Another ArcAAA Mod: TW0H w/ LTC3490

I finally got around to making something with the LTC3490 chips I picked up. These are minimal component 350mA constant current ICs that make it easy to power a LuxI. I had originally planned on creating a Solitaire Pill, but I got a couple of empty ArcAAA heads from cy and decided to change hosts.

I started by a quick etch of a test board to check the inductors I had on hand.


Since that worked OK, next I planned on a dead bug style pill. After working out the layout, I found out that pbarrette had done a very nice one. I used a modified version of his (Thanks pb). I had to choose between emitters - I had one hand a R3K (LuxI) and a TW0H (LuxIII). I tried them both on the test board, asked the wife and we both preferred the TW0H.

I polished the reflector and bored out the ArcAAA head a bit to make it a little "roomier" and assembled the pill (sorry, no photos ). I potted the pill in Arctic Silver to help conduct heat away from the components.

Here is a photo of the finished head.


A beamshot against the back of the sofa.


Beamshot compared to an ArcAAA with the LED replaced with a great performing CS LED. The mod had a hotspot about twice the intensity of the ArcAAA but the flood portion blows the Arc away. The colour is also much nicer. This beamshot does not do the output justice. I tried a ceiling bounce test and this is significantly brighter than three ArcAAAs (on at the same time) with the CS LEDs.
Left - ArcAAA(w/ a CS LED), Right - 3490 TW0H


What's the price for this increase in brightness? Runtime. This little baby is a real pig on batteries. It also gets pretty warm/hot. The LTC3490 chip is supposed to have a low battery cutout of 0.85V (when configured in single cell mode) so it should protect your NiMH collection. The chip is also supposed to PWM the output if battery voltage is too high and the current starts to rise above the setpoint. (good for Li-ions??).

Between 15-20 minutes runtime was typical using some oldish 750mAh Energizer NiMH AAAs.
Here's a runtime graph:


This was a fun mod but I have used up all my enjoyment making this one .

Paul

Nice work! :goodjob:

The 3490 is a nice driver for most any application. I'm still trying to get a Lux into a solitare... I did a 2AAA though.
I don't belive it will work efficiently at LithIon voltages, there was a post about that somewhere...
On the bright side, the regulated current can be increased by paralleling a resistor from the output cap to the LED. There is an internal 100mOhm and adding another one externally will increase current to 700mA! Efficiency does suffer though, (I should graph it out sometime). Just don't expect to run the higher current setting from a single cell source (1.2V or 1.5V that is), however, a Lithium 3.0V will work well.

Keep up the good work.

Thanks, Crux.

If I was going to do another one of these, I believe I might choose the MAX1797 chip that is used in the MiniPro. It's not a constant current chip, but it limits the battery current to 1 Amp. You also can get away with a low component count with this chip. I would expect runtime would double (if not triple) with that chip in a single AAA NiMH application compared to the LTC3490. 4sevens has taken this approach in a ArcAA mod.

There are some subtle differences for each chip that may lead each to be a better solution for different problems.

Paul

Just picked up some Li ion AAAs.
Here are some beamshots:

Approx 16 inches from wall.
LTC3490 TW0H (Li ion battery), Peak Matterhorn 3 LED Hi Power (Lithium battery), ArcAAA w/ CS LED (Lithium battery)


Approx 6 inches from wall (same order)


This thing has great spill!

Paul

I did another run-time with a Li-ion cell and fan-cooled the light. The regulation is very flat this time. I may run another test with a NiMH with fan cooling.



Paul

Cool, so by having such short runtimes, it's safe to say this is delivering 350ma to the emitter? Is it constant current?

Do you have a Fenix L0P to do a "shoot out"! That'd be interesting.

Sounds like a nice IC!

IsaacHayes said:

Cool, so by having such short runtimes, it's safe to say this is delivering 350ma to the emitter? Is it constant current?

Do you have a Fenix L0P to do a "shoot out"! That'd be interesting.

Sounds like a nice IC!

Click to expand...

Isaac, I would say that's a pretty safe assumption. (350mA current regulated)

I don't have a Fenix, but I suspect the Fenix would easily out throw due it having a real reflector.

I am very pleased at the level of regulation. The initial dip is due to the increased emitter temperature. However, once the temperature reached steady-state the output was extremely stable. I attribute the rise at the end of the run-time to the switching modes as the Li-ion voltage falling below the Vf of the emitter at 350mA. Recall that this is a TW0H with the H being binned at 700mA - at 350mA the Vf is likely in the G-bin range. The heat generated by the driver and the emitter would drop and lumen output would rise.

One of the great features of this is the ability to use any battery chemistry. It will work with a single AAA Alkaline, NiMH or Li-ion. This one may become my keychain light, however, I still like the extended run-time of my ArcAAA.

Paul

nice job!

Hi Chimo,

I'm glad my layout was able to help in any way.

Here is my runtime plot with the same circuit using an 850mAh Energizer Nimh:


This suggests to me that you should be able to get a longer runtime out of your circuit if you used a better inductor. I'm using the Toko A920CY 3.3uH inductors and have been very pleased. The Toko has a high current capacity, very low DCR, is small enough to fit inside the ArcAAA heads and is cheaply available at Digikey.

