battery charger topology

This is a good way to go. Since battery charging doesn't need ripple free current, you don't need the huge storage capacitors you would normally have to build into the design.

The down side of making a booster that produces a constant voltage is that you have to design for both the low line and high line cases. The high line case, sets the voltage you must boost to. The low line sets the current in the transistors.

At these high power levels, you may be better off letting the current peak mains voltage set the booster's output voltage.

I think they've set that point a little low. You can get some really amazing transistors these days. I would still question going over a few hundred Watts with just a simple booster design.

If you are going with a booster design, I would suggest that you go 3 phase or perhaps even 5 phase. More phases makes it much easier to keep the ripple current of the booster out of the input circuit. You are going to end up with some fairly large (physically) inductors in the design.

Within reason, changes in the mains frequency have no great effect. The lowest frequency is considered in the size of the storage capacitors,

At 24V the advantage of synchronous rectification is just about gone. If you do go this way, I suggest you put large schottky recifiers across the rectifier MOSFETs. This reduces the losses at the switch off point. Normally you either end up with the MOSFET conducting backwards briefly or its substrate diode takes a pulse of current. With the large schottky diodes, you can error in favor of turn the MOSFET off a bit too quickly and not have substrate recovery time causing losses.

You are just making a "variable power supply" and then varying it based on the battery.

Buy a face shield.

Thanks for your reply Moosefet,

Just to summarise what I think you said, and to clarify in my mind the topologies required

--- mains i/p

--- full wave rectification

--- res. caps to supply boost circuits

--- say 3 to 5 parallel active boost circuits to minimise ripple current and to perform PFC (with smallish res. caps)

--- I thought constant boost rail but you think variable

--- full H bridge circuit with transformer to step the voltage down, provide isolation (100kHz)

--- secondary side of transformer employing centre tap and two rectification devices (maybe synchronous rectification)

1) I am not sure what you are getting at, could you please explain. 2) don't you just design components for low line max o/p power, e.g. make sure boost choke doesn't saturate in this condition and also work out power losses, RMS currents throughout circuit to assist in component selection. Then let the control chip worry about keeping the voltage constant. I think i need to get under the hood of the control chip. 3) doesn't the boost rail need to be relatively constant to send to the full bridge converter? However I take your point about ripple current and battery's. 4) I have never used synchronous rectification in practice, but I thought as the high currents even with a shottkey power diode the losses would be big. Especially since in one design there is no fan. Why do you say at 24v the benefits are just about gone? 5) Squint before you turn it on and build a bunker more like!

Thanks again,

reggie.

Via EMI filtering

A bit more EMI filtering

No, the big capacitors are on the output of the PFC circuit. The bridge only has EMI capacitors on it.

Yes.

Yes, I don't think you should work hard making it constant.

100KHz may be a bit fast for the power level. Try to design the transformer and see what happens. Do you need a core you can't get?

Yes.

Don't forget the output side chokes. You are making a forward converter here.

Yes you work from the low line case to find the stress on the inductor. You also will find that the transistors are conducting their hardest in this case. Lowering the boost rail at low line can save a bit of power in the transistors.

Remember that RMS only works when things are resistive. On diodes it is closer to the average current that matters. On MOSFETs, the on voltage is very non-linear near their upper limit. Stay well away from that case.

Draw out the whole system assuming the control chip must be pure magic. Then think about what the control chip must do. Then find one or add circuits to one to do all the functions you need.

The batteries are not at the same voltage every time nor is the mains voltage. Why should the boost voltage be?

You have to apply a near perfect gate drive to the MOSFETs but there is a large S-D swing on the MOSFET. This makes the drive circuit harder to do. Schottky rectifiers only have a large forward drop if you don't use really huge ones. The MOSFETs will cost nearly what the really huge diodes do.

Squinting isn't nearly good enough. The can of a capacitor went past by ear so fast I didn't even know what happened. All I knew was that there was a loud bang and I could no longer see the circuit for the smoke. The insides were like a cloud of dust filling the whole room. They eventually settled to the floor and I could see that one capacitor was missing.

