ESD protecting inputs

[Rochey]

Hello folks,

I'm finishing a design here that takes a number of single ended analog inputs to a product.
It's a reference design for a high volume consumer product that will be made available through my employers in the next few months, to anyone and everyone to copy.

While we've got the silicon side of things taken care of, one of my customers asked about ESD protection on the inputs.
Everyone seems to have their own opinions on these things, and as a guy who likes data and opinions I thought I'd throw it out here.

So - a question to you all, especially those of you who have experience in real world quantity production and quality.

Whats the best way to add ESD protection to your product, as cheaply as possible.

My first thought was ferrite beads on the inputs. (that should slow down any high frequency bursts).
Then I considered TVS Diodes. They look like a good solution, however, some quick googling suggests that things like current leakage and capacitance could be a problem.
I also considered some cheap clamping diodes (1N4148's) that'd clamp any excess signals directly to the power rails, again current leakage and capacitance could be a problem?

This ref design is for a product that will likely resell for less than $80. Cost pressure is immense, so any solution has to be very generic, and multisourced from a million small manufacturers and vendors.

As always, I appreciate and look forward to any feedback!

Cheers

Rochey

 

[Rochey]

Some Extra homework done on this:
http://www.littelfuse.com/application/dvd-player-recorder/audio-in-out.html

Looks like they suggest either a TVS diode, or Multilayer Varistor

Protection Application:
Audio signals can have a maximum value of 12Vp-p, and a maximum frequency less than 30kHz. At this frequency, the capacitance of the suppressor does not need to be taken into account. The signal lines to be protected from ESD are the left and right channels.

 

[Samuel Groner]

A ferrite bead is expensive and doesn't help too much. A TVS tends to be somewhat costly as well and usually has very high voltage-dependent capacitance which will cause distortion with high source impedances. I'm not well informed about varistors but I'm sceptical about distortion contribution as well.

A pair of 1N4148 or similar is unlikely to cause any problems for mic/line level inputs. Just don't forget to add a bit series resistance in front of them to limit current. 100 Ohm is pretty safe, take 1k if noise is no issue and 10-47 Ohm for mic inputs.

On the other hand I wonder if this is really necessary; all contemporary ICs have on-board ESD protection. A bit of series resistance will aid things, but extra diodes are probably only necessary for overdrive protection.

Samuel

 

[JohnRoberts]

+1 to what Sam said.  Check with the IC design engineers for how much protection is already built into the ICs.

Then depending on the topology additional protection can be added with diodes and series resistors.

In an input is inverting, the clamp could be as simple as back to back diodes from the minus input to gnd. Non inverting plus input, clamp diodes to rails.

I found more problems in live SR business from outputs (probably bad power distro, not static), which can be protected with series Rs and more clamp diodes.

For consumer, you can use higher impedance on inputs and outputs, so easier to protect.

Static protection is sometimes required around physical switches as consumers will sometimes draw static sparks when touching the product to change some switch setting. IIRC this was a problem with early microprocessor control in consumer products before they were  better protected.

JR

 

[horvitz]

I don't know if it will help you, but I've been looking at the THAT1200 series for something else.  There's a piece in the datasheet about ESD protection on pages 9-10.  Might be a good read.

http://www.thatcorp.com/datashts/THAT_1200-Series_Datasheet.pdf

  Brian

 

[clintrubber]

Check with the IC design engineers for how much protection is already built into the ICs.

Hi,

All what is required should already be 'inside', unless very specific cases pop up where it's unpractical/uneconomical; then 'external help' is required.

Note that chips/naked-dies/IC's/... already need to be able to survive handling, packaging etc before they even reach the spot where application-assembly takes place, let alone user-abuse. So they better be rugged and have everything on-board themselves.

This even goes to the extend that during IC-processing, the lower (earlier) layers need to be safe already. It's not enough for instance that later top metal layers connect protections; the gate-oxide of a MOSFET can then be zapped already.


For additional illustration: IC-packages with multiple dies inside can have several inside links between those dies, so not even going to an IC-pin. ESD-protection is applied for all bondpads of the dies, so also those not seeing the outside of the package.

The kind of protection can differ per connection though, for instance a sensitive input without noise-requirements can have a larger series resistance while supplies & output-drivers will have nothing or less.

And one has to take care with multiple supply domains: power-on sequencing and ESD-diodes require a closer look.

Signals of certain amplitude, biased around or near VSS can go below VSS and may not forward ESD-diodes.

etc...


Regards,

 Peter  

 

[Rochey]

Peter,

I disagree. Depending on the IC design itself, a different ESD pad can be used for different pins.

RS485 devices are designed for a 15kV static jolt, whilst some CMOS I/O can only handle 2kV.

To quote an FAQ from the boss...

