DIY Electric Car Forums banner

SiC LLC modular charger design

75K views 272 replies 20 participants last post by  Tony Bogs 
#1 · (Edited)
Inspired by PStechPaul's idea of a modular charger and the availability of SiC and LLC devices,
I've decided to go ahead with a SiC / LLC based modular charger design / DIY homebrew build.

Design choices:

  • A SiC PFC CCM boost input stage, 390V DC out.
  • 3350 W out, (230V, 16A AC in) per module.
  • LLC transformer topology for isolation and additional boost up to 580V (for HV battery packs). FIXED FREQUENCY. Boost: by transformer turns ratio only.
  • CC mode only, buck hysteresis current control
The measurement of the LLC inductance (36.9uH) is an important step in the design proces.
 
#5 · (Edited)
I've had a quick look at reflex charging.
First impresssion: great for NiMH (memory effect) and lead-acid (loosen phosphate deposits).
I'm designing a charger for LFP batteries (obviously the best DIY choice ;)).

The high LLC resonance (transformer) leakage inductance
(cause: large distance between primary and secondary for isolation)
means that the Q factor will be extremely high
when a battery is directly connected to the LLC output.
This presents a major challenge for the frequency setting. Not practical.
Solution: a hysteresis controlled SiC buck stage between output and LLC.
Results in a high Rac load for the LLC. (Q=SQRT(L/C)/Rac)

The Q factor still remains high for LLC (~7 at max load).
Allowing ZCS doubles the bandwidth for frequency setting (gain>0.95).
To prevent high losses, it's going to be an all SiC design.

LLC frequency: 135 kHz
 
#6 ·
Yeh i have seen and used reflex charging for lead a NiCad, but i have only seen a hand full of articles on using it on Li chemistry, one used a ZVS inverter and the reports were quite positive so i think its something that needs more research into, someone should make a small one for testing 18650s and compare it to cccv over a few hundred cycles to see what happens.
 
#7 ·
And at least a few hundred batteries for a statistically significant result.
The papers I've read on LFP charging indicate that CV charging is not a good idea, especially at high voltages.
The charging strategy I am going to use is almost reflex, CC mode. With one big difference: no discharge cycle (can't see any positive effects for LFP).
The rest cycle is very short, only there for measuring the float voltage to determine the end of charging (eleminating internal resistance related errors).
At what voltage? Again, statistics. To determine the voltage at which a high SoC can be reached with only a very little chance (the statistics part) of overcharging a cell.
Then, for that tiny chance of a failing cell, the xS imbalance detection method can be determined, weighing cost, ease of use, mass etc.

SiC in ZCS mode. I've done the math.
Based on 2.5A pk at turnon, ~10% of the maximum tank current , since the ZCS mode stays close to resonance where the tank is mainly resistive.
Cree SiC C2M0080120D mosfet: approximately 10W switching loss per mosfet, IXYS Si IXFH44N50 more than 40W.
Both at 135kHz with SiC Cree diode C3D20060D antiparallel.
 
#9 · (Edited)
I've found the name for the charging method I'm going to use. It's: step current.

This paper describes the method and the underlying runtime model:

Min Chen and G.A. Rincon-Mora, "Accurate electrical battery model
capable of predicting runtime and I-V performance", in IEEE
Transactions on Energy Conversion, vol. 21, no. 2, pp. 504-511, 2006

users.ece.gatech.edu/rincon-mora/publicat/jrnls/tec05_batt_mdl.pdf

First time I saw "step current" was in this paper (Danish Technical Uni):

http://orbit.dtu.dk/files/9852018/Paper_EV_charge.pdf

Figure 3 on page 3 shows the pulse step charge / discharge curves.
The increase in internal charge resistance is very clear @ 85%.
That's about where the charging should stop.
The exact point for a multicell pack depends on LFP performance statistics (distribution).
Haven't found a reliable source for it yet.
 
#11 · (Edited)
Maybe I wasn't clear enough, but the last layout is for the PFC.
Looks similar to buck with the mosfet/diode combo and L/C.
Buck stage is the last piece in the design for the power section. Hysteresis control.
No MCU PWM or timer channels involved. I want to keep the MCU section simple. It's a DIY forum. No z-domain in SW. More Arduino style.
Next in the design proces is input stage: filter / inrush limiter (R + relay)/ auxiliary supply.
The schematics are still changing. It's an iterative proces. But I've posted the basic topology of a module.

Here's the first sketch of the input stage of a module.
 
#12 · (Edited)
A little bit PFC EE. The Infineon ICE2PCS01G needs compensation on both the I and V loop.
The design is ready for testing. Examples of the I loop transfer function are in the attachments: SPICE small signal AC model and bode plot.
Phase margin is more than 52 degrees.

Schematic PFC
All component values are on the new layout that I'll post when all checks are done (first PCB is in the waste bin).
The schematic can be found in the design guide for the Infineon ICE2PCS01G.
http://www.infineon.com/dgdl/Design...d24f8&fileId=db3a304319c6f18c011a336c455809ca

Small differences:

  • input cap (2u dc-link on bridge output)
  • IXYS gate driver added with a resistor divider input for Cree SiC
 
#13 · (Edited)
The waste bin PFC PCB. Components are not soldered on. Size 100x160mm.
Rsense in parallel (should be in series), diode and mosfet pins too close to cooler, toroid mounting points way off. Pads too small.

The first design is for 200 to 260V AC in, 390VDC, 3350W out.
Two modules are probably needed for 110V, 60 Hz in, 390VDC, ~ 3250W out.
 
#14 ·
The output of the PFC will be connected to a LLC converter in a boost configuration, converting 390V DC non isolated to 510V DC isolated.

