New EV Charger Design - Modular

I use KiCAD extensively for all my personal projects and like it quite a bit. It would be a good option for an open-source project as it is both powerful, free, and open-source itself so everybody can use it.

On the controller, bare-metal firmware is something I've done extensively since the 1970's. For a project like this, the firmware probably won't be terribly large, so a small 8-bit micro would work well. I've done a lot recently with SiLabs c8051fxxx family. If we went more extensive in options and functionality, I would want to build in a Cortex M3 or M4, and there are some very nice options there.

I also prefer building in touch-screen rather than push-button. I built this one into my BMS controller : 2.8" TFT with Touch for about $8. I used a C8051F850 (80 cents) micro to control the touch panel and provide an i2c interface for the main controller to read touch events. I use a SPI interface to an SD/MMC for the graphics data.

I suppose a next step would be to start nailing down the desired feature set for the system as a whole and break it up into modules. Maybe open up a collaborative workspace? Certainly for holding the working schematics and code, I prefer BitBucket, but git-hub works too.

I am also interested in building a DIY charger and might participate in the design process. I've never designed power supplies, just some motor and valve controllers with MOSFETs.

I would be building an onboard charger for an electric motorcycle so the charger would have to be compact and preferably isolated. A PFC is probably a must since I would use it in our parking lot which has 230 V and 10 or 16 amp fuses and I want to get all the amps to good use. 2 kW sounds like a pretty good amount of power for one module, in my case it could be also a little smaller. For an electric motorcycle, three phase charging is not really an option.

I think there definitely should be a braind board separated from the power modules. I have experience with Atmel's Atmega series and for novice microcontroller users using Arduino IDE is probably a good option. There is also the 32-bit Arduino UNO option.

I will also cast a vote for KiCAD. I've used it for some years and to my understanding it is the most popular open source schematic/layout program. It is still under development and there is not a perfect version easily available currently. I would suggest using the latest build from kicad.nosoftware.cz, not the old stable build from a couple of years back. The developers are working to release a new stable build this year, but also the current development versions are good.

Thanks for your interest. I tried Kicad, but probably over 10 years ago, and did not find it suitable for my needs. But it has probably improved a lot since then. Since I have the PADS suite, and am familiar with it, I plan to use it for my development efforts, but these will not be very complex boards and it should be easy to transfer the design to another PCB package. Mostly all that would be needed is a netlist and an XYRS file that locates the components on the board. I have done some work with the PADS parts databases and there are VB scripts that can extract much of the information from the files. The PADS designs also can be exported into an ASCII format which potentially could be translated to Kicad or other design software. I even started working on that when I was making decisions about what to use for my own work.

It will be good to have a number of people willing to collaborate on the design, and I envision this taking two or three iterations to get it reasonably close to production level. I want to come up with a preliminary schematic and BOM to make a set of boards fairly soon, probably by some time in May. I have some other projects I am working on now, but while I'm recuperating from spine surgery April 16, I'll be on light duty and designing PCBs should meet that requirement better than the machining work I am involved in now.

Please stay in touch and feel free to offer suggestions in this thread. I will probably use a Microchip PIC because I am most familiar with them, but it should be no problem porting to an Arduino environment if desired.

I am thinking about designing a charger myself.

So far i got a few requirements for myself:
-PFC is a must have, dont want any circuit breakers going out when pulling max juice. (planned implementation with a self regulating IC)
-Transformer design based smps as main conversion stage. Isolation from wall is required and desired.
- Output voltage range: 200-500V roughly
- Max power of 3,5 KW per module (standaard european socket)

I was considering running a dedicated PWM SMPS IC to do the main task of controlling the switches and monitoring the output voltage/current. Then to augment these values with an Master mcu, this would make the Coding side of the project alot easier and less critical.

However running the switches using the mcu should also be possible however, the control loop design needs to be very robust to keep the currents under control.