SU

er

ut

s s

Thanks again but i have more questions,

1) I take your point about EMI filtering, I cannot possibly describe everything in this forum format. Engineers couldn't be engineers without diagrams!!

2) My silly mistake about the res. cap. after the bridge rectifier; I know the boost circuit requires an un smoothed full wave rectified sine wave from the bridge rectifier so it can do its PFC thing and provide an input current waveform in phase with the voltage.

3)I have never used multiphase active boost circuits only single phase with parallel fets to reduce losses. What controllers would you recommend? Would you go for multi phase as apposed to multiple components single phase:

a) primarily to reduce ripple currents in the boost cap due to this high power specification; thus increase caps. lifetime. b) reduce emi due to reduced ripple currents in boost cap? c) distribute losses across many more components.

I know it depends on the application and there are many more things to take into account, but in this application what jumps out at you to go down the multi phase boost route?

4) You think variable boost rail. I take your point about battery's not needing constant V&I and tolerating high ripple currents, but as I don't know much about battery's, I am not sure what to design the PSUs output conditions to.

-For the full bridge o/p:

5) Should I design to deliver to the battery nominal 25A at 600+5%W thus 630/25=3D25.2 say 26V o/p voltage at min I/P V (thus max full bridge forward fet on time)?

I know it depends on battery charging time lets say (300AH/25A=3D12H charge cycle) ignoring the three stage charging for the moment.

6) With a variable boost voltage rail I am not sure what the max/min input voltage to the full bridge forward converter would be. Therefore I don't know the max volt seconds across the primary of the transformer thus I cannot work out the primary turns and thus the turns ratio.

Np=3D(Vl*dT/Ae*dB) =3D ((pri V ????- 2fet DS on drops)*Ton )/(Ae dB)

Vl =3D??? dB =3D set to say 250mT (using voltage feed forward ) Ae =3D (need to choose suitable core) Ton =3D set to max, at min pri voltage including dead time (also involved in core selection - size/frequency)

7) May be a silly question but if battery's don't mind high ripple currents and voltages then could I get away with a small output choke inductance and small output cap.? If this is the case then how would I implement the feedback loop? It would be all over the place. Sorry for all the questions, just I am not use to designing battery chargers.

8) I take your point about the cost, but I was more interested in losses and heat generation inside the box at these kind of output currents.

For a full bridge Duty Factor for each o/p diode=3D approx =3D (0.8T/2)/T =3D ton/T =3D 0.4 and for a flat top equivalent current pulse with its peak being the same as the o/p output current, the Iaverage =3D peak

*square root(DF) =3D 25* square root (0.4) =3D 15.8A say 16A this would give a loss in one diode of say (vf =3Dassume not checked =3D0.3)*(I diode av=3D16) =3D 4.8W if my maths is correct.

This doesn't seem too bad, but I would have to work out the total losses inside the box and see if I can maintain max component temperatures - de-rating values.

8) I think this is enough for now, but I am trying to figure out what conditions to set the individual topologies to and what output conditions are required for charging a huge 24v battery.

Thanks for your help, it is really making me think...

Reggie.

[....]

Further on EMI filtering:

The diodes in the input bridge produce RF. Their recovery time controls the frequency and the hardness of the recovery controls the power of this RF.

[....]

I would look at the Linear multiphase parts and if I couldn't find one to do the job, I'd look at the single phase controllers. If there was still no luck, I'd start thinking about doing my own design with, lets say, a PIC making the gate waveforms.

Also to reduce the ripple in the inductors to make the inductor design easier.

The ripple in the inductor is the EMI problem.

Yes to this too.

The power level suggested multiphase as it is a bit high for a non- multiphase booster is what suggested the direction. The nature of the load also sort of suggests it. The load can be fed ripple at 120Hz but you don't want 100KHz going out to it.

You need to read up on batteries a bit. You have to feed the battery the charging current at whatever voltage the battery needs for charging.

