This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Load Sharing of Power Supply, UC28710D

Other Parts Discussed in Thread: UCC29002, UCC39002, PMP

Good Day TI Support,

We successfully made a small SMPS based on PMP8363, using UC28710D IC. First, I want to thank you for this wonderful design. We tweaked it slightly and sold over 5000 power supplies already!

The problem is that this is a 5V, 2A power supply and we need no 4A. Surely, we can make one but that means changing to another flyback transformer. So, what if we can use the exact same PSU and parallel them using a load sharing Controller IC. TI make these.

I want to ask for information (application notes, etc.) to use an appropriate load sharing controller IC with UC28710, please?

Thank you in advance,

Solomon

  • Hello Solomon

    Great to hear that the PMP8363 design was useful !

    TI makes the UCC29002/1 load share device which you can use to parallel the modules.  Load sharing relies on measuring small voltages and requires careful attention to PCB layout - especially the grounding arrangements - for it to work successfully. The documentation for the device contains lots of useful information about this device - www.ti.com/.../technicaldocuments

    Regards

    Colin

  • Good Day Colin,
    I assume the UCC29002/1 and UCC39002 ICs will work with UC28710?

    Thank you for your quick response.
  • Hello Solomon


    I may have given you misleading information just now -  and I will have to think about this a little more -

    The problem is that the PMP 8363 doesn't have either remote sensing or a trim input - which means that the load share controller cannot adjust the output voltages of the module outputs being shared.


    I'll think about this a little more and see if I can come up with a sensible solution

    Regards
    Colin

  • Thank you Colin for coming back to me. I will await for your answer. Thank you.
  • Hi Solomon
    I missed the fact that the PMP8363 has no remote sensing or trim inputs so the UCC29002 load share controller cannot be used in this case.
    I think you have two simple options
    1/ Hard wire the two PSUs in parallel. You won't get load sharing but the output currents will add. The PSU with the higher output voltage will supply the load current when the load is less than 2A and the other PSU will operate at no load. As the current is increased, the first PSU will go into current limit - this will happen at approx 2.2A. As the load current increases further, the second PSU will supply the added current. This method is simple and will work providing you are able to accept the unequal load sharing. Increasing the output impedance of the PSU by adding a resistor can help load sharing at the expense of voltage regulation and power loss in the resistor, although this method does work it won't give you good load sharing

    2/ Redesign the transformer to allow operation at higher currents.

    I tried to think of ways to parallel two PSUs but this controller is a primary side regulated device and the schemes I looked at all got very complicated and frankly it would be simpler to redesign the transformer.

    What would I do? I'd look at hard wiring the PSUs in parallel - this will give you a quick solution if you can tolerate the unequal current sharing. If not, I'd change the transformer - simpler and lower overall cost too.

    There is a good summary of the techniques available in the document at www.ti.com/.../slup207.pdf;tisearch=Search-EN-Everything

    Apologies again for leading you astray in my first response.

    Regards
    Colin
  • Good Day Colin,

    I found an IC from another supplier which does exactly what I need it to do: LTC4370. It is basically a two-supply current balancing controller. I was about to design a circuit from scratch now, when I planned the layout, I thought to myself that just maybe someone else might have already done it. And somebody did. Please take a look at it and advise whether you have a similar IC. I Love to Support TI, because you support me!

    Thank you,

    Solomon 

  • Hello Solomon

    Interesting part !. Unfortunately TI doesn't have an equivalent so I would suggest you go ahead and use the LT part - (don't tell anyone I said so !)

    Now, TI does make a nice range of low Rds_on MOSFETs - The LT app note uses a 6mR part at 10A, if you are running at 2A (each side) then the Rds_on would scale to about 30mR - you could look at the CSD25201W15 for example - see the selection guide at www.ti.com/.../slit121d.pdf  for the full range. Of course the devices are being run in linear mode so Rds_on is important ONLY because it sets a limit below which the load sharing IC cannot reduce the Drain Source resistance of the MOSFET.