Paul have you continued developing the charger/chargers or has the EMW charger taken all your valuable brain time?

My work on the EMW charger, along with other projects, takes up most of my time, but much of it is a parallel effort because of what I am learning, so if and when I can dedicate more time to the modular charger idea, it should go faster. Perhaps we can collaborate on this, although being across the pond complicates things somewhat.

If isolation is needed, (and I agree that it is) there is the question of where it should be placed. One concept is to make an isolated DC source of about 350 VDC from 120 or 240 VAC line, and then the buck charger can be a simple non-isolated design. But for lower voltage battery packs converting from 350 VDC to, say, 60 VDC at 60 amps (3600W) involves rather high current peaks. It may be better to design the output using a transformer which can be wound with appropriate primary and secondary conductors for the intended output.

Thus a 3600 watt charger would have a primary of 350V and about 10 amps, while the secondary could be 60V and 60A. A 360V battery pack would require 6 modules and could provide about 20 kW.

It would be easy enough also to use a dual secondary for an alternate choice of 120V and 30A, and 3 modules would provide 10 kW for a 360V pack.

I have considered such a topology and I think I did a simulation for it at one time, but there may be a problem with the choice of the output inductor when used as a transformer. I found the following application note on the Flybuck topology and it is described as being useful for low power. It also has relatively poor efficiency (about 70-80%):

http://www.ti.com/lit/an/snva674b/snva674b.pdf

This circuit has a main output which is non-isolated, and drives a capacitor and voltage divider. The waveforms show it operating in discontinuous mode, which is probably necessary because there is no DC current path unless there is a load on the output. And even with that, DC cannot be passed through a transformer, so the primary and secondary must have a net zero DC voltage. The current, however, can be non-zero but it may lower the inductance. (At least, that is how I understand the operation of inductors, coupled or not).

The circuit in the link you provided is interesting, and obviously works, but it is not really isolated, and the secondary is only used for a low power control supply.

I think perhaps it may be better to use a transformer to produce a DC voltage proportional to the input voltage (which is nominally 350 VDC from the PFC stage), and then use another inductor for the buck current regulator. If the raw output is closely matched to the peak output voltage needed, the buck regulator will not need to handle much power and can be made quite small and efficient. It should also have little ripple at high current since it will be operating in continuous mode and the inductor will act as a ripple filter, whereas at lower currents and smaller duty cycle it may enter discontinuous mode with higher voltage ripple, but at that point the output capacitors can be used to better advantage.

These are my thoughts at present, but may change when I do more simulations or actually build a prototype. Matching the transformer output voltage to the battery pack may not be as difficult as it may seem. A high current high frequency transformer is best wound with multiple parallel turns of finer gauge wire, and four or even eight windings are not uncommon. Often they are connected in parallel right out of the transformer on the PCB, but they could each connect to their own rectifier and filter, which then could be connected in series or parallel to get the needed voltage or current. This may make it more difficult to market as a "universal charger", but for a DIY product it is not a big deal to require the setting of some jumpers, which may be soldered or use QC connectors for easy field changes. These chargers are usually mounted in the vehicle and married to the battery pack, so unless the pack is changed, there is no need to change the voltage, at least not by very much.

I just bough a teensy 3.1,

This thing is powerful guys any other arduino got nothing on this simple break out board.

http://www.pjrc.com/teensy/teensy31.html

But even beter is the chip itself. Dead time injection hard coded in registers This thing is going to keep me busy alot. Its 3,3 volts so compatible with Johannes inverter board if I want to try my hand at driving some motors. Also the system clock is crazy high compared to other "hobby" micros.

http://www.pjrc.com/teensy/K20P64M72SF1RM.pdf

Now i just need to finish off a schematic of a isolated DC/DC to test with.

Even some adavanced capturing of incoming channels with pwm signals on them. Duty cycle, period, Quadrature decoding ect.

Nice work on the EMW upgrade, Paul.

