The Texas Buck driver series, Q8 / Skyray King 2S/4S buck driver RELEASED!

Not yet, spare time is tightening up. But a bunch of those thoughts came from Ti docs, so I guess they make something, but don't know the specs. Given enough time, I'm sure I'll take an interest to look (and understand the high current trade offs). At the rate you're going it may be irrelevant, which is great.

http://intl-outdoor.com/ld4b-24a-17mm-buck-driver-3v16v-p-817.html

hard to see exactly what's going on there. It's "only" 2.4A. But it seems the inductor is lofted over some of the components, and they made use of the hole in the middle. That might even be a nice driver.

That is a common way of mounting the inductors but it adds to the height and also those are not shielded coils. Both of which would cause issues that I think would be better avoided on the 46mm version of the driver. Once we have this size working that is an option to consider for the smaller versions if they are deemed a good option.

I haven't had much time but I did get to review that synchronous buck article more closely and the related IC's. It seems not so exciting. I had the right idea with I^Vf vs I^2Rdson as the main losses to compare, but that didn't give the picture. The main difference, and it makes the synchronous worse, not better( saw that wrong), seems to be in the afterthought there, that in discontinuous mode, the inductor still reverses flow through the low side FET, but not through the diode. So that dumps energy and makes at least a too simplistic synchronous buck bad in moonlight.

At reasonably higher current, I^2Rdson is apparently a better deal than IVf and the synchronous buck can win a bit. I suspect at some even more higher current this flips the other way again but that may be irrelevant.

Anyway, maybe kind of obvious, but diode losses (or low side FET) matters twice as much when driving 1S as driving 2S just because you spend twice as much time freewheeling for the low voltage output.

I guess the synchronous issues will get worked out with smarter switching, but anyway, the IC's I found for that are also much more complicated to implement and appear to be quite a big larger. So this lead nowhere new.

I found an in interesting article about the importance of the input capacitor and having it very close to the FET. I guess the IC doc covers that though.

Right now I am only worried about high current performance, that is what we have always had issues doing with buck drivers. Low current is easy and can always be taken care of with a 7135 if needed.

When you say input cap, you are talking about input for the LED or for the IC? Not sure we would have room for an input cap large enough to make any difference for the LED.

Not 7135, now you're following my mispeaks. If a 7135 reduces current from the buck, then it reduces current out of the buck and the buck has to operate at low current, and we're current regulating it anyway. You've only raised the output voltage slightly, still with the same current, and dumping the extra dVI into resistive loss while not improving anything.

But I made the same slip on the last page. What you mean to say is for lower current you do LED PWM. That needs and extra FET, but not a big one because you start with low current. This was brought up by a couple of people in the "more efficient driver" thread.

Anyway, the cap I was talking about was input to the FET, parallel to the source. In principle that does nothing since it's hard tied to the source (and the batteries probably have pretty enourmous capacitance themselves) but this note mentioned stray inductance and needing this cap very close to the FET because of that. The claimed effect was very impressive. I lost track of the link. I'll try to find it again.

I started (not completed yet) my own calculator based mostly on this:

http://powerelectronics.com/site-files/powerelectronics.com/files/archive/powerelectronics.com/mag/606PET25.pdf

A very practical summary of many of the basics. It's probably redundant with your spreadsheet but I'll compare later.

Anyway started looking at inductors. It seems to me there's quite a bit in 22mm * 22mm * 22uH. My initial impression is that 22uH would be nice, but size is still a pretty big problem there. It looks like you've started with a bit smaller footprint, which is probably sensible if it can work. Going shielded certainly takes more space. Aparently shielding above 200khz can be conductive shielding (thin metal, maybe tape) and I don't think you need that much space to get away from the fields. The irony is by requiring shielding you force things well up over 200khz anyway likely by not being able to get as much inductance. An open inductor with shielding tape, maybe extra added, at 300 to 500 khz might not be a bad way to go.

Anyway, I've only just started looking, and just started seeing how actual numbers fit in the math. Math aside I've seen 22uH actually used in a cheap 10W driver. So it's not an absurd number, maybe just absurd for 15A. I've noticed though that some of these inductors, like Vishay can keep going well above their saturation current. They just don't go as well, so things will get more wild and less efficient, but this is pushing up into turbo modes anyway, so so what.

