Late UAV Power Supply Post

posted in: power | 0

I don’t think I’m going to go into the belabored details about how this power supply is designed.

Here is the link for the github source files.  Sorry you’ll have to generate the BOM yourself from the xml file.

As an IEEE project, I decided to do something a little above-ordinary.  I decided to pick the most complicated buck conversion chip I could find and just go with it.  So here it is, a working buck-converter.

More or less, it is over-engineered to power the equivalent of ten Intel NUCs.  It was more or less my choice to over-engineer the board because my dreams of flying supercomputers (which may or may not have driven this website in general) wouldn’t stop.  But hey, we can also use it to power the entire ground station if we were to spontaneously decide that we did not need to use our giant 6s lipos for anything else.


I decided to create this 12S to 19V switching regulator after realizing that Kevin was probably not going to complete a switching and backup system in time.  This is not to say anything horrible about Kevin; he is a capable and very intelligent person.  But he is relatively backed-up with making other priorities work correctly.

It helps that I was also very sick and rather upset, so I get to miss having a Christmas dinner with my loving extended family for creating this power board.  For whatever reason, I tend to show my best somewhere else when I’m extremely upset.  My guess is that it is a form of procrastination.

Oh yeah!  I also shared by thought about creating a yellow-brick Li-Po laptop charger (like the Li-Po phone chargers you find at Walmart) from one of our crushed yellow-brick Li-Pos to my lead, and he approved of my taking home one of our yellow-brick Li-Pos.  I’m not entirely allowed to talk about the exact process with which it was crushed, but I assure you that it was a “manufacturing accident.”

Buck Conversion

A buck converter is analogous to a down-converting single-winding transformer in a certain sense.  The difference with this “transformer” is that it operates with DC voltages instead of AC.

So, you can imagine that the converter scales the “transmission voltage” from the batteries (50.4 volts at times) to an acceptable 19V for the laptop input.

I will not go into detail about the exact operation of a buck converter (mostly because I still think that inductors are magic), and this is the nearest explanation I can provide.  I find boost converters a bit easier to understand, to be honest.

Design Process

I selected the buck converter chip on a dare from myself, to be honest.  It looked like a relatively complex chip to begin with, but all the calculations were given to me.  Along with this dare was the fact that it had the following features that would benefit my application:

  • Current Monitor Output
  • Current Limiting Hiccup
  • 3.5A Robust Gate Drive Current

It was basically plug n’ chug from the equations given by the datasheet.  Although I wanted to significantly reduce the effect of parasitic capacitance by using a lower switching frequency, this became nonviable after calculating exactly how much inductor current it would require for an exceptionally slow switching current from my chip.  This resulted in my first calculation redo.  With the second set of calculations, I opted to use something closer to what Ti’s datasheet recommends (125kHz).  They (Ti) say that it is a relatively close balance between efficiency and board size, so I went with it.  It’s more or less a maximum magnetic flux problem with the inductor.  If I were allowed an inductor if infinite size, I would keep my original minimum-frequency calculations.

After the calculations were done, I opted to buy stuff from Digi-Key.  For whatever reason, some part of me wanted to buy mostly Japanese components. (I may have been a victim of the capacitor plague a few years ago.)  Fortunately, I offered myself a budget extension in order to keep things nice.  Apparently, the cost to develop a working test board is close to or even above the cost of the evaluation board from Texas Instruments!

Mistakes and Hot Metal

The first selected toroidal output inductor ceramic did not stand up to the high switching frequency (I did not know that it was only rated to 1kHz).  It would heat up (it was pretty hot to the touch) and lose its inductive properties.  So, a lecturer of mine suggested using the T130-2 (toroid size 130, mix 2) core from MicroMetals.  I took his suggestion; he is incredibly intelligent (and apparently winds all of his inductors himself).

During this transition, I also did a board redesign and made fillets in my copper pours over the board.  Granted, I have pretty complicated copper pours compared to a lot of other boards, so this was not exactly a trivial task.  I also added LEDs for power on indication (and brownout indication… but I only found that feature after I powered everything on with my test bench).

Final Remarks

I don’t know whether using $120 (more than that) of my own projects money was worth it or not.  This may stem from my dissatisfaction with my current progress or may stem from the fact that I did not have enough time to integrate everything needed.  But you can certainly say that the whole project turned a few heads!  Especially because I probably could have used a more expensive Texas Instruments “simple switcher” module (and made my life a bit easier).

Leave a Reply

Your email address will not be published. Required fields are marked *