I'd like to introduce SiC and LLC resonant in my suggestions for a modular charger.
I need off line power for my EV diy projects, so I'll start a 3680 W 100kHz SiC boost (PFC) build as soon as possible.
That will get me up to 10kW from a standard european (3phase) socket.

I certainly expect that I'll be needing a flexible charger in couple of years once prices for batteries have plummeted.
Without batteries I have to use the power grid to supply my EV homebrew projects.
The stages of the grid supply are very similar to the charger: PFC and isolation.

PRIMARY STAGE 90 to 250V AC, 50/60 Hz, 3680W:
100 kHz SiC PFC boost, SiC does not suffer from high hard switching losses, so CCM is possible.
CCM requires a smaller input EMI filter than other methods.
A huge reduction in boost inductor size (volume) can be achieved as part of an overall reduction of volume and cost.
And there are more advantages. Cree has published a performance comparison, 10kW DC/DC, 100kHz SiC vs. 20kHz genIII IGBT :
http://www.cree.com/~/media/Files/Cree/Power/Articles and Papers/Power_Article_4.pdf

Boost inductor: iron powder, low cost, low inductance at high amps, high inductance at low amps, should be able to maintain CCM at all loads.
For 3680W I've selected: Amidon T-300A-26 toroid, app. 3 inch width, 1 inch height, 2 mm diameter wire (maybe Litz).
Apparently a popular toroid. Delivery will take almost three months. In the mean time I'll use a bigger toroid: T-400A-26.

INTERMEDIATE STAGE INTERLEAVING
Paralleling PFC stages with interleaved PWM reduces the size of the intermediate capacitor bank.

INTERMEDIATE STAGE ISOLATION
Second stage LLC isolation DC/DC converter:

• ZVS mosfet switching on the primary side, ZCS rectifier on the secondary side: low losses, SiC diode bridge on secondary side (single secondary winding, low conduction losses, reduced transformer size, simple design)
• capable of handling large input voltage variation on the primary side, mitigates PFC ripple requirements
• extreme high isolation: primary and secondary windings can be placed at great distance from each other, even on seperate legs of the transformer.
  The low coupling results in a leakage inductance that is an integral L part of the LLC topology.

In a LLC current mode charger there's no need for buck output stages. LLC outputs can be directly paralleled. The MCU sets the operating point for hysteresis control (basically the same priciple as the EMW LM211 bang-bang) Only one current sensor is required on the output.

System concept (as Tomdb suggested): use dedicated hardware for PFC and LLC. My choice for LLC: Fairchild FAN 7631, cheap and many (necessary) features.

For LLC, open loop (fixed operating point, hysteresis) under MCU monitoring for current sourcing should be OK.
Long periods of operation in the ZCS region are prevented.
Constant volt charging is not possible (as I mentioned, but maybe not explicitly).

Based on Hongmei Wan's (2012, Virginia Polytechnic Institute and State University) master thesis:
High Efficiency DC-DC Converter for EV
Battery Charger Using Hybrid Resonant and
PWM Technique

scholar.lib.vt.edu/theses/available/etd-05072012-141855/unrestricted/Wan_HM_T_2012.pdf

A very practical approach, the resonant part is a LLC at fixed operating point.

To substantiate my suggestions I'll soon upload a first design for a LLC setup, 230VAC input.
Controllers: Fairchildsemi FAN7631(LLC) and Infineon ICE2PCS01G (CCM PFC).
I'll build it and test it anyway. I need to power my inverter projects off line.
The LLC EPCOS N87 ETD49 core I'll be using is a leftover from earlier full bridge converter tests in 2007.
Up to 1 kW should be OK, since 2.2kW is the maximum power level for listed topologies in the EPCOS ferrite application manual.
Experience with low power LLC tests (SiC gate driver and 30W DC/DC converter projects) indicates that the combination of LLC resonant, low flux density, and forced air cooling might take it up too higher power levels. Otherwise it's a good opportunity to test LLC in parallel.
Derating at higher ambient temperatures (above 20 degrees C ambient) certainly is an option for my lab tests, but also for charging Li-ion batteries to prevent excessive degradation.
Actively cooled battery packs are not that common in the DIY scene AFAIK.
I've got two ETD49 core sets for testing LLC in parallel.
I've placed an order for the rest of the special parts.
Delivery may take several weeks.

