1kW Flyback converter for battery charging

[Ravi_H]

Please suggest me the techniques for voltage and current control of battery (48 V, 120Ah) at load, using a 360 to 500 V DC input and 57 V, 12 A output.

 

[pradhan.rachit]

A 1 kW flyback converter? This is too much power to be handled by it. Generally flyback converters are used to handle loads < 150 W.
A Full Bridge/ Push-Pull Converter seems to be best suited for your application and power delivery.

 

[treez]

For V and I control, have a sensor for current and voltage (eg pot divider for voltage, hall sensor for current)……also have an error amplifier for each of V and I control……..feed the sensed signal and the reference into the respective error amp and control like that. You can make the references variable, and then when you want to control only the current say , you can adjust the voltage reference so that the voltage control error amplifier doesn’t play a part.

So as youd expect, you have external opamp based error amplifiers which you use to do the control for you

 

[Warpspeed]

Flyback will be fine for this.

The only disadvantage of flyback at high power is the very high peak current.
At over 360v dc input, high peak current is not going to be a problem.
If you were trying to do this the other way around, convert 48v to 500v that would present much more of a practical difficulty for a flyback.

Output is 57v at 12A or 684 watts.
Input 360v at about roughly 2A average maximum.
Peak current in primary will be about four times that, say 9 Amps.
That's not too bad at all, and quite doable.

I would suggest the half bridge diagonal topology operating in discontinuous current mode. That has absolutely the fewest problems in stabilising the feedback control loop.

 

[Ravi_H]

Thanx pradhan.rachit , treez, Warpspeed for your valuable suggestions.

My major problem with Battery charging is that, when i charge the battery using constant current of say 'X' (A) from converter and after reaching the output voltage of say 'Y' (V), I want to hold that voltage. But when I hold the voltage at 'Y' (V), I want the current to decrease from 'X' (A), In my case the current takes a shoot i.e. increases beyond 'X' (A) and then start to decrease. I'm using PI control separately for voltage and current, without any inner loop.

Please guide me on this problem

 

[treez]

I would suggest the half bridge diagonal topology

Do you mean the two transistor forward converter?, which I have indeed heard called "diagonal" in the past

- - - Updated - - -

As you know, if comparing say a ccm flyback with a hard switched ccm bridge type topology, the peak current problem for the flyback is in the secondary not the primary. The bridge has an output inductor, which flows current continuously, so the secondary peak current for bridge vs flyback is lower (lower peak I for bridge smps).
As regards the primary, whatever bridge type smps you are comparing the flyback to, you can simply adjust the flyback Lp and Ls and duty to give you exactly the same primary-side current input waveform (trapezoid) for a ccm flyback as any ccm bridge smps

 

[FvM]

Usual constant I/V controller topologies would avoid overshoot. Either separate controllers with ORed controller output. Or cascaded structure, inner current, outer voltage controller with V-controller is pulling down the I setpoint.

 

[treez]

You are saying that when you switch over to voltage output control, the current suddenly spikes upwards then comes back down again.

I would look at the way that you change over from current control to voltage control…for example, if just before you switch to voltage output control, the voltage error amplifier is railed high, then indeed you will see that spike of current upwards,…..so what you need to do is soft start the voltage reference value, so that you get a soft start of the voltage output regulation.

 

[Warpspeed]

Do you mean the two transistor forward converter?, which I have indeed heard called "diagonal" in the past

Diagonal topology can be run either either forward or flyback.
Note the phasing of the secondary and direction of the output rectifier diode.

the peak current problem for the flyback is in the secondary not the primary. The bridge has an output inductor, which flows current continuously, so the secondary peak current for bridge vs flyback is lower (lower peak I for bridge smps).

Its far easier to find some very high current shottky diodes for the secondary than switch a similar very high peak current in the primary.
That is why high power flybacks are much more practical converting voltages down, than converting voltages up.

While I agree with you that a forward converter has far lower peak current than a discontinuous flyback, it also requires a filter choke and the result can be something that can be far more complex to compensate. Poor transient response and instability can often plague forward converters, and can often be very stubborn to fix.

The combination of current mode control, flyback, and discontinuous operation is pretty vice free, and always far easier to get working properly at the first attempt.

There are the definite disadvantages from the very high peak currents, no argument there.
But that problem is wholly dependant on the input and output voltages and power level. The very high peak current issue can range from trivial to horrendous.
In this case I see it as not being an obstacle.

 

[treez]

Thanks, your schem in the above post #9 is sometimes referred to as the 2 transistor flyback, as you know.
I am sure you will know that there is some situation with it, that has to be watched for, that is , something to do with if the referred voltage secondary to primary, is greater than vin, then it can go bang. Obviously, this is something that can be dealt with by a vin undervoltage shutdown circuit .

But yes I like the 2 transistor flyback, as it has less switching losses per fet compared to the high concentration of switching loss in a single transistor flyback. Also, it solves the RCD clamp dissipation issue which is bad in the single transistor flyback.

even better for the 2T flyback is that the high side drive can be a warts and all type , like the one in this thread...
https://www.edaboard.com/threads/348413/
...because as you know, even if the upper fet does linger on too long, it doesn't matter because the lower one is off.....and the "lower fet in the same leg" is not even a fet, its a diode, as your schem shows.