12A (plus ripple) will run 2S 2P at 6A per diode. That's already a BUNCH. And it will run 4p at 3A per diode which is also plenty high really and you can still crank higher with some slop and heat. I've been aiming for 15 saturation so far, but it's probably higher than needed. This is why I would buck a Q8 at 4s batt 2s LED. It's the only way that makes sense to me. Many factors to trade off though.

http://www.eetimes.com/document.asp?doc_id=1273212

That meets more with my intuitive understanding of input capacitance, basically saying it shouldn't matter for a perfect source. I'll try to hunt down the other note.

But even my link in the post above, page 49, bottom, calculates a value for the input cap somehow without any reference to the source impedance or other source qualities. That's a bit puzzling to me but similarly to the other reference that I presently can't find.

I don’t have time to read all of that but your summery is making sense. Using another small footprint FET to PWM the IC is an option, although it does add more components to the cost/build and complexity.

If we selected the right FET we could hopefully adjust the footprint to allow for a jumper to be used if someone did want moon mode lower then what the IC can do on it’s own (I still think it will be enough).

22uh is the best bang for the space I have come across, lower I don’t think will work at all. Higher is great but they quickly get too large to be practical. even at 22uh we are looking at close to 1mhz switching frequency according to the spreadsheet (I didn’t make it BTW, just found it online, no idea if it is correct).

Now for total current, I think we need to aim for 15 amps to start out with, it can always be lowered easily later but higher is much harder. Plus if it is stable at 15 amps then we know it will work fine at 10 amps.

For the Q8 we are going to be stuck with parallel LED’s on the stock MCPCB unless a drastic change was made and this driver became the stock driver for it (don’t see that happening). So high amps is a must. with 15 amps that is enough to drive 4x xhp35’s at around 3 amps + a bit of headroom. Or 4x xhp50’s at 3.75 amps (pretty low for them).

Like I said we can always go lower but higher is not so easy. For example with a custom 4S LED setup you would only need ~5-8 amps per LED which would be easy.

I would much rather design it for overkill and work backwards. Tis how I do things. Honestly if I thought it was possible I would be aiming for 20 amps on the Q8 driver. Just don’t think that we have enough room for that though.

Ripple kills LEDs so it’s a good idea to think about dealing with it otherwise you have to set current max lower.

Yeah, it is something to be dealt with. It is mostly an issue of parts selection though as this particular IC allows for basically whatever ripple you want if you select the correct parts.

That said space quickly becomes an issue, as to eliminate ripple you need large components and/or very high switching frequencies. both of which provide their own challenges.

With the numbers I am working with right now, it looks like ripple should be able to be kept under ~250 mA, which is less then 2%. What it will do in the real world is yet to be seen though.

I think we can get one thing fixed (and that helps).

So I have a full caclulator setup now using that reference I posted. I did make it, but I didn't make the equations. I did check some of them and looking into others.

I'm not sure I agree now that 22uH is a must. I actually cannot find anything in 22uh above 12A rated (you can find higher saturation, so maybe good cooling can push one beyond spec) smaller than 22mm. I did find this comprimise: http://www.digikey.com/product-detail/en/vishay-dale/IHLP6767GZER150M11/541-1287-6-ND/ at 15uH 14A rated 17.15mm 12A saturation shielded inductor. I'm not worried about the saturation. At high current you can live with less inductance. This might be the part. (darn expensive though)

But I think we need smaller than 22mm. Basically everything above 10uH above 12A, shielded, below 22mm is 17.15 *17.15 or close enough to not matter. So I would say that is the footprint. What do you think? There are options here. You can get a 19A 10Uh if you want high power or a 12A 22uH if you want better low power efficiency, so the end user can decide.

Now as for 10uH, Plugging some numbers through my calculator, at 10uH you get 2 to 4% inductor ripple current at high power (15A) at 1MHz or 8 to 15% at 300 khz, but lowside cap of 10 or 20uF easily shields the output voltage from this. The problem is in low power, 0.25A per LED you start to get in the the neighborhood of 100 to 300% ripple current or more in 2S depending on frequency (200 means entering discontinuous mode I believe), 3 times less still in 1S. But that's not the end of the world. Ok, so at low power you enter discontinuous mode. We know that and have discussed how to deal with it. I'm still only getting 7% total power loss (my gut says it will be twice that from realities not calculated, edit: gut was right, had an error in inductor loss equation.) at the threshold to DCM. The question is how low and at what frequency? From the equations there I'm not seeing big switching losses, but I want to review that more. 1Mhz makes me concerned we'll see unpredictable losses from many parasitics we can't easily predict, but this probably isn't a footprint issue anyway.

