EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: Saimoun on September 08, 2020, 05:09:01 pm
-
Hi
I am sorry if this question already has been asked many times, I have been googling it without really finding a clear answer (maybe there isn't).
I am wondering, for a professional project (that will be sold to customers at the end), how much ESD protection you should put?
I.e. is it really necessary to have a TVS diode on every single trace/pin that could be exposed? For example I have heard you should even protect plastic push buttons because high voltages will go through the plastic - do the electronic devices I can buy in a shop have ESD protection on every plastic push button?
And what if the device has a plastic case, then high voltage could go through the case and through the air to reach basically any part of the PCB, could it not?
Loads of questions as you can see ha ha :D - feel free to answer what you can/want ;)
Thank you!
Simon
-
That depends on the product. Some will require mandatory ESD testing, so protection is also necessary. What level of protection is necessary depends on the testing requirements in this case.
If it is just a regular consumer product, then protection is technically not necessary. Depending on your volumes you may put none and see the amount of returns. Then focus on the specific issues rather than covering absolutely everything.
-
As ataradov says there is not a universal answer to this question. In some cases there will be contractual or legal requirements. In most cases the only need is to keep customer satisfaction high enough and returns/repairs costs low enough. And there is no clear path to this. Depends on your circuits, packaging and markets. Many circuits have intrinsic protection that will be adequate for most instances. Circuits can also be arranged to reduce ESD susceptibility without adding TVS diodes. A debounce circuit also provides some level of protection for example. Same thing with impedance matching resistors. A device marketed to wet humid locations (humidity control devices, water level detectors) will tend to have low ESD exposure.
A few hours looking at commercial products should give you a lot of insight into what works for most applications. It isn't full protection of every pin.
-
How much do you want to rely on internal ESD protection?
Basically all chips have ESD protection of some sort, even if they don't specify it. 1kV HBM is the usual standard I think, and more is often seen.
But then, maybe they really didn't put in any ESD protection at all. You don't know. If I don't see it in the datasheet, I treat it as worst case. (In a recent design with a proprietary SoC that didn't put this spec in the datasheet, I had to ask their engineering dept -- what a bother, and so easily avoided.)
For stuff like membrane pads, and sure even pushbuttons, you will get static discharge over the surface. Even if it doesn't strike metal directly (through gaps in the enclosure, or defects in the overlay). A path to shunt this noise is highly recommended -- an excellent reason to use solid ground planes in PCBs.
Slightly conductive plastic can help, dissipating static rather than allowing it to spark -- this still won't eliminate interference from the odd spark nearby (e.g., discharge to conductive plane test), of course. May also be a comfort thing, so you don't have products being handled and generating static (triboelectricity). This can even be industrial-grade disaster when things are in continuous motion: material handling systems, conveyor belts and such. Neglecting to specify conductive plastic parts, or adequately grounding metal parts, can turn your system into a van de Graaff array!
Anyway, it's safe to assume that anything towards the outside of the enclosure, including wires, cables, ports, displays, buttons, etc., will be exposed to at least modest levels of ESD (a few kV equivalent). The reason is that an ESD discharge is an EMP blast, a propagating wave with a wavefront of 10s of cm -- it propagates through air, it reflects off surfaces, it transmits along wires. It's very fast, so if you don't need much bandwidth, it can be filtered out reasonably well. Because the voltage is high, the peak (direct or induced) current is also high, which is why it's so destructive to ICs. Because the voltage and frequency are both high, it can't be insulated away (it goes right through dielectrics, that's the whole point :) ), it has to be shunted away. Don't try to build a sea wall, the waves will just spill over it -- build a lighthouse that the waves effortlessly wash around.
You can use internal ESD structures to deal with this, and get away with it for infrequently handled items, or at low reliability, i.e., no expectation of long uptime, or it can be power cycled to restore functionality, or it has a short product lifetime, etc.
If you need high reliability, providing additional ESD control and filtering is a good idea. Choose interfaces that are specified with high ESD. Everything else, filter it as well as you can (only use the bandwidth you need!).
Using a shielded or metallic enclosure is also a good idea. And using it properly, i.e. filtering/protecting everything that penetrates the shield, where it penetrates. Especially if the EMI environment inside your box is expected to be very noisy (e.g., a big power converter?), or very susceptible to external interference (radio equipment?).
There can't be clear answers, because the environment is individual, and so is the solution. Lots of products do succeed with open PCBs in plastic enclosures; sometimes these are well designed and pass on the first go, other times they fail repeatedly and need several iterations to pass.
