New EV Charger Design - Modular

That original $200 charger thread was quite interesting. Seems like it grew from a simple design offered by Simon Rafferty in 2009 and then Valery used that as a basis for his project in early 2011.

It appears that this design has gone through innumerable changes and tweaks and was a learning experience for many people, but there are still some serious flaws (IMHO) and as I tried to investigate the hardware and firmware it became apparent that it was in need of a major makeover.

An isolated design is probably the way to go, and a PFC front end seems important also. But for now, I wanted to try a basic buck circuit with all the other circuitry that will probably be needed for any design. It may be better to use a single control board and then multiple power boards in parallel for higher power, although the control components are really not very expensive.

I like the idea of a transformer design where the inductive components are used to transfer power rather than store it and retrieve it as is the case with single inductor buck/boost/flyback designs. Once the needed voltage is available, then a rather simple buck circuit can be used as a current source for charging.

Thanks for the input. I'm not sure just where I will go with this project - I often take on something and get to a point where it is no longer interesting or challenging (or useful or profitable for me), but if there is enough interest and need from the EV community then I think this will be worthwhile, and the feedback will be an incentive to continue. I know that Valery is working on his own next generation isolated charger and has asked for people to join his design team, but I like to be "in charge" to a large extent. I am often frustrated with having other people make decisions that I don't agree with, but I'm always willing to discuss alternatives and make changes based on other people's knowledge and experience.

PS, the capacitor switching was an idea to put them in series for high voltage and parallel for low voltage. Similarly, a switched voltage doubler could be used. But the boost PFC may be even better.

I decided to try a simulation of a simple PFC type boost circuit, using a 220 uH coil and fairly small capacitors (100 uF and 300 uF), and I was able to achieve a 1 kW output at 97% efficiency and 67% PF. This circuit uses only a sampling resistor (0.1 ohm) and a single Op-Amp, and a BSC42DN25NS3 MOSFET (although a real circuit would need one with higher voltage rating):



I may actually build this circuit to see how it works. I may also try adding a secondary on the coil to see if the same design can be isolated. I already have a two-winding toroid inductor with 2x50 uH coils which is about 2" x 2" x 1" and cost only $1.25. If this works well I may buy a bunch of these (which are surplus). The same sort of coil at normal pricing would probably be $5-$10, although it would be easy enough to wind on a $2 core. All the components are probably no more than about $20 - not bad for 1 kW.

http://enginuitysystems.com/pix/240Sine-320DC_PFC_Boost.asc

Here is the isolated version. Simulation is very slow so I only display the first cycle of operation. Frequency is now 65 kHz with smaller inductors used as transformer, and other values were adjusted. I added an optoisolator and 300V of zeners to regulate the output. PF and efficiency seem not so good, but this is not an optimized design:



http://enginuitysystems.com/pix/240Sine-320DC_PFC_Boost_Isolated.asc.

It might be better to use a standard non-isolated PFC circuit with a dedicated IC for the DC front end, and then add isolation to the buck current/voltage regulator. Isolation will probably add some size, weight, and complexity, all things being equal, but if MOSFETs and higher frequency (50-100 kHz) can be used, size should actually shrink by a factor of 2 or 3.

The non-isolated boost circuit has the advantage that it it is essentially a pass-through when the input voltage peak approaches the output voltage, and it's only really working hard when it has to boost the lower voltage near the zero crossings. Larger front-end capacitors can reduce this at the expense of power factor.

[edit] I also ran the simulation at 100 VAC (142V peak) and it mostly holds about 280 VDC, but droops to about 160 V for about 4 mSec near the zero crossing. This is with only a 100 uF front end capacitor and 20 uF capacitor on the output. I used low values to speed the simulation somewhat. These designs use cycle-by-cycle current limiting on the inductor, which will prevent (or greatly reduce) the wasted energy and heating at saturation. Powdered iron cores are more forgiving due to their more random distributed gap, while ferrite is more consistent and predictable, but when it saturates, current rises very quickly unless the drive is immediately removed. The inductor current measurement also provides some current limiting, but this boost circuit only measures current in the MOSFET, and in general this topology cannot control the current once the input voltage exceeds the output.

Actually, the isolated version is protected because the MOSFET disconnects the inductor from ground and without any AC current, no energy passes through the transformer and the output drops to zero when the capacitors are discharged.

