Lets talk about input protection against ESD (human body).

Hi I am a ESD engineer dealing with ESD on IC level. Please do not mix IC level and PCB level ESD. PCB / Application level ESD is described in IEC 61000-4-2, which is almost a worldwide standard now. The Standards in the links above refers to IC level test which are not applicable to board design.

I have only minor knowledge in PCB protection but I know that you could add external ESD protections (Zener-clamps).

I guess the way of prevention is not that different at IC level and PCB level. Get the current dumped to GND , do not be scared about the 30A ESD current during an ESD event. It is just for a short time (ns-us) try to minimize the voltage drop due to parasitic resistors, this could cause much more trouble. And over all a mistake that all ESD "freshmen" do: Your device will never,ever see thousands of Volts!!! (Except you done something terrible wrong) It is just the capacitance inside the ESD gun which is charged to this level, ESD is a current problem! You may face high voltages if you force the ESD amps through some resistance but this is what I mean with doing something terrible wrong.

Here are some references that may help you in deciding which standard to design for and test to.

Hello Wolfgang,

Zeners are not widely used here. The usual trick of the trade is to employ double diodes a la BAV99 between IC pin and GND on one side, VCC on the other. Then a few 10k of resistance up front from pin to where the ESD gun can hit.

A BAV99 is a silicon diode so it has about the same U/I properties as the internal substrate diodes, meaning that the substrate diodes still share a large burden which could lead to latch-up or other grief. If that's a problem there are Schottkys such as the BAT54S. However, two compromises here. More leakage,plus they may cost an additional penny at large quantities. But with Schottkys the lion's share of the ESD related current pulse will flow through the external diodes because they start their current ramp much earlier than the substrate diodes.

Regards, Joerg

I haven't tried actual measurements, since this sort of stuff is not easy to measure. But I compared the IV curves for BAT54 vs. MMBD4148, and realized that at higher currents, the voltage drop through the Schottky exceeds that of the PN.

This was confirmed by my LTSpice models of ESD input circuits for HC logic.

I assume here that the input diodes look similar to 4148, but it is possible they are little lighter. In any case, I look to meet the IEC standards while also not exceeding datasheet specs even for static overload conditions.

That way I make few assumptions about the surge capabilities of the input diodes.

Check out this comparison of PN vs. Schottky (you will need LTSpice to use this file):

Since I did this I use MMBD4148 for input protections instead of BAT54 anymore.

Good day!

Hello Chris,

Can't agreed there. Ok, I only have the data sheet of the MMSD4148 but it should be pretty similar. At 100mA the BAT54 sits at 450mV. At that same current the 4148 is already above 700mV, meaning the substrate diodes are beginning to get punched.

I don't have LTSpice on this PC but are you sure the models are ok?

Regards, Joerg

Hmm, actually the model is the main question mark. I just did a DC sweep. There seems to be something wrong with the BAT54 model.

Here's a sim comparing 3 diodes including Fairchild's model parameters:

Hmm, their BAT54 also has more voltage drop at high currents, but the crossover isn't the same as the LTSpice built0in model.

Ok, anyway whatever the case, my ESD input protections use a resistor of about 100-220R between my external protection diodes, and before the device pin. That way the external protection diodes even if they drop a couple volts (I get a 10A, 6.4V spike across my diode with the 15kV IEC

I started another thread about this:

"Schottky diodes have higher V drop than PN?"

Maybe you can add some comments there too.

Good day!

Sorry about my absence from the conversation over the last few days. I had another pressing issue to attend to. It appears the conversation is buzzing along without me. All of you bring up interesting points.

I don't know how to do that fancy ASCII art so I decided to post an image of my circuits input structure. Please visit the following URL and critique the design.

I'd like to know how robust you believe this will be. Will it stand up to a 8KV or higher direct contact discharge using the standard ESD gun?

"What's the reason for pulling against +12V where your logic runs on

This is not so straight forward. I'm using a low dropout linear voltage regulator. I'm using a limited amount of capacitance at its output to conserve board space. If I draw too much current from the regulator with a low amount of output capacitance the regulator can become unstable. I need to keep branch current at 12mA to 20mA (depending on customer requirement) to act as wetting current for in-line connector pins and switch contacts. The proper amount of wetting current helps keep contacts clean, and therefore low resistance.

"If ESD is serious I'd take the circuit that Chris had suggested somewhere. Basically a large resistor value like 100K from the connector to the center of antiserial diodes, such as the BAV99. The diode clamps to GND and +5V. Then from that center point to the HC chip with another resistor which could be lower (maybe 1K)."

I'm considering doing this with other general purpose 5V logic inputs. Although I'll probably use 10K and 1K resistance.

However, for reasons listed about, I can't do this with my switch input. Branch current would be too low for proper wetting.

A question about using regular PN double diode clamps to protect the input to my micro. Something that was mentioned earlier in the thread. Say I had a 10K resistance limiting current to a BAV99 double diode clamp. Then 1K limiting current to the micro pin after that. The micro's internal clamps are Schottky barrier diodes. Thier forward voltage drop is much less than the BAV99's. Wouldn't they start conducting first and soak up the transient? Shouldn't I use an external Schottky clamp instead of a standard PN part like the Bav99?

