Monday
She had her coat on at the hook wall at 05:40 because the control room thermostat was still on its weekend setback, and because if she took the coat off she would sit down, and if she sat down she would start the morning report before she fixed line 11.
Line 11 said Call if any 11 over 2 s on a called point. She had written it in December. Two seconds had come from the EMS system timer, which was the number the vendor shipped, and on January 4 that timer had let 1.8 seconds of 11 go by without a sound. Then on Saturday Trey had called on a 00. Neither open nor closed, the breaker saying nothing for 1.3 seconds before it said open, and he'd called it because he didn't like the gap. Nothing on the note told him to.
She struck over 2 s with one line of pen, still readable, and wrote under it: Call if any 11, or 00 over 1 s, on a called point. Relay record says which part.
Trey's call sheet from Saturday was in the tray. 00 at 12:44:55.9 UTC, 01 at 12:44:57.2 UTC, 1.3 s. He had written -6 next to it and boxed it, because he was the first person on nights who did the conversion on paper instead of in his head. She initialled and dated his sheet and taped it beside the clause with two strips of blue tape. The line had his handwriting next to hers now.
A 00 for 1.3 seconds meant one of two things, and they cost different amounts. Either the breaker mechanism had been slow, in which case Hall 4's load had kept running on the interruptible MW for a second and a half after the QSE called it, and the bid credit for the 06:45 bin was wrong in the company's favor. Or the mechanism had opened on time and the 52b auxiliary switch, the little contact that tells the SCADA system the breaker's position, had lagged behind it. One was a market problem. The other was a linkage.
She didn't put it in a ticket comment, because ticket comments were where questions went to sit. She opened a work order for Abel, WO-PL-0117, priority 2, and wrote it so he wouldn't have to call her back:
Pull the protection relay event report for 12:44:55 to 12:44:58 UTC and give me current-zero time. If current zero is within ~0.1 s of the trip, it is the 52b aux switch lagging, adjust the linkage. If it is late, it is the mechanism, time the trip coil and check lubricant. Note relay-house temperature at the time of the call.
The front had come through at 03:00 Saturday. San Antonio had been 31 F.
The drafter had already written the morning report's first paragraph while she was at the wall. It wrote one every shift from the EMS event list, in the company voice, and it was usually right about everything EMS knew. Hall 4 PL BKR4 opened at 06:44:57 CST in response to the 06:45 QSE instruction; curtailment met. It had taken the 01 stamp, because the 01 stamp was the only time EMS had for the breaker opening, and it had folded the six hours in without saying so. It didn't know about the 1.3 seconds, and it didn't know the column was UTC. She didn't edit its paragraph. Editing it taught it nothing and hid what it had said. Under it, outside the paragraph, she added one line of her own: Bid credit held at low end of range pending relay record. Holding the credit low cost the company money in the morning report and saved it a settlement dispute in April.
Wednesday
Abel closed the work order at 16:12 Tuesday. He'd attached the relay event report as a PDF and written three lines of his own, the way he wrote everything, as if each word was billed.
Trip coil energized 12:44:55.87. Current zero on all three phases by 12:44:55.92, about three cycles. Relay house 37 F. The 52b linkage was stiff in the cold and didn't report open until 12:44:57.2. Adjusted. Test trip at 14:10 Tuesday, SOE showed 00 for 60 ms.
So the breaker had opened on time and the indicator hadn't. She read it twice. Then she opened the SOE pane, because Saucedo's relabel had gone into Tuesday night's build, and the column header that had said Time for three years now said UTC. That was the whole change. A word in a header. She checked one stamp the slow way, Saturday's 01, and the pane said 12:44:57.2, the number Trey had written by hand and boxed.
The build had fixed what the column was called, and that was all it fixed. On March 12 the local offset would go from six hours to five, and nothing in the pane would notice. She left line 8's -6 where it was.
