Building a 47mm OD, 13A LED BUCK driver

[matt304]

This is what I'm trying to have produced. (I simply do not know where to start looking, hoping I'll find luck here )

I need a circular driver PCB that is 47mm OD. This PCB will have a solder pad on the rear for a positive spring to be soldered. It will run on 2-3C li-ion (typically 3). So let's say a 6-13V input range. It needs to provide vF of 3.5-4.0V, and 8A to up to 13A to the LED (2-mode low/high would be preferable), via current-limited adjustment. (I imagined a 'Shark Buck 3 Amp' driver times 4, printed on the board, based on what the Shark can do for its size and height)

The entire driver needs to remain as low in height as possible, too, for the LED heatsink/seal to sit over. The custom light body/head is very large, and could dissipate the heat well.

 

[HarryN]

What you are asking for is possible, but challenging to design and build, especially if durability and reliability are concerns. Getting such a board to be thinner than 1 inch would be very challenging due to the required inductor capacity.

As a practical matter, such a board would take about $ 3-4K to design, proto and test, and then cost around $ 100 - 200 each to build if done in a broad temperature range version. A decent proto could be produced within 4 months, possibly less. A single mode version (but adjustable to be 2 different current rates) would be much easier than a 2 mode version, especially in that current range, just because of the parts available on the market.

It is mostly a matter of if you are willing to front the money for the project or not, knowing that sometimes things don't always work out the first time.

I did a healthy boost driver together with another engineer and it is non trivial to make it all fit in that volume.

 

[jtr1962]

Heat is going to be a major problem. Unless you use synchronous switching, you'll probably be looking at 90% efficiency or less. With the numbers you gave, that means the driver will be generating in the area of 5 watts of heat. Given the small size, you're looking at a large temperature rise. Even with synchronous switching, I don't think efficiencies much over 96-97% are possible. That still translates to ~2 watts of waste heat. This doesn't sound like much, but try putting 2 watts on something 47 mm in diameter. It'll probably get warm enough to boil water. As HarryN mentioned, designing something like this is non-trivial. I had heat issues with a 2 amp driver which caused failures in the long run and it was in a 1" square potting box.

Unless you plan to make hundreds to thousands, I'm not sure it's worthwhile to invest the money needed to design such a driver. I would personally make the design problem easier by trying to match the LED voltage more closely to the power source. For example, if you use LiFePO4 cells instead if regular li-ion, each cell more or less is good to power one LED. If you use 3 cells, then the light engine would use 3 LEDs in series. That brings the LED current from 13 amps down to 4.33 amps to deliver the same amount of power to the LEDs. Moreover, since the voltages are closely matched you can use a simple linear regulator. This will easily fit in the space required. It probably wouldn't be dissipating much power either since the cell voltage closely matches the LED voltage for most of the discharge curve.

 

[matt304]

I appreciate the advice. Thank you both.

What if the light required one large LED, like a given LED by Luminus since they have some high-amperage LEDs... How are people driving them in flashlights? What if you are using 3x 32650 Feilong cells, as an example. Now this light body is also very large, the aluminum thick, 15 large cooling fins. The aluminum body has room for copper "seal discs". A cooling compound/material was placed in between the two, so the thermal transfer to the copper went into this heavy light head. Say I sacrifice, and also say 10.5A @ 3.7V. Now, there are currently produced boards that deliver 1/4 the desired amperage at 1/7 the component mass--something that appears promising. As circuits can be run in parallel, why not use a method of spreading the heat across more surface and mass, than thinking of it applied to open air in a central location for example. That was what I am considering in the design. If the amount of mass is adequate. And there is a lot of mass (2lbs around the head area).

 

[matt304]

I guess what I'm getting at is, a design like this ( http://dx.com/p/adjustable-step-down-module-blue-dc-dc-12a-242230 ) couldn't be placed on a 47-48mm OD board for under $3-4K? That just sounds, I don't know, I guess really high compared to what guys have had made for rocket projects to deploy altimeter charges where the boards are made in low quantity (1-3), and do not cost this much in R&D work. Couldn't the same parts as the example linked just be remapped, and the body/head becomes the massive heat-sink required? I mean, they are going for $11 in a rectangular configuration, you can see the parts which could be moved/removed to fit the same components on a board this size. Especially if pressed to a copper disc, and as I said, this light head weighs close to 2lbs alone. The air cooling fins are similar in weight and surface area to what would be found on a 30cc two-stroke engine to cool it. I realize the concerns for a typical board in open air cannot exceed such temperatures, but that is far from the case here, where thermal transfer is a big part of the board design incorporation. I guess I should have stated the size of the head this is going in on my first post. Sorry for that being left out. This is a very specific build, if you will.

I ordered an arrangement of 3A drivers, which will stack (4x 3A) within the light in prototype form, horizontally stacked on the base plate, and heat testing under load will be performed using an oscilloscope to monitor the pulse waves, and temp probes to monitor heat, after the prototype drivers have been thermal potted to the copper and sealed up in prototype #1 which I have here next to me.

