The previous parts of this series focussed on design techniques which will benefit a large number of emissions and immunity characteristics whilst also improving signal integrity. This article finishes off the series with a number of issues where specific techniques may be required, and signal integrity may not be a concern.
The high voltages that cause ESD arise through tribo-charging, the natural process by which electrons get transferred from one material to another of a different type when they are rubbed together. Man-made fabrics and plastic materials are often very good at tribo-charging, so ESD problems tend to be on the increase. ESD is a very fast phenomenon, and very intense while it lasts (usually just a few tens of nanoseconds overall).
Machinery ESD occurs when isolated metal parts rub against insulating materials, or have a flow of insulating liquids or gases over them. The metal parts tribo-charge until they discharge with a spark into something nearby which was not previously charged, equalising their potentials. Sparks created in this way by machinery can be very intense, especially when the metal part being charged is large and so has a large capacitance, which can store a large amount of charge.
Furniture ESD occurs when metal furniture (or parts of furniture) such as chairs, tables, cabinets, etc., become tribo-charged by friction against insulating materials. This may happen when the furniture is moved across a carpet or plastic floor covering, or because materials are rubbed against it, for instance when a person gets up from a chair.
Personnel ESD is caused by people becoming tribo-charged, usually by walking around. Walking on plastic floor coverings, synthetic carpets, etc., is the usual cause of personnel ESD. Few people can even notice sparks from their fingers which are under 2.5kV.
Spacecraft ESD is not covered here, although many of the techniques described will be applicable.
All these three types of ESD are very important in the manufacture of semiconductors and the assembly of electronic products, and in these areas great lengths are taken to prevent the three types of ESD from reducing yields. Machine ESD can be a big problem for process control automation. But personnel ESD is the only type of ESD which we find in EMC standards harmonised under the EMC Directive. ESD causes EMC problems in three main ways:
· The spark voltages which get into semiconductors can easily damage them. Modern semiconductors use internal insulation which can breakdown and permanently short out areas of the device at just a few tens of volts. This is known as a hard failure.
· Most ICs are made with built-in protective devices to help prevent them from damage by ESD during handling and assembly. However, these internal devices can’t be made large enough to handle large amounts of power, and a significant ESD event can over-dissipate them, sometimes while leaving the semiconductor still functional. This is known as a soft failure, because the semiconductor usually fails a few weeks or months later.
· The intense transient electric and magnetic fields created in the vicinity of an ESD spark can induce voltages or currents into nearby circuitry and upset its operation. This does not usually cause direct damage, although the resulting malfunction can sometimes cause consequential damage of some sort.
The ESD simulator used for testing to EN 61000-4-2 is based upon the 150pF/330W human body model, and generates a waveform with a risetime of between 700ps and 1ns to reach a peak of several kV, which then decays to about 50% in 50ns. At a voltage of 8kV the peak current into a 50W calibration load is close to 20A. The frequency content of such an ESD waveform is flat to around 300MHz before it begins to roll off, so contains significant energy at 1GHz and above.
Some older test standards use an older human body model which only has a 5ns risetime, so its spectrum begins to roll off at 60MHz and it is not as aggressive a test as EN 61000-4-2. As high-speed measurement techniques improve, it appears that real ESD events may have risetimes faster than 700ps.
Testing to EN61000-4-2 (personnel discharge) involves the following:
· Air discharges of up to ±8kV (using an 8mm round tip to simulate a human finger) are applied to everything non-metallic which is normally accessible to the operator.
· Contact discharges of up to ±4kV (using a sharp tip which is touched against the product before the discharge) are applied to operator-accessible metal parts – and also to nearby vertical and horizontal metal planes.
Test voltages are increased gradually from low values, often using the settings 25%, 50%, 75%, and then 100% of the test voltage. This is because ESD failures are sometimes seen to occur at lower voltages but not at the maximum test level. The highest test level on an ESD test is not necessarily the one most likely to cause a failure (this is also true for other types of transients). Figure 6A is a sketch of the barest essentials of an ESD ‘gun’.
Bear in mind that in dry climates personnel ESD events can easily exceed 8kV. 15kV or even 20kV is not that unusual during freezing winter conditions when the air is very dry, especially in heated homes and buildings without humidity control. So, meeting an ESD test at ±8kV is not a guarantee of freedom from actual ESD problems in the field, and the environment and needs of the users should be taken into account when ESD testing to help produce reliable products.