A nominal 24V lead acid battery will be about 20.5V when discharged as far as is ok to do. You need to deliver nearly full charging current in this case.

It may be as low as just a few volts and mostly recover to good operation if recharged promptly. You don't want to charge with as high of a current in this case.

Fully charged the battery will be about 29V You need to provide full charging current up to about here.

See above about voltages. Full blast needs to be just a bit above the

29V so that you can regulate up to there.

Other than that Mrs. Lincoln, how did you like the play.

s many more components.

----------------------------------------------------------------------------=

--------------- thanks again,

There will be so many high dv/dt points producing common mode currents and di/dt points producing diferential mode currents, that i am sure there will be emi problems all over the place.

---What about points 6,7&8?---

6) With a variable boost voltage rail I am not sure what the max/min input voltage to the full bridge forward converter would be. Therefore I don't know the max volt seconds across the primary of the transformer thus I cannot work out the primary turns and thus the turns ratio.

Np=3D(Vl*dT/Ae*dB) =3D ((pri V ????- 2fet DS on drops)*Ton )/(Ae dB)

Vl =3D??? dB =3D set to say 250mT (using voltage feed forward ) Ae =3D (need to choose suitable core) Ton =3D set to max, at min pri voltage including dead time (also involved in core selection - size/frequency)

7) May be a silly question but if battery's don't mind high ripple currents and voltages then could I get away with a small output choke inductance and small output cap.? If this is the case then how would I implement the feedback loop? It would be all over the place. Sorry for all the questions, just I am not use to designing battery chargers.

8) I take your point about the cost, but I was more interested in losses and heat generation inside the box at these kind of output currents.

For a full bridge Duty Factor for each o/p diode=3D approx =3D (0.8T/2)/T =3D ton/T =3D 0.4 and for a flat top equivalent current pulse with its peak being the same as the o/p output current, the Iaverage =3D peak

*square root(DF) =3D 25* square root (0.4) =3D 15.8A say 16A this would give a loss in one diode of say (vf =3Dassume not checked =3D0.3)*(I diode av=3D16) =3D 4.8W if my maths is correct.

This doesn't seem too bad, but I would have to work out the total losses inside the box and see if I can maintain max component temperatures - de-rating values.

Reggie.

[... switcher charger design ...]

You start by trying not to make any more than you have to and then you filter and filter and filter some more.

If you let the boost voltage decrease for a dead battery only, then it isn't too hard.

If you are letting it vary with the line voltage, then you have to design the ratio for the lowest mains and highest battery and the total turns (ei: saturation issues) for the highest input voltage. You need to select your core for the highest magetization it will face.

Remember that this is a forward converter so the voltage on the secondary needs to be more than the battery voltage.

See the above. You may be thinking about this from the wrong end. The secondard side is what I usually consider first. Assume the needed voltage appears by magic on the secondary windings. Count up the voltage losses on the secondary side working up from a 30V battery charging at full current.

This tells you the voltage on the secondary winding. In a nearly ideal transformer, the volts per turn of the secondary is the same as the volts per turn of the primary. You don't need to worry about the primary while you disqualify the cores that can't work. You just have to assume that the secondary only can have 1/2 the winding area and that the I^2R losses are doubled.

Once you have thrown out the cores that can't work, you can try to design the transformer in the first or second one that it will fit into.

This may seem backwards but there is no point in computing the ideal design if you can't get the core for it.

For reasonable values of "small" you can live with small output filters. The needed EMI filter is part of your output filter. I assume you are charging the battery inside a shielded box so any RF that gets to it goes straight to the radio the guy is listening to.

Remember that capacitors have ripple current ratings. Stay well short of the rating.

Carefully very carefully :)

Your feedback loop is partly a current measurement and partly a voltage measurement. In both cases a short term average of the small ripple will clean it up. A fast reaction to an overvoltage or overcurrent can bypass the filter and just allow room for the ripple.