The most powerful simplification I can suggest is to put everything inside a Faraday cage, and filter/protect signals where they penetrate that shield. This is how RF circuitry is designed: assume every block sits inside an ideal shielded environment, with only N ports connecting it to the outside world. The impedance, bandwidth and amplitude of signals entering/leaving those ports can be evaluated, rigorously; there can be no other effects on the circuit, than what happens through those ports.
You can do topological transformations on that shield, changing its shape, opening holes, etc. Every time you open a hole, you're reducing its shielding effectiveness, and make the analysis more complicated. Now instead of N ports, there are N ports plus coupling to external fields, through those openings. Fields are more complex than ports, they are 3-dimensional; ports are 0-dimensional (at frequencies where they can be analyzed as ports, that is). This greatly expands the coupling matrix: rather than being a sparse matrix (mostly zeroes), now you have a lot of nonzero terms. Now you need to weigh whether those terms are still negligible (can be rounded to zero). If they can't, well, maybe you can still do some mitigation -- apply local shields, or add filtering or protection in the middle of the circuit as well as around the edges.
You can transform a closed Faraday cage, to an open PCB with ground plane, by making exceptions like these. This is why a ground plane is so important.
Tim
-
Most of the time ESD protection is done to avoid excessive warranty returns that can get very expensive if they get out of hand.
Most common chips have built in ESD protection to both make the product more resilient and to reduce the chance of the chip getting blown before it is soldered on the PCB. Sometimes this is enough, other times you need additional protection. Sometimes the chips datasheet will suggest the needed protection.
Typical practice is to focus ESD protection on things that are exposed to the user/enviorment/cables. So things like a touchscreen might have extra ESD diodes, metal buttons that stick out might have grounded bodies or resistors on the lines etc.. But most important are outside connectors. The user holding the conductive cable makes a very nice path for ESD to enter, the device itself might be charged and discharge trough the cable when plugged in or something might go wrong with the cable or the device that is on the other side that sends excessive voltage into some pins.
Product resilience is not only about ESD however. Some situations make it sensible to have reverse polarity protection on the power input, some situations might need transient clamping due to potentially dirty power input. Some situations require resilience against intense vibration or mechanical shock. All the ESD diodes in the world will not be any help at all if your product dies anyway due to a important component falling off the board due to vibration.
-
Thank you all that is a lot of wonderful information!! :D I learnt a lot reading you all, in particular the stuff about "[ESD] can't be insulated away" from Tim :)
And yes I did expect answers like "it depends" :P but that is fair :)
A couple of you speak about meeting ESD requirements - are there ESD requirements to get the CE of FCC marking for example? I thought it was mainly EMI (which - as I understand now - are still corrolated since ESD generates EMI :palm: ).
And just FYI the devices I want to make will be used in the audio (live concerts) industry - but feel free to stay general as I am sure this post will help others in the future ;)
-
A couple of you speak about meeting ESD requirements - are there ESD requirements to get the CE of FCC marking for example?
Not for CE or FCC. But for things like automotive applications and white goods you have additional safety standards. I don't know specific standards, I just know they exist.
Stage equipment may be subject to some severe stresses, depending on where it is used. If it is something attached to long cables, I would definitely think about additional protection. Especially if your equipment is on the critical path for the performance. You may not get a second chance if it craps out.
-
Thank you all that is a lot of wonderful information!! :D I learnt a lot reading you all, in particular the stuff about "[ESD] can't be insulated away" from Tim :)
And yes I did expect answers like "it depends" :P but that is fair :)
A couple of you speak about meeting ESD requirements - are there ESD requirements to get the CE of FCC marking for example? I thought it was mainly EMI (which - as I understand now - are still corrolated since ESD generates EMI :palm: ).
And just FYI the devices I want to make will be used in the audio (live concerts) industry - but feel free to stay general as I am sure this post will help others in the future ;)
Meeting ESD and EMI requirements will be required for CE marking although there are different levels of compliance for different classes of equipment and the degree to which the ESD or EMI affects the operation of the equipment. For your application you can probably get away with the minimum level of ESD compliance i.e. the ESD pulse doesn't permanently damage the equipment.
I suspect your bigger problem is going to be dealing with EMI i.e. interference from external radio sources into the operation of your equipment. Your clients are not going to be happy if they get break-through of spurious noises onto their audio if someone uses a portable transmitter nearby.
Dealing with ESD and EMI problems often go hand-in-hand. Techniques for dealing with EMI often help with ESD compliance and vice-versa.
-
Ok thanks that helps a lot :)
If I summarize:
* Insulation does not protect from ESD
* ESD can come from any exposed metal but also:
** any cable attached to the device
** any plastic push button
* Metal case is a good idea - helps with EMI as well
Still one question though: what about grounding? i.e. a lot of products nowadays are not connected to earth's ground, if for example I make a USB powered device or simply have an external power supply. Will it make a difference?
