LunchTrak Summer Pilot

The LunchTrak summer school pilot, unfortunately, did not go as well as I'd hoped, but there are some good learning experiences from it. There's a lot to cover, so I'll just start in chronological order. The pilot date was Tuesday, June 20, 2023.

Two Weeks Before Pilot

I meet Brian Casino, a tech guy who works at the district, and show off my project. He likes it and we discuss how to actually get it implemented for real. After the meeting ends, I email Mr. Dorosti and Ms. Catalano (who are summer school principals) about how I can distribute the tags. Funnily, I called Ms. Catalano by her first name in my email (I do that for my "business" emails) and she got really pissed at me. The lesson is to always call a teacher by their last name?

Also, the 75 3D printed cases are done around this time too.

5 Days Before Pilot

The LunchTrak tags finally came! Woohooo! I skipped the gym that day and spent a straight two-and-a-half hours programming the tags. I took a lot of pictures and videos to celebrate the exciting moment!

I was so excited that I even recorded myself setting it up. I have a video explaining how I did the first one, and then a timelapse of the rest.

I "tested" each tag by ensuring the pair LED flashed when I held the pair button. I kind of assumed that everything else worked, but oh boy was I wrong...

4 Days Before Pilot

I decide to play around with the wake up receiver just to make sure the antenna matching components are a good fit. I picked 5 random ones and they all worked. I go to sleep

3 Days Before Pilot

"Dylan, shouldn't you test all of the boards to make sure they work?"

My dad had a good point. I was feeling lazy, but this was an important thing to do. As I got deeper into my stack of tags, I realized something: more than half of them don't beacon correctly! This is bad... really bad.

I play around with software, review my schematics, and use every prompt imaginable with ChatGPT. Nothing works. I get really scared and frustrated and upset and I punch the metal frame of my bed before slowly falling asleep (this is at 1 AM, 5 hours after I started testing)

2 Days Before Pilot

I wake up really stressed out. I spend my first 4 waking hours doing more tests and writing more code to figure out the problem. Nothing works. It's also Father's Day today and we go to Flemings for lunch (which was pretty yummy!). My mom insists on buying a special item from Sephora, so we wait for her. Turns out, she's actually on an hour-long shopping spree! I sit outside in the Stanford Shopping Center stressing out about my project. I send Dave an email, knowing that he probably won't respond since it's Father's Day.

I get back home. Nothing works, but it's okay. I still have one more day to fix things.

1 Day Before Pilot

Uh oh, now I'm really stressed out. Dave hasn't responded still, so I text (not email) Ray and ask him to help me out. I thought that he was off of work since it was Juneteenth, so I had no problem asking lots of questions. I bombard him with questions and we go back and forth for about an hour and a half (10 AM to 11:30 AM). At that point, he asks me if I wanted to come over so that he could help me. I realized that he actually was working while I was texting him... whoops.

I'm feeling pretty good because I know Ray can figure out what's wrong with my boards, especially with all the equipment they have in the lab. I feel good enough to take a break and go to the gym.

It's 5:20 PM and I drive over to Atmosic. When I arrive, Ray is one of the only people in the lab. He seems to be really focused on something important. When I get there I see Tony leave and say bye to him; I think he was surprised to see me. Ray gives me a Joulescope and tells me to measure the current on my boards. Luckily, I have VBATT and GND test points on my board, so I soldered some copper wire to them and hooked up the Joulescope. As we analyzed the current, we noticed that the broken boards were not showing the beaconing pattern. We played around with BLE_adv settings and still could not find a beacon. We also pulled up a spectrum analyzer to confirm this.

We go to Ray's workspace and I hand him a broken board, my original prototype board, and an ATM2 module. He checks the schematics and probes around. When he goes to the solder station, he thinks that he shorted RFIO to GND. More testing showed that he didn't. That's when he realized the problem. There is supposed to be a DC block that will prevent current from leaking into the chip. There is supposed to be 3 megaohms resistance across RFIO to GND, but the broken boards show much less (around 300 ohms). This finding confirmed his suspicion that the boards had ESD damage, and he explained to me why ESD damage occurs. From my understanding, DC blocks (as the name suggests) do not allow DC to flow through (I think it is a capacitor). When you accidentally touch the pin, however, you might cause a sudden voltage drop or rise, meaning that there will be AC for a short amount of time. This causes the current to flow through the DC block and reach the chip, causing unrepairable damage.

As you can imagine, I was heartbroken. I didn't really know what to say to Ray except "Thank you for helping me". For 10 minutes, I sat in my car holding back my tears. I called my Dad and told him what had happened. I was lightheaded on my drive back home and I couldn't think straight. The pilot was tomorrow, what would I say to them?

As I got home, my parents looked at me and knew what had happened. I didn't say anything and went into my room. When I came out, I took a quick shower and ate dinner. I explained to my parents what had just happened.

I'm not going into the details of what happened next, but I think you (who is probably future me) already know what I was thinking and what I said to my Mom, and then to my Dad.

