When has a relay failed in a switching system?
It maybe obvious when a relay has failed, for example if the contact has definitively failed to close or failed to open through welding, mechanical failure, coil failure or some other mechanism. It is not always quite as clear cut as this though, and the answer may also depend on the type of relay involved and the users application.
Switching System Path Resistance
Switching systems are made up of a variety of parts which include the following components:
- Connectors which have conductors and mechanical contacts
- PCB tracks
- Wires to the PCB
- Solder joints
In most cases the cause of any path resistance variation is likely to be principally created by relays, though worn connectors or poor solder joints can create issues.
The contribution to path resistance of a relay is variable according to the type of switching system. In high current systems the relay may be a significant part of the path resistance because heavy wires are used to connect the front panel connector to the relay. In lower current solutions the relay is likely to have 30% or less contribution to the path resistance so any rise in the relay path resistance is masked by the path resistance of (predominantly) PCB tracks. Designs which are rated at more than 5A commonly use wires and very short PCB tracks unless the switching system is relatively large (for example as in some LXI systems).
Solid State Relays
In solid state relays the open or closed signal path is determined by the characteristic of a channel established in a semiconductor material. If the on state path resistance rises significantly above the value seen at the point of manufacture under the same environmental conditions (the path resistance usually increases significantly with temperature) then it is fairly certain that the relay is changing and headed for failure.
Solid state relays can suffer damage that appears as leakage current, if a solid state relay exhibits leakage in excess of its initial value (again under the same environmental conditions) it is likely the relay has suffered damage through excessive stress.
In most cases however it is very clear when a solid state relay fails, it will either be a short circuit (the equivalent of a weld) or will fail open and the case for replacement is obvious.
For reed relays the contact resistance of a new reed relay designed for handling low level signals (usually with ruthenium contacts) is typically in the region of 80mOHm to 150mOHm depending on the reed type used and on batch variation. Since switching systems using reed relays are primarily intended for lower current (1A and below) switching, the switching system resistance on reed relay designs is dominated by the PCB tracks used to go from the front panel connector to the relays in the switching system. The reed relay contacts are contained within a sealed glass envelope that make them immune to the ingress of contaminants and this gives the reed relay a stable contact resistance until such time as the precious metals on the contacts is eroded by either mechanical wear or hot switching events.
An increase in the path resistance of a ruthenium reed relay with use is attributed to a reduction in the usable contact area when the relay is closed. That might be caused by mechanical erosion but because reed relays do not usually significantly mechanically "wipe" their contacts together it is more commonly caused by hot switching eroding the surface and even creating small pits when a "soft weld" occurs. A soft weld occurs when hot switching causes the contacts to slightly stick together, an event which tends to increase the operating time but also makes a relay prone to further degradation as damage accumulates.
The plot below shows a reed relay being life tested (three samples of a 0.5A ruthenium reed relay being hot switched at their maximum rated power) with a number of hot switch operations.
As the number of operations conducted increases it can be seen that the relay resistance tends on average to increase. The rise is not entirely consistent (mechanical closure can flatten contacts and improve contact resistance for a while) , but in all samples the resistance after 5 million hot switch operations is higher than the initial value. All three samples are still operating consistently at the end of this cycle and none would be deemed to have failed.
If the resistance doubled then it implies the contact area has halved and it is probably reasonable to suppose such a relay is likely to struggle to meet its specification for hot switching or current carrying as the contact will be operating at a higher temperature than its design value.
As contact resistance rises then even on cold switching heat starts to be an issue. A reed relay whose resistance has risen to 1Ohm is increasingly unlikely to be able to sustain its rated current without the reed contacts and the reed materials reaching excessive temperatures. If a reed reaches a critical temperature it will lose all its magnetic properties and a relay which is normally open but has been closed by its coil will suddenly open until it cools down. A reed relay is (mostly) a small component (usually smaller than EMR's) and the reed relays primarily on its lead frame to remove heat, so the package cannot dissipate more than around 0.5Watt without concerns for the ability of the reed to carry its rated current.
In general if a reed relay has a contact resistance of 1Ohm it is likely it is well beyond its useful service life for users who wish to use it to its specification. The same will be true for users who need the switching system to have repeatable path resistance. A user who is only using the switching system to connect to a DMM in voltage mode may consider that more service life can be obtained but needs to be aware that that they may get increasingly unreliable results as either the contact area has markedly reduced or the precious metals have been eroded away.
A safer assumption is that if reliable operation is required is then a reed relay showing a doubling of resistance is becoming a risk to consistent operation of the switching system. That could be a relatively small resistance rise for the switching system.
There are some types of specialist reed relay which may have different resistances at low voltages than at high - those which are primarily intended for high voltage switching whose contacts are based materials other than ruthenium. Tungsten and to a lesser extent rhodium high voltage reeds can show these effects when measured with a DMM. The high voltage hot switching event can lead to variable and unpredictable resistance at low signal levels, but this does not impact their intended use for high voltage switching.
