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.

TPS23753A: Minimizing thermal impact of PoE PD + Controller ... simulations available?

Part Number: TPS23753A
Other Parts Discussed in Thread: TPS23755, PMP9175, PMP7355, UCC24610, TPS23754EVM-420, UCC24612

I have a design where we are presently using a PoE module with a TPS23753A onboard.
This design is suffering from thermal issues, even at relatively low load, 10W or less delivered power.

I am investigating the possibility of moving away from the module and implementing our own TPS23753A or TPS23755 circuit on our board, where some of the components can be controlled.
While TPS23755 looks like the easiest circuit to implement, I have a major concern that the internal MOSFET will limit the ability to solve dissipation issues or make them worse.

The goal is to deliver ~10W of output power. The final output should be 5v, but a second stage DC/DC, buck or boost, after the isolation stage would be acceptable if advantageous (reduce current in the flyback output stage?)

I'll also note that I am not a hardcore EE by any means, and I am searching for a resolution based on what I can glean from the datasheets and app notes.

The attraction to use TPS23753A is that we could choose the transformer, switching FET, and output diode.

Following slva305c, the MBRS540T3 seems to be about the best choice I can find for the output rectifier, in terms of forward current drop.

Use of a very high-performance MOSFET like BSZ900N15NS3GATMA1 seems like it could reduce both switching loss and resistive loss.
But given the relatively low primary current, maybe it would be better to select a part primarily to minimize switching losses and diode drop first?
Has the internal MOSFET for TPS23755 already been optimized for all 3 parameters?

The loss in Rcs seems like it may be as high as 330 mW based on the primary currents calculated in slva305c. However, think there is nothing that can be done about that, without the use of a very expensive isolation amplifier? That solution is probably not within scope.

Are there any SPICE simulations ( or any other package ... matlab maybe? Mathematica notebook of equations?) available for the reference designs for TPS23755 or TPS23753A?
There are some transformers with very low winding resistance and inductance (e.g. Moatel EP13) that would require a complete recalculation of the circuit, so having that simulation coded up could be helpful.

Any recommendations you have on how to implement a PoE PD and isolated converter in the ~10W range where every effort can be made to reduce thermal impact would be very helpful.

  • Hi apullin,

    To improve thermal performance, a efficiency optimized design should be used. A 2nd power stage will not be necessary.

    For 5V, we recommend PMP9175 using TPS23753A. Feel free to copy. TPS23755 is not intended for lower output voltages.

    This design uses a synchronous FET rectifier instead of a diode recitifer to improve efficiency. You can see the efficiency plot of PMP9175 in the test report & compare it to the efficiency plot of PMP7355 (diode rectified 5V).

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer
  • Ah, very interesting. It looks like a good candidate to implement. That is a lot of parts, but I could skip the 12V-to-5V conversion step with this configuration.

    The input active rectifier is interesting.
    Any insight if it would actually beat 2 diode bridge rectifiers selected for very low forward voltage drop and just enough DC blocking? e.g. SL310A or even SL36A.

    Additionally, if you or the PoE team have any recommendations about new modifications that could be added in (since that report is from 2013),

    Lastly:
    Any chance you could supply the native-format schematic and/or layout so I can import it directly and work from there?
  • Hi apullin,

    Difference between diode bridge & active bridge is ~2% efficiency. Active bridge is more commonly see at higher power levels.

    The design is up to date & we do not have simulation files.

    See attached files for PMP9175.

    For more info on flyback design see:

    Regards,

    Tom Amlee

    PoE PD Apps Engineer

    PMP9175.zip

  • Awesome. Those design files are extremely helpful to accelerate us.

    Yet another follow up:
    In most of the Test Data reports that I look at for the PoE reference designs, the thermal images indicate that the transformers are the major heat locus. Is there some potential to reduce the thermal impact of the design by sacrificing some board area?
    It looks like CoilCraft does have some options that are larger, but all with different winding layouts and inductances.
  • Hi Apullin,

    Generally yes, using a higher power rated transformer will improve thermal performance. I recommend using an EP13+ package. Also, using more copper for the primary & secondary ground planes should lower temperature for all components.

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer
  • Great, thanks for the tips.

    CoilCraft POE300F-50LB seems like a potential alternative. However, the winding topology is a little different, and there is no center tap on the secondary, so I cannot implement the output SR as it is implemented in PMP9175.
    Would replacing the SR with UCC24610 be a workable solution?

    Similarly, the different resistance and inductane of that transformer does appear to change the operating point a bit.
    Going through SLVA305C with the specs for POE300F, I end up with a very high secondary current, ~10A. Which would be thermally ruinous.
    Working through is a bit daunting for me, so any insight there would be really helpful.
  • Hi Apullin,

    We use the POE300F-50 in the TPS23754EVM-420 shown below. You can order an EVM and evaluate its thermal performance at ~13W if interested (as it can go up to 25W by design).

    The topology is a driven flyback; however, yes you can use a synch controller as you mentioned since the transformer does not have a gate drive

    winding. The UCC24612 has an EVM that helps retrofit it into existing power stages.

    http://www.ti.com/tool/ucc24612-1evm