Diagnosing Faulty Relays in a Switch Matrix
Relay switch matrix instruments do not generally provide a way of verifying the integrity of the relays that make up the matrix. So, how can you tell if there are shorts between adjacent paths? Or shorts to ground or some other voltage level?
Measuring continuity on a specific path does not guarantee that the path is not shorted to an adjacent path. How many matrix paths can you test at once to ensure isolation?
Instrument self-tests are usually not adequate to provide answers to any of the questions above. It is possible to use a DMM to find multiple failures (shorts on adjacent paths or to ground) but it is a very time consuming task given the exhaustive nature of the test. In addition, the amount of external circuitry/wiring required to switch the DMM to all the matrix inputs or the time required to execute such a test could prove inhibiting.
Teradyne's M9 Series Digital Test Instrument (DTI) provides a relatively easy way of detecting and isolating multiple types of failures in a relay switch matrix including:
- Stuck open
- Stuck closed
- Short to ground
- Short to in-between
This paper explains in detail how the M9 DTI can be used to isolate all of these types of failures to a specific relay in a switch matrix. A simplified version of an 8 by 8 switch relay matrix is used to illustrate the approach. Figure 1 illustrates the 8 by 8 switch relay matrix that will be used throughout the paper. The rows and columns are numbered from 0 to 7 and the switches connecting rows and columns are named k0 to k63.
Figure 1
8 by 8 Switch Relay Matrix
The Problem
It is possible for relay switches to get stuck in a given position. This is can be caused by sourcing too much current (over the current rating) through the relay contacts causing the contact to overheat welding the switch shut. The switch can also be welded shut by the arc generated by a high potential difference between the two relay contacts at the time the switch is closed (high voltage switching). It is also possible that the device stops working all together causing a stuck open failure. Careful design and programming can ensure that neither one of these conditions happen throughout the life of a given switch. However, even with careful design and programming, there still exists the risk that a Device Under Test (DUT) failure can cause a relay stuck failure or worse, multiple stuck failures. Figure 2 Illustrates a switch relay matrix with two stuck closed relays. Figure 2 also illustrates the conventions used to illustrate the four different types of failures discussed in the paper.
Figure 2
Two faulty relays in a switch matrix
The Solution
Teradyne’s M9 Channels can be independently programmed to source and sense signals in a given time period. This is known as patterns in M9 speak. A set of patterns makes up a test also known as a burst. The M9 can capture pass/fail information for individual patterns on any channel for a given burst. It is this capability that makes it possible to have a one to one wiring scheme between the M9 and the matrix inputs as shown in Figure 3 below.
Figure 3
Matrix to M9-Series DTI connections
Having connected the M9 to the matrix as shown in Figure 3, it is now possible to program the M9 to source a pulse or pulse train on a given channel while sensing on all other channels. By programming the switch matrix card to close individual paths the pulse can then be expected only on the single matrix path (channel) that is closed. The test (burst) is then repeated for all the columns in the matrix for the given path. Figure 4 below illustrates the approach on column 0, rows 0 and 1. This test collects pass/fail information specific to K0 and K8. The next test will be on column 1, rows 0 and 1, which collects pass/fail information on relays K1 and K9. The test is repeated until the last column is reached. Once the last column is tested, the M9 DTI will then move the source/sense channels one row down, in essence, “walking” the pulse down through all rows. Figure 4 illustrates the test setup for the first two relays (K0 and K8) in the matrix.
Figure 4
“Walking” a pulse down the matrix row by row, column by column
Collecting pass/fail information on specific relays as the pulse is walked through the matrix generates enough information to allow the software to isolate defective relays. For example, stuck closed failures are easily calculated given the failing M9 pin (row) and the column under test. These relays can be marked as stuck closed relays as soon as the failures are detected. Figure 5 below illustrates this scenario.
Figure 5
Faulty relays K24 and K48 are identified as stuck closed during column 7 tests
Stuck open failures require a little bit more analysis in order to identify the faulty relay. For example, if a relay is stuck open the M9 will report two failures, the first failure corresponds to the row where the faulty relay is located the second failure corresponds to the row that follows the true open failure. A simply approach could be to report the first open failure as a true failure and ignore subsequent adjacent open failures. However, the algorithm could be taken a step further to verify that the previous row did not fail open and that the following row failed open. This second step ensures that multiple adjacent stuck open failures are identified correctly. Figures 6 and 7 below illustrate how stuck open relays are identified.
Figure 6
Stuck open relays cause two failure reports (one true and one false).
The first failure corresponds to the actual stuck open relay.
Figure 7
Second (false) stuck open failure
Finally, it is possible for relays to fail in such a way that one of the contacts is shorted to ground or to some other voltage level. Programming the M9’s active loads and pin drive format correctly could identify these types of failures. If the relay the channel is testing is shorted to ground the M9 channel will not be able to drive that channel to the specified level causing the M9 to flag an error. Trapping for this specific type of error (MH) allows the software to identify shorted to ground relays inside the matrix. However, in this case, the software should verify that the row before the given relay also failed with a short to ground. This ensures that the correct relay is identified. Figure 8 below illustrates how shorts to ground and to in-between are identified.
Figure 8
Short to ground failure
Software Implementation
The software used to control the M9 and the Matrix switch card was written in C++ using an object-oriented approach. This simplified the overall program by encapsulating the details of programming the M9. In addition to simplifying programming the matrix card, object orientation made it easier to save failure information inside the matrix card object. This information was then later analyzed to determined true failures.
This approach also makes it possible to re-use the exact same test on many different types of matrix cards simply by re-writing the piece of the code that communicates with the specific matrix card under test.
A complete set of sample code for at least two different matrix cards (Ascor 4500 Series) can be downloaded for free here.
Conclusions
Utilizing the M9 to test switch matrix using the approach described in this paper has the following advantages:
- Interfacing the M9 to the switch matrix is a simple one to one connection.
- The M9 is able to detect multiple types of failures in a single sweep.
- The test can be re-used on many different matrix cards with minor modifications.
The main advantage of this approach is that the M9 collects all the data and the software determines the level of analysis to be performed.
