In: 21-32V

Out: 12V, 3A

Working Frequency: 300 KHz

Controller IC: FP5138 from Feeling Technology


Schottky Diode: MBR20100CT from ON Semi

The MOSFET and the diode are going to be isolated from the heatsink. Should I connect the heatsink to the ground via two screws that are shown below?

Should I connect four mounting screws in the corners of the PCB to ground?

I tried to minimize the length of the critical traces but the heatsink didn't allow me too much. In addition to that, I kept the area of these critical traces that can act as an antenna to a minimum -only as big as to carry the currents. Also, I tried to keep output capacitors away from the heatsink as they are electrolytic and they should be away from the heatsink.

It would be great if you can draw the main current loops!

Schematic Everything Top Bottom

  • \$\begingroup\$ Your heatsink is way overkill, IMO. With the given voltages and current you should be able to get over 90% efficiency, so that the complete regulator will consume around 3W. I'd pick smaller heatsinks for MOSFET and diode. \$\endgroup\$ – stevenvh Nov 14 '11 at 10:05
  • \$\begingroup\$ Well, with this IC and this design, an efficiency over 90% is a dream I think :) My guess is somewhere near 85% \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 10:08
  • \$\begingroup\$ With an LM3150 95% seems possible, see my answer. \$\endgroup\$ – stevenvh Nov 14 '11 at 10:26
  • \$\begingroup\$ Well, LM3150 is about 3$. We have chosen this IC because of its cost basically :) \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 10:44
  • 2
    \$\begingroup\$ Something is screwed up with Q3. It looks like you were trying to make a push/pull emitter follower, but not with Q3 as shown. Perhaps you confused emitter and collector? \$\endgroup\$ – Olin Lathrop Nov 14 '11 at 14:23

Added: Points raised are being addressed. Have left most in and tidied as example of things to consider and that have now been considered. Added PFET level shift comment at end.
[This comment is for anyone following this - NOT as an edit trace].

You do seem to be trying and have a general grasp of what is needed but, no rudeness intended, the circuit shows several signs of very major lack of design. You need to think things through MUCH more carefully.Fine detail cannot be looked at until you get the basic circuit details correct. As shown it will not work AT ALL for several major reasons.

Drive polarity wrong: The IC is able to be configured as a buck converter but as shown the output drive is of the wrong polarity if you use a P Channel high side switch and of the wrong voltage swing if you use an N Channel high side switch - see below. If using a P Channel MOSFET (which would be normal here) the output drive needs an inverter in it. As shown it will not work.

Level translator needed in driver: If you DO run the IC off a 12V supply (and Vdd max = 15V) then the inverting driver which you have not yet got also needs to level translate as the MOSFET is high side and gate drive needs to go to 30V or whatever for MOSFET turn off. While addressing that do also ensure that MOSFET Vgs max is not exceeded when driving.

FET is wrong type The MOSFET is nice enough BUT is N Channel (as befits the incorrect topology that you are using). An N channel MOSFET COULD be used there but the gate would need to be driven above V+ rail and you would need a gate drive supply. The overwhelmingly usual thing to do would be to use a P Channel MOSFET as the switch

Output diode The output diode is very nice but is "overkill". The high max voltage leads also to higher than necessary forward operating voltage. You can probably get a few % more efficiency end to end with a lower voltage Schottky.

At a glance without pouring over the details the IC looks competent and should be capable of good efficiency as a buck regulator. I'd expect 90-95% to be achievable once the circuit was correct.

Driving high side PFET.

Vin max = 32 V (specified).
Vdd Ic = 12V (user specified) or 15V abs max.

PFET will have a Vgs max. Above that you get magic smoke.
As PFET source is connected to Vin+ the Vgs is measure relative to Vin+.
PFET gate can be driven low BELOW Vin+ by Vgsmax - ideally a bit less.

FET's that are not logic FETS often have Vgsmax of 20 to 25V.
Most FETs are totally "enhanced" (aka turned fully on) by the time they have Vgs = 12V - see curves for FET of choice.
Let's set Vgs max actual = -12V relative to Vin+.
This means that when Vin+ = 32V, Vgs may range from about 32V (FET is off) to 32-12 = 20V (FET is hard on).
BUT available drive voltage at IC = 0-12V approx.
So a level shifter is definitely needed.

  • \$\begingroup\$ Thanks for the great info! I could (and probably should) have understood that NFET is the wrong pick when I simulated buck topology in the LTSpice, because it didn't work when I applied 12V square wave between gate and ground, if I had a better understanding of the working of MOSFET. \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 12:33
  • \$\begingroup\$ I tried to achieve 12V for IC with 78L12, it is at the most right. Maybe you should scroll in the big image? Or if you saw it, isn't this a good solution? \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 14:23
  • \$\begingroup\$ Hey, I have one question. When using PFET with the inverter, vgs goes up to vin. As far as I looked up, most of them have +-20V vgs max rating. Do I need a level translator? \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 15:48
  • 1
    \$\begingroup\$ @abdullahkahraman - see added anmswer re PFET level shifter. Summary - YES you need one as FET is up to 32V above ground but available drive is ~= 12V max. \$\endgroup\$ – Russell McMahon Nov 14 '11 at 22:19

You do realize that the regulator you've chosen in a boost regulator, meant for creating a higher output voltage from a lower input voltage? And that your input voltage is beyond the regulator's range (1.8V to 15V).
I think it's time to reconsider your regulator, and read the datasheet properly. Personally I never heard of Feeling Technology, but that's not necessarily bad, provided you can get the parts. However, I would go for more common manufacturers, like National or Linear Technology (the latter are more expensive).

Also, your heatsink is way overkill, IMO. With the given voltages and current you should be able to get over 90% efficiency, so that the complete regulator will consume around 3W. National's Webench has a design suggestion around an LM3150 with an efficiency of 95%.
I'd pick smaller and separate heatsinks for MOSFET and diode, so that the heatsink's shape doesn't compromise the layout. This model

enter image description here

is only 6.4K/W, so that the temperature won't rise more than ~15°C.

About the layout: the loop L1/Q1/D1/C7,8 is critical, and should be much shorter than you've placed the parts. I realize that this has to do with Q1 and D1 being mounted on the heatsink, but that's one more reason to see if you can't do without the heatsink altogether. Two reasons for having the loop as short as possible:

  1. Efficiency. There will be large currents flowing through the loop, and you want to keep resistance losses as low as possible, and
  2. EMC. Radiation won't be as high as with a regulator working in the MHz range, but even at 300kHz this may cause interference.
  • \$\begingroup\$ Actually, I didn't realize that this one is a boost controller. It do have a buck regulator example in its datasheet though. \$\endgroup\$ – abdullah kahraman Nov 14 '11 at 10:48
  • 1
    \$\begingroup\$ @Abdullah - Yes, I saw that one too. But a buck DC/DC converter with a buck regulator IC will probably have the higher efficiency. \$\endgroup\$ – stevenvh Nov 14 '11 at 10:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.