Great umich article, could save a lot of time, looks like CV charging is possible with LLC.
First conclusion: keep the fixed operation clear of the ZVS/ZCS transition region or use SiC mosfets and diodes (no potentially deadly high ZCS losses)

eem2am wrote:

http://www-personal.umd.umich.edu/~c....IET_final.pdf

"Optimal design of line level control resonant
converters in plug-in hybrid electric vehicle​

battery chargers"
ietdl.org

Click to expand...

Thanks for the link.
Main objective of the authors is improving efficiency by preventing ZCS (in CC mode) and high resonant currents in the primary circuits (in CV mode, ZVS, light load).
A great paper for a conventional mosfet LLC charger that stays powered for long periods of time (to keep batteries full at a constant voltage).

Not relevant for my charger design suggestions though since:

• there's no CV charging, the charger power circuits are turned on/off under MCU control between high levels of charge in CC mode (say 80 to 82%, hysteresis control, to extend battery life)
• the LLC runs in buck mode above the resonant frequency. The ZCS region is below the resonant frequency.

I'll keep the IXYS mosfets, but I've placed a comment in the schematic about replacing them with SiC for buck/boost operation. To prevent mishaps, it might be good idea the use SiC in all cases.

There is much good information in the papers presented here. The designs are too complex and critical for my purposes at this time, but if anyone wants to contribute such a design to my overall concept, I would welcome that. My main goal is to make a modular design consisting of low power (1-3 kW or so) that can be connected in series or parallel to obtain the desired output for EV battery charging (or perhaps other purposes). The topology of the charger modules is irrelevant to the overall design, but certainly important to understand, discuss, and implement.

At this point I am still working on the modifications to the EMW charger, trying to keep the same overall power design but using new control and driver circuitry incorporated in a single PCB. There will also be circuitry to deal with inrush current to the capacitor bank, and a self-discharge circuit to bleed off the charge in a reasonable time period.

Thanks for the discussion of possibilities that may enhance the efficiency and reduce size and cost.

Thanks Paul, the single schematic of the EMW charger you posted made the internal workings much clearer.
For instance the LM211 circuit: hysteresis (also called bang-bang) control.

LLC:
First of all, ZCS can not blow up mosfets if a controller is used with a correctly set up overcurrent protection.
And again, SiC mosfets/diodes do not suffer from ZCS losses, so efficiency is not significantly worse compared to ZVS.

The umich (ietd.org) paper assumes a feedback controlled LLC with variable frequency. I'd say: forget that for a charger.
If anyone likes a challenge, then the conventional feedback, buck/boost topology in the umich paper is the way to go.

Enter Hongmei Wan and MCU control.
Hongmei Wan introduces a fixed operation frequency point in the master thesis: at resonance (better: just above, guaranteed ZVS),
where the DC gain is constant (almost independent of the load) and the efficiency is at maximum.
Combine this with MCU control, where the MCU checks and sets the optimal operating frequency at regular intervals (timer ticks), always approaching from above resonance.
This way the LLC can never enter the ZCS region (is below resonance).

So how about V/I control, if the frequency stays constant?
Hongwei Wan integrates LLC with another topology to enable V/I control. That is indeed complicated and tricky.

Instead, a buck hysteresis controller, like the EMW LM211 controller, can be used to turn the LLC on/off (pulse skipping method).
If that doesn't sound trustworthy, what does?

Pulse skipping is a standard feature on the Fairchild FAN7631 advanced LLC controller chip.