Post Go back to editing

LTC3784 (LTC3787)- 1200W 4 phase FCC design - Hot inductors while idling, no load connected

Hi Linear-Staff,

we designed a 4-phase layout with the LTC3874 in a 2/4 phase arrangement. Everything seems to perform like expected but the inductors getting really hot when using FCC with no load connected.

The design is made with the following specs:

fsw: 230kHz

FET/Sync: Toshiba TPH2R506PL

inductors: Coilcraft XAL1580 3uH

DCR sensing

VIN: 10-45V


IOUTmax: 23A

POUTmax: 1200W

Using Burst-Mode or Pulse-Skip of course wont heat up the inductors. Any hint? I can't see any frequency-jitter, compensation seems to work fine as well (with rather low bandwidth of roughly 4kHz (70+dB PM) at 10V VIN). The loop compensation is design made with LTPowerCAD.

Iripple is about 13A per phase at VIN=10V, shouldn't be an issue for the inductors.

I haven't spoken to my local FAE as this is a proof of concept student project which may go into production in the future. (No time to market yet)

Any ideas?

Best regards, Christian -
  • Hi Christian:

    That is not a good inductor and the inductance is on the low side.

    You can take a look at the components used on our LTC3784 demo board (DC2061A).



  • Dear ,

    thanks for the reply. I suspect the gain of knowledge by just getting the EVM is zero to none as it doesn't address the actual problem. I studied the EVMs design and part choice while designing the actual implementation from the begining. Even our FAE had an initial look at it. Would you mind to explain why, in your opinion, the Coilcraft XAL series isn'ta good choice with this IC/implementation? The inductor series is listed as recommendation within LTPowerCAD as well. For sure, 3uH is a bit on the low end and can be increased in value. Even Coilcraft states that these are "optimal" for VRM/VRD application.

    I scoped the ITH pin for PS and FCC configuration at idle and 50W load. The AC-coupled results are shown below. (These measurements where taken with 4u7H inductors (I changed the XAL1580-3R0 -> XAL1510-4R7 to see the benefits on core and lowered Rcomp from 6k2 to 5k, everything else is unchanged)

    Beside this, is there any way to use Burst-Mode with more than 2 phases?

    Best regards, Christian

  • Out of curiosity i compared the XAL1580/1510-series inductors against the SER2918H-series used on the EVM.

    In general, Coilcraft states better loss performance for the XAL1580/1510-series as shown in the attached data.

    To see the losses in detail, reduced ripple current for higher inductance values must must taken into account. So i compared XAL1580-302 vs. SER2918H-332H in the 3u0H range, and XAL1510-103 vs. SER2918H-103. The results are shown in the following table:

    The results are based on Vin=10V, Vout=51V.

    While this is true for core-losses at idle, the XAL1510 will loose it advantage at full load current due to x3 the DCR resistance.

    Getting back to the 3uH class, how does the XAL1580-302 performs against the SER2918H-332? Well, they do pretty well:

    So again, why do you think the XAL1580/1510 series is a "bad" choice? Using Pulse-Skip this XAL1580 inductors should really do the job, no?

    Best regards,

  • Hi Christian:

    The XAL series has high core losses. That may be OK in some applications but it will result in higher temperature rise in your application.

    The inductors types specified in our data sheets and used on our demo boards have all been selected for low core and DCR loss.

    You can use LTC3784 in place of LTC3787 if you would like pulse skip mode in 4-phase design. The OV pin will allow the 3784 to run in pulse skip mode.



  • Hi Goran,

    please see my updated post above. As the Coilcraft tools always take frequency into account, i'd bet they take care of the AC-losses as well.

    Did anyone tried using the LTC3784 in Burst-Mode + Pulse-Skip for the slave controllers?



  • Hi Christian:

    Have you tried the two inductor types in your circuit?

    The 2918 is I think bigger part with higher core volume. The power loss calculation may not be comparing apples to apples.

    Also, Coilcraft prefers that customers use XAL type of parts because they are cheaper to manufacture and provide higher profit margin.

    The XAL part core is going into saturation from the very beginning which can be seen in inductance drop versus current graph.



  • Hi Goran,

    I haven't tried the SER2918H in our implementation yet, as i don't have the parts at hands.

    "Also, Coilcraft prefers that customers use XAL type of parts because they are cheaper to manufacture and provide higher profit margin."

    Might be, but i don't suspect them to provide wrong performance data in their evaluation tools just to place them in front. To my knowledge, the XAL-inductors are made as flat-wire aircoil-construction (which normally won't saturate) which then get "doped" by the surrounding composite.




  • Hi Christian:

    I don’t know how well the Coilcraft tool works for core losses. You can ask Coilcraft for details.

    Core losses are non linear and modeling the losses is not trivial and can result in big errors.

    If you run in pulse skip mode the core losses will be low. However, I think, at full load the core losses will be higher with powder iron core.

    In any case, I think that the only way to really find out is by comparative testing on the bench.