 |
| Automatic Antenna Selector |
This cable connects the Auto Antenna Selector to the AUX port of a K3. I bought two of these a couple of years ago, and I easily found one. But the second one eluded me. Without it, I couldn't test the Selector in Design C (Dual Radio) mode. I had done all my testing in Design B (Single Radio) mode.
Having the proper cables, I did my testing and found several bugs. Some of them were going to be difficult to fix. Part of the problem was my approach. I had organized my antenna selection table by the selections, then by band. This made it difficult to determine the choices available per band. So, I organized the table by band, then listed the choices. From this, I could determine which choices were available for a given band, skipping over those that would conflict with an existing port.
This greatly simplified the code for making the mode selections, which are very different in Design B (Single Radio) and Design C (Dual Radios). In Design B, the first selection is assigned to the first port, so there can only be three additional choices. In Design C, only one port is selected per radio, so we can potentially use all four selections (so long as there's no conflict with the other radio).
Design C bugs were all addressed, and Design B worked better than ever.
Before I retired, I'd been messing around with Anthropic's Claude Code. I was curious what Claude might think of the firmware, so I requested an in-depth code review.
Claude Code came up with a number of good suggestions. It found some dead code, made several type suggestions, a minor code re-organization, and way better comments.
What impressed me, though, was that Claude Code had a deeper understanding of the particular PIC processor that I did. It suggested I turn on the Watch-Dog Timer (WDT). That I change how Timer 0 (TMR0) was used. And, that I use the PMD register to power-down the modules that I wasn't using in the device.
I thought I was using TMR0 correctly. And I knew about the WDT, I just didn't think it was necessary. But the PMD register I knew nothing about. All of these suggestions seemed like a good idea to me, so I made them.
All in all, Claude Code did eight rounds of review of this code. It made a couple of suggestions that I knew were wrong, and it took corrections well.
The next day, I burned a chip and plugged it in the unit for testing.
It bricked, Nothing. Not even a single flash of the LEDs on startup.
Somewhere in the midst of Claude's suggestions, we had changed the software in a way that it no longer worked at all.
I made a new branch of the code and backed out some suggestions. I determined it had nothing to do with the WDT or PMD register. That left something amiss with TMR0. This could be a problem with Microchip Code Configurator (MCC).
I had switched TMR0 to use the HFINTOSC. It was programmed to generate an interrupt every 5 ms. Except it didn't work. I switched it back to us the LFINTOSC, also programmed for every 5 ms, and it worked as expected. I'm not sure why it didn't work with the HFINTOSC, but having several oscillators gives one options. I'm happy it works.
Once it was working again, I walked through all of the Design B and Design C test scenarios. Everything worked with flying colors.
AI is certainly useful, even for amateur radio projects. Just don't blindly trust it. Remain in the loop for all changes.
