2kW battery charger topology

Hello,

I want to build a 2kW battery charger for a solar system. I'll use a portable generator as a power source and a step-down/isolation LF transformer (former welding machine) to get around 30VAC.

As the load consists of a big battery bank (48VDC / 500A) I can't use an output voltage control thus I have to use a current control topology.

I was thinking of a regular CCM boost converter driven by a UC3845 chip. But, as the input/output current is quite big I prefer not to use a current sense resistor.

I've never used a current transformer before (hence I don't have one available) but I'm wondering if I could make a separate winding (few turns) around the input toroidal inductor to get a feedback voltage proportional to the inductor current.

I don't mind about output ripple and such, I just need a higher PFC to protect the gas generator.

Do you have any other suggestion?

I thought each panel would be only low power, and that you would just have a separate current regulated boost converter coming off each panel and pouring its current into the battery.
Or are you saying that there is an inverter already attached to the panels giving a 230vac output?
Because it seems a waste to have to go from the panel voltage, then thru an inverter to get 230vac, then thru the big iron step down transformer to get 30vac, then you are rectifying that and having to draw power from it with a high power factor to give you your current regulated output into the 48v battery.

Instead I would just come off the panel(s), through a booster(s), then to the battery.

I'm wondering if I could make a separate winding (few turns) around the input toroidal inductor to get a feedback voltage proportional to the inductor current.

Click to expand...

It's worth a try. I once wanted to detect a refrigerator turning on. I ran its juice through a single wire, and wound a few turns of it around a transformer (no particular kind). It was merely a makeshift, isolated hookup. By testing what came from all windings, I got sufficient power to light an led. Therefore the voltage is as high as 1.7V, and a few mA.

In the same way you might try winding a wire several times around a single conductor. It could work because it carries heavy AC current. It's a simple transformer. I don't know how well it will work.

you would have to wind an awful lot of secondary turns to get low enough current in the sense secondary winding.
And it just wouldnt be practical, -where woudl those turns go such that they were well coupled with the "primary", it would make a real mess of the power inductor.
You could do a separate unipolar current sense transformer for the boost fet current , and then say a hall sensor to regulate the output current.

BradtheRad said:

In the same way you might try winding a wire several times around a single conductor. It could work because it carries heavy AC current. It's a simple transformer. I don't know how well it will work.

Click to expand...

Many thanks for your confirmation, @BradtheRad! Actually, that's what I was looking for: a small voltage (low current) source for the feedback control.

The UC3845 current sense input has a voltage threshold of 1V thus I need to get a feedback voltage around this value for the maximum AC input current. I might wind few more turns to get an even greater voltage then use a voltage divider to "fine tune" the input current.

Anyway, I don't need a very accurate feedback voltage as I don't need a very precise charging current neither. I just want it to be pretty smooth as to not disturb the generator engine.

@treez:

I was talking about a back-up battery charger (when there's not enough sun, I have to use a gasoline generator to charge the batteries).

OK thanks, but all you need to do what you want to do is an off-the-shelf current sense transformer, and you can set it up in unipolar mode.
Another point is that this 2kw boost that you are doing is going to be power factor corrected i hope?.....as you know the rms input current from the generator to it would be horrendously high if this boost wasnt power factor corrected....and PFC boosts dont like CST's as they often have very high duty cycle where its difficult to reset the CST.

Or were you hoping to rectify and smooth the 30vac then use a simple boost to shovel charge into the battery from there?......that would mean a very high ripple stress on the smoothing caps for 2kw.

Above certain power levels, not having pfc becomes impractical aside from any regulations.....The PFC will bring down your RMS input current to manageable levels.

youll have difficulty finding a pfc chip that uses a CST input....youll prob have to roll your own with your own software PFC algorithm...virtually all OTS pfc chips use sense resistors....so what you would need to do is parallel enough pfc boosts so that sense resistors are practical in each of them.

and to reduce the current, get rid of the 230-30vac transformer, and do pfc from the 230vac onwards...then step it down tO the battery from there with a HF SMPS.

That's the whole point of this circuit: power factor correction.

By sensing the input current (hence the load current), a boost converter keeps the input inductor loaded with the preset current for the entire fully rectified input voltage range.

The UC3845's oscillator is starting the PWM signal at the beginning of every sampling cycle then the current sense input is terminating the PWM pulse if the monitored current reaches the preset value.

- - - Updated - - -

Btw, I'm not using any output caps; that big battery bank just don't care about a large current ripple.

Anyway, I'll choose a large enough inductor to smooth up the input current though, for the sake of that generator engine.

Right, so youve no big smoothing caps after the rectifier that follows the 30vac output LF trxformer.
And you dont have concern for sub harmonic oscillation either?
The UC3845 family can easily be converted to constant off time controllers to avoid subharmonic oscillation....and the decent off time you get with C.O.T garantees you reset your CST every cycle (if you are using a CST)
I think you will get very poor coupling if you try and jam a load of windings round the windings that are already going to be on the choke.
Youll definetely need an RC filter into the CS pin of uc3845.

By the way, the LT1245 is pin for pin compatible with uc3845, and is in the free ltspice simulatr, you could fairly exactly simulate your intended circuit to check it out first.