Here is the simulation with a 10 uF front end capacitor and 300 uF output, at 100 VAC:



The output is 792 watts with an input of 1133 watts for efficiency of 69.9%. 272 watts is lost in the MOSFET, which has 0.42 ohms RdsOn, so that can be improved greatly with a higher current device and a gate driver. It draws 14.7 amps at 100 volts or 1470 VA for a power factor of 77%. The output varies from 270 to 290 VDC. With a buck regulator, this ripple could easily be eliminated with a small amount of capacitance and a small inductor. The isolated PFC section can easily be modified to produce the highest nominal voltage needed for the battery pack, so the buck converter would only need to adjust it by a small amount, which will greatly reduce the size of the inductor and the power dissipation of the MOSFET or IGBT. Perhaps a 1 kW charger could be made small enough to fit in ones pocket, and maybe it could be called a "Pocket Kilowatt"!

I made a new simulation with a simple gate driver, but efficiency is still only 77%, and PF is 73%. But this is worst case, at 100 VAC input:



I made a new ASCII file as well:
http://enginuitysystems.com/pix/240Sine-320DC_PFC_Boost_Isolated.asc

I may use the Arduino, since it already has the libraries and functions to drive the existing display or the 1.8" module that is only about $5 and uses SPI. I won't be using the processor for any more than basic data acquisition, control, and reporting.

So the PFC section will be an autonomous unit which produces about 300-350 VDC over a nominal 100-250 VAC (or 140-300 VDC) input range, probably using an isolated boost flyback converter. and high frequency dedicated controller like the MC34262. It may have the ability to measure the input voltage and also be controllable to some extent. The same basic circuit may be used for my purposes of boosting 12, 24, 36, or 48 VDC battery packs to 120-600 VDC for DC link bus voltage for a VFD rated 208/240 or 480 VAC. It will also have an EVSE interface and primary connect/disconnect using a two pole contactor.

The charger section may be a simple buck converter that will monitor battery pack voltage and charging current, as well as temperature and BMS status, and perform charging according to a preset and programmable protocol or algorithm. The main processor will probably be located here, and it may communicate with a computer or tablet device via isolated USB or Bluetooth. If there will be a separate processor in the PFC unit, they may communicate using a serial connection over a digital isolator, or simple optocouplers.

Here, too, the PWM will be controlled using cycle-by-cycle sensing of the inductor current to avoid saturation problems and destructive currents when charging the output capacitor bank. But I don't think nearly so much capacitance will be needed. The current through the inductor (and into the battery pack) can be controlled so that it maintains continuous conduction with set-points perhaps 5% above and below nominal, and this small amount of ripple is insignificant. At the point of CV charging, the monitored current would just need to be set back to C/10 or whatever is required, and the output voltage can be monitored to assure that it is close to the ideal value.

I made a simulation using an LT1249 IC, with pretty good results:



This particular model uses a non-isolated output "OUT" as well as an isolated output "OUT2" with higher power. I got 0.92 PF with outputs of 37 and 369 watts or 406 watts with 112 VAC input and 192 volts on OUT2.

Here is the datasheet:
http://cds.linear.com/docs/en/datasheet/1249fbs.pdf

It's about $6 from DigiKey:
http://www.digikey.com/product-detail/en/LT1249CN8#PBF/LT1249CN8#PBF-ND/891322

I bought a couple of these similar devices that are less than $1:
http://www.digikey.com/product-detail/en/MC34262PG/MC34262PGOS-ND/1479304

I found that, for a 500 ohm load in place of the 100, the OUT2 voltage reached 400V and then throttled back to about 385, where with the 100 ohm load it rose to about 300. This would be OK for my purposes of generating about 300 VDC from a nominal 24 VDC battery pack. I made some other changes for this, where the primary of the transformer is 5 uH and the secondary is 200 uH, so I can get the boost with less stress on the components.



http://enginuitysystems.com/pix/PFC_1249_DCDC_24-240.asc

I am still exploring various approaches to these designs and there will probably be quite a few differences between the PFC front end and a high power charger, compared to my need for a DC-DC voltage booster. In general I feel more comfortable with a transformer design that just transfers energy from a low voltage DC source to a higher voltage output, using the turns ratio of the transformer and not the flyback effect.

My method for design and analysis is more intuitive than mathematical, and I am still trying to wrap my brain around some of the concepts. I'll try to explain my understanding of the difference between a transformer (energy transfer) design and one that uses an inductor for energy storage and release, like a flyback switcher.

For a transformer, I would select a core with rather high permeability and no gap, so that at the frequency being considered, relatively few turns would provide an inductance high enough to draw minimal current when unloaded. Thus, for a 24 volt input at 50 kHz, a 10 uH primary would present an impedance of 3.14 ohms and the applied voltage at 50% duty cycle would be about 12 VRMS, and magnetizing current would be about 4 amps. For a 1000 watt converter, this 48 VA is reasonable. As the load on the secondary increases, the current is reflected on the primary at quadrature (90 degree phase angle) and the magnetizing current becomes a small part of the total. The transformer will work until the ampere-turns create enough flux to saturate the material, at which point no more power can be transferred.

