17mm QX5241 5Amp 1-4 LED 2-4s Li-Ion Buck driver, WIP - I'm back in the saddle.

Custom Eagle parts makes sense. I’ll have to take a look at these PCBs when they arrive (since I know exactly what Eagle parts I used) and compare with real life 0805 components.

While looking at other drivers I recently realized that there’s no reason for me to fret over the need for a big inductor. I’ve been looking for wrapped toroids like we find on the DRY driver and original SRK driver, but I wasn’t able to figure out where to shop parametrically to find high-current toroids at the desired inductance. Turns out there’s no reason not to plop a (phsyically smaller!) SMD inductor right on top of all the other parts. It’s big enough to easily hookup with an air-wire.

This solves half of the crisis I was running into… the parts on the new-version DRY driver are probably spec’ed for 3-5A, it’s hard to say. zeremefico showed the driver running at 7A, but he did not discuss heat. I assume that at that current the diode and inductor are putting off some heat due to being under-spec’ed. I know people are going to want more current. A bigger diode is possible with a redesign and easy to source. We’ve already got the FET bit mostly figured out, that’s old news. So the missing link was the inductor. :slight_smile:

Big inductors still cost a few bucks and the corners start to poke out past 16mm or so, but I think this is doable.

Just realized that I’ve been developing this driver on a 16mm PCB, not 17mm.

FWIW the lfpak56 is about half the size of the larger fets and is rated for 100A if you find board space tight. There’s an even smaller “33” rated for 60A. The design is only a few years old so it might not go out of production soon like the Vishay did.

Thanks RBD. I’m going to look into that package, space is tight if I try to put an SMD inductor on the board.

The particular DPAK Vishay product we were buying on eBay was no doubt out of production long before we started buying it. Powerful DPAK/TO252 FETs will probably be produced until after we are dead. There should really be no fear of the supply of DPAK FETs drying up.

Well, thanks again RBD. That was a great suggestion, and together with making my work area an actual 17mm circle instead of 16mm (!), that freed up a lot of space. The pin configuration is also different, in a good way. I’ve done a total rework on the board.

  • SMD inductor. There is space for approximately an 8x8mm inductor. I will rework the pads to better accommodate various landing patterns.
  • ~1.1mm uncovered via to add a big wire-wound toroid inductor instead of the SMD one. This is just an option in case that makes sense for a build.
  • Large +/- LED pads on more or less opposite sides of the board.
  • Swapped diode to top of the board. Lots of space for the diode. You should be able to fit various ~5A diodes on those pads
  • Moved MCU to the bottom
  • LFPAK56 / SOT-669 / Power-SO8 style FET - there should be many options for this
  • Three 1206 sized sense resistors, still on bottom.
  • The bottom now has a smaller component height. You should be able to use a standard Nanjg 105c spring (5mm x whatever) or a standard 5x2mm brass peg. I’ve removed the wings from the 5mm BAT+ pad though, so in order to install it you need to put it on before the sense resistors. There is a large copper pour between them, just heat the pad using the sense resistor pads.
  • I kept the offtime cap, but I also brought out PB2 on top of the board. It’s available with a covered trace between inductor and FET - just scrape off the resist. This can be used for an e-switch, but like I said before this driver will have a really high parasitic drain and is a bad choice for an e-switch. I’ll probably remove most or all of this trace when I rework the inductor pad area anyway.


So here are the gotchas:

  • While the new-version DRY board has been run up to >7Amps by zeremefico, the diode will probably generate a lot of heat and could fail. AFAIK diodes in this style of package are only available up to around 5A, other packages such as DPAK are used for larger diodes. Therefore the current revision of this board is intended for ~5A or less operation.
  • I don’t have a complete grasp on inductor selection. The QX5241 does not have a fixed operating freq, the freq varies due to various factors: Vin to Vout ratio, drive current, inductance, maybe other stuff. The maximum the QX5241 can handle is 2Mhz. With the space limitations of a 17mm board, I think the highest inductance we can hope to achieve is 6.2uH with what I assume is a relatively expensive Coilcraft inductor. For cheap options we are limited to 4.7uH or less. Those inductors are readily available in the $0.65 to $2.00 USD range in values that can handle >5A without burning a lot off as heat.

