Here’s a write-up from Tony addressing some of the important and common issues with the early ADV’s, cleaned up and organized a bit courtesy of ChatGPT haha:
The entire team is watching and taking notes. Yep, the first batch turned out to be bumpier than we thought. I’m very sorry about this, but we’re already working on all the fixes. Here’s some progress on fixing those problems:
1. Battery Drain Issue:
The current battery drain is 0.008A, which is not good. This drain rate means your board will drain in 41 days, not longer than a winter season. Luckily, we have BMS protection, so it doesn’t damage your battery even if it’s left uncharged for too long. An easy fix is to plug it into a wall and leave it during winter. Another fix will require swapping out the LCM module for a new one. I’ve already made it drain at 0.0004A on my boards with a modification on the LCM. That means it will last about 800 days fully charged sitting there. Still observing for any quirks with that mod, which involves replacing the LCM bottom R6 resistor with a 1206 200K resistor, requiring a hot air station. I understand most people don’t have one, so I’ll just order 600 units of revisioned LCM for you to swap out. That’s how I plan on solving the slow drain issue. I found it using liquid nitrogen, dipping the whole ESC+LCM, and using a thermal camera. That resistor is the only hot thing during poweroff, and I replaced it with a 200k and the drain is gone. I designed the LCM with the IC’s datasheet requirement (10K). Turns out it’s too low for real-life situations, but thankfully the cost isn’t astronomical, so we can afford this recall.
2. Footpad Sensor Ghosting Issue:
The issue only exists in the first 80 footpads with the first batch of boards. The problem is the layers being assembled using the wrong type of adhesive, causing water to dissolve the adhesive rim around the sensor layers. We’ve shipped all 80 customers new revisioned footpads. You should be able to get them this week. The new footpads use a PET base adhesive, so they are water-resistant. I’ve tested the new ones underwater for over a week, and they still work great.
3. BMS Issue:
Yep, we need a more advanced BMS that can talk with the VESC in case of a battery pack malfunction. At least it should slowly push back before cutoff. For this to work, we’re now working on a new BMS and a new cable harness. This will allow the BMS to communicate with the ESC. Other than that, it will need individual cell monitoring with an app, active balancing, protection parameters set up via the app, and it needs to be a charge and discharge BMS for sure. Charge-only is not ok. In the event of an ignorant rider leaving the board uncharged for 3 months while the battery drained to 0, and they plug it into the wall, the next thing the whole house is burnt down and we’re on TV. So, I disagree with a charge-only BMS. Use this only if you know your battery stuff. We have a working BMS evaluation prototype that already does active balance+individual cell monitoring via the phone at hand. The problem we have is that this BMS is massive. We haven’t figured out a way to put this in our board. Probably need to do a 3-layer board-to-board-to-board design to fit it in. It’s a massive project, even bigger than the Floatwheel Adv itself. So allow us some time. We’ve seen 2 events of BMS cut off during a ride. It was in the first 80 boards. One vital QC process is not carried out during first batch production, which is individual cell drain monitoring. We now let the board sit in the factory for a week, test for any cell drain issues, so this will be stopped in the factory and no longer be an issue in the future. But like I said, a new BMS is being designed, and we’ll make sure it’s backward compatible with first batch boards.
4. Brittle Plastic:
Simply put, it was us being stupid chasing that matte surface finish and dumping too much fiberglass in the mixture. The first 300 boards will receive a whole refresh set of plastic shells (bumper set + fender base set).
5. Motor Making UFO Noises at Lower Speeds:
An easy fix is to tune your zero vector frequency up or down 1kHz, and the noise is gone. We think this has to do with the current sensor not being accurate enough at lower speeds with lower battery voltage. Can’t say for sure, but once I tune the zero vector frequency up a bit, it’s gone.
6. First 200 Motors Too Torquey:
They are wound to the same configuration as the hypercore, actually. The reason it feels slow compared to the GT is that it uses magnets that are way too high spec. On the GT, you can take one magnet out to measure, it’s around 1750Gs, which means it’s a N30-N35 magnet. The Adv is a N42H magnet, which measures 1929Gs in the center. On top of the higher-grade magnet, this thing is thicker, longer, wider than the hypercore’s magnet. So, the flux is 1.6x hypercore. With this magnet, it comes with great torque, but the magnet also generates massive BEMF. So, it doesn’t run fast without field weakening. Field weakening is basically injecting current into the Q axis to create a magnetic field to fight against the BEMF generated by the motor. So with 80A field weakening, you’ll find the motor can actually spin faster than the GT but sacrifices efficiency at higher speeds. The first batch of motors will be really efficient around 20mph, which is the speed I ride longer distances at. I purposely set it this way so the range is good even with a bigger ass motor. But since you guys like higher speed, it doesn’t hurt me to swap to 135 new design motors. It doesn’t really hurt your torque, though. It just needs 20% more current to generate the same torque as Batch 1 motors. So, expect the ESC to run hotter than the Batch 1 motors. In terms of accelerating performance, the new motor will feel more like a GT, which picks up speed quite fast but with higher torque to keep your nose from scraping.
That is a really long essay. Thank you for reading this. Now I’m going to get back to fixing the first batch problems for my supporters.
DISCLAIMER: this is not official, consult at your own risk. It’s a Wiki, so please contribute