All the design techniques described in the previous parts of this series help a great deal in improving the immunity of circuits to electric and magnetic fields, and so help circuits cope with the brief but intense bursts of wideband disturbances from ESD events. However, they are not usually enough on their own. The two main techniques for preventing ESD sparks from upsetting products are:
· Dielectric protection (insulation)
· Shielding (metal or metallised enclosures)
Dielectric protection is the preferred technique, but where it cannot be used for an entire product ESD problems can occur with both internal and external connections. These are discussed below. Apart from dielectric isolation, many of the techniques described below will also be useful for protection against other conducted transients and surges, which have not been dealt with in this series of articles as a separate topic.
This is the best ESD protection method. By not allowing an ESD spark to occur at all, not only are sparks prevented from getting into sensitive circuitry, but no bursts of electric and magnetic fields occur either.
Plastic enclosures, membrane keyboards, plastic knobs and control shafts, plastic switch caps, plastic lenses, etc., are all pressed into service to insulate the product (especially the operator-interface areas and controls). A 1mm thickness of common plastics such as PVC, polyester, polycarbonate, or ABS, is usually more than adequate to protect from 8kV of ESD (check the breakdown voltage rating of the material in kV/mm of thickness). But since no practical enclosure is without seams, joints, and ventilation, the achievement of adequate creepage and clearance distances becomes very important. Creepage is the shortest path that a current would have to take if it ‘crept’ along all available surfaces to reach the vulnerable part, while clearance is the shortest path to the vulnerable part through air (metal parts encountered along the way counting for zero distance regardless of their dimensions).
Clearance is the easiest to deal with, because the breakdown voltage of air is usually around 1kV/mm. So as long as the distance from the tip of the ESD gun to the vulnerable part is at least 8mm (preferably 10 or 12mm to give a design margin) an ESD spark can’t occur.
Creepage is more difficult, because the surfaces of plastics are always contaminated with mould-release chemicals, fingerprints, dust, etc., which attract moisture from the air and form a variable conductivity surface. Sparks from the tips of ESD guns are often seen to follow a random path over the surfaces of plastic enclosures, displays, keyboards, etc., sometimes for as long as 50mm as they follow the path of least resistance through the dirt on the surface of the plastic, eventually ending on a metal part. (Painted metal surfaces often show similar long random spark tracking, usually leading to a pinhole defect in the paint that it takes a microscope to see.) So it is very difficult to specify an adequate creepage distance which will protect from an ESD test, although more than 50mm is probably adequate except for polluted or wet environments.
Figure 6B shows a combined creepage and clearance design issue. A joint in a plastic enclosure could allow an ESD spark to travel along the surfaces of the plastic, then through the air inside the enclosure to terminate on a vulnerable PCB track. Figure 6B shows that it is usually a good idea not to line up PCBs with seams or joints in their plastic enclosure.
LCD displays, membrane panels, and tactile rubber keypads can be very good at preventing ESD if a few basic precautions are taken. Although their surfaces are ESD-proof at least to 15kV, they can have problems at their edges. ESD sparks can track along the dirt on their insulating surfaces, and go around their edges to reach vulnerable internal tracks.
LCDs often dealt with this problem by using large bezels which prevented fingers from getting too close to their edges. Insulating sealant and similar materials are now more likely to be used these days. Another method is to surround the LCD panel with a metal bracket that ‘catches’ the spark before it gets to any sensitive parts, but then something has to be done to remove the charge from the metal surround without it discharging itself into some sensitive part.
Membrane keypads and panels have internal conductive tracks, sandwiched between glued layers of plastic. If these tracks get too close to the edge of the panel, and if the glue has an airgap in it, sparks can track from the front surface (where the air discharge tip is applied), around the edge, through the void in the glue, and into the internal track, giving a false keypress if nothing worse. So whilst all attempts should be made to ensure there are no voids in the glue, it is still best to keep internal tracks at least 12mm from the edge of the panel (much more if possible).
Tactile rubber keypads also suffer from sparks that track through their surface dirt around the edges of their rubber mouldings and into the vulnerable keypad tracks behind. Unlike membrane panels, they usually don’t have the benefit of glue to provide insulation, so it is important to extend the rubber edges of the tactile key moulding for far enough out, whilst keeping the tracks on the underlying PCB far enough in, so that any sparks have too far to go.
When a plastic enclosure has an internal shielding coating applied to meet RF emissions or immunity requirements, this can compromise dielectric isolation measures. For the conductive layers to make a connection across enclosure seams they must extend at least a little way into the seams, and may even be fitted with a conductive gasket. This can compromise the creepage and clearance distances that had existed on the unshielded version, and ESD tests on such enclosures often find that when the tip of the ESD gun gets anywhere near seams and joints in the enclosure, a spark flies from the tip and disappears into the seam or joint to meet the internal shielding layer. In this situation it is usually very hard to achieve dielectric isolation for the whole enclosure, and the shielding method described next may have to be employed instead. Where possible it is a good idea to plan ahead so that plastic cases are designed to allow internal shielding to be added later without compromising the dielectric isolation ESD protection. This can be difficult to achieve, especially on small products.