Would it make more sense to produce a carry handle "fat/short" batt tube (like the 3x 26650 lights which are side-by-side batt configuration), run 1S3P @ 4.2V charged, to get close to the actual LED voltage? Problem is, as the batteries drain, they may run at a lower voltage than the LED, requiring a slight boost instead of buck. The LED would run very close to 3.8-4.0V.

Any thoughts on this idea, anyone?

 

[HarryN]

If a person is doing something as their own hobby, then the time involved to do it is zero cost. If you are buying something for your hobby, then normally you have to cover the cost of their time, which isn't zero. Heat can be dealt with using some special circuit board designs for dissipation, but those are not "normal" low volume shops and board designs, more of a specialty item.

You can't really compare what DX is selling to what it costs to build things normally. When I have added up their bill of materials from pretty high volume electronics parts suppliers selling "real" parts, just the parts price comes out higher than what DX is selling their stuff for.

Of course there are ways to deal with the heat of the driver conversion and making it round, but perhaps another way is to re-imagine your light in a different shape. For example, 4 series x AA or sub C cells are not that far from your desired voltage, and you can easily find a battery pack maker (for example cheap battery packs) that can make packs like this for you with very nice terminations.

A few packs like this in parallel can easily hit your 10 - 15 amp needs.

With some care, a group of pretty low end drivers, perhaps even just some decent resistors can take you close enough to the desired current flow.

In theory, packs of 3 in series x AA Energizer Li primary cells can hit a high enough voltage. If you run some wires through the center of the light, you could carry the needed current to the head. Running a high power light on primary cells can be a bit pricey, but at least it is simple and cheap to build.

While the head is often conveniently round, the light body does not need to be round, perhaps light with a battery box is viable for you ?

 

[RetroTechie]

If a person is doing something as their own hobby, then the time involved to do it is zero cost.

Ehm... NO.

Everything worth doing takes time.Time isn't free, the amount you've got is limited, and it's about the only thing you can't buy no matter how rich you are. So yes if you're doing something as a hobby the "bill" for your time reads $ 0. But you still need to ask yourself why you do it & what you get out of it.

(which eg. is one reason I hate spammers to the core - because they waste other people's time, on a massive scale. If you do the math in terms of people's lives wasted, they're like serial killers on the loose)

 

[The_Driver]

The "Ampere!"-driver vom the German company pcb-components has been tested up to 11A (default 9A or 5A - you need to switch out the sense resistor). It's only 25mm in width and 14mm high.

 

[HarryN]

Hi Retro - yes of course you are more correct, I was not trying to enter a philosophy discussion about what a hobby is, I was just trying to explain why the cost is not "essentially zero" to have other people spend their time on "your" hobby, such as trying to design and build a custom driver.

The Driver - the Ampere driver that you link to is really interesting and one of the few ways a person can make it that small. The price is also quite cheap for that much power management. I will be honest though, the way it is made, I have a hard time believing that it will be mechanically or thermally robust, and that un-shielded inductor combined with a variable frequency setup is going to be an RFI powerhouse.

I could be wrong, that is just what I see based on my limited experience building drivers.

 

[alpg88]

idk why you want to use luminos. unless you using the kind that has no dome and die sits on copper board, why not use mtg2, similar die size, more efficient, and run at lower current but higher voltage, and unlike sst90 there are copper direct path stars made for mtg, it is easyer to deal with voltage than with current. but that is just my imo

 

[The_Driver]

The Driver - the Ampere driver that you link to is really interesting and one of the few ways a person can make it that small. The price is also quite cheap for that much power management. I will be honest though, the way it is made, I have a hard time believing that it will be mechanically or thermally robust, and that un-shielded inductor combined with a variable frequency setup is going to be an RFI powerhouse.

I could be wrong, that is just what I see based on my limited experience building drivers.

First oft all: 35€ is a lot of money for an led driver. There are only very few on the market that are more expensive. While I am no expert on the details of driver design myself, people in the German TLF forum have noted that it is a very high quality driver an that it can not be sold for less money when made in small quantities.

idk why you want to use luminos. unless you using the kind that has no dome and die sits on copper board, why not use mtg2, similar die size, more efficient, and run at lower current but higher voltage, and unlike sst90 there are copper direct path stars made for mtg, it is easyer to deal with voltage than with current. but that is just my imo

The Luminus leds all reach a higher intensity (more throw) than the MT-G2. I agree with you on the heatsinking part. I don't agree on the part concerning voltage. There aren't many drivers that work with MT-G2. The MT-G2 also has an uneven surface on the DIE (little Dots), which can sometimes be seen in the hotspot when smooth reflectors are used.

