LM2576 / external MOSFET / buck converter

Discussion in 'The Projects Forum' started by takao21203, May 15, 2012.

  1. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    Hello I am looking for some assistance/suggestions how to improve this circuit.

    It is a LM2576 simple switcher (3.3V), wired to external p-ch MOSFET + diode/coil configured for buck converter.

    Gate drive is done with a small transformer, which on the input side has 47 Ohms dampin resistor. Turns are about 3:1

    It works so far, having seperate supplies for LM2576/buck input voltage.
    This voltage I can drive higher than 50V, using a dc/dc module, wired to 12V electronic transformer. Another transformer same model is also used for the LM2576 supply.

    There is no input capacitor, only a large ferrite coil.
    Which will emit noises, if I increase the input voltage, and also output voltage.

    At lower output voltages, it remains quiet.

    One improvement was already made, adding a reset loop to the gate drive transformer.

    Any advice how to improve the gate drive/the circuit?

    The transformer likely will emit spikes upto 100V, I tested that in LTSpice.
    Adding the reset loop, noise in the filter coil (not seen in the photo) is greatly reduced.

    Are there additional circuits that I can add to the transformer drive?
    Any good websites?

    The proto circuit is built that way in order to make changes as required, without too much effort. Isn't the way I build circuits in general.
     
  2. allibut

    New Member

    Jul 18, 2012
    21
    0
    It would probably be very helpful if you posted a circuit diagram rather than your photo of a bundle of wires. ;)

    I don't really see the need for the transfo. , why not just use the 2576 to drive a resistive load and power a MOSFET directly?
     
  3. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    http://forum.cytron.com.my/viewtopic.php?f=30&t=11648

    there are some additional photos incl. schem.

    I have posted elsewhere since I did not receive replies here.

    However, it is not so much important for me, since I use a TL494 circuit now, for upto 12A.

    Driving PNP isn't possible directly, because the LM2576 internal NPN can not withstand the high voltage.
     
  4. allibut

    New Member

    Jul 18, 2012
    21
    0
    thanks for the update. Looks like you did not get much help there either.

    I don't understand why you moved to PNP, that will cause a lot of unnecessary heat waste.

    Indeed , why pMOS to start with. The idea of using a transfo is surely to provide a floating output. This means you could use nMOS, cheap, powerful , minimal R_on.

    That would also mean you could reduce the current you need to drive the transfo.


    Max V_in for 2576 is 45V so you could use much higher than 50% divider chain and have much less heating in the resistors. Tho' I could not find the current required by the chip :confused:

    If you have access to a 'scope, you could evaluate the inductors by measuring the ringing frequency of an LC pair with a know capacitance.


    I may try this sort of cct. I have an adjustable LM25xx somewhere and those sort of coils form PC motherboards.

    In fact I think you are way too small there and that maybe why you are heating the diode. That really should not happen , it's only to suppress spikes.

    I used such a coil with a 25xx reg. to drop 18v to 5v, but it was supplying very small current. To do what we are both aiming at here, which had much higher output current, a much bigger L is needed.

    I suggest finding an old ATX PSU and digging in its guts. Those are designed for this sort of current so there's more chance of being in the right ball park.

    As usual , nothing beats actually designing what you're trying to do. I thing the LM spec sheet should tell you how to calculate the required inductance. At least if you know what you need you may have a better chance of finding it (even if you don't have an L-meter).


    I had not thought of using a TL494. Could you post your cct ?
     
  5. allibut

    New Member

    Jul 18, 2012
    21
    0
    subscribing myself to this thread.
     
  6. allibut

    New Member

    Jul 18, 2012
    21
    0
    really, subscribing myself to this thread.
     
  7. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    OK I made a thread here for TL494.

    It's giving 12 Amps or something.

    And PNP I think I know what I am doing.

    A little heat less or more- who cares? It mainly depends on the storage coil. And this is not as easy as to pick some ringcore from a PC mainboard!

    The LM2576 is not very good for this, because it has more than one current limiting systems. But you would see this yourself when you connect large 12V bulbs.

    Designing a SMPS is much more than doing some maths.
    It also has to be cost efficient, small, and without bad noise. No clickering, whistling or clunkering, or whining and all this.

    As for me part-time SMPS designer/experimenter, 12 Amps is not too bad, and without large cooling system.

    Let me know if you get somewhere near that.
     
  8. allibut

    New Member

    Jul 18, 2012
    21
    0
    I, for one. I'm designing an MPPT regulator for solar PV , the aim is to create a more efficient converter not to waste electricity by heating components.
    That is indeed what I said above. I hope I made useful suggestions.
    What I would see is presumably the same thing you observed. I was suggesting that your interpretation of why that was happening may not be correct. The spec sheet does not say how much it can dissipate, simply that it is "internally controlled". It appears from your cct that you may be creating too much heat by driving the output through a 47R resistor. This will dissipate more when the heavier load causes a higher mark/space ration. Especially at start up. I think this is what you are seeing.