For an inductor-based topology, a material with low permeability may be used, or the effective permeability may be reduced by adding an air gap. In this case, you want to apply voltage and store energy in the magnetic material of the core, and then switch off the applied voltage and allow the stored energy to be released. In buck mode, the current increases when voltage is applied, and decreases when removed. This is ideal for a current source, and the output voltage will be lower than the input. For a boost topology, the input voltage is applied until a certain current (energy storage) is attained, and then the drive is removed. In this case the energy in the inductor will be applied to the load, and the voltage will be higher than the source.

The limitations of a single inductor buck or boost is that the inductor must "work harder" under conditions of large ratios of primary to secondary. To boost 24 VDC to 240 VDC, at 5 amps, the inductor needs to have 50 amps built up, and then the energy will be dumped into the load at a higher voltage and lower current. In continuous mode, the inductor is "recharged" before all of it energy has been transferred, so the ripple is less, but higher levels of DC are maintained. Using coupled inductors in a buck or boost can reduce the wide levels of current and voltage swing, and can provide isolation. But an isolated converter requires twice as much wire for the windings, so the size and weight are increased.

Feel free to critique and correct anything I have said here. This is just my understanding of the principles and there is obviously much more to it than my brief explanation.

Now that the 12kW charger thread is borked, I'll add a bit to this thread. Here is a simulation of my basic voltage-controlled current-regulated buck converter:



Pretty simple, really, and probably could benefit from some improvements. There seems to be an anomaly at lower drive levels, where an increased drive voltage actually decreases the output at 100 and 200 mV and then it rises as expected from 500 mV (1.77A) to 2.5 V (2.68A). This is into a 1 ohm load, and frequency is about 25 kHz.

With a 5 ohm load, the output increases for all values of input drive voltage, but there is a big jump from 500 mV (1.46A) to 1 V (1.92A), and at 2.5 V it only rises to 2.03A. This is largely due to the input voltage limitation where 12V into 5 ohms is 2.4 amps.

With 36 volts input, the output is non-linear until the control voltage is 1 V (4.14A), and then it has steps as expected to 2.5 V (4.68A). It also now operates at 46 kHz. With a 1 ohm load, the frequency drops to about 18 kHz and the output is 4.8A at 1V and 5.5A at 2.5V.

I need to analyze why this is so. The basic theory of operation is that the control voltage turns the op-amp U2 on, which applies voltage to the IGBT gate, and the positive feedback from the output through R10 sets the (+) input to Vctl+(Vout-Vctl)*(R11/(R10+R11)). If Vctl = 1, and the op-amp is rail-to-rail on a 5 VDC supply, the setpoint will be 1+4*0.068 = 1.274V. The (-) input is ten times the voltage on 0R2 sense resistor R2, so at a current of 1.36A it should equal the value at the (+) input and the output should turn off, which resets the value of the (+) input to Vctl*(R10/(R10+R11) = 1*0.931 = 0.931V, which corresponds to a current of 0.466A. Thus the average current should be about 0.91A.

I notice that I had changed the inductor value to 50 uH rather than 200 uH as I have in my hardware prototype. But for the purposes of explanation, at 36 volts applied, the current should rise at a rate of 36/50 = 0.72 A/uSec. The simulation shows current rise of about 0.54 A/uSec, due to the lower voltage because of the output being at about 5 volts which is 5 amps into the resistor load. The (+) input of the op-amp has a brief low setpoint of 940 mV and rises to 3.98V when the output goes high. This may be due to the action of the 22pf capacitor and the high values of the control and feedback resistors.

I'll run it again with different values.

OK, this seems more reasonable. With the previous values, the PWM duty cycle was very low and the frequency rather high so the IGBT was only turned on for a few microseconds, and also the amplified signal from the sense resistor was hitting the top rail. So let's try this:



The waveform is shown at 300 mV control signal. With the output off, U2(+) is 200 mV. This turns the output on which is actually 3.5V with a 5V rail and the optocoupler drive. So U2(+) becomes 0.3 + 3.2/3 = 1.37V. The simulation confirms these values. With the 0.02 ohm sense resistor this corresponds to a current of 6.85A, and the simulation shows the comparator changing state at inductor current of 7.04A. However, there is a 17 uSec delay before the IGBT actually switches off, and this is probably due to the relatively slow PC817D. So the inductor current actually overshoots to 9.58A. The output voltage is about 5 volts so the voltage applied to the inductor is about 31 volts, which for 200 uH is a rise of 0.155A/uSec. The simulator confirms this closely as 0.144.

