The $2,300 San Antonio Solar Reinstall That Changed How I Spec Every Pylontech System
June 2019: The Job I Was Sure I Could Do in My Sleep
San Antonio, June 2019. 104°F in the shade. We had three days to pull a 6kW rooftop array, add four panels, and swap in a Pylontech LFP battery bank for a customer who'd been running lead-acid for years. His daughter's wedding was that Saturday. The panels had to be back up by Friday night.
I'd been running solar installs for four years at that point. I'd done roughly this same reconfigure twice in 2018—same roof pitch, same inverter model, similar bay. When the customer asked, "Can you get this done by Friday?" I said yes before he finished the sentence.
That's where I should have slowed down.
He also asked a question I remember verbatim: "Can I use one of these batteries in my car?" I laughed. Told him no—automotive LFP packs go through a completely different set of UN 38.3 transport and abuse tests, and the BMS in a stationary Pylontech module isn't designed for the vibration profile a vehicle sees. He nodded, seemed satisfied. I moved on.
The Part Where Overconfidence Started Billing Me
Here's the thing about solar panel system diagrams: when you've drawn enough of them, they stop being documents and start becoming habits. I pulled up a diagram from a similar job in 2018, changed the panel count, updated the customer name, and called it good.
I didn't recalculate the combiner box rating.
Now, I need to be specific about the mistake because this is the part that costs money. The Pylontech US2000 modules I was installing run at 48V nominal with a CAN/RS485 BMS communication line to the inverter. The Pylontech combiner box in a setup like this has to handle the DC string side and pass that monitoring signal cleanly. My old diagram had the smaller monitor breaker from the 2018 job, because I'd upsized the array but defaulted the combiner spec to "basically the same thing."
It wasn't basically the same thing.
For the record: I've logged every significant mistake I've made since 2016 in a running file. The total at the time of this job was around $9,400 in wasted parts, rework, and credits across four years. This one added about $2,300 to that number. The file is not a source of pride. It's a source of checklists.
The Day It Fell Apart
Tuesday, 11 a.m. Panels off, rails prepped. I sent my apprentice to grab the combiner box from the rack we keep on site. He came back empty-handed. I'd ordered the wrong model number—the one on my diagram was the older revision, discontinued, and we didn't have the replacement in stock.
Time pressure decision. Normally I'd call the factory rep, confirm the current spec, and wait out the lead time. I didn't have that luxury. So I went to a local electrical supply house in San Antonio and bought what they had: a generic combiner that "should work."
Should work is not the same as works.
By 6 p.m. we had the array back on the roof, the four new panels installed, and the Pylontech battery bank wired in. First power-on looked clean. Then the inverter started throwing fault codes every eight minutes. Not a hard failure—a reboot loop. The kind that looks fine while you're standing there, until you walk away and the homeowner calls you at 10 p.m.
What the Fault Codes Actually Meant
I spent Wednesday morning on the phone with the Pylontech technical line and with the inverter manufacturer's support team. Two separate engineers, two separate conversations, same conclusion: the generic combiner box I'd substituted didn't have the right DC current handling spec for the monitoring loop between the battery BMS and the inverter. The Pylontech modules were fine. They were fine the entire time. The combiner was the weak link.
Per Pylontech's published specifications (pylontech.com, accessed January 2025), the US2000 series is engineered for 48V low-voltage integration and specifies CAN/RS485 communication for inverter pairing. That's the detail that matters here. A combiner box in this configuraiton sits between the string side and the inverter, and if it can't pass the monitoring signal cleanly, you get exactly the reboot loop I was hearing. That's not a Pylontech problem. That's me not reading the compatibility documentation before I signed the quote.
One more source worth knowing about, because I should have had it on the wall of the truck: NEC Article 690 (Solar Photovoltaic Systems) sets the code baseline for DC conductor sizing and disconnecting means in these installs. The local AHJ in San Antonio was running inspections against the 2017 NEC at the time, and the permit packet I'd filed reflected that—so my paperwork was clean. My parts list was not. Passing the inspection doesn't mean your system actually works. That's a separate test, and I failed it.
The Fix, and What It Cost
We pulled the generic combiner Thursday morning and drove to a distributor that actually stocked the correct Pylontech-compatible unit. We lost four hours. The reinstall pushed to Friday afternoon. The wedding happened. The panels were up. Barely.
The damage:
- $740 for the correct combiner I should have ordered in the first place
- $380 for the wrong one I bought from the supply house and couldn't return
- $960 in lost labor hours from the two-day delay
- $250 store credit I gave the customer for the disruption
Roughly $2,300 when you factor in the extra crew day. On a job I'd quoted at $1,400 in labor margin.
That number is the only reason I remember this job so clearly. I've made bigger mistakes in dollar terms. But this one compounded—every hour of delay made the next hour more expensive, because the crew was booked on another job Monday and I had to pay overtime to squeeze this one in. I've come to believe that compounding delay is the true cost of a parts mistake. The parts are $1,000. The delay is everything else.
What I Actually Learned (and What I Keep on My Checklist Now)
Three changes came out of this, and they've stuck for six years:
1. No diagram gets reused without a fresh load calc. I don't care if the job looks identical to one I did last month. The combiner rating, the breaker sizes, the wire gauge—all of it gets recalculated. This sounds obvious. It wasn't, because I trusted my memory more than I trusted my calculator.
2. I check the official spec sheet before I sign the quote—not before I order. The Pylontech US series, the Force series, the high-voltage stackables—each has its own compatibility matrix with the common inverter brands. Pylontech publishes this. My mistake was treating my 2018 diagram as a source of truth instead of treating Pylontech's documentation as the only source of truth.
3. I write down the customer's side questions. That guy asking about putting an LFP module in his car wasn't a distraction. It was a signal he trusted me and wanted more information than he knew how to ask for. Now I keep a running list of "things customers asked that I didn't have a clean answer for," and I research them the following week. That list is currently 11 items long. Two of them became standard pre-quote checks.
Where Pylontech Ends and I Begin
If I were writing a Pylontech battery review for another installer, the honest summary is this: the modules are solid, the cycle life is a genuine step up from the lead-acid I used to sell, and the compatibility documentation is better than most—if you actually read it. But the battery is one part of a system. The Pylontech combiner box is another part. The diagram, the sizing, the permitting, and the interconnect are mine.
Here's the thing about talking to Pylontech's technical team: they're specific. They'll tell you what the battery will do with which inverter, at what voltage range, with which protocol. What they won't do—and what I respect them for—is pretend to be a full system design shop. When a customer asks me "can I just buy the whole system from Pylontech and skip the integrator?" I tell them no. Pylontech makes the storage. Full system design is not their lane, and they don't claim it is.
I'd rather buy from a manufacturer who knows their lane than one that claims to do everything. Same reason I tell customers when a job needs a structural engineer or a licensed electrician I don't have on staff. Same reason I told that guy in 2019 not to put a stationary LFP module in his car.
The vendor who says "this isn't our strength—here's who does it better" is the one I keep calling back. Pylontech has been that vendor for me since 2020. I just had to learn the hard way that they were never the problem—I was, for assuming I could reuse a system diagram without doing the homework.