    Please don't feel offended by any suggestions, I have no idea of your background or level of knowledge, as you do not have of mine.

    Feel free to give reasoned arguments as to why what I have said does not apply, but do not take anything as an insult and get irritated.

    Sometimes doing some maths can help. For the rest I agree totally.
    Yes, it's a good start, that's why I'm interested in what you have done.
    I think this sort of converter should be possible in 90-95% efficiency range. that is what commercial products quote. This would imply minimal cooling requirements.

    Running a 60W load with 10% losses is ~6W that probably does not even need a fan.

    If something is getting hot, it's a good place to look for design errors.
     
  9. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    It is not loading the LM2576 much, not even near enough to heat it up just a bit. Remind this is driving the primary winding of a small toroid transformer.

    The current limiting, there is more than one, seems to monitor the transitions obtained from the feedback, and if it is too much off what it expects from the waveform it outputs, it will actually shutdown to 1/6 frequency.

    12V bulbs have 6x inrush current...I managed to play some tricks so they can be started up, but was not easy, and there is not unlimited potential let say to control 50A with a LM2576.

    Yes OK. I am a programmer-type personality. I remember some years ago I wrote programs, well today they are no longer useful for me.

    This is not correct. Most circuits (for smaller currents) only archieve 80%. For higher efficiency, different technology than storage coil is used. As well synchronuous rectification. And for instance a gate drive IC.

    All this is absent here!

    Using MOSFET or Power PNP, only gives improvement of 4% to 5%.

    Large Power PNP have very low resistance!

    But, above 200V, there are not many PNP MOSFETs or transistors.

    For the TL494 I use resistor divider, and simple PNP base driving method. It is quite usual for lower currents to do it that way.

    96%...well show me a high-tech ATX supply that can archieve 96% or even 90%. They are quite lossy, and they heat up. Even if they use a flyback transformer and driving transformers as well.

    My circuits are very simple remind that. If I can archieve 80% that's more than enough.

    For a SMPS that is adjustable over a wide range, and is using a storage coil, and let say 10A or the like, show me a circuit that can get above 90% without some very expensive exotic components.

    Microchip's small MCP1640 can archieve 80% or a little better for some voltages, but that's it. And they are modern chips, MOSFET technology, 500 KHz.
     
  10. allibut

    New Member

    Jul 18, 2012
    21
    0
    LM2576 claims 88% Vin=30V , I load = 3A ; Vout=15V

    Lower outputs are worse.

    What I was commenting on above are commercial MPPT converters which often claim figures like 90-95% I mentioned.

    Obviously I don't know what they have in their guts...

    I am hoping to approach that kind of figure. If I'm not going to get beyond 75% I'll not waste time designing an MPPT.
     
  11. allibut

    New Member

    Jul 18, 2012
    21
    0
    I think the key point to getting dissipation down in switched mode applications is the transition of the conduction zone as well.

    this is where driving the large base current of one of those power PNPs with an inductance may be causing heating.
     
  12. allibut

    New Member

    Jul 18, 2012
    21
    0
    There is the thermal shut down and the output current limit sensed off the first output transistor.

    I see no indication of any other mechanism in the data sheet. That suggests that either they are hiding some covert "features" or you are mistaken.

    I'll let you decide which is the most probable.
     
  13. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    Current through the primary of the gate transformer is milliamperes.
    Very thin wire.

    Current through the IRF9540 is Ampere's.
     
  14. allibut

    New Member

    Jul 18, 2012
    21
    0
    Ok , so you're back to pMOS and I don't have the actual the circuit we are discussing.

    But how do you think 2576 senses the current passing through that device? Are you still suspecting this is some undocumented feature?
     
  15. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    I have read the datasheet carefully. More than one time.
    There seems to be frequency reduction in place.

    Also the pMOS circuit schematic is included in the thread on cytron.

    I have developed not only one SMPS circuit but 4 or 5, the LM2576 one of them. Main project = TL494 power supply.

    But the LM2576 circuit also works quite well, only needs a small $1 VGA cooler, for 4.2 Amps output current. I also have a photo in the thread where I have actually rebuilt the LM2576 circuit, but much smaller. I have considered it as AUX supply, but also the LM2576 needs 12V to work. Yes it could work from 30V, but then the amplitude through the primary of the gate transformer becomes much higher.

    I tried transformer gate driving with a larger pMOS in the TL494 circuit, but that did not work out at all. Actually one pMOS was destroyed. Smaller IRF MOSFET fries up instantly for 2x 12 bulbs. So I think TIP36C (32 Amps rating) is more robust than MOSFET.

