|
|
|
dock decommissioning, 2025
Tuesday, November 18 2025
location: 940 feet west of Woodworth Lake, Fulton County, NY
I got out of bed at about 5:00am this morning and began my day. I made myself a french press of coffee and got the woodstove going. Temperatures inside the cabin had dropped from the 50s into the upper 40s overnight, but after a few hours they were up into the 60s. After doing my usual cabin morning rituals, I began a conversation with ChatGPT about what to do about the weird new data coming from the WiFi dongle of the SolArk inverter. I started by posting the Arduino sketch that I'd used to successfully parse the old data, a passage of old data, and then a passage of new data. ChatGPT seemed very upbeat about the situation, and kept talking as if we'd just solved the whole thing. But the fact of the matter was that the new data format was very different from the old one and lacked the same bits of text to search for to know where to begin looking for the values that needed to be parsed. I soon found myself bumping up against the essential problem of communicating with a large language model (at least the ones that exist now): they pay attention, but their ability to prioritize the value of what you've told them is poor. So I found myself constantly reiterating things like the details of how my parser works and the fact that the old and new data formats are completely different on multiple levels. ChatGPT kept trying to get me to go off an implement its ideas, but it was good that I held off, because they were all pretty hare-brained. This went on for hours while I also listened to or watched YouTube videos.
By the late morning, I needed a break from all that. A completely different sort of task was the seasonal decommissioning of the dock on Woodworth Lake, something that very much needed to be done. And it was best not to wait until too late in the afternoon, as working in the snow after dark is not anything anyone wants to do. So I grabbed a beer and my camera and walked down to the lake. (There wa a single bufflehead duck down there when I arrived.) Everything else I needed to do the decommissioning was already there, including two hand-powered winches and two cables, each running over pulleys placed high in trees near where the floating part of the dock needs to be dragged ashore.
The first step in the process is always to disconnect the floating part of the dock from the rest and then paddle it over to where it will be beached. This all went well, though in the process I managed to lose of the cotter pins securing the piece of pipe that serves as a hinge pin connecting the floating part of the dock to the half-floating segment. (The pin flew off as I jerked it out and disappeared into seven feet of water.)
Normally at this point I would've then jacked the up the half-floating part of the dock out of the lake, hanging it from a temporary support pole used only for that purpose. But I delayed doing that so I could use the chains that are part of the over-winter rigging as temporary attachments for the winches I use to pull the floating dock out of the lake. I already had the two cables I use for winching set up from last year. All I had to do was fix the thimbles and grips on the end loops, which dated to before I knew the best practices for assembling those. Soon I was winching the dock out of the lake using two different winches operating from two different angles (about twenty or thirty degrees apart). This time I was less concered about the cable failing and flailing at me, though I made sure to look away from the cable as I winched. Near the end of the process, there actually was a hardware failure; one of the quick links hadn't been secured sufficiently and had been distorted into a yawning C-shape. But it hadn't completely failed.
With all that out of the way, I could put away most of the chains and quick links and then use the smaller of the two winches to hoist the floating end of the half-floating section of dock up out of the lake. This all went very smoothly. One advantage I had today with all of this work was that the lake level was about as high as it ever gets, meaning I didn't have to expend as much work getting everything out of the water. That said, the decommissioning still took about three hours. When it was done, I cracked open that beer I'd brought, which was now very cold from having lain for three hours in the snow.
Back at the cabin, I soon turned my attention to unloading the solar panels from the Forester. Since the panels are so bulky and their sharp aluminum corners so gouge-inducing, I decided it was best to bring them into the basement via the bulkhead entrance, which I hadn't used in over a year as I worked on covering its walls with old CDs and DVDs. So first I had to clean up that entrance, removing all the pieces of wood used as makeshift braces while gluing CDs in place. There was also plenty of scrap spray foam, scrap CDs, and other plastic detritus, all of which I stuffed through a hole into the void beneath the bulkhead steps, where it will help beef up the insulation that keeps the slab at the bottom of the bulkhead entrance as close to average subterranean temperatures as possible. Then I could bring in the panels one by one. That was an exhausting job and I felt it in my body after I'd done it. Once all the panels were safely in the basement and the bulkhead was closed, I could pack the panels away in a way that used a minimum of space.
Later I boiled up some Nasoya dumplings, watched some YouTube videos, and started drinking booze before, at some point, overcoming the procrastination keeping me from resuming work on the SolArk Copilot. When I resumed work on that, I used some of the information ChatGPT had given me to resume my visual analysis of the data coming from the SolArk WiFi dongle. It still included hex codes, but they were broken up by passages of such as "I (130690242) MB: <1> Characteristic #3 10s_2 (2) value = (0x00000000) read successful. : 16." And the hex codes also includes memory addresses and the codes' equivalents as actual characters, like you'd see in a hex editor. ChatGPT assured me that this was a normal output of the ESP32's BLE (bluetooth low energy) functionality, though in this case none of this data had ever been transmitted via bluetooth. As I looked over the data and paid attention to the actual values from the inverter (using the data reported via the SolArk cloud), I thought for awhile that I'd found the byte holding the battery percentage data. That's the easiest one to find, since it changes slowly and only occupies a single byte. But then as more data came in, it seemed maybe I was wrong. I frustration, I opened a trouble ticket at SolArk as follows:
I had been intercepting serial data being communicated between the WiFi Dongle and the SolArk inverter for use in real-time automation on-premises (that is, not using the cloud) but a firmware update seems to have completely changed the format of this data. I have tried to write a parser for the new data but have not had success. I am asking to confirm whether such a data format change happened from a firmware dated Oct 15 2025 and whether anything can be done to restore the old firmware.
The response I later received wasn't especially helpful:
Thank you for reaching out, and I appreciate the detailed explanation of what you are trying to accomplish.
I do want to clarify a few important points regarding the Wi-Fi dongle and the data stream you were previously accessing.
- The firmware update did change the internal serial data format Yes, beginning with the October 15, 2025 firmware release, the internal communication protocol between the Wi-Fi dongle and the inverter was updated. This update included packet restructuring, encapsulation changes, and data-handling improvements. Because this is an internal communication channel, the format is not guaranteed to remain fixed from one firmware version to another.
- The internal serial stream is not a supported interface The UART/serial data exchanged between the dongle and the inverter is considered an internal protocol. It is not documented, and it can change at any time as part of normal firmware development. For this reason, engineering does not provide or support access to that stream for third-party automation.
- Firmware rollback is not available We are not able to roll dongles or inverters back to older firmware versions. All units automatically move forward in order to maintain system stability, cybersecurity, and compliance.
This is the kind of world we live in now, when even after you spend thousands of dollars on something and many hours customizing it to suit your life, the company that built it can reach across the void and fuck it up completely. If I ever get my data parser working again, you can be sure I will also change the WiFi information so the WiFi dongle can no longer phone home.
The dock after I'd floated away the fully-floating segment. Click to enlarge.
The floating part of the dock just before I'd winched it onto the shore. Click to enlarge.
One of my winches pulling a cable running to the floating dock via a pulley high in that sugar maple a third of the way from the left. Click to enlarge.
After I'd jacked up the half-floating part of the dock. Click to enlarge.
Sunset this evening with some lingering beech leaves in the foreground near the cabin. Click to enlarge.
Peck Lake is now visible through the trees from the cabin. Click to enlarge.
For linking purposes this article's URL is:
http://asecular.com/blog.php?251118 feedback
previous | next |