I will only run this charger for 1-2 hours per day (if I really have to).

As the big battery bank is the only load (and it has proved to tolerate a lot of stress) I have no real concerns about harmonics and stuff.

Till now, I was using that big transformer without any PFC at all (just a rectifier bridge). The batteries were happy enough, but the generator engine was screaming like hell (the load was surging higher currents for short periods only, at the sine wave peaks).

Anyway, I guess the circuit is just to simple - a classic among classics - to worth running a simulation. In fact, it's a lot easier to actually build and test it.

kathmandu said:

Hello,

I want to build a 2kW battery charger for a solar system. I'll use a portable generator as a power source and a step-down/isolation LF transformer (former welding machine) to get around 30VAC.

Click to expand...

Is this welding machine of the constant voltage (MIG) type, or the constant current (stick, TIG) type ?

It was just a dummy stick welder but I've completely removed the magnetic shunt (and I rewound the secondary) thus it's just a regular transformer now.

I know what you mean, the magnetic shunt and the transformer itself, with its lousy windings has been somehow helpful, but its PFC function was questionable.

It sounds like you are off to a pretty good start.

A bridge rectifier with a choke input filter will eliminate the very high current peaks, and present the portable generator with a quite acceptable current waveform.
That will get you some bulk dc to play with.

I used to design large industrial battery chargers, and I just had a look in my "little black book" design notes from many years ago.
For a 48v 40amp supply, the filter choke would be about 3.24 mH and the electrolytic 28,000uF to get 5% ripple at full output.
Not at all critical, but a pretty good starting point.

Whatever topology you finally decide on, two of them run interleaved at 1Kw each would be more practical, and offer less ripple at both input and output.

actually i forgot the uc3845 is 50% Duty cycle max so not so bad for subharmonic oscillation

I've never used a current transformer before (hence I don't have one available)

Click to expand...

For 40 amps, how about a Hall effect sensor ?

These can output a dc signal of several volts, and can be easily scaled to work directly into any type of PWM control chip.
The advantage is that the current monitoring is dc coupled has wide bandwidth, and is fully isolated, and you do not need to worry about noisy grounds or voltage drops.

I know, I've used Hall current sensors but as I don't need accurate current control I thought I could use a simple extra winding on the input inductor.

Btw, i came across this schematic:



What's the role of that extra diode? To by-pass the boost converter when the input voltage is greater than the output? Or it's for start-up purpose only?

Regarding the exact topology, I was thinking of the CCM one:



From the diagram above, looks like it needs a variable sampling frequency? I thought of using a constant sampling frequency (UC3845) and design the input inductor accordingly to get a CCM working mode.

yes the diode just stops LC ringing between the boost choke and output cap..from the input surge, as you know, in most boosters, the vout is usually zero to start with, but in your case, it would be the battery voltage.
Yes pfc can be done ccm with a constant frequency.
As you know, your pfc wont be perfect by allowing the uc3845 to switch away like discussed, but i suppose its not the end of the world.
As you know, constant frequency pfc chips like LT1248 have additional features to get you a really good power factor, but they need to be used with a sense resistor....

- - - Updated - - -

As you know in a proper pfc chip, the error amplifier voltage , and the signal indicating the instantaneous voltage of the mains, are fed into a multiplier block, and the output of this goes into a pwm comparator which also looks at the actual current flowing…the multiplier output thus provides a variable reference voltage into the pwm comparator as required…and the variability is sinusoidal due to the fact that one of its inputs was the (half) sinusoidal bus voltage.

You can see it in the LT1248 datasheet.

You obviously wont have this sinusoidal reference voltage into the uc3845 pwm comparator, because the uc3845 has no multiplier, but you will still get a reasonable power factor…better than before anyway.

Got it, it's a boost pre-charger diode:



- - - Updated - - -

Thanks, @treez. Then I'm going to use UC3845 (as I already have it). I'm just reading some app notes from On Semi/Infineon about PFC boost implementation and I hope to start building it in a very short time.

The attached LTspice simualtion of it dosnt run too badly.
Schem also attachd in pdf
As long as you do add in the features shown to prevent overcharging and accidental turn on when load disconnected.

Thank you very much for your time, @treez! I'll try to customize the component values to suit my needs. I hope I could run it at a higher frequency (100kHz) to minimize the inductor size.

Thanks for the protection tips, too. Anyway, I already have a dedicated circuit to monitor the batteries state. I set it up to disconnect the transformer from the mains if an overcharge occurs (then, after few minutes, it turns off the generator, too).

Btw, if I will not be satisfied with that extra winding for current feedback I'll use a Hall effect current sensor instead.

ok. try putting extra winding on the sim boost inductor and see what happens. There is a general rule with simulators and ccts like that....if you cant get it working on the sim, you will never get it working on the bench.
The sim will show you all those gotcha's that we all forget about.

i think Allegro do a good line in hall sensors at high current...you need high bandwidth as possible for monitoring peak current as i beleeve u r doing.........youre 100khz frequency of switching may make hall sensor not possible...it just wont adequately depict the current peaks at that frequency, you may need to come right down in frequency if you use a hall....but if you use CST, then no problem 100khz