    Biggest problem is the storage coil heating up, so I use two in parallel. I tried many toroid cores, most did not work well, means too high current, no matter the frequency. Wrong magnetic material I guess.

    I could put more effort into it, order even more toroid cores, and to try them, but it's not needed currently, and 12Amps current is more than enough!
     
  16. allibut

    New Member

    Jul 18, 2012
    21
    0
    I think it would be quite easy with that transformer idea to exceed the Vgs max. , probably depends on how fast the IC is switching the output. I was actually suprised you have not already hit that problem.

    A power PNP will actually sink rather a large base current, so the voltage will get dumped by base junction. You could regard that as being robust. You would normally drive those devices in totally different ways. So I guess one has to be more suited to what you are doing. That's just the luck of that draw , not robustness.

    I think the FETs would be robust if they are driven correctly.


    You should be able to switch a MOSFET with tiny current, just keep the Vgs in range (circa 15V) There's no guarantee that charge flow will be identical in each direction so the is a possibility of Vgs creeping up and it staying partly in conduction.

    That is why I asked it this was an established technique or just an idea you were trying.

    I don't think this would be used in a production environment because of cost and space but since the bits are recoverable free, I'm intrigued as to whether it could be a solution to high side driving nMOS devices.

    A large number of fine secondary windings should mean you could get enough of a pulse to flip a MOSFET with minimal primary current. May need a pair of back to back zenners to protect the gate.

    If the main coils are getting hot they are probably going into saturation. Also in such close proximity there will be magnetic cross coupling between them. I would not like to guess at what that will be doing.
     
  17. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    Maybe we mix up features from the circuits now. But I developed them together, partially to try out some things.

    LM2576 is MOSFET, driven with gate transformer.
    TL494 is PNP (TIP36C), driven with resistor divider. Also IRF9540 and large p-CH MOSFET have been used.

    Yes MOSFET gate current is zero, but you need to charge/discharge it very quick, so you need current anyway.

    The transformer has not enough capability actually to damage the MOSFET gate. This is why I have the 47 Ohms resistor in place (on the primary). Many windings well maybe 20... When I tried it on the TL494 circuit, without primary resistor, actually the MOSFET fried up. It did not work at all, I guess due to saturation. The LM2576 is driving it in a way it does not saturate. Otherwise it could not work with good efficiency.

    In the TL494 circuit all of the cores heat up (Output 4A - 12A). Not immediately, but if there is no airflow it slowly creeps up. This is partially to the thin magnet wire. So the loss inside the core is maybe 1W or 2W. No problem to keep it at bay with some airflow.

    And most cores = wrong magnetic material, let say input current 2 Amps.

    Using the large yellow cores, I get it down to about 1.3 Amps. So any core that will result in more than that, is bad.

    Interference over air between coils is a non-issue at 100 KHz. Placement on PCB causes some minor differences, but even 10cm wire does not matter.

    I have been bending the toroids actually while under full load, to see if there are any crosstalk effects. So far I have not been able to observe any flucations for input current.

    I have also tried for instance to rewind cores with thicker wire, however, it seems to be they simply don't have the right properties, except the yellow one's I am using now.

    And I have tried for instance 2mH common mode chokes, not giving good results...

    Testing all the coils was a lot of effort actually. Transistor/MOSFET - only minor difference, some %.

    I must also say before I made these circuits, I did not know properly how to use power MOSFETs. I have learned it while working at the circuits! I think I understand the basics of it now.

    Previously I used various switcher ICs, some small LT1073? or 1043?, MCP1640, LM2576, LM2675, MC34063.
     
  18. allibut

    New Member

    Jul 18, 2012
    21
    0
    It sounds like teh 484 switches faster. That would give greater dI/dt and bigger secondary voltage spikes. This is what I was saying about exceeding Vgs.

    I don't think 30% losses is particularly "good". If you are running 50 or 60W bulbs, that 30% means a lot of heat is going somewhere.
     
  19. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    good for me means the heat can be controlled, best if the cooler remains cold, or only heats up a little.
     
  20. allibut

    New Member

    Jul 18, 2012
    21
    0
    Just noted this in the 2575 datasheet:

    It may be that the warm-up surge on the 60W bulbs is pushing the unknown valued inductors into saturation.

    I think you may be starting to saturate already which is why you have relatively low efficiency and a fair amount of heating in the coils.

    BTW if you don't need the driver power the 2575 drives 1A output and only costs about a buck.

    Properly designed driving of a power device should make that quite a simple and cost effective regulator.

    I'm still digging for ideas but I think I'll order a couple and start testing a few ideas.
     
Loading...