A Vin that exceeds Vout will produce a current output that is generally greater than 350mA on this circuit. Also, the efficiency of the converter goes to hell when this happens. The circuit will try to limit the current output as much as possible, but it does so at the expense of efficiency. I have found that when Vin is just barely lower than Vout, efficiency is well over 90%. But the second that Vin exceeds Vout by even the slightest amount, the efficiency drops to nearly 50%. This is clearly shown in your runtime plot. A ~350mAh Li-ion cell putting out ~350mA to the LED should have a runtime of around 1 hour. Your 30 minute runtime suggests an efficiency of ~50%.

This circuit is not made for Li-ions, but they should be relatively safe to use for short bursts of 1 minute or less. Remember that at 50% efficiency, not only is the Lux producing a full watt of heat, but the converter is also burning a watt itself. The low capacity and relatively high internal resistance of the AAA Li-ion cell is a good thing in this case.

Anyway, I think you should try again with this circuit and take a look at the Toko inductor. I think you'll be pleased with the performance.

pb

EDIT:

I forgot to add that the Luxeon used in my runtime plot is a RY0J Lux-I. Since you used a TW0H Lux-III with a much lower Vf, your runtimes should be able to beat mine with a better inductor.

is this circuit alkaline friendly?

Thanks for the kind comments,

pb, I hear you on the inductor. I believe I ended up using a 3.3uF Coilcraft MSS7341 series inductor.

The efficiency isn't great on Li-ion or NiMH. I was, however, very impressed with the "regulation" on the Li-ion.

The first test I ran with the Li-ion was very interesting. I did not cool the light so the thermal regulation on the chip kicked in a few times. It was a rather "saw-toothed" looking as it cycled thermally.

When I put this together last October, I didn't have a very great selection of inductors on hand. Since then, I have picked up some more CoilCraft and Wilco inductors. The driver has been coated in thermal epoxy to improve heat transfer so I don't want to disassemble it. I'm also out of 3490's. Perhaps I have to get some more... I would like to see run-time between 45-60 minutes. Cheers,


Paul

greenLED said:

is this circuit alkaline friendly?

Click to expand...

It will work with alkalines, NiMH and Li-ions, Lithium, and the OxyWhatever :thinking: one.

However, it is very unfriendly to a poor a little AAA alkaline cell because the current draw is very high. The other chemistries have a lower internal resistance and give better regulation where as the alkalines decay more steadily and have a longer tail. There is a run-time graph above that includes an alkaline run - the light was not cooled for these tests.

Paul

I've combined 2 Li-ion graphs, one cooled and one not cooled. Lots of thermal cycling going on on the non-cooled plot.



Paul

Hi Chimo,

Yeah, the thermal cycling can get really bad without cooling of some sort. I also noticed a lot of jitter at the end of the battery life which is probably the IC trying to settle on an output current with a battery that barely has any juice left. My graph was done with data points every second and the test was done with the light and the sensor in a sealed cardboard box, so you can really see the cycling.

The inductor on your test board looks like a 22uH coilcraft which seems like a really bad choice, but I certainly understand the need to use what you have on hand. I've used the MSS7341, the MSS5131 and the Toko A920CY all in 3.3uH. The Toko wins hands down. It's about the size of the circle on top of the 7341 and a lot shorter, so there's more room to fit it and a decent heatsink in the Arc. It also has a lower DCR which means better efficiency.

If you look at my graph, you'll see that I got ~40 minutes to the start of the major drop off point and ~45 minutes to 50% with my RY0J Lux1. I have some TW0H Lux3's, but I still have some ArcAAA heads that need parabolas cut into their reflectors before I stuff them into a light and test them out. The lower Vf on the H bin Lux3 should mean a longer runtime.

If you plan on using a Li-ion, it might make more sense to use a Lux with a higher Vf. That would make it more efficient which should make for a better runtime. The problem is finding a high flux Lux in a decent color bin with a high enough Vf.
pb

pb, that is a 22uH on the test board. I had a 3.3uH in the same series - I don't recall which one I ended up using on the actual pill. I may try to disect the pill and replace the inductor.

I checked Digikey and found the Toko, but Digikey did not seem to have the LTC3490. I checked Linear's web site and they wanted >$30 for shipping!

Paul

Hi Chimo,

I've had difficulty sourcing the LTC3490 as well. Neither Digikey nor Mouser seem to carry it and, like you said, Linear's shipping cost to non US addresses is insane. If you have a friend in the US who could re-ship them to you it will save you a lot of cash. Linear's shipping is as low as $3-4 when they ship to a US address.

pb

What was the voltage of the Li-ion at the end of the run? Was it still at a safe voltage? I like the performance of the Li-ion better than nimh, but I'm afraid of overdischarging the Li-ion.

chimo said:

I've combined 2 Li-ion graphs, one cooled and one not cooled. Lots of thermal cycling going on on the non-cooled plot.

Paul

Click to expand...

baylisstic said:

What was the voltage of the Li-ion at the end of the run? Was it still at a safe voltage? I like the performance of the Li-ion better than nimh, but I'm afraid of overdischarging the Li-ion.

Click to expand...

I remember checking right after a run, but I don't remember what the voltage was. However, I do recall not being worried about it being too low.

Thanks, chimo.

I will run Li-ion on mine. Runtime and regulation on nimh isn't that great.

The 3490 can run on a single cell(NiMH, NiCD, Alks, etc.).

Something to consider when using rechargeable LiIon's.