 

[HarryN]

Ok, well I am not going to argue about price points, what I am saying though is that it is pretty hard to build a driver with that much power capability, rohs compliance assembly, that much efficiency, that size, and still sell it for 35 euro. If you doubt it, buy one and try just looking up the price for the individual parts, even in qty 100 or 1000. Honestly, it is a heck of a bargain if the specs and performance are real like you believe.

Building up a home made light with significant power can easily hit 300-500 euro in parts, some that work out, some that don't. It isn't cheaper to make your own light, just fun and educational.

Part of the education process is selecting battery + power control + LED + optics combinations that give good performance at reasonable prices. Perhaps your battery selection will need adjustment to lower the driver cost ?

 

[jtr1962]

I don't agree on the part concerning voltage. There aren't many drivers that work with MT-G2.

In this case you're right because using an MT-G2 would require a boost circuit. In general though if you need to deliver a certain amount of power to LEDs it's always better to increase the voltage of the LED string, at least up until the point where it gets close to that of the power source. Buck drivers get more efficient as the output voltage gets close to the input voltage. Moreover, you're decreasing the current the inductors need to deal with. I still say the best solution for many high-powered LED lights is to use LiFePO4 26650s (or even 32650s) and a linear regulator. Compared to smaller sizes, LiFePO4 in larger sizes don't have as much of a capacity deficit relative to LiCo. More important, each cell is more or less a close match to an LED's forward voltage. A simple linear regulator can take up the excess voltage with little waste heat production and no RFI.

 

[alpg88]

there are few very good\reliable boost drivers, just from the top of my head, taskeld hyperboost, hb flex, maxflex

if round shape is not needed, luxdrive flexblock will work just fine with 36v mtg2, it only needs 0,2A iirc. not a big deal for any of the mentioned driver.

if mtg2 are 6 and 9v, than you don't need boost driver (op stated he will power it with 3 li ion), almost any 3\2a buck drivers will work, taskled offers b3 flex, hccflex h6flex. derwitchel driver, 2,8A driver, and i'm sure there is more.
it is a lot easier find a boost driver that puts out 36v at 0,2A than any driver that puts 4v and 9-14A.

 

[The_Driver]

Ok, well I am not going to argue about price points, what I am saying though is that it is pretty hard to build a driver with that much power capability, rohs compliance assembly, that much efficiency, that size, and still sell it for 35 euro. If you doubt it, buy one and try just looking up the price for the individual parts, even in qty 100 or 1000. Honestly, it is a heck of a bargain if the specs and performance are real like you believe.

If a German company (basically any non-Chinese company) produces nothing but flashlight drivers then I trust them to not lie about the specifications. There are some reports though (in the German TLF forum) that some drivers were defective and needed to be replaced by the company.

Building up a home made light with significant power can easily hit 300-500 euro in parts, some that work out, some that don't. It isn't cheaper to make your own light, just fun and educational.

Believe me, I know. I have hardly any stock lights anymore because they don't meet my requirements (concerning light quality, heat sinking, how good they are regulated, beam form and quality etc.).

Part of the education process is selecting battery + power control + LED + optics combinations that give good performance at reasonable prices. Perhaps your battery selection will need adjustment to lower the driver cost ?

In this case you're right because using an MT-G2 would require a boost circuit. In general though if you need to deliver a certain amount of power to LEDs it's always better to increase the voltage of the LED string, at least up until the point where it gets close to that of the power source. Buck drivers get more efficient as the output voltage gets close to the input voltage. Moreover, you're decreasing the current the inductors need to deal with. I still say the best solution for many high-powered LED lights is to use LiFePO4 26650s (or even 32650s) and a linear regulator. Compared to smaller sizes, LiFePO4 in larger sizes don't have as much of a capacity deficit relative to LiCo. More important, each cell is more or less a close match to an LED's forward voltage. A simple linear regulator can take up the excess voltage with little waste heat production and no RFI.

This 5A 3-mode buck driver works great with the MT-G2. I have a light using this combination. You just need to use quality batteries and put a lot of effort into the heat sinking.

 

[matt304]

There is a person belonging to the other light forum who did this last year I found out. I was told the driver was also meant for a very large custom host, it is bucking 3S to over 11 amps to an MTG2. Or was doing that in the prototype test and the driver was specd to hit over 13A but was tested at lower amperage he said. It's not made to power an LED with that much current normally but is built with reserve power for a colored luminus LED that uses over 12 amps I was also told. Apparently the driver will be powering an Xpg2 S4 and the pictures of beamshots I got were mind boggling. 1,700,000 candela in a custom lens host and the aspheric lenses were high-end coated custom design that I have never seen anywhere like Thor or Edmund, not super domed like everywhere around but shallow, using a collar to recycle light. The lenses for as much as I can say about them looked beautiful. I thought Thor coatings looked good or maybe the best, until I saw these lenses.

I figured if I didn't post a photo of the driver no one would believe me so here is the photo I was sent (it's on the right side):