 

[Warpspeed]

, something to do with if the referred voltage secondary to primary, is greater than vin, then it can go bang.

Never heard of that before, or ever seen it happen.

If its running at very heavy load, and the load falls off, no damage will occur, as all the stored flyback energy is returned to the input via the two clamping diodes.
Its a very trouble free, bullet proof circuit.

In fact you can run it at full maximum duty cycle with NO load and it will run fine like that continuously.

 

[treez]

I am sure you don't have the problem because you correctly set the turns ratio so that voltage referred from secondary to primary , is less than vin. If some one did not do this, then it would go bang............but of course, if some unusual thing happened such that vin suddenly dropped, then it could happen then.

If you have the book “switch mode power supplies “ by Basso”, the situation of the potential damage spoken of here with the 2 switch flyback is covered on page 615 of Basso’s book……
“yes you have guessed it, if you reflect more voltage than the input voltage then your colleagues are going to applaud at the first power on”

 

[Warpspeed]

I have quite a few switching power supply books, but not Basso.

I have also used this diagonal half bridge topology without any secondary winding at all for several rather unusual applications, and nothing has ever gone bang.

It can NEVER reflect more voltage than the input voltage because its clamped by the two diodes. All energy is recirculated back into the main dc bus capacitor.

This voltage clamping and energy recycling feature is extremely useful for high current inductor testing, and battery desulphation to name two off beat applications.

 

[treez]

It can NEVER reflect more voltage than the input voltage because its clamped by the two diodes

yes this is indeed the problem, you can't clamp two different voltage sources to each other. At least , not recommended to try to due to it might smoke

 

[Warpspeed]

That's total nonsense.

The flyback energy will distribute very well without blowing anything up.
It can feed both the bulk input capacitor (via the clamp diodes) or the output capacitor (via the rectifier).

Flybacks with multiple secondary windings are extremely common, and the cross regulation is pretty good too.

The good Mr Basso is barking up the wrong tree with that one.

- - - Updated - - -

My major problem with Battery charging is that, when i charge the battery using constant current of say 'X' (A) from converter and after reaching the output voltage of say 'Y' (V), I want to hold that voltage. But when I hold the voltage at 'Y' (V), I want the current to decrease from 'X' (A), In my case the current takes a shoot i.e. increases beyond 'X' (A) and then start to decrease. I'm using PI control separately for voltage and current, without any inner loop.

This sounds like a classic controller wind-up problem, its pretty typical behaviour when one control system takes over from another control system.
Its common at the end of soft start to see a voltage spike, and see a current spike when constant current control takes over from constant voltage control, or the reverse.

Consider a standard PI loop, when it is in command of the the output, it will be operating happily within the active part of some limited control range.

If its not in control of the output, (the output being controlled by something else) the output will go to some extreme value and lock up at the limit of its control range.

For instance, when the current control loop is in command providing constant current output, the output voltage will be either above or below the voltage set point. The voltage set point has no relevance in constant current mode.

If the output voltage happens to be below the voltage set point, the voltage PID will ramp up to absolute maximum, but nothing happens because we are in constant current mode.

When you switch over to constant voltage mode, the voltage PID instantly takes over from the current PID. When that happens, the output immediately spikes to full absolute maximum, until the voltage PID can pull it back to the control point.

So seeing massive spikes and discontinuities when one control system takes over from another control system, is a fairly common problem.

Its not easy to fix, but the first thing you should look for is try to keep the amplitudes of the PID output within the normal control range.

Suppose some mythical PWM modulator uses a zero to one volt ramp and a voltage comparator.
The dc control input from 0 to 100% will require an 0v to +1v control output from the PID loop.

But suppose our PID is powered from +/- 15v supplies and can swing over the whole -15v to +15v range.

When we switch over to this PID the output (from something else) it may be locked solid at either the -15v or +15v extreme limit.
The system output will immediately go to full 0% or 100% and sit there for some time until the PID can slowly recover back into its normal control range.

What it may need is a voltage divider between the PID output and the ramp comparator to limit the +15v to -15v output to 0v to +1v. It will probably still spike a bit, but recovery will be much faster.

Another "trick" for very smooth changeover is to slowly ramp the input references to the PID's not just suddenly switch between the outputs.

By very slowly ramping the voltage reference, and current reference you can get a smooth changeover.

 

[mtwieg]

While I agree with you that a forward converter has far lower peak current than a discontinuous flyback, it also requires a filter choke and the result can be something that can be far more complex to compensate. Poor transient response and instability can often plague forward converters, and can often be very stubborn to fix.

In CCM both the flyback and forward have a second order response, with the flyback also having that nasty RHP zero. From a control standpoint a buck derived converter like the forward converter is generally the easiest to deal with.

- - - Updated - - -

That's total nonsense.

The flyback energy will distribute very well without blowing anything up.
It can feed both the bulk input capacitor (via the clamp diodes) or the output capacitor (via the rectifier).

Flybacks with multiple secondary windings are extremely common, and the cross regulation is pretty good too.

The good Mr Basso is barking up the wrong tree with that one.

I don't understand what failure mechanism Basso is referring to there, having the figure he refers to might help.

But keep in mind the diagonal flyback will fail if duty cycle is pushed past 50% due to saturation, similar to the forward converter. If the controller doesn't automatically limit duty cycle then a drop in Vin would definitely be a problem.

 

[treez]

top of page 7 of this is talking about this particular failure mechanism of a 2 switch flyback...ti.com don't call it a failure , but just say that transformer operation is interfered with.....
https://www.ti.com/lit/an/snva716/snva716.pdf

post#5 of the below gives FvM's opinion on it, 8 years ago!...
https://www.edaboard.com/threads/136067/


the below document , however, makes no mention of the problem of reflecting more voltage than vin from sec to pri in a 2 switch flyback...probably because most flybacks are offline, and step the voltage down, and have turns ratios which just never reflect more voltage back than vin.......but then again, what about those freak incidents where vin suddenly collapses.
https://www.eetimes.com/document.asp?doc_id=1273232

 

[Warpspeed]

In CCM both the flyback and forward have a second order response, with the flyback also having that nasty RHP zero. From a control standpoint a buck derived converter like the forward converter is generally the easiest to deal with.

All very true.
But I suggested flyback, with discontinuous operation, and there is no reason to exceed 50% duty cycle with a suitable transformer turns ratio.
This avoids ALL of the above problems of RHP zero and second order responses.

But keep in mind the diagonal flyback will fail if duty cycle is pushed past 50% due to saturation, similar to the forward converter. If the controller doesn't automatically limit duty cycle then a drop in Vin would definitely be a problem.

Just about any topology falls on its face if you hit the saturation wall.

No reason at all to go past 50%. Just add a few extra secondary turns to ensure flyback energy cannot turn on the primary clamping diodes at minimum design input voltage. If that happens, there is no real drama, except the dc output voltage drops out of regulation.
If we stick to discontinuous operation, although the peak currents are higher we avoid the control discontinuity of shifting into CCM.
Its just a matter of giving it sufficient air gap to stay discontinuous at full maximum output power.

Not saying this approach is always the best one, but for this project it offers simplicity and fewer problems taming the control loop without introducing any significant disadvantages.

- - - Updated - - -

top of page 7 of this is talking about this particular failure mechanism of a 2 switch flyback...ti.com don't call it a failure , but just say that transformer operation is interfered with.....

All this confusion about something really simple !!

When forward conduction ceases, the energy stored in the core generates a very fast flyback voltage with equal volts per turn in EVERY winding.

The voltage rises very fast until it is clamped by one or more of the diodes on one or more of the windings.
The flyback power goes mostly into the diode that turns on first at the lowest voltage. If there are multiple secondary windings, the most heavily loaded one gets most of the flyback energy.

Provided all the secondaries are sufficiently loaded, and the dc input voltage sufficiently high, the flyback volts per turn can never rise high enough to turn on the primary clamping diodes in a properly designed circuit.

If there is insufficient secondary load, or the dc input voltage falls, the primary clamping diodes will conduct some, or even all of the stored flyback energy.

There is no mystery, no vanishing energy, no smoke or flames.

Apart from avoiding obvious problems such as excessive leakage inductance and magnetic saturation, the only "trick" if you can call it that, is to arrange the turns ratio so that the flyback voltage in the primary is always less than the forward voltage. That is easily done by adding a very few extra turns to each secondary.

 

[treez]

Thanks, I will now be very cringeworthy and bring up one application that the 2 switch flyback can’t manage because of its problem if it reflects back more voltage than vin……..the “single stage, offline , PFC’d Flyback”…..for obvious reasons, this cannot use the 2 switch flyback, which is a darned shame as its power handling capability would have made it very good otherwise.
(I am still wondering what Basso meant by the Applause as in the above posts #12 and #14)

 

[mtwieg]

All very true.
But I suggested flyback, with discontinuous operation, and there is no reason to exceed 50% duty cycle with a suitable transformer turns ratio.
This avoids ALL of the above problems of RHP zero and second order responses.

Under steady state conditions, and nominal line/load conditions, yes. Not a robust design philosophy to ignore transient conditions and edge cases though.

Also battery charging is usually such a low bandwidth application that the change from DCM to CCM shouldn't be a challenge, at least not compared to the battery's own unpredictable behavior.

- - - Updated - - -

Thanks, I will now be very cringeworthy and bring up one application that the 2 switch flyback can’t manage because of its problem if it reflects back more voltage than vin……..

Again, the reflected vout cannot exceed Vin, it can only be approximately equal to it.

the “single stage, offline , PFC’d Flyback”…..for obvious reasons, this cannot use the 2 switch flyback, which is a darned shame as its power handling capability would have made it very good otherwise.
(I am still wondering what Basso meant by the Applause as in the above posts #12 and #14)

Agreed, it's not good for things requiring a very large duty cycle range like AC-DC converters.