As for the diode it does look like it will be the biggest loss (edit: in some situations), but it's ok and hardly an issue at all in 4S to 2S.

I'm not worried about 4p personally. Rewiring the LED board is simple, but I see your point.

Updating this post for the records:

Calculations are here: https://budgetlightforum.com/t/-/41130/124 (post 124)

Inductors: (done but not decided)

The three inductors to beat, all 17.15*17.15mm: SELECTED FOOTPRINT

http://www.digikey.com/short/39nz84. IHLP-6767GZ-11 IHLP6767GZER150M11 15uH 14A rated, 12A saturation, price $5.95 each SELECTED (but the other two in this size are viable options and fit)

http://www.digikey.com/short/39nz8h PA4344NLT Series Datasheet 22uH 12A rated, 18A saturation, 26.5mOhm, price $4.55

http://www.digikey.com/short/39nz8b IHLP-6767GZ-11 IHLP6767GZER100M11 10uH 19A rated, 17A saturation, 9.3mOhm ,price $5.95 each

If you can fit 22mm on the Q8 I definitely would. You can bring 26mOhm down to 7 (BIG efficiency savings, few watts in 3.5V output ) and get a full 15A rating!

http://www.digikey.com/short/39nz8r

Caps: (done) see post: https://budgetlightforum.com/t/-/41130/132 (post 132)

1210 10uF, for input and output. Maybe include 2 or three in parallel for input, (to save about three percent loss in 14V output)

I can find some cheap ($.30)

http://www.digikey.com/short/39nzjt SELECTED. (for input and output)

and some individually specced (rare, reading through these cap sheets is a pain) up to high frequency, even if not amazingly low ESR/tan delta/dissipation factor/high Q (average).

http://www.digikey.com/short/39nzj3

Diodes: (in progress, not quite given up yet)

Things to consider, primarily Vf, and max reverse voltage, forward current, but thermal performance is also important, including reverse current at high temperature.

A number of diodes in exist in POWERDI5060-8 like this: SELECTED FOOTPRINT

http://www.digikey.com/short/39n297 SBRT20U50SLPQ

http://www.digikey.com/short/39nzj8 SBRT20U50SLPQ

http://www.digikey.com/short/39nzjz STPS30L30DJF

This one has a particularly low leakage even when very hot (Vf might be a little worse though)*:

http://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeySTPS30M60DJF-TR

But there seems to be more selection in TO-263 (D2PAK) like this:

http://www.digikey.com/short/39n299

There are also TO-220 through hole packages like this:

http://www.digikey.com/short/39n299 (mouser has many)

Which may provide different heat sinking options.

Diode heatsinks (impact solder footprint and space):

The POWERDI5060 and D2PAK can proabably fit on the same pads if made a little cleverly and extra large pads are documented to help cooling anyway. The 2-wire To-220-2’s like the one above might be able to span the same footprint too if we really want a do-it all pad with through holes. Most of the TO-220’s are three wire though with the anode in the middle, which becomes a different issue.

*0.1A leakage current is 1.6W times duty cycle so 0.8W at 2:1 or 1.6W in 4:1) Some of these get up to that level when very hot! Not a big deal when using 84W output but keeping leakage to 30mA is better. Low reverse current seems worth probably 0.1V in Vf to put them in some balance.

MOSFET:

The trick here is rdson and gate charge divided by voltage it's measured at. Every 30 nC/10V (3 nF) produces about 0.11W of minimum switching power at 1Mhz even in low modes. Keeping this below about 6nF is very helpful, 3 is better

30mohm of Rdson however produces in the ballpark of 9W of power at 15A in 14V output, so getting this below 10mohm is very helpful, 5 would be better.

TA found this:

http://www.digikey.com/short/39nh5h

Which I think is probably fine, and the footprint seems easy to find things in, but I'l look a bit more later.

Update: gate charge is way to high.

This seems much better,

http://www.digikey.com/short/3bj1wp STL30P3LLH6

and this is probably a better compromise yet:

http://www.digikey.com/short/3bpr78 BSC084P03NS3 G (I'm liking this) SELECTED 8.4 mohm Rdson, 6nF gate capacitance.

Something seemed, wrong, was missing the square in I^2R in my inductor power loss.. Now this matters, getting up to 18% power loss total in worst case (high power, 1S, so highest output current) seems more right.. still working bugs out. This is actually a pretty big deal though and makes the other options more attractive than that particular 22uH inductor at least.

Percent power loss for the 22 and 15 uh inductors and the percent inductor ripple current:

2S 1S
1 MHz 300khz 1 MHz 300khz
Value High P Low P High P Low P High P Low P High P Low P
15 uH 14.4mOhm P_L % 2.67 0.40 3.17 1.82 6.44 0.72 7.10 1.79
delta I_L % 2.84 68.06 9.45 226.85 1.54 23.09 5.13 76.97
22 uH 26.5 mOhm P_L % 4.79 0.52 5.41 1.92 11.70 1.13 12.51 2.30
delta I_L % 1.93 46.40 6.44 154.67 1.05 15.74 3.50 52.48


Diode and switching losses bring that closer to 20% for 1S high power, plus whatever losses I haven't considered.

High power there means 15A, ~53W for the 1S and is 12A ~84W for 2S. Low power is 3.5W total in all cases.

1S is tough. In non-shielded inductors or in 22mm footprints, there are considerably lower resistance options available though.

3.5W is discontinuous mode at 15uH 2S, but 1Mhz keeps it alive, and maybe this fancy IC will anyway.

All very good info. It is possible to fit a 22mm inductor onto the Q8 driver but obviously any smaller drivers will not handle that.

While I want it able to handle 15A I am also realistic that we will most likely be overdriving something to reach that.

I am not sure what you mean when you say that rewiring the Q8 MCPCB would be easy? More like impossible? Now you could use individual stars as a replacement but that kinda of ruins the point of the nice mcpcb. It is an option for the extreme among us though.

Honestly my biggest goal is as an XHP35 driver with 4s input and 4s output and able to provide the ~3 amps per LED that they can handle with a low enough ripple that they don’t die.

If we can reach those goals with a bit of headroom it should handle anything else we may desire. XHP50’s would be best run with an FET+2 for the amount of power they need, same for the 3V emitter options. This is why I made the Texas Avengers for those setups.

Now I do plan to make smaller versions of this driver for other lights but it will obviously have much lower current needs as well. Just not sure what lights would be able to use them besides things like the L6.

First Cap footprints: Main cap, 35 to 50V rating, requires 1210 (or 1206) specifically this:

http://www.digikey.com/product-detail/en/samsung-electro-mechanics-america-inc/CL31B106KBHNNNE/1276-6767-1-ND

Maybe use the same cap before the FET.

If you can fit 22mm on the Q8 I definitely would. You can bring 26mOhm down to 7 (BIG efficiency savings) and get a full 15A rating!

http://www.digikey.com/product-detail/en/wurth-electronics-inc/74435582200/732-4237-6-ND

Worry about smaller ones for smaller drivers. We’ve found a few, easy to come back to them. Further I would add through holes to allow optionally mounting unshielded toroid inductors unless there’s really no place for them to exist. It will give options for testing/learning/improving and that might help down the road.

It’s not quite true that 2S is a slam dunk after doing 4p. Yes I^2R and IVf loss/heat get easier, but inductor ripple is maximum at 50% duty factor, about 3x worse than for 25% duty factor, so you enter discontinuous mode earlier and either need to start pwm earlier (not so bad), or you need more inductance or higher frequency, either probably at the expense of some losses for the same footprint, but cheaper than the same game at 1S. Probably PWM earlier is as good of a solution.

Like you, I don’t see output ripple being a big issue. As I interpret the capactor dissipation factors, I’m getting values of next to nothing (1 mOhm) for ESR at 300khz. I’m inclined to stick 20mOhm in the math just in case, which still leaves output ripple under 2%, actually a tiny fraction of that at high power. The hard part here is likely good connections and traces.

