^ It is being reported that the "engineer applied the emergency brake" moments before it derailed. It sounds like a manual activation, when you're describing something that sounds automated in regard to the brake being activated.
The way it works is:
1. The signal system 'fires' electromagnetic pulses at a certain Hz frequency through one of the running rails (occasionally switching sides to the other rail every so often at an insulating joint). Even idle like this it can tell which switches are thrown and whether there's a break in the rail just by where that pulse travels between insulators.
2. The train's axle passing over this pulse completes the 'circuit' by passing it through the other running rail.
3. Signal system detects the completed circuit on the return end.
4. By detecting the completed circuit the signal system can ascertain the position of the train to within whatever area the circuit is in (i.e. between insulating joints). And in some cases crudely estimate the speed based on how many insulated circuit sections it passes over in a given time.
5. The signal system can 'fire' a change in signal based on position of the train, such as signifying that the next block is occupied or telling the train to slow down. Basically, every single function the signal system performs is based on where those circuits do and do not complete themselves.
^^Even non- cab signal lines with just the wayside lights have track circuits (most of them...the T still has a couple truly ancient installations like Rockport and Reading that don't have circuits at all and just switch the lights through pole-mounted telephone wires by side of the track). On a wayside-only system the track circuits only control the signal lights, not the train.
The cab signals are an
extra layer on top of the track circuits. The existing track circuits fires the pulses, and the train has an antenna mounted on its underside by the wheels that detects those pulses (all while its wheels are making the 'circuit'). The engineer's controls have a readout that interprets those pulse patterns as "codes". The codes convey the same exact information that the patterns of wayside signal lights do: basically "clear" (full speed ahead), "restricting", and a couple of intermediate categories. The engineer is still controlling the train, and still has to know the speed limits on a given stretch of track. But when the code gets changed to something more restrictive, a buzzer sounds and they have so many seconds to get in compliance with the speed limit before a brake penalty is applied.
If the X seconds of warning expire with no engineer action applying the brake to slow down to the new speedlimit...the signal system steps up the penalty by cutting out. "Cutting out" means what it implies: the signal shuts off entirely, and no longer makes a circuit. Train loses the signal...it goes into
emergency. Loud alarm goes off at the console, and if you do not hit those brakes in a split second the train forcefully does it by itself and a bunch of passengers get a rough jolt and start swearing at your conductor co-workers...who then chew you out. Old Colony, Stoughton, Fairmount, Worcester commuters: you've probably been on trains that have done the rough WTF? slowdowns when the engineer gets slapped by the cab signals.
That's the fail-safe. The train treats it as if there's a problem with the signals and that they've gone totally offline. So you stop, and you can't proceed at >20 MPH until the train picks up a new signal telling it that it can do something else. The simplicity and robustness of the system is that the "Engineer bad!" punishment is the same as the "Hold up! I think we might have a track problem ahead; I can't see the signals!" alert like a broken rail.
The type of PTC system ("A.C.S.E.S.") that Amtrak is installing here--and which has had running New Haven-Boston since '01--is a
third layer on top of the track circuits and cab signals. Both of the pre-existing layers still perform all their vital functions, but the RF transponders allow for full 2-way computer communication with the trains to fine-tune things much more precisely and close the last loopholes. The old tech is near-perfect at what it does best, so this flavor of PTC is strictly an
augmentation that adds new safety features. If the PTC computers go on the fritz, then you still have the full force of the cab signals backing everything up.
-----------------------------
When they use the term "train went into emergency" in these news reports...that means the engineer blew the signal, did not respond in time, got slapped with an "Engineer bad!" drop-out, did not respond NOW, then the train went into automatic stop. But 106 MPH was way way too fast for the braking maneuver to slow the train enough before it jumped the tracks. 3 seconds, -4 MPH, and then Amtrak's data feed to the locomotive's black box cut out...presumably that was the start of the derailment. The shudder that the was felt throughout the cars before they went flying was probably that emergency braking hitting the start of the curve and tearing out the rail.
You can probably do some complicated maths based on the position of the signal that sent the last penalty and the stopping distance to figure out what kind of upper-bound speed that signal's placement was designed to stop in a worst-case scenario. It certainly was designed for a hell of a lot less an upper bound than 100 MPH...Physics 101 will tell ya that much.
So the question is...how in the hell did this train accelerate to 106 MPH so quickly to elude the
previous signal from applying the harshest braking penalty? Because the previous interlocking by North Philadelphia station imposes a 60 MPH limit for a short distance that would've involved a hard penalty. And we know this train wasn't running flagrantly afoul back there. Then in the short distance of 70 MPH territory after N. Philly station but before Frankford Jct. there's 1-2 more intermediate signals that would've started tightening the noose before this last one cut out. You're talking, like, 20-30 MPH acceleration in a third- to half- mile to elude the signals if they were functioning. All while pulling 8 coaches roughly 40% full. If a Sprinter locomotive has ever achieved that kind of acceleration before, it was probably on a Siemens test track. Somehow, some way, this engineer managed to pull that feat off on an exceptionally curvy stretch of the NEC.
Yeah. I can see why that guy's lawyering up to his eyeballs right about now. I mean...wow. We're so cosmically beyond
[*puts on Desi Arnaz voice*] "You've got some 'splaining to do" here.