    My initial caveats about layout would still apply - watch for the sensing connection to the load sensing resistors. One thing to be careful about is the connection to the MOSFETs. Most MOSFETs are intended for switching applications and work well when used as such. In linear mode one of the main concerns is that the device is operated in its Safe Operating Area - the other is thermal runaway. Finally, keep the connections to the MOSFET gate and source as short as possible so that it doesn't pick up noise.


    Good Luck

    Obviously I can help you only with queries that involve the TI parts you may choose to use

  • Thank you Colin,
    I was busy looking at N-MOSFETs when I received your reply. In the nick of time! I will take a look at them now and order some samples to test.
    Thank you once again!
    Solomon
  • Good Day Colin,

    As you can see, I have not given up yet...on TI. I looked at my problem slightly different. TI is making a ORing FET Controller IC: LM5050.

    Looking at it, the LM5050 use almost the same setup as the LTC4370, but one need one LM5050 for each power supply. The IC is small and cost effective enough to make this choice. 

    NOW: The LM5050 senses the output voltage and effectively changes the forward voltage drop of the "ideal" OR- "diode" (The NMOSFET used in this case), by changing the Gate Voltage of the NMOSFET. This is exactly what the LTC4370 is doing. So, by applying the parallel scheme of the LM5050 the load current should be shared almost equally among the two power supplies. The great thing is that one can continue to add power supplies in parallel with this setup, whereas the LTC4370, can only handle two power supplies.

    Will you please confirm my understanding?

    Even if the load current sharing is with 10% error, it will still be an acceptable solution. 

    Thank you,

    Solomon Smit

  • Hello Solomon
    Interesting. Let me think about this - I may need to talk to some colleagues in Santa Clara so it may be a day or so before I get back to you.
    Regards
    Colin
  • Thank you Colin,

    I was about to order a few LM5050 Evaluation boards to test it out. Maybe someone at TI has already tested it out. I am positive that it would work, provided that the voltages of the two PSUs are within range of the Forward Voltage drop of the body diode of the NMOSFET. If it is too far out, than it is just OR'ing. In our case, we have made testing equipment to test each power supply to be between 4.9V to 5.2V. Thus, it will be OR'ing and Current balancing to some degree.

    Thank you,

    Solomon 

  • Solomon
    This is what I got back.
    'The LM5050-x will control gate drive to maintain some minimum drain-to-source voltage (i.e. Vin – Vout) through the Rds(on) of the mosfet. See the ‘Vsd(reg)’ spec in the datasheet. However, the LM5050-x min/max limits for ‘Vsd(reg)’ are a bit on the ~sloppy~ side. So I would not expect two LM5050-x stand-alone devices match very well.

    Since the gain of the Vsd(reg) comparator is relatively high, the behavior of the gate control at in this condition tends to be a bit choppy, as the gate discharge (off) is very fast and the gate charge (on) is much slower, so the Vsd(reg) gate drive looks like a saw-tooth. Some users reported this Vsd(reg) gate drive behavior, mistakenly, as an oscillation.

    Unlike LTC4370, there is no allowance for variations between two different input voltages. The LM5050-x is always OR-ing controller so a difference of approx. >28mV, Vsd(rev), between two input supplies will force the lowest supply to zero current.

    Finally, there is no way (that’s obvious to me) to connect multiple parallel LM5050-x circuits together so that they are aware of their relative current share.

    All-in-all, I seriously doubt that LM5050-x can be used effectively to balance currents from multiple input supplies.

    The good news is that multiple LM5050-x EVMs can be wired up to test this out. '

    It may be possible to degrade the gain with some well chosen resistors between the LM5050 and MOSFET gate.

    If I had to summarise this - the best thing to do is to get some LM5050 EVMs and see how they actually work.

    Let me know how you get on.
    Regards
    Colin