Develop a QCA7000 board?

[johu]

My modem draws 300 mA on the 3v3 rail plus maybe 200 mA for the ESP. The linear reg heats up to 60°C, maybe a switching regulator right down to 3v3 is better.
Option 1 still dates back to the first oi design where even the gate drivers were powered from the 5V rail.

You haven't asked but I'll still put it forward: the switching element for the port relays should NOT have a freewheeling diode to 12V but rather a TVS to GND. You so often end up in situations where the freewheeling diode reverse powers your circuit

[muehlpower]

This module is my favourite, but regardless of that, there is a suggestion for the power supply and the transformer in the documentation.

[uhi22]

Thanks, yes, of course a solution is to use an existing board and hook this together with controller. But somehow it feels like a nice challenge, to avoid the multi-board-approach, and just populate all to one board. I do not say, that this makes sense or is the best solution. It is just one possibility, and has a challenge level that feels good. The approach may fail, and in this case, the carrier boards are a nice backup. The software will be the same for both ways. The amount of time spent, knowledge gained and smoke produced is different. That's the motivation.

[uhi22]

The linear reg heats up to 60°C, maybe a switching regulator right down to 3v3 is better.

I was even playing with the idea, to spent independent linear regulators for the controller and the QCA. To avoid that the extrem noisy controller supply disturbs the sensitive RF circuits. This does not reduce the total dissipated power, but splits it to two spots.

[muehlpower]

8Devices does the same thing with the two 3.3V regulators on their WhiteBeet

[uhi22]

Option 1 still dates back to the first oi design where even the gate drivers were powered from the 5V rail.

and do you see drawbacks? It is my favorite, because the accepted input range is more robust (40V instead of 24V). And it gives slightly more power, could be helpful to drive my chinese 5V-coil-relays for testing...

... the switching element for the port relays should NOT have a freewheeling diode to 12V but rather a TVS to GND. You so often end up in situations where the freewheeling diode reverse powers your circuit

Good point. I missed the port relay outputs completely on the todo list

[catphish]

Regarding the power supply: We need something that converts the 12V (9 to 30) from the car to 3.3V. I found two options in the openinverter projects. Both are using a step down from 12V to 5V and a linear regulator from 5V to 3.3V.

I'd encourage you to ave a look at the design I use (attached).
JLC parts:
* C311983 (TPS54302)
* C408412 (Inductor)

In this case I was running it at 5V with a TLV75733PDBV linear regulator because I needed both rails, and I often prefer to do this anyway in RF applications, but it's easy to reduce the voltage to 3v3.

This TPS54302 supports up to 28V input (30v absolute maximum).

[uhi22]

Is the 68pF intended to be on the switched output? The data sheet says, it should be after the L:

[catphish]

Is the 68pF intended to be on the switched output? The data sheet says, it should be after the L:
image.png

You're absolutely correct
See attached the generated example from TI PowerDesigner with the correct design.

Looks like I only build one board with this error and fortunately it wasn't for a customer!

[johu]

and do you see drawbacks? It is my favorite, because the accepted input range is more robust (40V instead of 24V). And it gives slightly more power, could be helpful to drive my chinese 5V-coil-relays for testing...



Good point. I missed the port relay outputs completely on the todo list

Oops
No drawbacks, its just larger because of the low frequency

[uhi22]

The QCAs arrived. Thanks @projectgus for finding a reasonable priced source and organizing the shipping.
Although there is a small risk that they ship the wrong parts, everything went well. At least from outside they look original to me. The lower one is the original in the Ioniq CCM, the upper is the freshly arrived.

[uhi22]

And later a clear fail:

... and some rework ... and finally:

One surprise was, that the "pins" of the QCA do not like so much to be tinned. It needs an amount of flux and patience, and a hot iron, that they get tinned. I pre-tinned them before placing on the board, because I did not trust my soldering process.
This is how the "pins" look like. Left tinned, right untinned:

[catphish]

... and some rework ... and finally:

Great work, I hate prototyping with QFNs!
I guess the current board just exposes the SPI bus for testing. Did you have any more thoughts about what processor to pair it with? I'm a huge advocate of STM32, and I would imagine this could be a very near setup with just the QCA7005 and a STM32 to manage it and provide a CAN interface. However I guess it comes down to what you're comfortable porting the code to.

Thanks for putting so much effort into this!

[uhi22]

You convinced me, (and the ESP is crap ), I installed the CubeIDE, had a look to the different derivates of the STM32. I ended up with an nucleo F303RE for the first steps, and would plan the F103RE for the board. Does this make sense?

[uhi22]

I hate prototyping with QFNs!

I never did this before, and was quite insecure whether I have a chance to do this. But after watching some videos I understood it is easier then expected, when knowing the tricks. Flux is the key. And pre-heating the board to 100°C. And then hot air, and the part will float by itself to the correct position. Really amazing to watch. Ok, my first attempt went wrong, but nothing broken, just again flux, preheating, hot air, tip from the side, and it exactly jumped in the correct position. The next time will be faster I guess.