EMR's can be a little more complicated than reed relays to understand, the contacts are not in the benign environment of a hermetically sealed glass envelope and the signal levels they handle can be far higher than reed relays. This all serves to introduce the complications caused by non-linear contact behaviour and the ingress of contaminants which is highly dependent on the details of the relay design.
For relays that handle lower currents (generally 2A or below with 60W or less hot switch power) contacts are generally of gold, a soft material that does not oxidise and has good electrical conductivity. The main cause of path resistance increases is that either foreign materials appear on the contact (for example outgassing from plastic packaging of the relay case) or the gold is eroded away by hot switching. Foreign materials are usually able to be wiped away by using a contact design which includes a deliberate wiping action,pushing the surface materials away and flattening the contact areas. These types of relays have consistent path resistance at all signal levels they are rated at until contact degradation starts to occur. Increases in path resistance are typically due to the erosion of the gold or the build up of contaminants that cannot be cleared by the wiping action. Contact resistance is usually less than that of a reed.
The plot below shows some life tests for EMR's at low signal conditions. The life tests are accelerated, the relays are operated at a rapid rate in order to shorten the time that the test took - real mechanical lifetime is often higher when switching takes place at more sedate rates, particularly when higher loading conditions are tested.
The plotted EMR's use gold contacts and the contact resistance during their life to beyond 100million operations is stable. Two traces show more erratic resistance after 100 million operations, but at no stage is a catastrophic failure present. For some applications (for example use within resistor modules or to switch resistor module in and out of a system) the performance above 100 million operations is not good enough. For other applications the test system may accept this level of performance.
The resistance might fall under heavier load conditions because the erratic resistance almost certainly indicates that for those relays significant areas of the gold contacts have eroded, but other materials may provide a lower resistance as the current is increased.
For switching systems designed for lower signals levels the resistance of the relay is likely to be just a part of the switching system path resistance and the relay effects will be masked by other factors, such as PCB trace resistance. A doubling of the path resistance would indicate that significant change of resistance has occurred in the relay, perhaps by an order of magnitude.
For EMR's rated at higher currents the situation can be more complex. The contacts may have a gold flash but this intended only to protect the contacts in the manufacturing process and storage, as soon as hot switching occurs this soft material tends to get worn away and then the contacts operate with materials such as silver and nickel which more easily react with other elements but more strongly resist erosion due arcing or mechanical pressure. This can create a situation where the contacts become non linear - the path resistance varying according to whether the relay was or hot or cold switched and the amount of current flowing in the contact. The non linearity is both with signal level and with time. A relay carrying a small current can have a significantly higher contact resistance than when it carries a high current (which tends to heat the contact and often evaporate away volatile materials), and often that contact resistance will reduce with time.
The issues can be complex and below is a test that was performed on a high current relay that had been subjected to hot switch operations and was then tested for contact resistance on 4 samples following a hot closure at 12V 100mA, the horizontal axis plots time from closure of the contacts.
What can be seen it that for three of the relays the resistance was very consistent, but one relay shows a marked resistance versus time dependency. That relay would be considered to serviceable, if the relay had been tested at much higher current then the resistance would have been lower. That relay would not be considered a failure by its manufacturer and is perfectly usable in the majority of intended applications. But is also shows why we strongly recommend to users that they use relays designed to switch the signals that are present. Using high current signals for DMM voltage two terminal resistance measurements is not a recommended design practice. To complicate perception further if the relay with the strong time dependency was retested following a high level hot switch operation the resistance may become lower, and with time those relays with a consistent resistance might have shown more time dependency.
The plot below shows a different vendors single high current relay which was hot switched at different signal levels and its resistance measured over time (horizontal axis).
What can be seen is that the path resistance is highly dependent on both time and the signal levels applied. The contact materials on this relay are different to those used previously. Under the lightest load condition (the trace with the highest resistance) the relay resistance had a general downward trend but also has a few transient upward resistance changes. At higher loads the resistance becomes much more consistent.
Again the relay vendor would not deem this relay to have failed because their test condition measure resistance at high signal currents when the contact has settled. For a user this may make the relay unsuited for a given application, but the origin of the issue is that a higher current relay has been used.
It is for this reason that Pickering Interfaces switching system test tools, eBIRST and BIRST, do not support high current relays since the tools measure path resistance at relatively low signal currents (up to 30mA). A DMM measures resistance at even lower currents (usually 1mA) and is even more problematical.
Although some types of relay degradation are outright failures in other cases the situation is less clear. In particular there may be a dependency on the signal levels present and the sensitivity of the application to path resistance. Tools such as eBIRST and BIRST avoid testing these types of switching system for this reason, the use of DMM to test the rely closure is also unreliable as the signal levels used are even lower. Users may need to consider using high current levels which are similar to their application requirement to judge whether a relay continues to be suited to their application.