EX-G Wired Trackball Retrospective

14 Feb 2026

This post is a look back on Converting a Wireless Trackball to Wired. I wanted to capture some thoughts while they’re fresh in mind.

Duration

I initially wrote down the high level plan on 2025-12-14. I’m writing this post on 2026-02-14, exactly 2 months later. Like many people, I under estimated the time it would take. I was thinking it might be done around the new year. I was hoping to take advantage of the winter holidays to meet this new year deadline. I had an overdue IOU to replace the floor in our laundry room. I underestimated the flooring effort and it ate into the time I could spend on the trackball during the winter holidays.

Blogging

One of my goals was to blog about the effort. I am not a fast or efficient writer. Most blog posts took 3-4 hours. I would be lucky if I got through the writing effort in under 2 hours. Blogging about the journey likely doubled the time cost of completing this task.

The goal of blogging was to practice writing and communicating my thoughts. A large part of writing is skill. Like any skill, it takes practice to get better. I could benefit from taking a writing class or two, the act of writing is something I can do on my own time and at my own pace.

I tried a couple of approaches to the writing. Some posts were written after the fact. These posts are usually written entirely in the past tense.

Some posts I wrote as I did the work. I would write a plan for the next small step, I would do the task, then I would write up results. Rinse and repeat. I’m not sure how well they read to others, but I enjoyed writing those posts more. It broke up the writing. More importantly, it helped me write out my thoughts beforehand. Often my thoughts or hypothesis would be wrong. The past tense posts would often have me omitting some of the mistakes I made to get to my result.

What I Learned

I’m sure there’s more, but these were ones I thought to write down as I went:

The first topics are pretty well covered in their own posts.

Schemdraw

Schemdraw was a nice find. I really enjoy making programmatic diagrams. I often use mermaid.js diagrams where I can. The programmatic nature of the diagrams can feel a bit constraining at times, but it lends itself to quick modifications and edits. They’re also fairly concise for sharing.

I was able to leverage the:

I tried to use the pictorial schematics where I could. They provide a depiction that’s close to the real world, but are much cleaner than any picture I would try to capture.

The source for all of the Schemdraw diagrams are available in the diagrams directory of this blog’s GitHub repository.

Giving Up

There were a number of times where I hit a roadblock that made me question if I should cut my losses and end the activity.

After disassembly and realizing all components were attached directly to the circuit board.
After finding the PMW-3320DB-TYDU data sheet and seeing that it lacked sufficient information for communicating with the sensor.
After finding out the Atmega32U4 was 5V and the PMW-3320DB-TYDU was 3V.
After discovering the ESP32C6 didn’t support USB mouse

Disassembly

I had mentioned in Using EX-G hardware switches how I had hoped the switches used wiring so I could rewire easily. I also mentioned in hindsight how this was a false hope. My limited experience with small electronics seems to indicate that most of them put as many peripherals directly on the circuit board as possible.

The feeling of giving up was fairly fleeting for this problem, but it was still there.

PMW-3320DB-TYDU

The optical sensor was a big unknown. I didn’t do any research prior to the tear down, so had no idea what I might be getting into. I was fortunate to run across the ADNS-3050 data sheet. If I hadn’t found the ADNS-3050 data sheet, I likely would have given up. Not even trying to reverse engineer with a logic analyzer.

5V vs 3V

When I found out that the PMW-3320DB-TYDU couldn’t handle more than 3.6V I must have been in a bit of lull for the overall effort. It wasn’t a major roadblock, but it was enough that I was considering shelving the project.

For whatever reason I had latched onto using the Atmega32U4 and was looking for a 3.3V version of it. They make 3.3V versions, they’re just a bit harder to get one’s hands on. It took me a while to decide that I should try a different board and chip. With the difficulty I had getting the ESP32S3 into the EX-G case, I think it’s a good thing I didn’t try to use the longer Atmega32U4 board.

ESP32C6

I had chosen the ESP32C6 to replace the Atmega32U4. When I found out the ESP32C6 didn’t support mouse outputs over USB. I had to look yet again for another solution. The ESP32C6 was already my second board to try to complete this project with. I wasn’t sure I wanted to keep trying more boards. I had a better idea of what functionality to look for in a board, but it was still pretty frustrating and made me consider ending the project.

I also considered trying to use the ESP32C6 as a bluetooth peripheral. I think this would have greatly increased the scope of the project. I would have needed to figure a way to power the board from the one AA battery the EX-G provides. I also would have needed to figure out how to properly put the board to sleep and wake-up based on the PMW3320DB-TYDU interrupt. This idea twirled around in my head for a while before I finally decided it wasn’t going to be a good path. If I had tried down the bluetooth path, I’ve got a feeling the extra effort and time likely would have resulted in me shelving the project before completion.

The Failure Posts

There were two posts that were me failing to do what I had set out do.

Often I end up communicating the good parts, or the final result. I have a habit of not communicating failures until after I’ve gained some emotional distance from them. Failing is a big part of learning. I struggle with embracing failures for the benefit they give. These posts were a way to force myself to write directly after I failed, or while I was failing. During the writing of the posts it was tempting to not publish them.

Giving the failures their own blog posts also helps to convey that failures aren’t always fast. Sometimes they take a while to realize them for what they are. It’s not uncommon that the work necessary for the correct solution only takes the last 10% of the time, with the failures taking up the other 90% (those are made up ratios).

I’m glad I took the time to write and publish those posts.

Cost

I had to put in this section, because someone was giving me a hard time about the cost of this project.

description	cost
EX-G	$39.99
Atmega32U4 (3pack)	$17.99
logic analyzer	$19.99
ESP32C6 (3pack)	$22.99
ESP32S3 (3pack)	$26.99
total	$127.95

The table shows monetary cost. The time was quite a bit. Taking the low estimate of 3 hours to write each post. I imagine the work the post was writing about was likely close to 3 hours as well. There were 25 posts prior to this one. That implies that it was at least 150 hours of effort. About 20 hours a week for 8 weeks.

Summary

The biggest take away for me is I finished the project.

It’s very easy to take on a side project and get a little into it and then lose interest. I was able to keep pushing through this project and come out on the other side with a working solution.

I also learned some new tools and insights along the way. It’s nice to learn things and be able to apply those learnings right away.