And can ESD damage anything if someone who has a lot of (electrostatic) charges built up touches the metal case connected to ground?
-
Ok thanks that helps a lot :)
If I summarize:
* Insulation does not protect from ESD
* ESD can come from any exposed metal but also:
** any cable attached to the device
** any plastic push button
* Metal case is a good idea - helps with EMI as well
Still one question though: what about grounding? i.e. a lot of products nowadays are not connected to earth's ground, if for example I make a USB powered device or simply have an external power supply. Will it make a difference?
And can ESD damage anything if someone who has a lot of (electrostatic) charges built up touches the metal case connected to ground?
Insulation can help but you need quite thick insulation and there must be absolutely no gaps in the insulation - the ESD pulse will find and sneak through even the tiniest gap.
Grounding is very important. The most effective way of dealing with ESD is to divert the ESD pulse currents to ground rather than having it go through your sensitive circuitry. The ESD pulse can be >10kV peak voltage and >20A peak current so it is essential that the path to ground has very low impedance (especially low inductance) and that it is sufficiently well separated from sensitive parts of the PCB so that there is no possibility of flash-over to adjacent tracks.
If you have a USB connected device make sure that the discharge path is connected through the shield on the USB cable to provide a safe ground path for the ESD pulse to flow through rather than through the USB signal and power leads.
If you don't have a defined mains ground, such as when you have an external DC plug pack, then the discharge path has to flow back through the power lead to the plug pack and then through the stray capacitive coupling in the plug pack to the mains supply. The high voltage of the ESD pulse can put a strain of the insulation of the plug pack although it is usually not a great problem with switching regulator based plug packs. These typically have an RF interference suppression capacitor coupling the mains side and output and this bypasses the ESD pulse round the PSU circuitry and back to the mains supply.
-
Earth helps you none, here. Earth is only a galvanic shock hazard, low frequency (mains and DC) thing. Earth is only needed when mains voltages may be around, and when double insulation cannot be provided, or isn't sufficient for whatever reason.
Consider a situation: say you have a device, at the end of a cable, with a metal front panel and a PCB behind it. The metal panel is earthed through the cable. The PCB is powered from other wires in the cable. Say it's mains powered, and it's a traditional 3-prong cord.
At the instant an ESD spark hits the panel, it gets charged up to several thousand volts. The pulse propagates over the surface, coupling into nearby conductors. (Meanwhile, the pulse propagates backwards down whatever was charged, be it an ESD gun with ground return strap, or an unfortunate human.) This in turn induces several thousand volts on the PCB, but not as much as on the panel; effectively, there is a capacitor divider, from panel to PCB to surrounding space. Thus, even if the PCB is built with a solid ground plane, its components and surface traces see the electric field of one or a few thousand volts on the panel, with respect to local ground.
A reminder that voltage is only ever relative. At any instant, a component only knows what voltages are across it. If a PCB ground happens to be thousands of volts above its surroundings, but there's only a few volts dropped across nodes in circuit, which one does a component see? Just the few volts.
So the fact that there's a few thousand volts difference between the panel and PCB, couples some fraction in turn into circuit traces and components. This might not be destructive anymore (it might be 10s or 100s of volts worth of ESD per trace; most traces are thin and not very long, so don't have much cross section to the electric field), but it's still enough to upset digital logic (e.g., accidentally strobing SPI or parallel bus devices (memory?), desyncing them, or corrupting a few bytes?).
Some nanoseconds later, the ESD pulse propagates down the cable, still coupling with the power lines in it. The change in impedance, between the bulky mechanical stuff (probably a lower impedance?), and the relatively thin wires (~80 ohms?), probably dulls the waveform peak a bit. (Meanwhile, the wave propagating over the charged body, reflects off features on it -- there pay be peaks and dips corresponding to its shape.) Eventually, the ESD pulse finds a meaningful ground return (say, a metal conduit box bolted to metal studs/beams), and the discontinuity reflects the pulse back up the cable (somewhat attenuated). The fraction in the power lines continues on into the mains network. When these pulses return, the panel voltage oscillates negative, and the PCB-to-panel voltage isn't driven by much so it follows again by coupling. At this point, most of the high frequency energy has dispersed (radiated away, lost in the wiring* or scattered by the differing impedances and velocities of the panel-PCB system, the cable, and common/differential modes), and the waveform rings down somewhat more gently.
*PVC is a common ingredient in wiring, and isn't a very good RF dielectric; I'd be willing to bet it accounts for a good maybe 5% of loss here.