Ahh, and the 7005 is no real QFN, it is QFN with "wettable edges", this means we can inspect the solder connections. The real QFN is harder I guess, because you see nothing.

I guess the current board just exposes the SPI bus for testing.

Yes, exactly. The SPI, the interrupt pin (which maybe is not necessary) and the reset, and some test points to be able to program the Flash.

Something is wrong with this ugly board: No smoke up to now. It just talks to the ccs32berta. This went a little bit too smooth.

[projectgus]

Flux is the key.

Glad you got these OK, and powered up already! Good flux is extremely helpful, especially with these low volume parts sales that are often so-called 散新 "loose new" (ie production run leftovers). If they've been sitting around and not stored well, they can have a thin layer of oxidation on the pads that needs fluxing off.

If you're keen to do more QFN and you already have hot air, then I'd suggest looking into solder paste and stencils. At this point I'd much rather build a single board with QFN and quality fresh solder paste than a single through-hole board!

[catphish]

You convinced me, (and the ESP is crap ), I installed the CubeIDE, had a look to the different derivates of the STM32. I ended up with an nucleo F303RE for the first steps, and would plan the F103RE for the board. Does this make sense?

All STM32 chips are good. The important thing is just to ensure that it meets your requirements. With that said, some designs are much newer than others. The STM32F303 chip you're looking at seems to have everything needed but I suspect based on its clock speed, it's an older design.

Personally I use the STM32G4 series in all new designs. Compare the chip you're considering with the STM32G473RC.

[uhi22]

I'm going on with the STM32F103RE, this seems to have everything we need.

Regarding the contactor driver, seems I have too many options and too less knowledge, any ideas welcome. I see the following options:
(A) Just a logic-level or 12V digital logic output, which can control an external circuit.
(B) ULN2003 which provides some power, but maybe not sufficient for the real power contactors.
(C) A more powerful output, I guess also including shortcut detection / tolerance.
And these options could be combined with PWM, to reduce the dissipated power in contactors which do not have integrated economizers.

What would be a good starting point for the next prototype PCB?

[crasbe]

I think the ULN2003 is pretty "outdated" at this point in terms of power dissipation due to it's high internal resistance.
ST (and others) offer pretty awesome 3.3V Logic Level Power MOSFETs, which aren't toooooo expensive.

Some examples would be the STD120N4LF6, STD155N3LH6 and STD95N4LF3. All of them are capable to drive a contactor directly and won't become exceedingly hot when doing that.


AND you can combine that with option A if someone does not want/can not use the internal circuit.

[catphish]

I always drive contactors with a NCV8402. These are MOSFETs with added protection to allow them to drive an inductive load directly with no additional components. However, they are VERY slow to switch, so are likely not suitable if you want to do PWM.
For PWM (and with an appropriate flyback diode), pretty much any modern logic level MOSFET will work. I would likely use AO3400A, but only because it's in JLCPCB's basic parts list and provides a good resistance value at 3.3V gate drive.

[uhi22]

Very good. Datasheet found here: https://www.mouser.de/datasheet/2/308/1 ... 317670.pdf
- The NCV8402 can drive 2A (shortly more, 6.75A are mentioned as continuous current if the heat sink is good and 25°C ambient).
- It has an internal protection against overheating and is able to handle shortcut properly. Much better than a simple FET.
- Switching time is slow (30us on, 200us off). This forbids fast PWM, but I guess the contactors would also be happy with a 100Hz PWM, but would need an external free-wheeling diode over the contactor coil, to avoid heating the NCV with the 42V clamped free-wheeling voltage.

[muehlpower]

The BMW LIM could serve as a template for the range of functions. This would make it easier for many (including me) to upgrade from LIM to OI modem later.

RGB diode outputs, 2x contactor, control and feedback of the locking actuator, CAN interface, evaluation of CP and PP, switchable resistors for CP (2.7k, 1.3k).

Optional:
Analog input for HV voltage sensor at the charging port, temperature monitoring of the DC pins(PT1000), fuel flap switch, button to start/stop the charging process, HV contactor feedback.

I think you can do without the control of the central locking!

[uhi22]

Yes, this makes sense. To be as much as compatible with the LIM from interface point of view, and also to inherit some ideas from the internal implementation. Do we have the details already somewhere stored? Or somebody to have a detailled look and provide the information?

Edit: Found the pinning/wiring in the wiki. This leads to the next questions: Any idea which high-side-driver to use for the LED outputs? I was thinking of supporting the WS2812 RGB LED (which is 5V, serial controlled, full RGB in one part), but if we want to be LIM-like then we need three separate outputs to control each color with a dedicated pin.

[muehlpower]

this is the link to the wiki:
https://openinverter.org/wiki/BMW_I3_Fa ... LIM_Module.

if you need anything else, get in touch. I have one open on my desk and can also identify components if necessary.

[uhi22]

The high-side drivers for the LEDs, and the full bridge driver for the connector lock motor would be interesting. Although I think there may be better alternatives today, I guess the LIM design is ten years old.