Not a gotcha, but the overall size of v020 is similar to that of a Nanjg 105c - this board is about 1.0-1.5mm taller depending on inductor selection.

On the FET’s that I have the drain is the big tab and bottom plate, the gate is the single pin you have as the gate, and the source(ground) is what you have connected to the diode.

Well that’s bad. Let me go back and look.

I had the part right, it’s as you say. I got those backwards when I hooked it up in the schematic. Grr.

Better now than later. :open_mouth:

Absolutely. It’s actually no big deal, there’s so little going on on the top, and so much space, that it’s just a non issue. I’m wasting a little time making things decent, but I’m not going for really nice at this point. I’ll post a new version in a few.

I am surprised mattaus has not poked his head in yet. :stuck_out_tongue:

looks good so far to my untrained eye.

Thanks! Maybe he doesn’t like the cheap, barely documented Chinese IC this thing relies on? :wink:

Here’s the update:

  • LED+ got a 3mm pad with room to scrape extra. LED- would fit the 3mm pad, but looked pretty hacked up because of the other tstop items. I just left the whole pour bare.
  • I rotated the inductor and put large covered copper pours around the pads. A little scraping should allow for any pad layout an inductor this size can have.
  • I moved the covered switch pad (PB2) to be located right below the gate pin on the FET, it has kind of an L shaped area you can scrape if you want that pin.


There’s a part connected to pin 7 of the mcu that looks to have both pads connected to ground. Is that correct?

Nope. :bigsmile:

I (almost) wanted to redo those GND vias anyway. They were bothering me.
EDIT: scratch that. I’m totally not “re-doing” the ground vias right now. I did remove the offending vias. As you can probably make out now, that’s the middle of the voltage divider for battery monitoring.

If the space is available, I will spin the QX5241 90 degree clockwise which will shorten the trace.

What parts get hot from losses?

Ideally none ;-). In practice, only parts on top of the board should get hot under any circumstances. I’d expect problems to hit in this order:

  1. diode
  2. inductor
  3. the FET won’t get hot, but that’s next on the list

Good suggestion. I tried this before, but I was unable to make it work. Since you suggested it, I looked and you are right: that trace seems much too long. So I tried again, this time I made it work.

  • rotated QX5241 90° as Microa suggested
  • Moved QX5241 closer to sense array
  • moved smoothing cap for QX5241 to left
  • moved offtime cap to top of board (so PB2 is automatically available on top now)

Obviously several traces and vias moved to facilitate the changes. Everything passes DRC (other than all the intentional overlaps & stuff). Minimum trace width on the board is now 13mil (between the ATtiny13A pins), DR allow for a minimum of 6mil.


So! Good news on two fronts.

One is that I finally got my brain-cells lined up and did the freq calculation. There are three main factors we control which affect the frequency (remember we are capped at 2Mhz):

  1. inductor L (inductance in units Henry)
  2. input voltage minus output voltage (a wider gap means higher freq)
  3. output voltage (higher means higher freq)


Clearly when building a driver one must consider the #2 and #3 factors when choosing what to use for the inductor (#1).

From what I can see, the absolute minimum inductor we could use while driving a single XM-L (call it 3.4v) from an 18v source would be ~1.414uH. Working in the other direction, I find that if we use a 3.3uH inductor that puts us at about 856kHz with the same XM-L and 18v power source.

Driving an MT-G2 at 8A it seems to me that a 3.3uH inductor will give us a frequency of about 545kHz.

Two is that as far as I can tell, with some help over here, there are perfectly good Schottky diodes in the 8-15A range in a package called POWERDI5 which will fit on this PCB just fine. I borrowed an Eagle package from Robert Starr and added some legs to it to better represent what we’re actually laying down on the board. I think everything fits quite nicely. (At 5A many packages are available, there’s plenty of space here for any with this general shape.)