If the power cord is about 2m long, the propagation is about 6ns at the speed of light, and a full reflection is both directions, so we'd expect the ringdown (standing wave) to be around 80MHz. The ESD rising edge is a few ns, so contains energy up into the 100s of MHz.
We can draw the equivalent circuit of the system:
The ESD strike acts like an ideal switch between a source network and the EUT.
The source is typically drawn as a series capacitor and resistor, but there is a complex RLC (or transmission line) equivalent, which gives a somewhat double-peaked waveform. In an ESD gun, this is comes from the wave reflected from its ground strap. From a human, this is the reflected wave from their overall body length. Skin resistance isn't terribly low, hence the ESR, and the velocity factor in flesh is a bit lower than in air, so the ESD gun needs a somewhat longer (usually around 2-3m from what I've seen) strap than the length of the human body I think.
The EUT panel and PCB might be modeled as a few capacitors with respect to "free space": this is questionable, but can be justified when the panel and PCB dimensions are small (smaller than the pulse risetime). There are capacitors from each node to ground, and one between them. The capacitances to space are also radiative (lossy), depending on frequency; this represents coupling to free (propagation) modes.
The cable might be modeled as an LCL network, the number of terms (LCL(CL..)) depending on how much bandwidth we want to model. A better model is a pair of transmission lines, one representing differential mode (power lines with respect to the earth conductor), one representing common mode (power and earth with respect to free space). The two lines have different velocity factors (DM has lower velocity, as DM is mostly in dielectric while CM is mostly in air), and their losses will be different, because of the dielectric vs. radiation loss. The CM line is grounded to one side (representing the ground return to surroundings -- the assumed larger metal structures), and the DM line is terminated in part (representing the continued propagation through the mains network -- this might not be a good match, so some RLC equivalent might be justified).
A note on how these TL sections are wired: ideal (SPICE or otherwise) TLs have no common mode current flow themselves, i.e. the two ends are ideal ports (two pins with equal and opposite current flows). In the common mode, we are modeling a TL stub, so the far end is shorted (and for SPICE, referenced to ground just to avoid a singular matrix error), and the near end, one terminal connects to the panel, the other end connects to earth. The DM line is connected between the panel and PCB at the one port, and the other port is terminated (and again, referenced to ground only to avoid an error; its actual absolute voltage is meaningless).
Note that this makes frequent reference to some sort of imagined ideal ground: "free space". We don't really have any way to invoke fields at infinity, without modeling the whole damn thing -- this is, after all, an equivalent circuit. So we assume there's some kind of ground at infinity and assume that anything that would propagate out that way, is lost to the circuit -- that is, burned in a resistance.
We could of course model near fields, like reflections off metallic surfaces coupling back into our system, but it's really not necessary as the ESD pulse is very sloppy to begin with (30% tolerances in IEC 61000-4-2, IIRC); once our simulation is within this margin, we really can't care less about finer details like that. (Or like if the human body has arms and legs -- these would be modeled as open TL stubs, and have the effect of adding small lumps to the pulse.)
Anyway, for ESD protection purposes, local ground is all that matters. Obviously, we wouldn't have so much voltage between panel and PCB if they were grounded together locally, rather than through the mains network (which is to say, not really grounded at all -- in the above story, I've assumed several thousand volts between them). If they were bonded at one point, then the panel-PCB system would act as a tuning fork resonator, struck by the ESD pulse; for a roughly handheld sized object, this would be in the 100s of MHz, and might still be worth a thousand volts or so (if it's a bit smaller, it will actually be in the cutoff range where less energy is present; but if bigger, like a tablet PC or laptop sized thing, not so much). If we add multiple ground points, we can short out that mode as well, and so on and so forth; as we spread ground points around the system, we push the resonant frequency up (into the GHz), or eliminate resonant modes entirely (by making a fully closed shield). Now we might have mere 100s or 10s of V between panel and PCB, or for a metallized enclosure it might be less than 1V -- a >80dB attenuation ratio!
Which means, we would have good reason here to use a mains-isolated power supply in this device (so we don't care about surges up the power line, or ESD down it), and to ground the secondary side to the enclosure, and earth it. (Which also only needs Basic insulation in the power supply -- i.e. if a single fault occurs in its isolation barrier, fault currents will be shunted to ground. If the enclosure were not earthed, it would have to be Reinforced, i.e. a double layer of insulation is used, so that a single failure of either does not expose the user to hazardous mains voltages.)
Tim
-
Thank you both for the very helpful explanation. Tim I did not understand the whole thing but I am definitely smarter on the subject than I was before :D Thank you!