My friend Mihail Disanski and I are currently building a pair of electric skateboards from scratch.
Our goal is to build as much as we can by ourselves. From the deck to the battery pack. We also aim for decent quality.
This allows us to be creative during the design process and add innovating features to our board.
All individual components of our electric skateboard are discussed on this page.
At first we thought we could DIY a great eskate like the Boosted Board (1500$) for cheaper than a chinese board (500$). It turns out chinese boards are incredibly cheap and there is no way we can compete in their price range.
Nevertheless, the quality and performance of the chinese electric skateboards are also much lower than what we want to build. Also, it does not require much more time for us to choose or build superior components.
We calculated a very rough power requirement of around 1100W to go from 0 to 30km/h in 3 seconds, for a 75Kg person. From this value and our initial research, 1500W was chosen as the power value for all components. 1500W should be sustainable for at least 3 seconds.
0 to 30 in 3 seconds is ludicrous and we'll probably never need this speed or power.
To prove how insane 1500W is, the following data was taken on my 750~1000W 400$ chinese board ("Rated motor power: Dual motor each 250W but maximum motor power: each 375~500W" - Wowgo 2S)
Here is a plot with GPS speed value for a short acceleration to the maximum speed of around 31km/h. Start to top speed is under 2 seconds with a fully charged battery. I didn't hold max power because I was carrying a laptop, arduino and GPS receivers with antenna! (Sorry the data is in knot*100 at 50Hz (WTF GPS use knot unit as speed!?) I will get better plots when our build is complete).
Skate Metric has better data comparing most popular boards. My data on the wowgo is way different because the board was fully charged and I am light.
The battery pack is the most costly component in the system. The following parameters are all closely linked. Many solutions are possible and we did not have many constraints except the 1500W requirement. Weight, volume, energy density and cost are all important and we chose the parameters that seemed to optimize the overall build.
The battery configuration is defined by how many cells we put in series and in parallel (Explanation on the S and P notation).
Modern eskates use 10S LiFePO4 (10 x 18650 cells in series). It's what most eskate motors and chargers use are built for. Electric bicycles use higher voltage so we looked to use 12S, 14S or more but decided against it simply because the prices of the higher voltage components are significantly higher for higher voltages. Also, the energy loss associated to more current is not much in our case.
A 10S pack charges at 42V (104.2V) and has a nominal voltage of around 37V (103.7) for LiFePO4 18650 cells. Finally, if the project is abandoned, with a 10S pack, we can reuse the pack on my chinese board.
Cheap battery packs usually are 10S2P. For our 1500W power requirement, the pack would need to output 35A in a 10S2P configuration. With two cells in parallel, we would need cells delivering 20A each. Those are not cheap and stressing the cells is bad if we want them to last long.
Initially, we planned to do 10S2P with power cells, like the Sony VTC6 (15A to 30A) or the LG HG2 (20A). Although the cells would probably heat a lot at 1500W, this pack is an OK configuration. The problem with a 10S2P at 1500W is heat from the busbars connecting the cells together (discussed below).
10S3P is also better for the battery health of the cells. The difference between 10S2P and 10S3P is 50$ and 500g, 1.33x the range and it is surely safer for the batteries.
There are many possible arrangements for cells. For our use case, our final design (number 5) optimizes for height, while keeping the pack small We minimize height because of longboard deck flexibility (shown below). We don't want the pack to touch the ground. At the same time, the pack can't be too long or the battery case could curve and break. This last point is negligible because our battery case will have clearance between case and deck connection (explained later).
Keeping one row of cells minimizes height and that is our most important constraint here. Other designs were considered for minimal volume but we think one row is the way to go.
Final layout design:
It might be possible to split each 3S subgroup in it's own case and add space between each for the board flex to apply to the battery pack.
Battery connectors an interesting component of the system. The material dimensions and welding type are important design choices. We chose to use copper busbars without soldering or welding. We clamp the batteries with poron foam.
Most people doing DIY packs use nickel strips that they solder or spot weld to the cells. The nickel strips used are not safe at all at higher currents! Their ampacity (continuous current rating before overheat) is very low. Some people use nickel strips that have a current rating of 4A. Some packs use copper, which is much better because of the lower electrical resistivity of copper (almost five times lower than nickel).
Here is a table with approximations of the ampacity of copper and nickel for different metal strip sizes. Green means it is safe for 20A continuous.
You can see using Nickel isn't ideal. Even if we stack them together. That's the main reason why chose to use copper.
There are also busbars made of a copper base with nickel connectors to the batteries.
Copper with nickel busbars pictures are from this seller.
If we wanted to spot weld nickel, we'd need to get a spot welder (see image above). We can use regular soldering but it heats the terminals much more and possibly hurts the cells (need source). Cheap spot welders cost around 200$ online. There are DIY kits and some people build them with microwave parts. The problem with copper is that we can't spot weld it properly with cheap spot welders.
Tungsten inert gas welding (TIG) welding can also be used for copper but that's no cheap welding machine. TIG welding is better because the welded parts are protected from oxidation during the process (great video). Unfortunately, a decent but used machine, able to weld 0.5mm sheets, is around 15K$. Interesting thread on welding copper here.
An other option is to bend copper plates or strips to a form that accommodates the battery terminals (Copper sheets are sometimes called copper shims on sites like McMaster). No spot welding required but we need to compress the batteries lengthwise. This is dangerous if we don't have proper surface contact to the cells. The connectors could heat at the point of contact and cause a failure.
This option requires much less work. We don't need a spot welder and the cells are removable! We simply need to find a way to clamp the cells to the busbar, without compressing them too much or block the safety features...
To clamp the cells, we use a layer of resilient polyurethane material called poron. Even after extended compression, poron is supposed to keeps it's compression properties (available on McMaster).. This material is sandwiched between the case and the copper and we use it to cushion and apply pressure evenly to the copper and the cell terminals.
Poron is used for battery pad applications in cellphones. It's dielectric constant is 1.71 so it won't store much charge and resistivity is almost 0.7 teraohm-meter so it won't compete with the copper. Other electrical properties like dielectric strength and dissipation factor are better than some epoxy coatings they put on high power copper busbars. Our usage of poron isn't for electromagnetic
With the previous parameters in mind, we started looking for lithium ion cells. Usually, people use 18650 in their packs but Tesla says their Sanyo 21700 is the future (18650 means 18mm diameter by 65mm length). So we parsed all of the current available cells in sizes 18500, 18650, 20700, 21700 and 26650 into a spreadsheet and sorted them by energy density and removed those priced higher than 2$ per Ampere hour. We filtered out all of the cells that we considered obsolete considering their weight and current ratings. The winner is the Sony VTC6, priced at 5$ each for 3000mAh and rated for 15A continuous discharge (they can even go up to 30A if temperature is regulated under 80°C !)
We bought the cells from liionwholesale.com because it is the cheapest place with shipping. Their customer support answered my question by email in less than 5 minutes!
To protect the cells, the battery case should be water resistant. We can't have air vents so to lower the internal temperature, we will use a metal case acting as a heat sink for the cells.
The innovation here is that compared to regular eskates, the battery case on ours is not in direct contact with the board:
A system like this has two advantages:
1) Almost doubles the surface area of the heat sink. Also cools by redirecting air between the deck and the case.
2) Vibration dampening using screws with a rubber part to connect to the deck. Often sold as "ignition box vibration mount".
We will probably use a metal press to shape the case
The job of a BMS is to make sure the series groups are kept at their tolerated voltages. Cell parameters are not always perfectly equal so they don't discharge and charge at the same rate.
A BMS would not be required if we used a balancing charger. Unfortunately, balancing chargers cost much more than a BMS.
I would have really liked to design a BMS based on my final year project but it would cost way too much and honestly wouldn't be worth the time. There are plenty of chinese BMS that will do the job almost as good as an intelligent BMS.
The remote used with our ESC simply uses UART communication.
The ESC communicates with the remote using UART transmission.
Bluetooth requires more energy but we would be able to control the ESC with a phone. It also sucks. Most hobby RC applications use 2.4Ghz remotes. We could use 433Mhz/915Mhz modules like the RFM69W/HW but they cost a bit more. We already have remotes using 2.4Ghz so we will design one using this technology.