"Also, keep in mind that physically small resistors might arc over when

hit with a few kV. I have seen one turn from the usual off-white ceramic to green bubbly glass. You may have to place several in series."

Something Wolfgang Kemper mentioned earlier. ESD is a current based phenominon, not voltage. That is as long as you keep the series resistance of this current path to a minimum. If you place a 100K resistance in series, the resulting voltage drop will be substantial, possibly enough to arc. Now, what about the caps I have right at the input pins? Will these absorb some of the transient current so the injection current into the proposed series input resistor can be limited, therefore lowering the voltage drop across the chip resistor?

Trying to picture exactly what happens.

Thank you so much.

Gerb

Hello George,

What's the reason for pulling against +12V where your logic runs on 5V? Also, I don't trust zeners for this application. Too much tolerance and they cost too much.

If ESD is serious I'd take the circuit that Chris had suggested somewhere. Basically a large resistor value like 100K from the connector to the center of antiserial diodes, such as the BAV99. The diode clamps to GND and +5V. Then from that center point to the HC chip with another resistor which could be lower (maybe 1K).

You could also pull up with about 1M if needed (but to +5V, not +12V). A cap from the center of the diode pair to GND can slow down transients. However, then your HC should be a Schmitt input.

Also, keep in mind that physically small resistors might arc over when hit with a few kV. I have seen one turn from the usual off-white ceramic to green bubbly glass. You may have to place several in series.

Regards, Joerg

"Maybe use a switcher instead of the LDO?"

The word "switcher" is taboo in the greater Detroit area automotive electrical design community. EMC nightmares... Also, a switcher sounds expensive compared to my $0.17 iron clad vreg.

"You could apply the wetting current right to the connector and from a source that has enough caps to weather the ESD surge."

Could you please re-phrase this thought? I'm having a hard time following you. Are you talking about adding a second 5V source to provide the pull-up for my external switch? Why not use 12V, its free?

"And a resistor that doesn't fry from the 8kV."

OK, are we talking about peak pulse power limitations or adjusting package type to prevent arcing? Remember, I'm not real familiar with what happens above 100V.

"HC behaves like normal diodes."

That is what I always thought. However, Freescale Semi insists that their internal clamps behave like Shottky barrier diodes and have a lower Vf.

Please look at this updated drawing. I added some more detail about the 12V side of the circuit to see how ESD transient current might affect things. I also changed the uC pin protection to something more along the lines of what we have been talking about. As long as the input impedance of the micro is >> than the 100k series resistance I should be OK.

Thanks a bunch,

Ge0rge

Hello George,

Maybe use a switcher instead of the LDO?

You could apply the wetting current right to the connector and from a source that has enough caps to weather the ESD surge. And a resistor that doesn't fry from the 8kV. Then go from there with 100K -> double diode -> 1k -> HC logic.

HC behaves like normal diodes. Anyway, the BAV99 will have cut it down to, say, a volt or two beyond the rails when pushed hard (depends on what resistor is in front). Now there will be less than 2V across the 1k resistor which results in a small current surge that HC can easily tolerate.

It's still a voltage thing when an external cap however small is charged to 8kV and then discharged. Input caps directly across an input may not be so good since they could fail. A resistor may see an initial 8kV across its body. ESD tests like this are often degrading component performance over the number of tests conducted on the same circuit.

It might be worth to invest in a book about ESD. Can't give you a hint here since I kind of grew up with this stuff and don't have a book. In our case it's medical where Hipot and defibrillator testing are done. A defibrillator test is nothing to sneeze at. 5kV roaring down an inductor bank from a 32uF cap can literally disintegrate things. Once the TUEV guy wanted to see it live. When I was ready to press the big button he moved his chair several feet back ;-)

Regards, Joerg

Hello George,

At which qties? 17c is a lot of money if you are talking about a few million units a year. Also, make sure that whatever LDO you are using will not go on allocation. Or at least that your company is high up in the pecking order should that happen. And check it out for stability inside out. These things can be rather fickle.

I'd have reliability concerns with any heat dissipating semis in cars. EMC nightmares? No pun intended but maybe they should worry a little more about the keyless entry systems or ignitions computers ;-)

I remember someone keying an FM radio mike and a little car next to them slowed down everytime they did that. They could render it unable to pass them.

Ok, you can use 12V if it's before the 100K like you have in your drawing. But when the switch is open you'll see 7mA or so flowing at all times, more if the switch is pressed, of course. If the circuit is powered down when not in use that may be ok. Else the valuable customer sits there one winter day in his fancy vehicle and the engine won't turn over :-(

That is something you need to talk about with the resistor manufacturer. They are all different.

Freescale? HC parts? Mine are spec'd from ONSemi, not Freescale. I'd be surprised if they sold any. From a specsmanship point of view (familiy specs, ESD etc.) Philips is really good but their web site is a pain.

Anyway, HC is a regular plain clothes CMOS process. Look at the family specs.

That MLV cap is suspicious. Could fry on ESD. Also, the 0.01uF and 100K create a 1msec time constant. HC doesn't really like that, you should use a Schmitt such as the HC14.