She went back to Monday's report and wrote, beside the drafter's paragraph, 06:44:57 is the indicator, not the breaker. See WO-PL-0117. Then she released the hold. The 06:45 bin got credited from current zero, 12:44:55.92, which was 06:44:55.92 Central, not from the 01 stamp. The load had left the bin 1.3 seconds earlier than the SOE alone said. She reran Trey's range from that start and got 0.1 MW more at the low end. She wrote it on his sheet under his numbers. She didn't replace them.
Under line 11, in pencil: Jan 15: 52b, not mechanism. Credit from current zero. Relay record beats SOE.
Then she texted Trey, and noticed as she typed it that it was the first message she'd sent him that wasn't a question.
Relay says you called the right point for the wrong part. Call it the same way next time.
He sent back a thumbs-up and, eleven minutes later, from what she guessed was his kitchen: what part was it
The part that tells us, she wrote.
Thursday
The San Antonio day desk had a lunch lull from 11:30 to about noon, when the forecast desk went to the taqueria on Commerce and left the phones to whoever was covering. Thursday it was her, on a day swap, at Rubén's desk with Rubén's card taped to the monitor bezel.
Her sticky from December 8 was still under the card, its corner gone soft from people pressing it back down. EMS vs revenue meter >60 s on open/close/shed: keep line takes the meter. MC 12/8.
Six weeks ago that had seemed like the careful version. Sixty seconds was a disagreement you could see from the chair. But on the fifteenth, EMS and the meter hadn't disagreed by sixty seconds. They had disagreed by 1.3, and neither of them had been right about when the breaker opened. The relay had been. Her sticky would have sent the question to the wrong instrument, and done it with her initials.
She opened the DCR she hadn't opened in December, DCR-SA-0031. She raised it against Rubén's card, not her sticky, and copied Rubén as the owner, because the card was his and he'd earned the right to say no. She didn't rewrite his line. She added two under it.
Keep bin: revenue meter. Breaker open time on a called point: protection relay current zero, from the relay event report, not EMS status and not the meter.
If they differ by more than 1 s, say which one you used and attach the other.
She attached Trey's call sheet, Abel's work order with the relay PDF, and Wednesday's rerun. She wrote superseded by DCR-SA-0031, MC 1/20 on the sticky, peeled it off the bezel, and stapled it to the DCR printout, so the reason it changed would stay with the change.
Then line 8, which had been bothering her since the build. In pen on the hook-wall copy, which she'd brought in her bag: SOE is UTC. Local = UTC -6 until Sun Mar 12 2028 02:00 CST, then -5 (CDT). Write the offset you used next to the time.
She added one line to Saucedo's conversion change request, which had a due date of March 1: Hook-wall line 8 now carries both offsets by hand. If the pane converts on its own after Mar 12, tell me which build, so I can strike my line the same day.
Rubén answered at 12:10, from the taqueria, judging by the photo he sent of a paper plate. Two words.
Approve. Finally.
Friday
She'd emailed Abel on December 23 with one question and he answered it four weeks later, at 22:40 Thursday, from the truck, which was when Abel answered things.
PL relays and the RTU both take IRIG-B from the satellite clock in the control house. One clock, good to about a microsecond while it has sky. The revenue meter has its own GPS puck on the meter cabinet, separate antenna, and its event log only keeps whole seconds. Checked them against each other Tuesday after the test trip: relay current zero 14:10:03.412, meter event 14:10:03. Agree to the second, which is all the meter can say.
Line 7 had needed exactly this since December. At 06:20, under it, in pen:
7a. PL relay + RTU = one clock (control-house GPS, IRIG-B), ms. Revenue meter = its own GPS, whole seconds. EMS status = scan, 2–4 s late. Same UTC, different grain. Pick the finest clock that saw the event, and write which one.
She read the email again and stopped on while it has sky.