ESD susceptibility varies from product to product, even when designed on the same wafer fabrication process. While some processes may be inherently more robust, the electrical requirements of the various pins of an IC may demand different techniques for ESD protection. The design of these ESD protection circuits is a specialized combination of science and art. As the state of the art improves, later products developed on the same wafer process may have improved resistance to damage to ESD. ESD protection is generally a sub-circuit (ESD cell) connected to each pin of an IC. It is designed to pass desired signals with negligible interference while protecting the primary circuit functions from static discharge. Different ESD cells may be used on the various pins of an IC to suit the specific input/output requirements of that pin. Texas Instruments has many different bipolar processes, many different CMOS processes and each process may have many variations. With the differences in processes and the products made on these processes it is impossible to provide a typical numbers for ESD tolerance according to process type. Current design standards for new products call for a minimum ESD tolerance of 2kV, human-body model. This standard may be relaxed if available ESD protection techniques interfere with the design requirements of a particular product. Many recently released analog ICs often provide tolerance in the 4kV range. Some products with special requirements may be designed for even higher ESD tolerance. ESD test results for individual products are available in reliability reports available from the quality department of Texas Instruments.

I'm doing some homework here to find out what our ESD tolerance is on the devices used, however, assuming the worst, that's it's 2kV Human Body Model, then I need to decide if that's enough for a commercial product.

/R

 

[clintrubber]

Peter,

I disagree. Depending on the IC design itself, a different ESD pad can be used for different pins.

Ehh, aren't we saying the same ? 

Regards,

  Peter

 

[Rochey]

forgive me, having reread your post I understand what your saying.

I read:

All what is required should already be 'inside', unless very specific cases pop up where it's unpractical/uneconomical; then 'external help' is required.

And figured that you were saying that *Most* devices are strong enough.

My concern is that most devices are strong enough to survive the manufacturing process, but may not be strong enough for real world connectors etc.

I'm still digging for more data on our devices, and will report back with my findings (for those of you who are interested in such things).

/R

 

[clintrubber]

My concern is that most devices are strong enough to survive the manufacturing process, but may not be strong enough for real world connectors etc.

Hi,

I have no idea which part of the life of 'the die' is actually the hardest. Being connected to the actual application-circuit might both enable nasties reaching the IC and kind of take part of the nasties away... "so it depends", which is of course an unsatisfying thing to base an eventual addition of more protection-components on.

Bye,

  Peter

 

[Rochey]

I think the silicon manufacturing and product manufacturing process are probably easier than the real world, as they are typically tightly controlled. (ESD labs etc).

In the real world, someone walks across a carpet, in dry air, with rubber shoes on, and connects the TV to the Set Top Box.... That's a lot of charge on those poor pins, unless they are protected.

Just a thought

/R

 

[JohnRoberts]

I wouldn't get into a debate about who's job is more difficult.

When ICs fail the IC manufacturer gets blamed, even if the final product design is responsible, so chip makers have made great strides in making the ICs more robust.

Anyone who worked with the original opamp ICs that cost close to $100 and would blow up if you looked at them cross eyed is thankful for the efforts made by the chip designers.

My first exposure to electronics back in the '60s (a technician gig) was working on development of a DC to DC switcher based around a very early IC VR (LM100). As i recall those were much too easy to kill, and we ended up switching to the LM300 (consumer grade) to save money as we blew them up routinely.

Modern voltage regulator ICs by comparison are just about indestructible, while there are always customers who can kill anything. .

JR

 

[Samuel Groner]

I think the silicon manufacturing and product manufacturing process are probably easier than the real world, as they are typically tightly controlled.

However once the parts are soldered to a PCB the surrounding low-impedance areas (ground planes etc.) plus stray capacitances are of great help to suck the charge. As far as I know most ESD damage happens during the manufacturing process (PCB assembly, not IC manufacturing), not in use.

Samuel

 

[clintrubber]

Yep, as said, the application-circuit will likely assist in protection.

Those harmless new-born dies encounter a lot of things before they make
it to a package. I already mentioned the processing-steps.

In addition they may/will be tested on-wafer (pre-dicing) and/or may be tested as
sawed-but-still-unmounted samples etc, all in an environment where each millisecond of
testtime adds cost.

So things have to be done quickly, so need to be robust.

I think the silicon manufacturing and product manufacturing process are probably easier than the real world,
as they are typically tightly controlled. (ESD labs etc).

In the real world, someone walks across a carpet, in dry air, with rubber shoes on, and connects the TV to the Set Top Box....
That's a lot of charge on those poor pins, unless they are protected.

Indeed 'attacks' during all phases of life, but don't mistake a processing-fab & a test-factory for a friendly environment!

That said, even early-development-samples without the full-ESD-story on them happen to survive quite a lot of evaluation-tests
(unless we really go for it  ;D )
 

But like JR said, if it's the one or the other, they have to survive all!


Regards,

  Peter