It was about 9 PM when I realized that not all the boards were damaged. I spent an hour testing all of the boards and found that 26 had functional RFIO. Of those boards, only 15 had functional WURX. I was happy to have had enough to actually run a quick test with my own Living Skills class. Still, I felt the lingering effects of what I had thought to be the end of my project.

I went to sleep a little more optimistic about the pilot that would happen tomorrow.

Launch Day

This is the big day, the day I've waited 6 months for. I woke up naturally at 7 AM and immediately ate breakfast, brushed my teeth, and then ran some quick, last-minute tests with the system; it was all ready. I drove over to Gunn and went to the cafeteria. When I got there, I said hi to the lunch workers and set up my LunchTrak system on their computer. They didn't have any AC adapters, so I had to go to the tech office to ask for some. Syed drove me around campus in his golf cart and I actually got to see the server room in Gunn; he found a couple of old AC adapters and USB extenders there. I set the system up and went to class.

I think the teacher forgot about my thing (even though I reminded her at the beginning of class) because I didn't get to talk to the class about the project until 20 minutes before brunch. When I did, I gave a whole speech like "Who here hates long lunch lines? I'm working on a project to fix that" blah blah. The students were initially confused but were willing to test it out. I handed the tags to most of the class (though some were the "no wurx" tags) and I helped them go through the pairing process. 5 minutes before lunch, everyone was set up, and I left class early to set up my camera in the lunch lines.

Brunch Time

Now the interesting part. I will show the video here before I talk about it (the real thing starts around 2:00).

I can't stand rewatching it for more than 10 seconds; it was a disaster. The biggest issue on my end is that my scanner would scan students in immediately as they entered the room, meaning that the lunch worker wouldn't see their names pop up as they left the room. This problem is hard to fix due to the unstable nature of RSSI and Bluetooth. The next problem that you might have seen is that numbers would combine into something like 95000000950000001 when someone is typing their number and a LunchTrak test is right behind them. This is a software issue on their end, but it is still something I have to consider when developing a new solution.

Takeaways

I was very disappointed but also surprisingly relieved that the pilot was over; all of my worries and stress just disappeared. I knew that I would need to update everyone who helped me with the project, so I tried thinking of ways to phrase how the pilot went. Here's the email I sent out.

Hi All,

Just wanted to give you a quick update on my LunchTrak testing yesterday during summer school.

Due to a problem during manufacturing, many of the tags were damaged and only 26 of the 75 tags worked properly. I was at least able to hand the working ones out to my Living Skills class. Because there weren't that many tags, we ended up mixing the lines with both LunchTrak testers and students who typed their IDs. The problem is that when a student types their number into the software, there is a short delay before the software registers that number. If a student with LunchTrak is right behind, the two numbers might be combined, leading to something like 9500000195000002. We had to ask the students to retype their numbers. Another problem is that a couple of students were mistakenly picked up by my scanner upon entering the room, causing confusion when they tried to exit because their names would not show up again. I am still looking into why this problem occurred and how I can prevent it.

Though the pilot did not go exactly as I planned, there are some positive takeaways. The students who were scanned in correctly had no delay and enjoyed the convenience of not needing to enter their ID. At the very least, this experiment helped confirm the idea that skipping the keypads leads to time and convenience benefits. I just need to focus on ensuring the above issues do not happen.

Overall, I will be brainstorming and tinkering with potential fixes to solve the issues mentioned above. Thank you all for your help with my project. I will keep you updated on my progress.

Thanks,
Dylan

So yeah, admit your faults early and then focus on the positives! (see my post on the book How to Win Friends and Influence People)

Conclusion, Lessons, and What's Next

I spent 6 months building a product that didn't work. But it wasn't wasted: the students who were successfully scanned in loved the product and its speed + convenience factor). My "Lean Startup" Spidey-sense tells me that I had just confirmed my value hypothesis, but I will need to pivot on the underlying technology. If I had read the book earlier, I would know that 6 months is a wayyy too long turnaround time to run an experiment like this. I should have at least tested it with my friends earlier on.

Seeing the results of this experiment could easily make me depressed if I don't try seeing it from the positive side. For a first product, being able to demo it to PAUSD administrators, get them to pay for it, actually go through the manufacturing process, and have everything on time for a pilot is actually very encouraging news. I have the ability to get my ideas and products out there, I just need to focus on getting it to work next time (this should come easier as I learn more and get better). The fact that I have my own LLC should make me happy enough. I can give it a new name (maybe LuWang LLC, or probably something else) and make LunchTrak a DBA. Then I can make all my small businesses a DBA to my parent LLC. Dude, I have a business bank account and credit card. Why should I be upset?

I'm not sure exactly what the future looks like for LunchTrak, but I do not intend on giving up right away. I will, however, take a small break from it to explore some other projects like my 3D Printed Guardian Sword, Gunn Alumni (complete pivot), and Gunn Elimination. I have a lot of cool things to work on this summer and I look forward to all of them. Okay, maybe not so much on my college essays.