Uh oh... Driving 2S2P array with a 7.1Vf ?

[MrNaz]

I am working on a dive light, and my benchtop power supply just arrived. I have just tested my 2S2P XR-E R2 light head array, and have found that 7.1V is needed to push 2.0A through the array.

Trouble is, I only have 2S3P 18650 LiIon battery cells to do this.

When fully charged, the LiIons will deliver 8.4V. The buck driver will presumably need 0.1V overhead in order to keep regulating, so I need 7.2V. That's great, except that I'll only get to use 50% of the batteries' capacity. 7.2V means I only take each cell down to about 3.6V from its fully charged state of 4.2V, and nowhere near the depletion charge of about 2.5V.

So I guess I need a buck/boost driver, in order to fully utilise these cells. Alternatively, I can just leave it be, and know that my cells are always well protected. However, how much runtime can I expect? I was hoping to get about 3 hours out of the design, but if I'm only able to use 50% of a 3P array and am discharging near 1C, I only get about 1.5, maybe 2 hours out of it.

Do any of the smart people here have anything they can add to help me out? Is there a solution to my predicament?

 

[Justin Case]

Drive the LEDs with less forward current to reduce the forward voltage.

 

[MrNaz]

Drive the LEDs with less forward current to reduce the forward voltage.

I want that forward current due to the brightness requirements of the application.

 

[Justin Case]

Wire the four LEDs in parallel and use a 2xShark Buck 2A sandwich.

 

[MrNaz]

Wire the four LEDs in parallel and use a 2xShark Buck 2A sandwich.

Hmm... is that the only option? $120 is a little steep for a driver circuit.

 

[Justin Case]

A Shark Buck 2A is $27 each. You are looking at an assembled Shark Buck with Remora.

You could also make a 3xSOB1333 sandwich. Each SOB blank is $16. The .05/.15 sense resistor pair is 75 cents and you need three pairs of those. Total cost is $50.25. You have to solder on the sense resistors yourself, as well as assemble the sandwich.

 

[MrNaz]

How about the Blue Shark? 2.7V-25V in and up to 980mA out at up to 32V. Looks like I could wire my LEDs in series and use one of these, right?

 

[Justin Case]

Blue Shark is a boost board with 1A output. You could wire the LEDs in series. You should be well within the 4A max input current spec for the board. As long as your battery pack can deliver about 2.5A, you should be ok.

 

[MrNaz]

I know. I was thinking I could put the LEDs in series, requiring around 15V, and then use my 7.4V LiIon pack with it.

 

[HarryN]

It is kind of a dangerous assumption to think that you can pull that much current through a driver with only 0.1 volts of drop. It might work, but many buck circuits need more like 1 - 2 volts to run at full output.

There are a couple of slightly off the beaten path approaches:

1) Series driven - voltage + resistor regulated
- Put all of the die in series
- Get a boost voltage driver (many on the internet) and put it in series with a 1 - 2 ohm resistor + all of the LEDs
- You will be surprised at how efficient it actually is, since voltge drivers are a lot easier to make than current drivers.

2) D2DIM + Resistor
- Configure the die to be 2S + how ever many parallel
- Put a taskled d2dim + slight resistance in the strings (will take more than 1)
- It will keep going (with reduced output) as the cell voltage drops, but at least you can keep running as the voltage drops.

Maybe it is not such a bad thing to have a dive light start to "fade" rather than "shut down" as the cells start running low?

Harry

 

[MrNaz]

1) Series driven - voltage + resistor regulated
- Put all of the die in series
- Get a boost voltage driver (many on the internet) and put it in series with a 1 - 2 ohm resistor + all of the LEDs
- You will be surprised at how efficient it actually is, since voltge drivers are a lot easier to make than current drivers.

2) D2DIM + Resistor
- Configure the die to be 2S + how ever many parallel
- Put a taskled d2dim + slight resistance in the strings (will take more than 1)
- It will keep going (with reduced output) as the cell voltage drops, but at least you can keep running as the voltage drops.

Maybe it is not such a bad thing to have a dive light start to "fade" rather than "shut down" as the cells start running low?

Harry

Where are good suppliers of drivers? DX and Kai have a few, but they're all cheap and nasty, and Taskled's drivers are all way too expensive to use in a retail destined product design. I'm looking for any driver solution in the $10 range that's reasonably efficient and easy to assemble. AMC7135 boards suit me due to their low cost and the near impossibility for the board to screw up; it's just an IC on a board. Other drivers are susceptible to poor design and bad workmanship.

If you know a good one stop driver shop, please point me in that direction.

How's this for another option:

Connect the LEDs in series (~14.8Vf) and then boost the battery pack (5V - 8.4V) using a Fatman? That way each emitter sees 1A. Sound reasonable?

 

[HarryN]

The fatman looks like a good option.

 

[Justin Case]

I would hesitate using the Fatman. P. 6 of the LM2700 boost IC datasheet shows a switch current limit for 12V out and 8V in of about 2.3A (closest conditions I could find in the datasheet to match your setup). Assuming 90% driver efficiency, 3.5V Vf for each emitter (14V total), 1A forward current to each emitter, and 7.4V Vbatt, I get a calculated Ibatt of about 2.1A. If your Vf is just a little higher so that the total drop is 15V, Ibatt~2.25A.

