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.

UCC256404: 300W LLC Converter for lower input voltages

Part Number: UCC256404
Other Parts Discussed in Thread: UCD3138A, UCC256403, PMP3162, PMP20657, PMP22519, UCC2897A, UCC2897

Efficiency is of high concern for my current design, hence the choice of a resonant LLC.  Despite the datasheets not providing any evidence a low input voltage (say, 20V+) isn't valid, I see a lot of chatter that says this isn't the proper topology, and the WebBench tools don't allow for sim at that low of a voltage.  Can someone elucidate the reasoning behind that limitation?

My design: Input = 27-36V (28V nominal), Output = 12V @ 25A

  • Clarification: All voltages listed above are DC... there is no AC involved in this design.

  • Hi Daniel,

    For LLC, there is a circulating current on the primary that doesn't contribute to the power transfer but is there to drive the ZVS transitions. For higher input voltage, this circulating current is ultimately beneficial as the ZVS payoff is substantial compared to the higher conduction loss penalty from the circulating current. At lower input voltage, the efficiency benefit from ZVS is diminished but now the circulating current is going to really start adding to the conduction loss on the primary, especially in a half bridge LLC. A full bridge LLC would be much better suited to handle lower input voltage ranges because the voltage across the primary winding is 2x compared to the half bridge LLC and the conduction loss would go down but UCC25640x is not capable of supporting a full bridge LLC configuration without significant glue logic. We do have a reference design for full bridge LLC with a digital controller UCD3138A that may be of interest to you:https://www.ti.com/tool/PMP22519

    Regarding getting UCC25640x to work with 27V to 36V input, this would be too small to successfully achieve HV startup (the HV pin needs at least >40V at the pin in order to successfully charge up VCC to 26V). But a non-HV startup design would certainly be doable with UCC256403. You could do a small external circuit to step down the input voltage slightly going to the VCC pin. 

    For lower input voltages such as this, other soft switching topologies might be better suited for achieving your efficiency targets. I would suggest taking a look at PMP3162 which is an active clamp forward converter:https://www.ti.com/tool/PMP3162. It is not quite "on the dot" for your input/output requirement but it should give you an idea for what is achievable with this topology. The reference design achieves almost 95% efficiency.

    You could also take a look at phase shifted full bridge which is another soft switching topology:https://www.ti.com/tool/PMP20657

    Best Regards,

    Ben Lough

  • Thanks for the response, Ben.  I'm not a power engineer (digital), and topologies beyond basic buck/boost are outside of my realm of knowledge.

    I should have mentioned in my OP, full isolation is a requirement here... a quick glance at the schematic leads me to believe that knocks the PMP20657 option out of the running, possibly the same with the PMP22519 option.

    As far as the UCC25604 is concerned, wouldn't providing the input voltage (27-36V) directly to the HV pin provide enough voltage for the initial charge to VCC?  If I understand the chip's operation correctly, the HV pin's functionality as far as AC inputs go will be unused in a DC-supplied design, essentially turning it into little more than a bootstrap for VCC.  I'm specifically looking at the VCC_StartSelf spec (or VCC_StartSwitching for the '03 variant)... arguably, the 28V max for that spec is higher than my possible minimum input voltage of 27V, but in principle I would think that overall design is still valid.

    I will take a closer look at the PMP3162 option you mentioned (though feel free to correct any incorrect assumptions I made above on the other options)... if I can get the necessary power out of it for the slightly lower input voltages I'm working with, it could be a winner at this point in the game.

  • Taking a closer look at the PMP3162 setup, it looks like it's not fully isolated, either, so that's a no-go :-(

  • Hi Daniel,

    PMP3162 is a fully isolated power supply. The circuitry on the primary are referenced to "PGND" and the circuitry on the secondary is referenced to "GND". There is a safety rated Y cap (C2) that is there connecting PGND to GND. This is quite common for EMI reasons. Could you let me know what is giving you pause regarding the isolation of this design?

    Best Regards,

    Ben Lough

  • Hi Daniel,

    The current source between HV and VCC is a high voltage JFET that will require some voltage headroom in order to successfully and reliably charge VCC all they way up to the VCCStartSelf threshold. As you mentioned, you will likely find some devices would be able to successfully charge up VCC while HV is less than 40V but it wouldn't be possible to reliably guarantee this over temperature, process variation, etc.

    The active clamp forward controller in the PMP3162 reference design (UCC2897A) has a similar pin called VIN to perform HV startup. The recommended range for this pin is 18V to 110V so your input voltage range would pair well with this controller.

    Best Regards,

    Ben Lough

  • My bad, I read the schematic without my glasses on ;-)  I thought N_DRV and P_DRV were crossing the boundary (I was looking at secondary side VDRV flags), but it seemed like a really odd drive method.

    I will take a deeper dive into the PMP3162 design...

  • Hi Daniel,

    If I could offer a few words about the gate drive, there are versions of active clamp forward that use an N channel MOSFET for the clamping FET but these are generally more expensive because they require connecting the clamp to Vin instead of to gnd so you would need a floating gate drive. A P channel MOSFET will give you the correct body diode orientation and you can avoid needing a floating gate drive. The driving circuit is effectively a cheap way of using a capacitor to "level shift" the gate signal to apply a negative voltage on the gate of the PFET in order to turn it on. 

    Let me also share with you these slides on active clamp forward. It may help with getting more familiar with the topology:https://www.ti.com/lit/ug/tidu185/tidu185.pdf?ts=1597858311005&ref_url=https%253A%252F%252Fwww.google.com%252F

    Best Regards,

    Ben Lough

  • Ben,

    One last question (I hope)... any advantages to using the UCC2897 (active clamp) versus something like the UCCx808 (push-pull)?  WeBench gives initial estimates efficiency ~89% for the 2897, but ~91% for x808 designs.  At 300W, my design appears to be on that edge between multiple topologies, with no one topology providing greater advantages/disadvantages than another, but I'd be remiss to lose out on even a couple of percentage points if I can easily get them.

    Thanks.

  • Hi Daniel,

    I would say there are a couple differences to keep in mind between push-pull and active clamp forward. Push-pull is a double ended topology and will deliver energy to the output every time the switch is on. Active clamp forward only delivers energy when the main switch is on. So there are some benefits in terms of better core utilization for the transformer over active clamp forward. Push pull is a hard switching topology, so there will be more switching loss on the primary FETs and you may need snubbers across the primary FETs to limit max voltage stress, but with 27V - 36V this may not be too bad. Active clamp forward is quite easy to do self-driven synchronous rectification on the secondary which would be more important for higher output current designs. This is not possible to do in a push pull. I don't believe  Webench is considering this aspect.

    In my opinion, I think an active clamp forward with synchronous rectification would give you the best opportunity to maximize the efficiency for your power spec.

    Best Regards,

    Ben Lough