Richard Vale. He was on his phone talking quietly, but his eyes tracked Malcolm across the room. Malcolm did not acknowledge him. Dr. Reed Daniel turned from his workstation.
The three stable runs I found two consistent differences. Fuel temperature at turbo pump inlet was 1.7° warmer in all three cases. And the restart timing interval, the gap between engine cutff and restart initiation was between 0.3 and 0.5 seconds longer than in the unstable runs. Malcolm stopped walking. He turned around.
Show me the timing intervals for all 14 runs. Exact values not rounded. Daniel pulled them up. Malcolm leaned over the workstation and looked at the numbers. 11 unstable runs.
Three stable. The timing intervals for the stable runs 8.34 seconds, 8.41 seconds, 8.29 seconds. The unstable runs ranged from 7.668 668 to 8.1 seconds. Malcolm straightened up. The three stable runs weren't stable because of what your team did right.
He said they were stable because of what the operators did differently without knowing they were doing it. The restart timing fell outside a critical window by accident. He looked at Daniel. You're dealing with a frequency locked resonance. The engine has a natural oscillation frequency that gets excited by restart events at a specific timing window. land inside that window consistently and the oscillation builds.
The software corrections Veil recommended didn't fix it. They tightened your timing precision. You started hitting the critical window more reliably, which is why the last two simulations were the worst ones. The room was quiet. Daniel's mouth was slightly open.
From the corner, Richard Vale ended his phone call. He stood up. He was a large man built like someone who had played football in college and kept enough of the physicality to make it part of how he occupied space. He walked across the room with the unhurried confidence of someone who had made this particular calculation and decided it was worth making. Dr.
Sloan, he said, addressing Rebecca and not Malcolm. I wasn't informed we were bringing in an outside consultant for this session. You weren't required to be, Rebecca said. Vale looked at Malcolm then for the first time directly and the look carried something in it that Malcolm had seen on his face before eight years ago in a different facility at a different moment that also hadn't gone the way Vale had planned. Read Vale said Vale you've been in the room for Vale checked his watch with deliberate precision 11 minutes and you're already telling my team that the work we've done for the past 6 weeks made the problem worse.
I'm telling the what the data shows. Malcolm said the timing correlation is in the numbers. Anyone can see it. My team reviewed those timing intervals. Your team averaged them.
Averaging removed the pattern. Malcolm looked at Daniel. Pull up the chart with the raw timing values. Daniel did it without waiting for anyone else's approval. The chart appeared on the central monitor.
Malcolm pointed to it. Below 8.2 2 seconds, the oscillation builds. Above 8.25, it has a chance to damp out. 300ths of a second is your stability boundary. He paused.
The firmware update veil recommended improved timing precision by plus or minus0.05 seconds. It locked the system into the unstable window. The silence in the room had a different quality now. Not the silence of people absorbing information. The silence of people who were doing math in their heads and arriving at an answer they hadn't expected and didn't entirely want.
Vale crossed his arms. That is a remarkable conclusion to draw from 11 minutes of looking at someone else's data. Then show me where the reasoning is wrong. Malcolm said, "Not aggressive, factual, the flat, patient tone of someone who knew the difference between an argument and an error check and was offering the latter." Vale held his gaze for 4 seconds.
Then he looked at Rebecca. "I'd like to discuss the protocol for outside consultant access. This is a critical path program and we have established review procedures." Richard Rebecca's voice was quiet and completely final. He's here at my authorization.
If you have concerns about protocol, send them to me in writing. She turned away from him. Daniel, let's get Dr. Reed set up with full telemetry access for the past 14 simulations. Clean data, no filtering, raw sensor output.
Daniel was already moving. Vale stood in the middle of the room for two more seconds. Then he walked back to his corner, sat down, and picked up his phone again. Malcolm watched him go and then turned back to the monitors and the watch. Veil was doing from the corner was constant.
The way weather is constant. Something you're aware of without looking at it directly. Malcolm spent 4 hours in that room. He didn't eat, didn't take a break, spoke only to ask specific questions to Daniel, to a structural analyst named Priya, who had the turbo pump vibration data on her workstation to a fuel systems engineer named Torres, who had the cryogenic thermal logs. He moved from workstation to workstation methodically. the way he moved through a complex repair job, not attacking the symptoms, but mapping the system, building a picture of every component and every connection until the shape of the actual problem emerged from underneath the noise.
At 3:47 in the afternoon, he stopped. He went back to the central workstation where Daniel was still sitting, still pulling data whenever Malcolm asked for it, showing the particular patience of someone who is learning something important from watching how it's done. Malcolm picked up a marker. I went to the whiteboard at the far end of the room, the one covered in six weeks of equations and diagrams from the previous analysis efforts, and without hesitation drew a single horizontal line through the middle of everything. Several people in the room straightened.
He wrote three equations below the line, working quickly, the marker moving with the decisiveness of someone who has been running these numbers in their head for hours and is only now writing them down because the paper is a formality. He stepped back. He said the oscillation isn't in one system. It's a coupling resonance between all three simultaneously. Chamber pressure wave, turbo pump, rotational frequency, and cryogenic feed line thermal expansion.
The frequencies are incommensurate individually, but at restart under the current timing sequence, they achieve a brief coherence. They sync up. That synchronized moment is your instability window. He drew a quick diagram. It lasts about 0.8 seconds.
If the resonance builds enough energy in that window, it sustains itself beyond the point where the system can naturally damp it. That's your runaway oscillation. He capped the marker. Your software corrections didn't work because you were trying to suppress the oscillation after it appeared. But by that point, it already has momentum.