I’ve got an SS24 diode on there right now, as well as SO-8 FET and a diddly little inductor. 2A is about as high as I want to go until I install better parts. I’m still assuming that this can safely scale up to 5A+ using appropriate parts (based on results from the “new DRY” driver).
I need to expose the GND ring on the top of the PCB for the next revision of the 20mm board.
That is good news wight. Congrats! It was my felling that getting a larger board working and then trying to make it smaller was more likely to lead to success than trying to make it small from the start and there are plenty of hosts that can use the larger size. While there are hosts that can use a larger board it’s not feasible to design one in a larger size and then make it smaller.
Last night I robbed a couple of parts and tried to build a higher current 20mm board:
I used an R400 sense resistor. Needless to say, it didn’t work for crap or I’d have mentioned it last night. It was getting late and after the reflow it didn’t function at all. A quick check showed I’d installed the diode wrong. I swapped that around and got only dinky current (and it was clearly not doing buck regulation).
I had some suspicions about what went wrong by the time I went to sleep, but I wasn’t sure. Today I measured the board populated in the final config from post #78 (stock SKU 20330 config) to see the switching freq. It was approximately 222kHz, a totally safe number for the QX5241 according to the datasheet. I think proceeded to measure what I think might be the frequency of the build I described earlier in this post. It was difficult because the scope image is pretty crazy, so I’m not certain that I got it right. It looked like about 5MHz, way above the maximum freq for our controller.
My math backs that up, more or less. The formula from the datasheet gives about 4.4MHz for that configuration based on my set current (and some Vf numbers I made up). I think if we consider that the set current is never being achieved then 5MHz seems reasonable.
Bottom line, I’m now going to try bumping the set current up to… 5A? That seems like it would put me at <1Mhz up to maybe 10v Vin or so for a single XM-L.
FYI for my calculations in this post I’m using the arbitrary figure of 3.5v for Vf. [When it seems relevant I can and will adjust that to a more accurate figure, but for what I’m doing at the moment it’s fine.]
Good news! I used two 1% R100 in parallel for 4A and got good results.
This produced about 833kHz switching freq with 7.18v at Vin.
I got frustrated trying to stack an R250 in the middle of the two R100, gold bricks style. I was having a little obsessive streak and I figured I’d better quit before I damaged the resistors with excessive heat. I double checked my math and 4A drive current looked OK, so I went ahead and did the test.
The FET and diode remained cool or warm, but the inductor did become hot quickly. The inductor I used was BOURNS SRP7030-3R3FM.
Rough numbers:
3.22V (Vf) - measured with a DMM before the current shunt DMM
4.1A (LED current) - measured with a HF DMM w/ ~4in 16guage leads
7.00V (Vin) - reported by buck PSU
2.41A (input current) - reported by buck PSU
The input current is possibly questionable, the input voltage is likely accurate. Anyway if these numbers are correct we are looking at about a 78% efficiency at 4A with a scavenged diode (and an FET that won’t fit on the PCB ;))
That’s good news. Do you think the efficiency will improve with better parts?
Probably not a lot, but we’ll see. The inductor isn’t too bad I think. The FET is good. The diode is the primary suspect. It might be the one heating up, it’s really hard for me to tell. I don’t have the appropriate equipment to measure component temps.
I thought the scope image of the output looked OK at 4A. I went ahead and tried to run 6A but ran into problems. My little CC/CV buck PSU couldn’t supply enough current to deal with the buck driver’s input spikes (3A limit) at 15v which was as high as I wanted to take the driver with these components. I switched to the 5V line on an old computer PSU, which was a cross-load since I had no load on the 12v line. I have no idea how the 5v line behaved, but I definitely insta-poofed the emitter. I doubt that the cross-loaded PSU had anything to do with it, but now I wish I’d just hooked up a known-OK 12v PSU to the driver instead of a questionable 5v supply.
There’s little doubt that the emitter damage is from over-voltage, so I think I need to re-examine the scope trace at 4A and see how high that’s actually getting.
I’ll order some better parts ASAP.
EDIT: Regardless of what better parts do for efficiency, I expect that higher in/out voltages will improve efficiency. So driving an MT-G2 or a string of XM-L2’s should show better efficiency. Also, reducing switching freq will probably improve efficiency, so a higher value inductor would help but I don’t think we can fit a higher value inductor w/ a high current rating.
I briefly had a 17mm board functioning at 0.5A, regulated. (!!!)
Something I’m doing is killing… something. Not really certain what’s up, but it’s encouraging that the 17mm layout may be OK after all. I think I’m damaging either QX5241 chips, FETs, or both. I should have taken notes, I forgot exactly when it started working or when it stopped working. The important part is that it worked with the board populated in a fairly normal way. I tested it multiple times in the working configuration with from 6 to 12v inputs feeding an XM-L2, regulated pretty much rock solid at 0.5A using an R400 current sense resistor.
I’ll come back to it in a bit.
Looking forward to updates from the front 
No real updates on this, but I did play around a small amount over the weekend.
