PMP8740: NO Load Dissipation For PSFB MOSFETs

Hi Aneesh,

1) All MOSFETs from primary side of the full bridge should have smaller losses at light load compared to full load, even if ZVS is lost for Iout < 20% of nominal. If you experience these temperatures, you have probably shoot through or not enough dead time in these MOSFETs. Please double check by watching the switching nodes (center point of each leg) and the gat driving as well. If this happens, you might have the dead times on primary side dependent on the load (which in my PMP8740 I don't have, because they are fixed). So, please, disable the dependency from the load and/or increase the dead time. Also check out the dead time between primary and sync-FETs.

2) The PMP8740 works in peak current mode: this is necessary to balance the flux in he transformer and avoid saturation. It is possible to add a capacitor in series to the shim inductor, but only if you change the modulation from peak current mode to voltage mode; if you add a capacitor and still running peak current mode, there might be slight unbalance in the voltages during the two half cycles (Ton), resulting into subharmonic ripple on output. In addition to that, the capacitor balances the flux in the transformer and also the peak current mode modulations wants to do that; this might result into fighting against each other.

Best regards,

Roberto

Hi Aneesh,

You wrote:

"Is my MOSFET heating because of more dead time added?, Because as per Figure 29 and Figure 30 in datasheet Current sense voltage will reduces the dead time right?. So if dead time is less i should get more problem at high load"

In my board, the CS (current sense) signal is added by means of resistor R8. As you can see, this resistor is "DNP", so is not populated. This setup eliminates the dependancy from the bridge current to modify the delays AB and CD. Do you have the R8 populated?

If you also have R8 as DNP, then the total delay will be defined by the formula (3) in the datasheet Tab_set=5*Rab/0.26 (when CS=0), where Rab is in KOhm; that means, at Rab=82KOhm, the AB delay is 1.577 microseconds. This delay is very long and, and I do not suggest to keep it. Maybe, since the delay is so high, the switchnodes are swinging up and down and the low side FET is turned on when it reaches the highest value (involving high switching loss). Please double check how the switchnodes at the bridge look like.

BTW, as starting point, I suggest to employ the values I used in the PMP8740.

My board works in peak current mode. The way to change from peak to voltage mode is explained in paragraph 7.3.11 of the datasheet.

Best regards,

Roberto

Hi Aneesh,

What kind of light-load problems you experienced when you used 49.9K on DELAB and DELCD resistors? Were spikes in the current? This might be coming from lost ZVS and hard switching. Anyway this is normal, and is reduced by using the latest CFD or CFD2 technology from FETs. You can check out if the converter works in zero voltage switching or not, by watching at the same time the VDS and the VGS of each MOSFET: if it is driven when the VDS is zero...it´s ZVS :-).

So, I suggest to work with lower delay times, and solve the issues with noise (by filtering, for example).

Now "Connect ADEL Pin with a resistor divider to CS Pin(Mount R8 In PMP8740" means that the higher the load current, the lower the delay times, because the higher the primary current, the faster the switch nodes are swinging from ground to Vin, and the lower is the delay needed at the MOSFET.

In addition: your points 2.1 and 2.2 are correct. For sure working in peak current mode by setting the controller with Rsum resistor, and adding the series capacitor to shim indutcor is not recommended.

Regards,

Roberto