Custom FOC board for wind turbine generator 200W

Hey SimpleFOCers

As some seen over the past few months we are building a “small” wind turbine that utilise SimpleFOC to actuate two stepper motors, which in turn controls the pitch of the blades. I/we developed a custom board, which we have been super stoked about! The SimpleFOC library has been really helpful.

Currently we are using a VESC 4.20 Mini from Flipsky as ESC. This is controlled over serial from an ESP32.

I would like to investigate the option to replace the VESC mainly to reduce cost, but we might also eventually try to gain additional performance from lower control latency & optimised speed tracking.

The requirements to the controller is that it’s able to work with DC voltages between 10 and 40V nominal.
The motor current requirement is 15 Amp continuous.
The electric rpm is 19000 eRPM.
I’m thinking that it’s best to measure at least two phase currents, but I’m not really sure about the performance difference between 1 and 2 current sensors.
Ideally based on ESP32 (since we are most familiar with that platform and we want wifi connectivity anyway).
Currently I’m inclined to use 3 standard hall sensors as it’s fairly cheap to get those sensors included in the skateboard motor/generator we buy. Would perhaps upgrade the code to use timing for improved angle estimation. Alternatively the two linear hall sensors mounted off axis outside of the motor could be used in an improved version.

I’m currently thinking about basing the design off DRV8313PWPR since it’s currently in stock on JLCPCB and has over the past year been reasonable available.

For the layout I would probably start by looking at the HackJammer.

For the current sensing we would probably use the ACS712ELCTR-20A-T.
Cost wise the goal is to hit max 50 USD (with the current sourcing prices) - of cause the cheaper the better :smiley:

I would be curious if you have any input or alternative routes you feel I should consider.

BR
Andreas

1 Like

@aokholm

Interesting. So you target $50 per board, does that include all shipping, and import duties or just out of the factory?

You cannot work with 1 current sensor, you need at least 2, better use 3.

Never done any designs with esp32. I use ESP32 only for connectivity and drive my design with STM32. If you do an integrated design, it may be best to use an integrated 3-rd party MCU board footprint instead of hard integration like I did with STM32 HackJammer. It’s a lot easier to pop the mini-board with the ESP MCU on and off when programming and testing since your turbine will have the power board connected hard to the metal and you need to reprogram the ESP32 multiple times, or use a few boards with different programs to test. Just my opinion. Also it will be a lot easier to replace the MCU later if you need to, just modify the footprint, and you can use a BluePill or BlackPill, or some of the new Arduino Nano, or even an RPi board.

That driver is only 2.5A max current, you need 15A continuous, so your only choice would be discrete components.

Thank you. However the HackJammer (and the very expensive JackHammer brother) are using discrete components, and you said DRV8313PWPR so I’m a bit confused. Which way you want to go?

There are a lot cheaper current sensor than ACS ones, if cost is an issue. You may want to check the Cross Chip CC6903SO line of sensors, such as CC6903SO-20A. It’s $3.00 vs. $0.80, multiply by 3 and you saved $9.00 vs. $2.40, that’s $6.60 you can spend on something else. I just saved you over 10% of your board budget.

Let me check my database for some 15A components. At that low current you may get lucky.

Cheers,
Valentine

1 Like

Let me open a new one for the suggested components:

MCU: As I said, keep a footprint and just buy off-the-shelf ESP32 Wroom and pop it in ($6.70).

MOSFETs: RU30J30M by Ruichips ($0.20), it’s an integrated half-bridge, 30V/30A continuous, plenty of voltage and power. If you want higher power, go for DMNH6042SSDQ-13, 60V/20A 2 N-Channel in a single body ($0.98), some routing required to turn into half-bridge.

Current Sensing: CC6903SO-20A ($0.20)

Driver: SLM7888MD by Sillumin, ($1.08)

You will also need a bunch of resistors, diodes and capacitors and buck converter, may be an LDO, let’s say about 10 dollars.

So your BOM would look like

6.60+0.6+0.6+1.08+10.0 = $18.88 for basic components per board
$17 for board and SMD (sunk cost)
$15 for components ordering and SMD arming (sunk cost)
$40 for express shipping
$9 discount
An Order of 5 board would cost you about $157 for 5 board to your doorstep using DHL Express.
That’s $31 per board including sunk cost and shipping

Now the easy part, designing the board :rofl: :rofl: :rofl:

Cheers,
Valentine

1 Like

My bad. The chip I really wanted is the DRV8323 - gate driver. I didn’t check the DRV8313 datasheet well enough. I though it was just the previous model with the 2.5 Amp in the title were the specs for the gate driver ><.

I was surely imagining to use external MOSFETs

Thanks! That’s why I were considering using a single or just two phase current sensors in the first place. $0.80 seems much more reasonable :smiley:

The Driver: SLM7888MD also seems suitable even if it’s intended for much higher voltage drives. Thanks.

MOSFETs: Regarding the MOSFETs I was thinking of spending a bit more in order to reduce the cooling requirements C538034 IRFS7537TRLPBF 60V 173A 3.3mΩ@10V,100A 230W 3.7V@150μA N Channel D2PAK MOSFETs ROHS

I’m right to assume that the approximate heat dissipation in the MOSFETs (board) can be calculated as

Highside: I^2 * R_on
Lowside: I^2 * R_on

Total: Highside + Lowside = 2 * I^2 * R_on

We have I = 15A. For the mosfets with R_on of 35mOhm you provided that would result in 15W of power dissipation.

The IRFS7537TRLPBF which is 140% more expensive would result in 1.5W of heat dissipation.

I current goal is to keep the total power dissipation below 5 W in order to allow the board to be passively cooled without a heatsink within a slightly ventilated white plastic enclosure.

The $50 per board is before shipping and duties. - The 50 USD is the break even point (also considering the engineering effort required). If we couldn’t design something cheaper, then we might as well continue buying the stock Flipsky VESC. If we can achieve 35$ including shipping and duties that would be great!

MCU
Programming the MCU in place in not really a problem for us. I have a test stand where it’s easy to plug into the MCU directly with serial, but we also have the capability to do OTA updates with the ESP32 firmware, which would be an added benefit.

For many of our other board I we also started out by using breakout-board for the MCU. It’s a little quicker at first, but then it becomes a hazel to solder all the connectors ><. I also looks cooler to have it integrated!

Let me try creating the schematic :slight_smile:

1 Like

If you are looking for a cost-effective and cool discrete MOSFET then you may want to look at HSBA6074

0.647$/unit and R = 1.7mΩ@10V, 20A

It’s in stock, too.

1 Like