I had two 3.3uH SMD inductors. One was a Bourns SRP7030-3R3FM, I already installed that one on a 20mm board. The other was a Wurth Electronics 74437349033. When I reflowed the Wurth everything looked fine, but then I grabbed it and waggled it to make sure that it felt secure…and it broke in half. The encapsulating material just felt… clay-ey and crummy. The Bourns one looks and feels much harder. Anyway 74437349033 still works as far as I can tell (it’s on a driver that doesn’t function yet). The windings are still hooked up and look fine. That said, I doubt I’ll order any more of that particular one. Now that I check the datasheets again, Trise is much higher on the Wurth model. It’s also way taller and the top on mine was smudged!
In other non-news, I’ve had a Mouser cart open on a home computer for a couple of weeks, building up an order… that computer died, and the cart was tied to the session on that web browser. Crap!
I broke a buck converter on one of my HX=1175b's. So I ordered some QX5241 chips in the hope that they will be compatible with the circuit and have the same pin out. I'll report back on what happens.
They don’t have the same pinout, that much we can be certain of. I’m moving this conversation back into your thread though (#67), this info is more useful there! Short version: it’s probably QX9920.
Thank you wight.
I added a 22uF cap to the output and made things worse! (I installed the cap using an airwire between the positive and negative LED leads)
I'm using a 20mm driver w/ two R100 for a set current of 4.1 Amps. With no output cap I get a peak voltage of about 6.5v according to my scope. With the cap it looks like over 7v! The entire waveform seems to be at a slightly higher voltage with the cap installed and the ringing or whatever is definitely bigger. (both taller and longer)
My test XM-L2 on a Noctigon did survive this.
I picked up a cheapy LCF meter from eBay. The one I purchased (which I don’t actually recommend so far) is http://www.ebay.com/itm/121228917349 from xinyangshi. Identifying markings both on the PCB and in the firmware’s splash screen are:
2011.05.27
XLDZ
The screw terminals seem to be a problem, so I’m hardwiring the inductors now. I pulled two pins from a 2.54mm pin header and soldered them onto the bottom of the LCR meter PCB, then soldered each inductor onto the ends of those pins.
I measured:
Edit: redid some measurements.
keep goin and let us know when we can buy some!
I’ve decided that I really don’t like the Power-SO8 footprint I’m using here. Somehow I ended up using a footprint that’s 8mm long - that’s huge compared to a ~6mm long Power-SO8 package, including the LFPAK56 descendant. When you reflow the FET tries to pull away from the end where the gate pin is and just barely hangs on. The recommended footprint for LFPAK56 is 6.8mm long, I could go even shorter here I’m sure. Also less wide. This would increase the density of the board and make building it more of a sure thing.
I still think I’m damaging these controllers somehow - maybe just with sloppy work, I’m not sure. I’m going to do a little more testing today at ~3-4A and see where it gets me. If nothing else I’d like to release the 20mm board soon so people can play around with it.
I have some of These
They don’t look like the same larger board but I wonder if it has the same buck ic on a different layout and could tell us anything. The second pic in the link shows a large cap that might be 47microf. If you want one to look at or for parts or simply want me to open one up for pics let me know. I don’t like them as they have the 5-mode program I eschew.
It would be interesting to see pics of the larger board. The smaller board is just an FET PWM modes board, you should be able to remove it and still operate the driver as a dumb single mode driver. It seems that 9 times out of 10 (figuratively speaking) the controller will be either QX5241 or QX9920. If there is a sanded SOT23-6 on the board compare it with the pinouts I noted in post #104 in the HX-1175b thread. They are easy to differentiate: one controller uses a middle pin to control the FET (DRV pin) and a corner pin next to that for GND. The other controller uses a corner pin for DRV and the middle pin next to it for GND.
I’m almost beginning to think building some breakout boards for these components and putting on a prototyping peg board might save ALOT of frustration…once R&D with the breakout boards is done and working good, then build up the actual driver board in one shot…this way at least you can build multiple components and mix/match w/o all the daggum soldering and worrying about overheating/damaging the components.
Like this but OSHPark
You guys are kicking so much butt, but it bothers me that you guys are getting so close and stupid stuff like soldering and re-soldering sensitive components just seems to be tripping you up.
Would a slew of very simple breakout boards w/ header pins and/or solder points help at all?
I’m hope you guys don’t think I am trying to poopoo your work…but maybe peg boarding it up might help with the R&D aspect of building an experimental driver
This guy built a high current motor controller for a skateboard…his prototyp
e
It’s a good suggestion, and a valid point similar to what RBD has indicated a couple of times with the “start with a larger PCB” thing. Also maybe flawed in some similar ways :-(. As I’ve pointed out before, we have a functioning schematic. That is the schematic from the datasheet. We do not need to develop a schematic. Component values may need to be adjusted, etc, but I’ll be very surprised if the schematic actually changes.
Motor controllers have some fancy stuff, but they are also very much just high current PWM drivers (no step down).
Unfortunately I think that there are a couple of things that prevent us from using a breadboard for this:
So that’s what I think about that. 
I think you look over a lot more material concerning development of projects like this in forums and videos than I do. If you see other people doing this kind of thing (step up or step down driver development where the driver operates on the order of several amps) please do let me know. I’m definitely willing to be wrong about this.
OTOH here is the current news: