Where to Add Current Feedback

[DBaron]

I have a question I hope some of the converter feedback experts might be able to help with.  I'm building a constant current, constant voltage output switching power supply.  I've successfully got the constant voltage feedback loop implemented as a Type 2 (Venable) using the LTC3723-1's internal error amplifier.  See attached schematic.  This converter is a 28V, 10A output push-pull topology, using peak current mode control (pulse-by-pulse).   I'm starting to close my constant current portion of the converter loop, and was wondering where in this circuit should I add the feedback from the current sensor to regulate constant current?  I could see adjusting the voltage feedback by pulling down the top of R16 or even adding it to the sensed pulse current as part of the pulse-by-pulse current sense portion, or abandoning the internal error amplifier and using external op-amps to implement this and putting both feedback inputs directly into the PWM comparator via the COMP pin.  I think any of these could work, but I'm concerned I'm missing something, and that there's probably a recommended best practices way to do this.   I'm planning to use an Allegro ACS713 as the output current sensor if that is important.

The internal drawing of the LTC3723-1 is on page 8, here:  http://cds.linear.com/docs/Datasheet/372312f.pdf

Any help as to the best place to put this second feedback input would be very helpful.  Thanks in advance.

[amspire]

I am not sure the current limit will work the way I think you are intending.

When you say you are designing a constant voltage/constant current supply, do you mean that these values are fixed, or are they variable and that the supply works in either a constant voltage or constant current mode? I just want to get this clear, as a fixed voltage and fixed current supply is a much easier design task.

The the regulator's current limit circuit is designed to limit the peak pulse by pulse current in the inductor, which does not correspond to a true current limiting on the output.

I think you need to implement the CS input to control the peak inductor currents anyway using a local sense resistor as shown in the data sheet. This is the primary protection for the supply itself rather then the load. If you want  true current limit on the output, that will have to be an additional circuit that could limit the supply by either overriding the voltage sense in, or by adding an additional offset to the current sense circuit. The regulator's current sense input it not really designed for a precision variable current limit on the supply output.

Is there any reason why you are going to a hall sensor for detecting the output current instead of a resistor? A resistor has advantages that it is more linear and the only way it can fail is by going open circuit which is a safe failure. If the hall effect sensor dies, then you could go to maximum current.

Richard

[DBaron]

The design is for a fixed output voltage with a fixed output current limit.  I certainly will keep the peak current mode control to protect the switches and to take care of flux walking, and this is already implemented and working well using a pulse high frequency current transformer.  My question was more related to the nuts and bolts of how to implement the override of the voltage feedback.  Precisely where do this and then how to properly compensate to ensure a clean switch over between the two feedback loops.  Based on your response and another day of considering in the meantime, I think I understand how I should proceed. I'll draw up what I've been thinking and post it this evening.

As for my thinking regarding the hall effect, at 10A of output, I'd assumed the losses of the sense resistor could become somewhat high, and also require a differential amplifier or high side shunt chip (INA138/168, perhaps).  I'm certainly not opposed to a shunt method, I'd just found the hall effect and thought it could work and never really evaluated the shunt.  I will consider this as well.  Are hall effects particularly known for being less reliable than the high side shunt monitor chips?

Thank you very much for your response.  I really appreciate your feedback (pun most certainly intended).

[amspire]

My first question then is exactly how precise do you need this output current? Not sure of your actual switching topology here, but if the inductor is large enough so that at 10A output, the peak inductor current can be set to 12 A, then the regulator's standard current limit would provide a 10A to 12A current limit on the output.

That is usually good enough for most switching supplies.

Even if you add a super precise extra current control circuit, on application of a short circuit, the output filter caps will first dump all their energy which will be in the hundreds of amps. Then whatever current is currently in the inductor will flow into the output. If at the point of shorting the output, the inductor had 20A flowing, then after the capacitors have dumped their energy, the inductor will supply 20A.  The bigger the inductor, the less overcurrent it can deliver. It could be that adding an extra precise current limit circuit does not really achieve what you were wanting anyway.

Richard.

[DBaron]

I'm looking for an output current regulation in the 2-5% range.  The converter will operate in current limitation in normal operation, not just during faults.

While looking into this, I stumbled upon an Analog Devices article (and subsequent data sheet) that happens to cover the topic fairly well.  I don't know why my particular combination of google keyword happened to reveal this while several prior similar searches did not.

http://www.analog.com/library/analogDialogue/archives/31-2/liion.html

It's an obsolete product, but the information and configuration is useful.  I've got some reading and digesting to do.

[amspire]

I'm looking for an output current regulation in the 2-5% range.  The converter will operate in current limitation in normal operation, not just during faults.

If you make the inductor big enough, you can get the 2%-5% current regulation without any extra circuitry. Something like a 100uH 10A inductor, if you can find a big enough torroid core.

[IanB]

The converter will operate in current limitation in normal operation, not just during faults.

You seem to have some problems with definition here. A DC power supply can either operate in voltage limited mode or current limited mode, but never both at the same time. If the voltage or the current is limited to a precisely controlled value then it is called by convention a constant current or constant voltage (or constant current/constant voltage) supply. But these terms are convenient fictions--there is no such thing as a truly constant current supply or truly constant voltage supply. The values are merely upper limits.

So in particular, a supply that operates in current limitation in normal operation cannot be a constant voltage supply, since the voltage will vary while the current is regulated at the limit. I have not followed the links you posted, but you might find your searches to be more fruitful if you look for a constant current supply.

[amspire]

I suspect that Lowell is going to use this to charge batteries, so that it will normally be in current limit mode. If so, it will only be in constant voltage mode at the end of the charge.

[IanB]

I guess so. So if you are building a CC/CV supply for battery charging, is it easier to build a CC circuit and put a voltage limit on it, or build a CV circuit and put a current limit on it? Or does it make little practical difference?

[amspire]

I guess so. So if you are building a CC/CV supply for battery charging, is it easier to build a CC circuit and put a voltage limit on it, or build a CV circuit and put a current limit on it? Or does it make little practical difference?

It actually ends up being the same. In constant voltage mode, the current control is doing nothing. In constant current mode, the voltage control is doing nothing. The only exiting time is at the crossover point.

[DBaron]

I suspect that Lowell is going to use this to charge batteries, so that it will normally be in current limit mode. If so, it will only be in constant voltage mode at the end of the charge.

Sorry, I should have been more clear regarding the application, but didn't want the thread to get mired down in "Why on earth are you building such an odd voltage output battery charger?" questions.  In actuality, I just picked a somewhat arbitrary application and am doing this design for learning and experimentation.  Voltages and powers were selected to be within the range of the test gear I have available, while still being significant enough that I'm forced to learn and consider many factors to get something that doesn't immediately release blue smoke.  I'm trying to do things that challenge me and poke at areas where I'm technically weak, and the feedback theory needed some poking.  Nothing motivates me to learn quite like building something and having to make it work.

I'm not looking for definite answers, just nudges in the right direction, and I think I've got a much better idea now of how to proceed.  That crossover point worries me, but I guess I'll cross that when I get a constant current output control working. 

[amspire]

I'm not looking for definite answers, just nudges in the right direction, and I think I've got a much better idea now of how to proceed.  That crossover point worries me, but I guess I'll cross that when I get a constant current output control working. 

If you use the chip's own current limiting, the crossover point shouldn't be a problem.