At this point there are a few shenanigans from the demons in the op-amp, but it settles down. The energy stored in the inductor now is applied to the output capacitor and load resistor. The energy transferred is the peak of 0.5*200*(9.58^2-0.99^2)=9080uJ (uw-sec). With load power of 25W, This would take 365 uSec. The simulation shows this time as 215 uSec, which is explained by additional energy being stored in the capacitor.

This may or may not be a viable and practical circuit for a surge current limiter or buck mode battery charger, but it shows that the topology can be realized with simple analog components and controlled by an applied voltage level. It also illustrates the effect of seemingly minor items such as the 17 uSec delay and overshoot. Hopefully this may also provide some insight into the operation of the PWM buck converter as implemented in the EMW charger.

A major difference is that the current mode buck converter uses a lower frequency as the load voltage drops, resulting in lower switching losses. Of course, as the output power increases, the frequency also increases. For instance, with a 1 ohm load (about 5 amps and 25 watts) the frequency varies from 3.3 to 3.9 kHz with control voltage of 100 to 600 mV. With a 5 ohm load (about 5 amps and 125 watts), the frequency varies from 6.3 kHz to 5.6 kHz over the same control voltage range.

Here is the simulation showing the waveforms at control voltage of 600 mV:



The output power is 122 watts with input of 128 watts and efficiency of 95%. The MOSFET accounts for 1.3 watts and D1 contributes 4 watts.

If you want to play with the simulation: http://enginuitysystems.com/pix/Buck_Current_Limiter_1.asc

I have made a prototype of this circuit and it seems to work pretty well. I am using an Arduino Uno to generate drive voltages using PWM and an RC filter. Here are some data points:

Code:

12V input, load = DMM 10A
PWM  I(in)  I(out)
 0   0.01   0.00
10   0.43   1.77
20   1.09   2.61
30   1.58   3.70
40   2.43   4.65
50   3.49   5.62
60   4.26   4.76 (unstable)

24V input, load = DMM 10A
PWM  I(in)  I(out)
 0   0.05   0.01
10   0.26   1.80
20   0.41   2.75
30   0.81   3.70
40   1.22   4.67
50   1.68   5.54
60   2.18   6.33

24V input, load = 2 ohms
PWM  I(in)  V(out)  I(out)  W(in)  W(out)  Eff
 0   0.05   0.01    0.00    1.20   0.00    --
10   0.36   3.43    1.71    8.64   5.86   67.9%
20   0.72   5.22    2.61   17.28  13.62   78.9%
30   1.22   6.99    3.50   29.28  24.46   83.5%
40   1.85   8.68    4.34   44.40  37.67   84.8%
50   2.13  10.12    5.06   51.12  51.20   ?

Apparently the resistance of the load may have increased because I was dissipating about 25 watts in a 10 watt resistor (two in series, one 10W and one 25W). The 200 uH 10mOhm inductor got just barely warm, while the IGBT got fairly hot. At 10 amps, the inductor resistance accounts for about 1 watt. The IGBT is an STGB10H60DF which is rated 600V 10A, and has a voltage drop of about 1.5V at 10A, so that accounts for as much as 15 watts. The losses observed at PWM=40 are about 7 watts, and thus the IGBT is by far the limiting factor.

I had planned on using this for higher voltage input, and thus lower current. The prototype may not be safe at higher voltages, and in fact the output capacitor is only rated for 50V. It is a buck converter, so that might not be a problem. However, I might change the IGBT and instead use a MOSFET such as the HUF75645, of which I have quite a few, and it is rated at 100V, 75A, and 14 mOhms. With that I could make a higher current buck converter, with an output of perhaps 30 amps, at which point the inductor might exhibit about 9 watts of resistive losses and the MOSFET perhaps about 15 watts, while the output could be into a higher power 1 ohm load of 900 watts with an input of 48 volts and about 20 amps.

My power supply is only 5 amps (300 watts) so I will have to limit bench testing to that. I might be able to use a lower resistance load of about 0.3 ohms for 270 watts output (30 amps and 9 volts). But eventually I'd like to see if this inductor is capable of being used for a 1000 watt module. I don't have complete specs for it, except that it has two separate windings that in series measures 200 uH and 10 mOhms. It is about 1.75" OD and 0.70" thick (with windings), and the core is a light greenish-yellow, 40mm OD, 22mm ID, and 15mm thick. They are available from Surplus Shed for $1.25 each, so if this performs well, I may purchase a large quantity. It should be simple to add a microcontroller (Arduino or PIC) to perform the other functions of a battery charger, and perhaps the cost of a 1 kW version could be well under $100 or even $50. If they can be connected in parallel and driven from a single control board, it should be possible to make a modular charger as I would like to do.