To communicate with the ESC and to read instant stats like speed and power, we can use a simple Bluetooth serial connector with a cellphone or a laptop. There are apps for the VESC on the android play store. I think the Bluetooth serial board like the HM-10 can be plugged at the same time as the PWM control.
I would love to build this watch with radio builtin and use it to see stats from the ESC: GoodWatch21 It would be cool to see stats on the board on Casio calculator watch.
TX (remote) 3V microcontroller is the Arduino Pro Mini 3V3. RX 5V microcontroller is the Arduino Nano.
The receiver is powered with 5V from the ESC UART port. We use a NRF24L and NRF24L shield to properly convert the voltages because the Arduino Nano's 3V output is shitty.
The transmitter (remote) is powered by a 18650 cell and also uses a NRF24. A TP4056 board charges the cell and protects the devices from overcurrent and undervoltage of the 18650. Uses 5V USB to recharge the cells at a maximum rate of 1A.
If the NRF24 communication is unreliable, there are more powerful options like the NRF24L01+PA+LNA.
Overall, our DIY remote kit with TX and RX will cost 15$, including the 18650.
Special instructions for the TP4056: "The charging connector can not be too small wire too long. Such large connection resistance!"
The wii nunchuck hack is cool and doesn't cost much. We bought 2$ nunchucks to can gut them, reuse the joystick and add some buttons and holes to see the TP4056 LED states.
An other possibility is simply 3D printing a perfect case for all of the electronics. This one is well known.
Simply reusing software is the fastest option. We will use a modified version of this simple Arduino library, which uses this Vesc Uart Control library. Example of UART communication with the VESC can be found here.
We will also add a voltage check in the remote software to make sure the 18650 cell does not go lower than 3V.
The ESC software will need to be configured with the required settings to control the board from UART.
Did you know 1/16in Canadian maple wood is only manufactured for skate decks? All information on different materials and techniques can be found here.
We use a kit of 7 layers of maple, with some Titebond III. Glue the layers together and compress with the desired curve. Wait and cut to shape.
Some people use anti-spark switches for the on/off switch of their boards.
After discussing with an electrical engineer friend who has knowledge in power system design (merci Francis), I learned sparks aren't that bad in our case.
With eskates, sparks happen when we connect the ESC to the batteries and the capacitors of the ESC charge.
The spark is visible when connectors are connected or disconnected and current has to pass through air. If it happens close to an electrical board, it could spark on the components and break them. We have more than 10cm of wiring between the switch to any board so that's probably not a problem.
In the eskate world, popular anti-spark switches are Vedder's anti-spark switch or simple XT90-S loop key. Costs around 50$.
1) Vedder's uses mosfets. Many mosfets are used in parallel to handle all the power. The advantage of this system is that we can use a fancy switch. Costs 60$.
2) The XT90-S loop key is a anti-spark XT90 type connector with a pre-charge resistor inside. We solder a wire to loop the connector (Or any copper piece). Someone also designed a 3D printer part to add some paracord. It makes it easier to remove. The loop key is really like any key. The skate won't start if the key isn't connected. This is a security feature. Costs around 10$.
The pre-charge resistor is 5.6 Ohm and it won't survive if the key is inserted partially for too long.
User jackjetful on endless-sphere took one apart:
3) Use a mechanical relay with a switch. Simpler than Vedder's but more complex than the loop key.
We decided to use the loop key. Connected to the battery - wires.
Most eskate ESC are based on the open hardware project called VESC.
I really wanted to DIY this but it would cost way too much in time and money. We bought a cheap ESC that is based on the VESC hardware version 4.12.
The difference between the more recent versions is simply price and heat dissipation. It is not worth it with our specs to get the more recent VESC.
The motor should be a sensored brushless motor with specifications matching Michel's simulations.
We bought the motor but plan to maybe DIY a motor someday if we find proper sized stators and time. Machining stators costs too much.
3D printed and designed by Mitch.
We simply bought the cheapest set online. Both the trucks and wheels are from the Meepo Board store.
I am interested in DIY the wheels.
Most references are posted as links in the text.
Christoph Bolsinger, Matthias Zorn, Kai Peter Birke, Electrical contact resistance measurements of clamped battery cell connectors for cylindrical 18650 battery cells https://doi.org/10.1016/j.est.2017.04.001