A satellite clock that lost its antenna didn't stop. It kept running on its own oscillator, and it kept stamping relay records the whole time, and nothing on any screen she owned would tell her it had happened. She wrote back before she hung the note: If the clock loses GPS, does anything tell us?
He answered at 06:41, so he hadn't slept. Contact output on the clock, wired to the RTU as PL CLK ALM in 2019. Never mapped to EMS. Nobody asked.
She opened DSP-1244. Map PL CLK ALM to the PL display and the DR-CALLED group; a called-point time is not usable while it is in alarm. She attached both emails and copied Saucedo, because the conversion CR touched the same tag table and she didn't want two builds fighting over one row.
Last line on the note, 7b: If PL CLK ALM is in, the ms are guesses. Use the meter and say so.
Trey came in for turnover with his coat still zipped and read 7b over her shoulder.
"Nobody asked for five years," he said.
"Somebody asked yesterday. Me, at six-twenty."
Saturday
DSP-1244 was queued behind the conversion CR, which meant that until PL CLK ALM showed up on a display, the only way to know whether the clock had sky was to go look at it.
At 05:50 she walked across the gravel to the control house with Abel's phone photo of the rack. The air smelled like wet caliche and the cooling towers. Inside it was warmer than the control room. The clock was where he'd said, second unit down: a green LOCK lamp, 11 SATS, the front panel reading UTC to the tenth of a second and counting.
The RTU had never asked it anything, so she scrolled its event log with the arrow keys. Three entries since 2025. Two were power cycles during a firmware update. The third was the one.
2027-12-02 21:14 UTC to 2027-12-03 00:38 UTC. HOLDOVER.
She remembered that night. Lightning had taken the antenna surge arrester. Three hours and twenty-four minutes with no sky, and the clock had stamped every relay record the whole time.
She called Abel. He was in the truck; she could hear the turn signal.
"Oven oscillator," he said. "Datasheet says under a microsecond an hour in holdover. Ten a day if it's cold. That night it re-locked and corrected two point one microseconds. I wrote it in the log."
She did the arithmetic standing at the rack, on the back of his photo. A microsecond an hour. Ten a day. A month of holdover in the cold would still be under a third of a millisecond.
Line 7b, which she'd written the day before, said that if the alarm was in, the milliseconds were guesses. It was wrong by three orders of magnitude for anything shorter than a month. The December 2 relay records were still good to the millisecond. So was anything a holdover night would ever stamp, unless nobody noticed for weeks.
She had written 7b because while it has sky sounded like a warning. Abel hadn't meant it as one. He'd meant it as the condition on a number. She had heard a hazard where he'd stated a spec.
Back at the hook wall she struck 7b with one line, readable, the way she'd struck line 11 on Monday and her own sticky on Thursday. Under it:
7b. PL CLK ALM in = clock on its own oscillator. Ms still good for days. Check the front panel at turnover; if HOLDOVER is over 7 days, use the meter and say so.
She added a row to the turnover sheet, CLK: LOCK/HOLD + SATS, and filled in the first one herself. LOCK 11.
Trey read it with his coffee. "You crossed out your own line."
"It was scared of the wrong thing." She capped the pen. "The clock's fine. It's the EMS that's two to four seconds late."
He looked at the hook wall, at the three strikes in one week, two in her pen and one stapled to a DCR in San Antonio. "You going to do that every time?"
"Every time I'm wrong about which clock saw it."
He wrote LOCK 11 again on the line under hers, because he'd checked it himself on the way in, and the two entries sat there in two hands, nine minutes apart.
She went back to DSP-1244 before she left. The ticket still said the alarm should make a called-point time unusable. That was Friday's fear, and it had made it into a ticket, where somebody would build it. She added a comment so it wouldn't get built that way:
Revise: alarm should page at 24 h of holdover, not at first loss. First loss is weather. Called-point times stay usable in holdover; turnover sheet logs LOCK/HOLD. Front panel event log is the record.
Then she took her coat off and sat down to write the morning report.