Anyway, Rufus, ripple kills LED’s? and PWM doesn’t? I don’t believe it. 6A modes with 10% current ripple kills LEDs. That’s not exactly the same as “ripple kills led’s”. Of course we're talking about ripple voltage though, and that translates to more in current, still it's overvoltage/current that kills it.

As far as PWM, actually it probably doesn’t need another FET at all, just use the PWM port on the IC as it was designed. You just need one trace and a spare mcu pin, no component. Then it's up to software.

Ok, making 2S is probably somewhere between easy and impossible, I haven't looked. You should be able to cut traces (no?) and solder onto the top connections of the LED’s (after scraping the coating). Apparently that’s a pretty big pain after they are already on a heat sink. Or just sand off the trace coverlays and solder to the board. This is really too bad though. Why not have jumpers/solder pads on the LED boards? Is it really too late to do that? I doubt there can be a more capable driver with present tech, than a 4s to 2s buck. You can get melt-down modes, and suck the life out of the batteries to maintain intermediate modes, although moonlight will probably end up slightly more efficient on the 1S. so like 500 hours instead of 350.

I just started searching for schottkys. I way underestimated Vf. At 0.55V (ok 0.45 at 15A, less if it warms up well) ... schottky losses are looking very big in 1S output, about 9.5% vs 3.1 in 2S. I did find a synchronous IC that could work (bypassing this loss), but it was constant voltage, divider style, only. Might be dimmable with some hack.

There are 10uf caps available in 0805 as well, I have some that I ordered for C1 at some point.

22mm should be able to fit but would leave very little extra space for other components so we need to make sure they are all in order first.

Far as 1s vs 2s vs 4s. In the Q8 (what this driver is really meant for), it comes stock with 4x 3535 LED’s in parallel and another 4x 5050 led pads as optional, also in parallel (assuming that Thorfire uses our design).

The MCPCB design is setup for parallel only and switching it would be much harder then simply installing new DTP stars.

So with the setup it comes with the 1S 3V emitters it comes with are best driven by an FET as they will be needing 16-20 amps of current, while this buck driver could do that it would be a lot of extra cost and effort for very little gains unless the LED’s were rewired in series with individual stars, that is not worth it except to the extreme modders.

The 5050 pads are basically for XHP50’s, they are once again best driven by an FET driver due to the fact they could easily use 20+ amps. Once again this buck driver would work but if max light output is your goal we all know that the FET will win.

Now the last major mod idea is XHP35’s, they are obviously 4S voltage only and can not be driven with an FET due to the fact they like to fry like that. So that is where this buck driver comes in, it is to allow XPH35’s to be dropped into a Q8 for a high CRI 8000+ lumen monster of a light without a donut hole issue.

So 4s input to 4s output is the primary concern, the others are great to play with later and if it works with 4s to 4s it should work with any of them given component changes. The XHP35 is the only setup that requires a buck driver, the others would just be for the extreme modders that want the very best, even if they are spending as much on the driver as the light itself. Although most of them will want an FET anyways.

I had thought the PWM port on the IC should work but figured I had missed something.

I'm sorry, yeah, I misread (maybe a few times) as 4p being primary concern I see (didn't really know about xhp35's and hadn't looked yet). Well I still think 4s to 1s is nice even if 2s might be "impossible". Max output is not all that matters. The buck will give you controlled output much longer (in anything other than 4s to 4s at least). Can that light really sustain more than 50W or power without burning your hand? Turbo shine is neat, but for long use at say 30 to 50 W without battery sag getting in the way, a buck still seems useful. I'll have to get my hands on the light. It's hard for me to imagine how it could be so hard. Worst case you resolder the LED's with their bottom pads grounded down and use the top contacts to connect wires, unless their is really no space for a wire. Mildly complex mod but so is assembling one of these drivers.

Anyway, I will add 4s to my scenarios. I put in component specs and get out performance specs for as many use cases as I want. Still, all these scenarios have different difficulties. I don't think any one makes all the rest optimized. In 4S the diode loss won't matter at all for example, in 4:1 it's a big deal.

As for the caps, yes, I've found many in 0805. I didn't find any with high voltage rating. I guess they should be 30V minimum, but I definitely could have missed them.