Signal Anomaly Investigation on a Data Acquisition Board

A data acquisition board can pass placement and visual inspection and then show a drifting zero, excessive noise, or readings that differ between channels. Replacing the amplifier, the converter or the reference is the usual first reaction and usually the wrong one, because the analog signal path is sensitive to component parameters, to the state of the joints, to contamination and to the way the board is connected to the test equipment.

Three Questions Before Removing Anything

The first question is whether the fault appears on one board or across a batch. The second is whether every channel is affected or only a particular one. The third is whether the problem is present as soon as the board is powered or whether it develops after the board has been running for a while. Those three answers divide the possible causes into groups that can be investigated separately.

If a single channel is out, the input protection, the precision passives, the amplifier and the connection to the converter on that channel are examined. If every channel has shifted, the attention moves to the reference voltage, the supply, the grounding and the conditions of the test itself, because a change that affects all channels at once is almost never a fault in one channel”s components.

When the report is written, it is worth including the anomalous waveform, the input signal, the supply conditions, the fixture, the ambient state and the number of boards involved. That is far more useful than a statement that the measurement is inaccurate.

Separating a Parameter Problem From a Soldering Problem

The precision resistors, capacitors, amplifiers, references and converters on a data acquisition board are checked against the bill of materials for the full part number, the tolerance, the temperature coefficient and the package. Two parts with the same value but a different grade will not look different on the board and will produce a different measurement, particularly when the temperature changes.

Where a project is imported, the customer is asked for the Gerber data, the bill of materials, the coordinates and the assembly drawing, together with a note of any analog device that must not be substituted. When a temporary substitution is made, the reference designators affected and the material batch used are recorded, so that a batch does not end up containing two different parameter sets that cannot be told apart afterwards.

Before any repair is attempted on an anomalous board, the fitted part number is verified. Repairing first and identifying the part afterwards is a way to lose the information that would have explained the fault.

Using the Print Record and the Joints

The analog front end packages are small, and a short deposit of paste can produce a joint that is intermittent while an excess can produce a bridge. SPI measurement of the paste area, volume, height and offset after printing gives the investigation a record of what the process was doing before the fault appeared, which is why the data are kept rather than discarded once the board has passed.

If the same channel and the same pad keep failing, the stencil opening, the support under the board and the placement position are examined. After reflow, optical inspection and examination under the microscope confirm the offset, the bridging, the short deposits and the wetting of the joint. A device with joints beneath its body cannot be judged from the surface alone, so whether a further examination is warranted follows the failed location and the customer requirement.

Contamination on a High Impedance Input

Some acquisition boards accept a high impedance input or a very small analog signal, and in that case residue on the board, moisture, contamination in a connector or an unstable probe contact can change the measured value. The first step is therefore to exchange the cable and the fixture for ones that are known to be good, and only then to look at the board and the interface.

If the result returns to normal with a different set of test connections, the amplifier and the converter were never the cause. Whether the board needs cleaning, by which method, and what the acceptance condition is afterwards, all follow from the components fitted and the customer requirement, because not every device can be handled the same way.

Narrowing the Fault Along the Signal Path

A functional test can inject the specified signal at the input and observe it in turn at the protection circuit, the amplifier, the filter, the converter input and the digital output. When one stage is correct and the next is not, the range of possible causes collapses to the components and joints between them, which is the whole purpose of the exercise.

After a repair, the original test conditions are repeated and several channels are compared rather than only the one that was touched. If the fault is concentrated in one production batch, the material, the print data, the reflow record and the repair history are reviewed together, because a shared cause is the only explanation that fits a shared symptom.

Our SMT assembly line builds the analog front end, PCBA testing runs the functional test along the signal path, and the records are held under quality management.

Working From a Narrowed Boundary

`r`n

Once the fault has been narrowed, the remaining work is much smaller. If the deviation is confined to one channel, the components between the input protection and the converter on that channel are examined one at a time, and each is compared with its counterpart on a working channel rather than judged in isolation. If the deviation moves with the board but only when the board is warm, the thermal behaviour of the amplifier and the reference are considered, and the measurement is repeated at a controlled temperature so that the observation is reproducible.

`r`n

If the deviation follows the fixture rather than the board, the board has been cleared and the fixture has not, which is a result worth recording because it prevents the same investigation being run again on a similar board. A test that is only carried out once, without a comparison, cannot distinguish between a board fault and a condition of the test.

`r`n

Batch level anomalies are handled differently. When several boards from one batch behave the same way, the material lot, the print record, the placement position and the reflow profile are reviewed together with the repair history, and the point at which the deviation started is identified. That is the difference between a corrective action and a repeated repair.

`r`n

Finally, the repaired board is returned to the same test conditions and the channels around the repair are compared with the ones that were not touched, so that the repair is shown to have solved the original problem without introducing another.

FAQ

Is a drifting reading a sign of a failed converter? Not necessarily. The reference, the supply, the grounding and the test connections are all capable of producing the same symptom.

Why does the factory keep the paste data? Because a fault found later can then be compared with the print condition at the time, which is the difference between a diagnosis and a guess.

Can a board with a high impedance input be cleaned? It can be, but the method and the acceptance condition depend on the devices fitted, so the material and the customer requirement are checked first.