Shielding attempts to divert the (very large) ESD currents away from internal circuitry. In general it is not as good as dielectric isolation because it exposes all the external conductive connections (and possibly internal circuits too) to indirect ESD injection via ‘ground lift’.
When an ESD spark occurs to a metal enclosure, for the first few microseconds the enclosure will be at a much higher voltage than any protective earth it is connected to. This local ‘ground lift’ decays as the charge on the enclosure leaks away through the inductance of any protective earth connections (usually several tens of mH). Where an enclosure is not connected to protective earth the charge on the shielded enclosure leaks away slowly through ionisation currents in the air around it, conduction through humid fabrics, and similar mechanisms.
During the beginning of a ground-lift event, internal circuits may still be at their previous voltages and sparking may occur between the enclosure and internal parts. This is known as ‘secondary arcing’ and it can be as bad for semiconductors and signals as the original ESD spark.
As long as the internal circuitry can cope with the sudden change in potential of its enclosure they don’t care whether they are at protective earth potential, or 8kV relative to it. So it is quite practical to make battery powered or double-insulated products withstand ESD events (even though they are not connected to earth), although their isolation from earth may create greater problems for their external interconnections (discussed later).
So, on its own, having an external shield is not enough. One solution to secondary arcing is to bond the internal circuits to the enclosure shield, using connections that have a low enough inductance to maintain a low voltage between circuit and shield during an ESD event. These connections are often direct bonds from the 0V planes in the circuit boards to the enclosure, but they could be capacitive connections instead. Another solution is isolate the internal circuits from the enclosure using materials (or air spaces) that will not break down due to the voltage overstress.
Having dealt with secondary arcing we need to address the problem of transient current injection into the internal circuits.
The circuits within a shielded enclosure are all exposed to the internal fields created by the momentary ground-lift of the enclosure, before the internal potentials have had time to equalise. Because all the different PCB traces and components (even lead frames and bond wires) have different amounts of stray capacitance to the enclosure they each experience different amounts of injected stray current. Because the risetime of the ESD event is so fast and contains such a high frequency content, even very small stray capacitances can inject quite large currents. These different currents can create differential signals which can upset circuit operation. Figure 6C tries to show how this problem occurs.
Solutions to the transient current injection include:
· Bonding the PCB ground plane(s) to the enclosure at very frequent intervals, to reduce the bonding inductance so that the ESD voltages equalise as quickly as possible. Quicker equalisation = lower internal voltage differences = lower values of transient current injection.
· Shielding sensitive circuits or ICs with PCB-mounted metal boxes bonded to the local PCB ground plane. These may be thought of as intercepting the stray capacitances from the components and PCB traces to the enclosure, diverting their transient currents into the ground plane where they will do less harm.
The problem illustrated by Figure 6C is often made worse where an enclosure contains a number of interconnected PCBs, as it is so difficult to ensure that during an ESD event they all charge up to the enclosure voltage at the same rate. If one circuit board has a low inductance connection to the enclosure, whilst another has a high inductance connection, then there can be a substantial transient voltage difference between them. This would inject a pulse of current into the boards’ interconnections, causing cause signal corruption if not actual damage. So it is always a good idea to bond the 0V planes in different boards together using a number of conductors (one reference conductor for every one or two signal conductors may not be excessive) to help prevent large internal voltage differences.
Where galvanic isolation of circuits from their enclosure is used to protect sensitive signals from differences in earth potentials, usually where long external cables are involved, the isolation is usually only needed at mains frequencies so it is often possible to capacitively bond across the galvanic barrier (with suitably rated capacitors). This equalises transient voltages during an ESD event without compromising functionality. Where this method is unsuitable, an internal shield over the isolated circuitry, bonded to its local isolated reference plane, may be required.
Even tiny gaps or joints in enclosure shields are weak spots, because they divert the very large fast currents from the ESD spark as they flow around the enclosure, causing locally intense pulses of electric and magnetic fields to be emitted through the shield and into the enclosure. Looked at from the frequency domain, we would say instead that the high frequency components of the ESD event find gaps and joints useful as slot antennae, radiating into the enclosure. So it is important to keep all gaps and joints in shields to a minimum size, as described in Part 4 of this series. Even if they are very small, sensitive circuitry should be kept well away from them. |
