth limits they come with.

[Baliszoft]

You wish it was a million dollar question. The answer is no of course. Why would i do that with a fun project?! It is being abused just enough to make sure these wont let me and my friends (who are already using it) down.
Go and design your own, and put it full with diodes!

[Baliszoft]

If my datasheet says a processor can contiuously clamp 20mA on a pin (more than the 16+ years old automotive uC) without any damage, i will believe it, then test it, and use the hundredth of that value (without any worries).

It will also say

Stresses above those listed under “Absolute Maximum Ratings” can cause permanent damage to the device. This is a stress rating only, and functional operation of the device at those or any other conditions above those indicated in the operation listings of this specification is notl implied. Exposure to maximum rating conditions for extended periods can affect device reliability.

I have yet to see a Microchip data sheet with with clamp diode current ratings where functional operation of the device is implied - so you are winging it. I remember reading somewhere they were going to start characterising and adding specifications for some parts.

If you look at an Infineon C166 series datasheet (the one I looked at today anyway and it wasn't an 'automotive part') you will find lots of specification of allowed clamp diode current where functional operation is guaranteed which they call overload. Limits of both polarities, different limits for analog and digital pins, multiple pin combined overload current limits, and coupling factors indicating what proportion of overload current might leak into adjacent pins.

An ATMega datasheet I looked at didn't even specify an abs max clamp diode current rating.

From abs max ratings Infineon parts would appear to be less robust than Microchip parts but as far as tolerance of clamp diode current with Microchip I have no idea. I know for sure that clamp diode current does affect the operation of some Microchip parts, starting with analog functions such as ADC, comparators and RC oscillators, PLLs are probably sensitive. How much current on which pins has how much affect on which functions is unknown and a 20mA abs max rating tells you nothing, you are left to guess or test and hope the next batch of parts isn't any worse - in other words wing it.

Im on holiday at the moment, but as i will be back at home again, i will overload and destruct a dspic to find out this limit.

[free_electron]

(puts pinkie in mouth a-la dr-evil) riiight ....

So you have built 2 or three of these hobby thingies, which have not gone through any kind of (automotive or other) certification and industry standard testing as per the relevant specs. The design is based on something you saw other people do without knowing the details...
You also claim that this is perfectly fine and the way it should be done.

Someone who raises questions to, and warns about the problems with, this approach is labeled 'full of shit' ...

riiight ....  have a great holiday... don't step in any ...