One advantage of using a transformer is that the secondary can be wound specifically for the target voltage. The primary can stay the same, for a nominal 350 VDC from the PFC boost front end. Another possibility is to use dual windings for both primary and secondary. Thus for a 120V input the primaries can be in parallel, and switched to series for 240V. Similarly, the secondary could be parallel for up to 180V at 10A, or series for 360V at 5A.

A toroid transformer is fairly easy to wind, especially for lower voltages. I would expect a core capable of 2 kW would have about 5 volts/turn at 50 kHz, so 180V would be just 36 turns. It may even be best to use 6 output windings of 60V each that can be connected for 60, 120, 180, or 360V. This would cover most battery packs from 48V to 320V, and thus a 2 kW charger could provide 33 amps from an ordinary 120V outlet. The output voltage selection can be done with a matrix of slip-on terminals on the PCB. It might be necessary also to switch the filter capacitors, rectifiers, and chokes, but having six small circuits may be more cost-effective than one with larger components. It is also easier to make a PCB with multiple traces for 5 amps than one for 30 amps.

Once you have smooth DC from the input rectifiers, capacitors, and PFC circuitry, very little output smoothing will be required. The isolation transformer creates a square wave which, when rectified, has only some small high frequency components during the switching. Here is the rectified output of my DC-DC converter at 10 kHz with a 48W resistive load:



This is the prototype. It uses a ferrite E-core transformer with 1:10 ratio for a 10x voltage boost. The trace above is for 24 VDC input.



This design is probably good for at least 1000 watts, but I may need to use a higher frequency, and more or larger MOSFETs. Here is the basic circuit:



A simulation:



http://enginuitysystems.com/pix/Half_Bridge_24V-320V.asc

Note that the output inductor is only 20 uH. It may need to be larger for a buck current regulator. This is really a DC-DC voltage converter. But it will probably not need a very large inductor if the output voltage is fairly close to the raw DC, which can be matched pretty closely with the series/parallel arrangement of the transformer secondaries.

What you really need is the adjustable buck current charger, which I have also built and simulated:



http://enginuitysystems.com/pix/Buck_Current_Charger_36-24V.asc

Using a transformer, the primary can be matched to the PFC-processed line voltage, 180 VDC for a 120 VAC line, and 360 VDC for a 240 VAC line. It can use 200V capacitors in parallel or series for optimum utilization. A three phase source does not really need PFC and the capacitors can be much smaller. Voltages as high as 600-800 VDC might be handled, but then the protective components (fuses, circuit breakers, and relays) become more critical and expensive, and other components such as MOSFETs and even resistors will become critical. And wire insulation and PCB trace spacing will be an issue.

The secondary can then be designed to produce close to the needed maximum battery voltage, and the wire size will be able to handle the maximum current according to the power specification. A buck switching regulator with a single winding on a choke, with a working voltage of 350 VDC, needs to be able to handle 34 amps for 12 kW, and if you want to charge a 180V pack at 12 kW you need 67 amps. So the single winding inductor design limits the charging power for lower voltage packs.

[edit] I tightened up the hysteresis and now this circuit runs at 20 kHz (rather than 10) and has better current regulation. It just involved changing R11 from 50k to 10k:



The maximum here is 486 watts output and 558 watts input for 87% efficiency, which could likely be improved greatly by some optimization.

I had forgotten about the advantage of using three phase rectification to reduce the need for capacitance and PFC. So although three single phase charger power modules could be connected with inputs to each of three phases, they would need the usual capacitance and PFC. A three phase bridge rectifier will produce a DC voltage equal to the peak voltage of the L-L input, so a 3-phase 380 VAC source with 220V to neutral will produce about 536 VDC. This is dangerously close to the 600 VDC rating of common MOSFETs, IGBTs, and other components.

Here is a simulation, and it also shows what can happen when two capacitors of the same value but different parallel (leakage) resistances are used to obtain higher voltage rating. After one minute, the capacitor with 500k sees more than 80% of the full output voltage. In practice, the leakage may increase with voltage and thus balance the voltages, but it is much better to use balancing/bleeder resistors across each capacitor. This is IMHO another "potentially" serious problem with the new V14 EMW charger.



http://enginuitysystems.com/pix/3_phase_380_L-L_FWB.asc