Railroad Forums
Moderators: mtuandrew, therock, Robert Paniagua
Testing found that a spurious high-frequency modulated signal was being created by parasitic oscillation from the power output transistors in the track circuit module transmitter. This spurious signal propagated through the power transistor heat sink, through the metal rack structure, and through a shared power source into the associated module receiver, thus establishing an unintended signal path. The spurious signal mimicked a valid track circuit signal. The peak amplitude of the spurious signal appeared at the correct time interval and was large enough to be sensed by the module receiver as a valid track circuit signal, which energized the track relay. This combination—of an alternate signal path between track circuit modules and a spurious signal capable of exploiting that path—bypassed the rails, and the ability of the track circuit to detect the train was lost.I think Alstom's insurer just had a very bad day....
farecard wrote:The NTSB has issued an immediate warning letter http://www.ntsb.gov/recs/letters/2009/R09_15_16.pdf to WMATA and the other parties. They are concerned about the track modules going into oscillation and spoofing the receiver.Yikes, that is a pretty pathological problem. A circuit designer's worst nightmare. Not the non-linear issue we speculated about earlier, but it was one that only appeared at high signal levels. I am an electrical engineer and just about everyone of us has encountered a circuit with parasitic oscillations (often intermittent) at one point or another. Incredibly hard to track down and harder to predict. Actually there were 2 pathological/difficult to predict problems. First, the change in impedance bond required a larger output signal from the amp, and at the peaks of the signal the amp would go into oscillation. Second, this oscillation was somehow getting transmitted through a parasitic (unintended) path into the receiver. If the path was through the power supply, I'm not sure why you need to invoke the heat sink and the electronics rack -- the current spikes from the oscillation are often enough to cause effects in the power supply.
..
Yikes, that is a pretty pathological problem. A circuit designer's worst nightmare. Not the non-linear issue we speculated about earlier, but it was one that only appeared at high signal levels. I am an electrical engineer and just about everyone of us has encountered a circuit with parasitic oscillations (often intermittent) at one point or another. Incredibly hard to track down and harder to predict. Actually there were 2 pathological/difficult to predict problems. First, the change in impedance bond required a larger output signal from the amp, and at the peaks of the signal the amp would go into oscillation. Second, this oscillation was somehow getting transmitted through a parasitic (unintended) path into the receiver. If the path was through the power supply, I'm not sure why you need to invoke the heat sink and the electronics rack -- the current spikes from the oscillation are often enough to cause effects in the power supply.Does this validate my previous intuition that concentrating large amounts of signaling equipment in a centralized location can lead to more problems than if the equipment is parceled out in individual cabinets along the right of way?
Jersey_Mike wrote:Not really, to my mind. The only surefire solution would receivers in one building, transmitters in another. But there are many other solutions to the coupling problem; ones that do NOT require many more physical locations along the track.
Does this validate my previous intuition that concentrating large amounts of signaling equipment in a centralized location can lead to more problems than if the equipment is parceled out in individual cabinets along the right of way?
farecard wrote:From what I can infer from the NTSB memo, the system was oscillating at high (RF?) frequency, not audio. it would just do it at the peaks (postive and/or negative) of the audio-frequency track signal. One could say the RF oscillation was modulated (in a non-linear fashion) by the magnitude of the track signal. The oscillation was strong enough and of the right frequency to get transmitted by a parastic path to the receiver, unlike the regular audio frequency signal. So if you low-pass filter it or take the envelope (more or less which is what the receiver must have been doing), you would get something like the original track signal at the receiver, which was the fatal flaw. JS
I agree with pumpers that is a EE's nightmare; one far more likely at RF frequencies than audio. I recall struggling with various crystal oscillators that wanted to NOT oscillate; or oscillated wherever they wanted, but not where the crystal fundamental frequency was.
)Silverliner II wrote:I was in Washington on the 14th of January and did some riding around on a number of Metro lines in the course of going to a number of destinations scattered across the system. I noticed that manual control is still in effect, including manual opening and closing of the doors.Because WMATA feels that their operators aren't smart enough to know how long their trains are and to stop at the appropriate markers.
Is that the reason why all trains, regardless of length, are stopping at the extreme ends of the platforms, instead of more towards the middle as they would in automatic mode?