Regards, Joerg

Hello Marco,

Thing is it slows down the transition into the HC chip. Without a Schmitt input that is a concern.

I am more worried about his cap from the connector to GND. If it's big there won't be much ESD coming through but the switch will suffer. Bzzzzt. If it's small then the cap may go kapoof.

Well, yes. Some of the more luxurious sport utility vehicles in the US even have some hefty 115V outlets in the passenger cabin and baggage area. They pass EMC.

Then there are pull-down LCD monitors with back light converters in the ceiling, navigation aid screens etc. Just imagine, cruising down Highway

Regards, Joerg

- Are you planning to do early ESD prototype tests? If not, I would suggest to do early ESD tests on a prototype as soon as possible, because this story might cost a relayout! Not only the circuit and components but also the layout, connector and front are critical here.

- The 0.01 uF cap never sees voltages higher than say 8V so if it is cheaper you can replace it with a lower voltage type down to 8V. Ceramic caps can withstand quite a lot of overvoltage without breakdown.

- I'ld guess that the 1K / tantalum 35V cap combination may be the weakest link in the ESD test.

- I was wondering about the wetting current, and the reason why you cannot pull it from the 5V voltage regulator. If instability at 20 mA extra loads is a problem, you could probably solve that either by adding a capacitor in the regulators feedback network, or by using the right type of output cap. I mean you may get it stable without needing more boardspace. By pulling from 5V, you save the large 1210 resistor, and you don't stress the 12V caps so you make the ESD story lots easier to guarantee.

- You mention that switchers are an EMC nightmare in Detroit. Well here in the netherlands we did a couple of switchers, even on data acquisition systems, and were ALWAYS able to get EMC right. Early prototyping, as well as involving the component application engineers is essential. btw sometimes the digital guys screwed it up with their clock oscillators ;-) . Anyway, the currents you need for the other 5V clients seem fairly small so that makes building a low-radiation switcher easier.

Anyway if you'd succeed with a switcher, you can also save a lot of power dissipation compared to a 5V regulator that draws 20mA from 12V. Especially if you need to interface more multiple momentary switches.

Best regards,

Marco

"At which qties? 17c is a lot of money if you are talking about a few million units a year. Also, make sure that whatever LDO you are using will not go on allocation. Or at least that your company is high up in the pecking order should that happen. And check it out for stability inside out. These things can be rather fickle."

Qty = 800K (for now)

You think this part is expensive? Geesh, I looked long and hard for this regulator. It's very precise, has tons of built in stuff that eliminates external circuitry, etc...

Do you think a switcher could be implemented for anywhere near this price? I don't think so. But then again, I've never really looked :).

I picked a device that can be sourced from 3 different vendors if need be.

The regulator is stable given the loading vs. capacitance used on its output. I've tested it to death and have never had a failure.

"Ok, you can use 12V if it's before the 100K like you have in your drawing. But when the switch is open you'll see 7mA or so flowing at all times, more if the switch is pressed, of course. If the circuit is powered down when not in use that may be ok. Else the valuable customer

sits there one winter day in his fancy vehicle and the engine won't turn over :-("

The 12V source switches off with vehicle IGN. Key off load current is about 90uA. The acceptable limit is 300uA. I'm safe.

"Freescale? HC parts? Mine are spec'd from ONSemi, not Freescale. I'd be surprised if they sold any. From a specsmanship point of view (familiy specs, ESD etc.) Philips is really good but their web site is a pain.

Anyway, HC is a regular plain clothes CMOS process. Look at the family specs."

I think there is a little confusion here. The device is a microcontroller manufactured by Freescale (Motorola). Not a common logic gate. True, the HC in the part number MC68HC908kX8does stand for CMOS, but the microcontroller implementation goes a little further with its input scheme than a general logic gate. Or so I am lead to believe.

"That MLV cap is suspicious. Could fry on ESD."

This multi layered varistor (MLV) is made to absorb transient engergy from ESD. I would only hope it can stand up to a strike without frying. Take a look at its data sheet here:

Hello George,

Well, that's a lot but not a whole lot ;-)

If you can save other parts with it you may have the optimal solution. I just did that with a TPS77001 by "mooching" the reference off of it for other stuff. It is also an LDO but quite picky about the output cap ESR. It's like in a food recipe, not too much but also not too little.

Probably but it would require quite a bit of engineering and pricing research. It might not be feasible at 800K qties since you have to factor in the engineering costs per unit sold. Anyway, you also have to look at the caps and their prices. Switchers can sometimes live with lesser parts here. But they do need an inductor which adds cost.

That sounds good.

Sorry, I thought the HC08 in your schematic meant 74HC08. I would still be quite surprised if the uC would feature Schottky limiters. That would be like the golden faucets that royalty has in their castles.

I am surprised that a micro in the $3-4 class makes it into cars. Some of my friends from college work in automotive today. They always say that every penny gets turned around and around before it enter into the cost of goods. Sometimes it then has to be taken out again in a cost reduction effort.

Didn't see that since you drew a cap on your schematic, not a MOV symbol.

Regards, Joerg