Basically, it looks borderline. Small changes against your favor in driver efficiency, voltage drop of the load, and-or battery voltage may result in exceeding the switch current limit. For the extra $5 or so, I'd probably go with the Blue Shark if you want to go with a boost driver.

I personally like buck drivers since the current draw from the batteries is typically less stressful. But the price may not be all that desirable for you.

An advantage of wiring the LEDs in series is less chance of thermal runaway problems. That would be a downside to using the 2xShark Buck 2A approach. If your LEDs are unbalanced, or get unbalanced for some reason, one emitter could get overdriven and cook. Then the others could start cooking in turn.

 

[Justin Case]

Where are good suppliers of drivers? DX and Kai have a few, but they're all cheap and nasty, and Taskled's drivers are all way too expensive to use in a retail destined product design. I'm looking for any driver solution in the $10 range that's reasonably efficient and easy to assemble. AMC7135 boards suit me due to their low cost and the near impossibility for the board to screw up; it's just an IC on a board. Other drivers are susceptible to poor design and bad workmanship.

The AMC7135-based drivers have a recommended Vin max of 6V and an absolute rating of 7V, so your existing battery pack is going to cook the driver. Even if the driver didn't fry, the efficiency would be poor. You could wire your LEDs in parallel and build a 4.2A 4-board sandwich, each board containing 3xAMC7135 (or a 3.85A 3-board sandwich, with two boards each with 4xAMC7135 and one board with 3xAMC7135). Or if you didn't want to make a sandwich, you could piggyback solder the requisite number of AMC chips on a single board. The efficiency for an LDO linear regulator is essentially Vout/Vin. Thus, 3xD-NiMH in series would be a good choice wrt voltage match and mAh capacity for long run time. Still could be hard to make a 3h run time, though. I suppose if you went 3S2P with some D NiMHs, then you would easily make your 3h run time mark. The AMC sandwich driver would be inexpensive, assuming you can get the exact number of AMC boards you need. Otherwise, you'll have to spend ~$30 for a 20-pack from DX or KD (maybe 10-packs are available, in which case the price would be that much less; plus, you could in theory sell the extra boards on the Marketplace).

Another AMC wiring method is the one from download. See this thread and this one. Besides Post #1, see Post #110 in the second thread I linked.

 

[MrNaz]

I would hesitate using the Fatman. P. 6 of the LM2700 boost IC datasheet shows a switch current limit for 12V out and 8V in of about 2.3A (closest conditions I could find in the datasheet to match your setup). Assuming 90% driver efficiency, 3.5V Vf for each emitter (14V total), 1A forward current to each emitter, and 7.4V Vbatt, I get a calculated Ibatt of about 2.1A. If your Vf is just a little higher so that the total drop is 15V, Ibatt~2.25A.

Basically, it looks borderline. Small changes against your favor in driver efficiency, voltage drop of the load, and-or battery voltage may result in exceeding the switch current limit. For the extra $5 or so, I'd probably go with the Blue Shark if you want to go with a boost driver.

I personally like buck drivers since the current draw from the batteries is typically less stressful. But the price may not be all that desirable for you.

An advantage of wiring the LEDs in series is less chance of thermal runaway problems. That would be a downside to using the 2xShark Buck 2A approach. If your LEDs are unbalanced, or get unbalanced for some reason, one emitter could get overdriven and cook. Then the others could start cooking in turn.

By the time you wrote this, I'd already purchased the Fatman from Cutter, and I don't want to take it back for exchange again, as I already exchanged a bunch of Recom drivers for it. Luckily, I tested it with my array and it seems to work just fine. I've wired up my Fatman with a MOSFET, as you can see here:

The white lead in the lower part of the photo goes to the reed switch. I have my MOSFETs' gate and source connected with a 10kO resistor so the reed switch becomes a momentary switch, only on when the reed switch senses proximity. The MOSFET has been epoxied to the underside of the heat sink.

This circuit has been epoxied into the inside of my light, as you can see here:

The white epoxy is arctic alumina, so there is at least some thermal path between the driver/MOSFET and the outer case. I'm going to smear regular epoxy over the legs again to increase the strength of that bond.

In general, how does this look as a build? I'm rather nervous about this, as it's my first from scratch build and there are so many hack solutions to problems I've found along the way that I don't have confidence that it'll even work :candle:

- Naz.

 

[HarryN]

Hi Mr. Naz - I don't know enough to comment on the reed switch / mosfet aspect, but in general it is fine.

Just for future building, the finned heat spreader there is to increase the area exposed to air - for open air kinds of setups. In a closed setup, the fins really don't do anything useful. In a more perfect world, it would be better to directly attached the driver to the flashlight body wall. It should be ok.

If you just cannot quite get enough power from the fatman driver, then just turn it down a bit. The difference in light output at 800 vs 1000 ma is relatively small.