2001
Blood pressure monitor amplifier susceptibility
(IEEE EMC Conference)
A new non-invasive blood pressure monitor was an electronic version of the old 'cuff' method. While testing it for RF immunity it would fail to measure at all between 950 and 1000MHz. It was found that its pressure sensor was outputting misleading signals during the test, despite being a standard part that had been used for many years without problems (or so claimed its salesperson). The pressure sensor had a 6-pin package, with 2 unused pins marked "do not connect". Copper tape over the transducer and its pins made the problem go away. The problem was then isolated to just three of its pins, one of which was a compensation capacitor for the sensor's internal amplifier. An engineer working for the Japanese company that made the sensor said that he had seen the problem before in an automotive application.
The N/C pins were connected to the inputs of the internal amplifier and used for performance checks during production testing. The pins were acting as antennas, picking-up the external RF field and injecting it into the internal amplifier at its most sensitive point, where it would be inevitably rectified (demodulated) by the semiconductor junctions in the amplifier's IC and cause major shifts in DC operating points. Even with these pins cut off from the package the problem still remained - the amplifier was so sensitive that the internal leads and the bond wires to the IC still made an effective antenna at 950MHz.
Eventually the sensor was modified by the manufacturer so that it did not have this problem. In the meantime one year's worth of production of the new product suffered the additional cost of $20 per unit for a shielding can and its fitting.
2002
Incorrect data cable screen connection
I was asked to do EMC tests on a multi-channel digital location recorder designed and built by the R&D department of a well-known record manufacturer. The recorder was housed in a 19" rack unit and controlled by software running on a laptop computer, via RS 422. The audio results were said to be excellent, but they invariably had problems with the control functions. On the last recording session, the machine went into record mode as requested, but during the session, control of the recorder was lost. No command would allow the engineers to stop the machine or come out of record mode. The whole system had to be re-booted before they got control back. This was a classical orchestral session with 80 musicians, so the problem could have been expensive.
I placed the recorder unit in the EMC test chamber, connecting the system up normally, but with the laptop computer outside in the control area. This was to isolate the two different parts of the system. The recorder unit passed the basic emission tests when running in record or playback mode on its own. But when the RS 422 line was connected between the laptop computer and the recorder rack, the system failed the radiated emission test by a wide margin.
If the RS 422 cable radiated interference, it was very likely that the same cable would receive interference. I set up for the conducted immunity test. ... The recorder unit was put into record mode and I started the test. At first all went well. But, as the modulated frequency approached 8MHz, the time code display on the laptop screen stopped. All other controls seemed still to be working. However, at about 16MHz, a second event was detected, and the laptop had lost control of the recorder unit. The recorder was permanently in record mode! I put the EMC test system in pause, and rebooted the recorder and laptop. Restarting the test at 18MHz, everything was working properly until the modulated RF approached 33MHz (the bus/processor frequency of the recorder electronics). Multiple events were detected and control of the recorder system was lost once more and the laptop crashed. To cut a long series of tests short (similar problems were encountered on the other side of the recorder's processor/bus frequency) - the problem was obviously interference on the RS 422 data communication circuit. But how could this be the case? RS 422 is a balanced transmission system and the cable was shielded.
An inspection of the RS 422 connectors at each end of the circuit revealed the following:
The connector at the laptop end had the cable shield correctly bonded to the chassis.
The connector at the recorder unit end was an insulated component. The cable shield connection was wired directly to the logic 0V track on the printed circuit board.
The custom made RS 422 cable had the cable shield connected to the recorder unit end only. The cable was constructed in this way, because the engineers had found that hum was introduced into the recorder when the RS 422 cable was connected to the standard desktop PC used during the design phase of the project.
Thus, any interference current induced into the cable shield of the RS 422 data communication circuit was injected directly into the recorder unit's ground conductor, allowing interference currents to flow in the RS 422 I/O electronics, resulting in poor or bad data on the RS 422 communications circuit. The laptop (or any other) computer, and the recorder was, at the very least compromised, by any interference induced on the RS 422 cable shield.
2002
Differential amplifier susceptibility
(PC Design)
In another life I ran a company that made weighing systems for industrial trucks - i.e., scales to make sure trucks were within legal limits. They were portable, could be towed behind a police car on a small trailer, and used portable electronics that plugged into the car's cigarette lighter. They ran off the same electrical system the police radio did, and the indicators were often placed right next to the radio or on the car roof, right next to the antenna. Immunity to RFI was a significant design requirement.
We had just finished a complete redesign of our indicator family. We had access to a screen room facility and a technician through another company, so we went there to do the EMI/RFI testing. During the very first test, however, the indicator went totally off-scale! After two hours of tweaking we got some improvements and then hit a plateau. No matter what we did we could not quiet down the indicator.
The screen room technician finally spoke up and asked us what the input circuit looked like. We told him it was a high-gain differential amplifier, which then fed an A/D converter. He asked us what part number the amplifier was. We told him. It was a commonly available amplifier made by at least four or five manufacturers. He then asked us who manufactured the part. We told him. He then told us that particular manufacturer often had RFI problems with its parts and why didn't we buy the same part from a different vendor. We did, and the RFI problems almost totally went away. It took us only a few more hours to achieve the RFI objective and the product then successfully went into production.
There was no clue in any of the published specifications from any of the manufacturers of this part number that there would be differences in RFI sensitivity between product offerings. We had no reason whatsoever to suspect that part. We might have struggled with that design for months if that technician had not put us on the right path. There are two morals to this story:
A good technician with experience can be more valuable than someone else with all the university degrees in the world.
There can be subtle differences inside IC packages in otherwise identical parts that may only be determined by laboratory testing or trial.
More than one engineer has been 'burned' by a part that behaved unexpectedly. Sometimes, as in this case, there are simply differences in design or manufacture of otherwise "identical" parts. Sometimes a supplier changes a manufacturing process without telling anyone. Often this involves the implementation of an improved process, which coincidentally may offer faster rise times. Perhaps the manufacturer thinks that the change or improvement will have no particular consequence for anyone, and treats it as simply an in-line adjustment. But sometimes the faster rise-time results in timing or EMI problems that didn't exist in the user's design before. These can be particularly difficult to trouble-shoot, because people rarely equate the problems with a device, particularly a device that used to work just fine.
|