But for the test of the PFC I want to use cheap heaters. Two in series.
Since they are resistive, the transfer function of the PFC changes dramatically. The Infineon design guide assumes a constant power load (f.i. a buck stage).
If my math is right, the resistive load replaces a single pole at very low frequency (Hz) with a pole at the other end of the spectrum (MHz).
So for the test the V loop compensation has to be adapted.
A PFC needs a very small bandwidth. Upper limit should be well below the line frequency.
 
#15 · (Edited)
For a small scale test I'm going to use my 30W (60W pk) LLC protoboard, 325 to 450V in, 15V out, and an ordinary Li-ion 16850 notebook array.

In the picture: safety transformer (top), rectifier, protoboard, 25W Rload and the Li-ion batteries. The LLC transformer is on the opposite side of the board.

Goals:
- find out how LLC performs with a battery load
- is it possible to directly switch LLC on/off for buck hysteretic control
 
#16 ·
:D Great start (not really) of the 30W test: I connected the 18650 battery pack with polarity reversed to the output of the LLC.
Clearly, I've mixed up the usual wiring scheme when I built the 30 W LLC protoboard. Yellow = GND, black = +V.

Soon the smell of hot CuL wire filled the room. Seconds later the LLC controller protection kicked in.
No damage except for the transformer (shorted wiring) and some warmed up (empty) batteries.

PFC PCB. The second version is OK.
 
#18 · (Edited)
Next stage in the design: the mains (line) input module. On the PCB are:

  • EMI filter
  • bridge rectifier and cooler
  • auxiliary power supply (Recom RAC06 and Linear LT3082 LDO) with 24V and 20V outputs, one with on/off control
  • inrush current limiting resistor and bypass relay
  • high isolation optocoupler control (Fairchildsemi FOD8320, 10mm clearance & creepage) for compliance with safety regulations
  • auto reset overcurrent protection ( 2 x PTC) for inrush circuit and AC/DC converter Recom RAC06


What is further needed for the high P (3250W) test of the PFC module?

  • FANS for forced cooling
  • a dedicated 16A/230V AC outlet (is available)
  • plugs and wiring for the two 2kW heaters (done)
  • 2 mm CuL on the PFC toroid
Test is scheduled for this weekend. Assembly on saturday, test on sunday.
 
#19 · (Edited)
The FOD8320 optocouplers are not in yet, so I soldered on leftovers from a project in 2006 / 2007: Avago HCPL3120.
Bit messy, but it will do for testing.

I've modified the PFC PCB layout again cause there was a huge c*** up in the Rsense connections.

First test: repeat light bulb low power test with mains input PCB.

Relay and PFC controller switch OK. The scope clearly shows the differences between relay on/off , PFC on/off.
At low load the PFC is running in DCM.
SiC mosfet drain signal is very clean, no parasitic ringing at turn-off/turn-on. 396VDC DMM reading on the output.
The picture shows the MAINS INPUT PCB and PFC board. The PFC is upside down for better access to the SiC mosfet drain, exposing the Rsense fix (in green/yellow wire).
 
#21 · (Edited)
All OK at ~2500W (2 x 1250W) and ~3400W (both heaters at 2kW).
Well, all but the standard household mains switch that switched the heaters on.
The only part that is clearly underrated for 400VDC/9A.

CuL on the toroids is now 2mm. And that is about as thick as one can get it on.
Reduces the copper loss on the PFC toroid to ~7W in stead of ~30W at full load.
There is a slight increase in core loss, since 66 turns is the maximum that will fit (with silicone tubing for insulation).
Was 78 turns. So now 12 W core loss, worst case, = 2 W more.
I'll replace the mains switch with mosfets for further testing.

Soon: the resonant isolation LLC converter.
At the moment with an isolation distance of 3 mm (on the transformer).
 
#22 ·
I only just now found and caught up on this thread. Very good work and promising results! It may not be worthwhile for me to do much more on my concept if yours works as well as it seems to. I'd be interested in getting the boards and some of the components for my own use, possibly with some modifications. I still intend to build one or more electric tractors using lead-acid batteries of about 24-48 VDC, and will need a charger, although it would not need to be at this power level.

I would also like to explore the possibility of using a similar circuit to boost the 24-48 VDC to about 320 VDC (or even 640 VDC) for a VFD. As such, the output voltage would not need to be regulated, and just a simple 30:1 or 60:1 ratio would be fine.
 
#23 ·
Well, so far, so good. Glad you like it, but it's not there yet.

Lead acid and boosting sounds great for a light tractor.
I guess two 24V batteries, 200Ah, motor 20kW.
Speed is not important so even 10kW might do.
Seems feasible.

The LLC provides the high galvanic and CAPACITIVE isolation.
I'm pretty sure I can get the 10mm clearance and creepage distances.
And probably more.
The goal is a constant, load independent, LLC DC gain at series resonance under MCU control.
So effectively, the PFC acts as regulator.
 
#26 · (Edited)
The LLC (power stage) module has been assembled.
It is the key module, because it provides the necessary isolation for the simultaneous use of multiple power sources with different voltages.
Any voltage 90 – 260V AC 50/60Hz or DC 110 - 325V will do.
At high input voltages the modules can operate at full 3250W output,
but at lower voltages, for instance 110V, 60 Hz, power output has to be lowered.
The maximum input current for the shown modules is ~18A DC/ AC RMS.

For testing an ETD49 core is used.
The final version will have the 3.3 kW full power ETD 59 core with 10 mm clearance and creepage distances and a very low capacitive coupling.
It will be mounted perpendicular to the PCB.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top