Consider this:

D1 In ---+---->!------+------- Out ! ! ---/\\/\\-----+ R1 ! === C1 ! GND

It isn't a very good circuit but it makes the point. Assume that D1 is some sort of diode that has exactly 0.7V forward drop for any current. If IN is more than 0.7V above OUT, D1 conducts and pulls C1 up. If IN goes 0.7 above the regulation point, D1 causes the regulator to back off.

In real life, there will be some op-amps involved.

Remember that the battery has nearly zero impedance when it is charged. If you are watching the voltage on the battery, you will get less noise than a resistive load.

The current into the battery can be averaged to filter it. You don't need to react quickly except for in the case of faults. You regulation servo can have a gain crossover at 0.1Hz if needed because nothing interesting happens in less than a minute.

If cost is no object, go for it.

Use about 0.5. The diodes are highish voltage schottky. You have about 48V across them.

Also bolt the hot stuff to the case. This gets the heat to the outside quicker.

Hi Moosefet,

1) I have been doing some digging and have come across the term "follower boost" used to describe n On Semiconductor part NCP1653. It's a Fixed - Frequency, Continuous Conduction, Mode PFC Controller.

Is this the sort of thing that you were getting at with the input voltage following the output voltage? I suppose if you want only the PFC thing then why should you boost the voltage. I know you said multi phase and this is not a multi phase I was just referring to the follower boost concept. I have only ever seen boosts that well, boost a lot!

From the data sheet of the NCP1653

-------------------------------------------------------------------------------- Follower Boost Benefits The follower boost circuit offers an opportunity to reduce the output voltage Vout whenever the RMS input voltage Vac is lower or the power demand Pout is higher. Because of the step-up characteristics of boost converter, the output voltage Vout will always be higher than the input voltage Vin even though Vout is reduced in follower boost operation. As a result, the on time t1 is reduced. Reduction of on time makes the loss of the inductor and power MOSFET smaller. Hence, it allows cheaper cost in the inductor and power MOSFET or allows the circuit components to operate at a lower stress condition in most of the time.

----------------------------------------------------------------------------------

I know, it still has to boost because of the circuit configuration, but what would you set the amount of boost to be? Say if you had a

100Hz full wave rectified waveform with a peak of 90*square root (2) what would the output of the boost follower be then? Say 30V ontop of 90*1.414 = 157V?

2) Is that because you need to limit the current drawn from the PSU in this condition thus the voltage dips, I am not sure what you are getting at with this comment. 3) I agree, also max fet on time on the full bridge circuit. 4) Agree again!

5) Only by one diode drop, say 31V on the secondary winding and 30V output to a 24V battery; unless I am missing something? Maybe 2 diode drops if I include a blocking diode to stop the battery pushing current back into the PSU.

can't > work. You just have

6) This is not backward at all I think its clever! 7) Understand.

8) I figured that charging a battery over say a 12 hr period the feedback loop wouldn't need to react quickly. When someone turns on into a fully discharged battery the current will try to jump quickly, but I will employ a soft start to gradually build the current up over a few seconds. 9)I may be missing something obvious hear but where do you get the 48V from. Reverse recovery? When one diode is on the other is conducting in a centre taped secondary. In the dead band the output inductor tries to drag to maintain its current direction, thus reverses its voltage and both diodes conduct clamping the voltage at the diode choke junction to ground. I have never used centre taped secondary's, so please correct me if I gave the wrong idea. 10) I will learn more about battery's and design the output choke to limit the ripple current to the output capacitor and battery and choose the capacitor based on that ripple current and lifetime. 11) One last point on the power train, I think I am gaining a better understanding on what I need to do for the forward converter minus the points in this post, thanks for your help on that. I just need to understand how to approach the boost side now.

If as you suggested I use multiphase boost converters. If I use 4 converters I reckon they will each supply 833W/4=208W [(25A*30V)/0.9 =

833W total power from boost to forward]. Is this power output ok for a multi phase boost? Could you please point me in the direction of a suitable application note or datasheet for a multiphase controller. Thanks very much for your help so far I really appreciate it.