Simple boost circuit build. Need a little help.

Mike S

Newly Enlightened
Joined
Apr 29, 2011
Messages
132
I've been playing with this circuit on and off for a while now. It uses the Linear LTC3490 driver IC. Input voltage can be between ~0.8-3.2V and it outputs a constant 350 mA on fresh cells. I've had a couple of problems getting it to function correctly.

The first light it was going to be used in was an inexpensive single cell light. It originally had a type of joule thief circuit that powered a 5 mm LED. It would draw a little over 600 mA at the tail cap and the LED must have had 5 or 6 dies in it because current at the emitter was close to 250 mA.

When I put the LTC3490 circuit together, I soldered a couple of wires to the positive and negative pads so that it could be tested outside of the flashlight. A NiMH or lithium primary would be pinched between the wires and the XP-G lit up perfectly. Current at the LED was a fairly constant 340 mA. Current at the cell would start at roughly 800 mA and then climb to about 1600 mA, which is what the datasheet shows as typical. The problem was that the driver IC was getting hot after a couple of seconds. This didn't happen every time. It seemed to occur mostly when the connection to the cell was poor. Sometimes the wires would slip as they were being pressed to the cell. If a good connection was made on the first try, there was no heat, or so it seemed that way. I figured once the driver was in the light, the switch would take care of the poor connection. It didn't. Actually the IC heats up every time the switch is on, but there's also a new problem. There is way too much resistance in the tail cap.

With a single cell light, the datasheet says 1 ohm of resistance has a significant effect on the output. At the LED, I'm seeing 220 mA instead of 340 with the switch in place. With the switch bypassed, the output goes back up to 340 mA. Is there anything that be done to this cheap switch other than bypassing the spring with a piece of wire? It looks like it's pressed and might have to be drilled or cut out.

Anyways those are the two main problems. Heat and tail cap resistance. Actually, later on I killed this driver when I was screwing down the tail cap. The switch must have been on because the LED flickered a few times. Once it was tightened down, the LED would no longer light up. I've never worked with an IC that was so sensitive to poor connections, but then again, I've never built a boost circuit like this before. Just the occasional joule thief.

Here are some photos of the first hand soldered PCB. The output cap is too far away from the IC, the traces were too thin and the pull down resistor is way too large. I wonder if the length and width of the traces had anything to do with the heat?
boostCircuit.jpg


This is the light. It's my brother's flashlight and was originally modded with an AMC7135/14500 set up. He's too careless to use a lithium ion cell and constantly allows the protection circuit to trip. That why we're looking to AA boost circuits so that he can use a more stable power source.
lightTest.jpg


tailcap.jpg


This is the latest build. It's a 15 year old incan minimag. It was modded earlier with a Nite Ize drop-in that made it a whiter light, but not a brighter light. The tail switch was reused for this 1W boost circuit.

The 14 mm inner diameter made it difficult to fit all the components onto one board. Since that plastic spacer needs to be removed and takes up 8 or 9 mm, it made sense to use a double layer PCB and move the inductor to the top. It fills the gap perfectly. The IC on the far left is the LTC3490 in a "DD" package. Way too small to work with. After several failed reflow soldering attempts I switch to the SOIC-8.

reflowSoldered.jpg


pcbCompleteClose.jpg


circuitTest1.jpg


driverTestFit.jpg


Current at tail cap with 2 NiMH Enercell's charged to 1.32V.
currentTest.jpg


I still have to make a small heat sink for the XP-G or XP-E and figure out which reflector to use.
lightComplete.jpg
 

georges80

Flashlight Enthusiast
Joined
Oct 23, 2002
Messages
1,262
Location
Sunnyvale, CA
You are basically pushing the smelly stuff uphill with the setup you have.

ANY appreciable resistance in the tailcap/spring/contacts is going to really hurt you big time with a boost driver in this situation.

Consider you have nominally 1.1V at the cell (AA nimh under load and somewhat discharged) and you need say 3.3V at the LED. So without taking into account efficiency losses you have a 3:1 boost ratio. So, with 350mA at the output you're look at 350 x 3 = 1050mA at the input (again assuming zero switcher/inductor losses).

Every 0.1ohm of resistance losses means you lose 1050mA x 0.1 = 0.1V loss in input voltage to the driver. With less voltage at the driver it will draw even higher current, higher current means higher input voltage drop and so even higher input current.... i.e. things go downhill rapidly.

Again, with such high input current requirements (trying to use a single cell nimh etc), you need to get rid of ALL contact resistance paths.

Now, I've used the LTC3490 (in the DFN package), but I used a primary lithium cell as the source. Efficiency at 2V - 3V is MUCH better with this chip.

Look at the datasheet, it shows 50% efficiency driving from 1V (which is quite realistic in your situation). That means the above calculations actually imply 1050 x 2 = 2100mA at the input, so resistance drops will hurt you at 0.21V per 0.1ohm. With 350mA x 3.3V = 1W to the LED AND you have 50% efficiency that means 1W heat losses in the driver. That's a LOT of heat to dissipate, especially in the SOIC package since it only can dissipate through the leads.

The SOIC package has a Qja of 150C/Watt, that means 150C temp rise of the junction for 1W of power loss!!!
The DFN package has a Qja of 43C/Watt, so 43C temp rise of the junction for 1W of power loss.

Of course even with good thermal conductivity of the package to the PCB, the PCB needs a GOOD thermal path to the body of the light to dump that excess heat.

Anyhow, in summary, so great DIY effort, but I'm afraid your light, switch, contacts, thermal path of IC and PCB, battery voltage (single cell nimh) is going to doom this to failure unless you can correct most of the above issues.

cheers,
george.
 

Mike S

Newly Enlightened
Joined
Apr 29, 2011
Messages
132
I don't know how to help you, but your homemade PCBs look wonderful, was the traces etched?

Thanks, yes these were done with the toner transfer method and etched with hydrogen peroxide/muriatic acid. (3:1 mixture)

You are basically pushing the smelly stuff uphill with the setup you have.

:)

Thanks George, I needed to hear that from someone like yourself. I love reading your driver threads like the recent b3flex.

You answered all my questions, so I guess it's back to the drawing board. Making this driver work in a single cell light really isn't going to be possible for me. That's a bummer. It was such an easy circuit to put together.

If I was able to find a light with a low resistance tail cap, how would you improve the thermal path of the SOIC-8 package? I have some DIP-14/16 heat sinks that could be cut in half, but they're almost 5 mm tall. Does the heat transfer out through the pins and into the traces? When I was testing it, it seemed as though most of the heat was coming directly off the front and back of the device. Even the bottom of the PCB was getting quite hot.
 
Last edited:

georges80

Flashlight Enthusiast
Joined
Oct 23, 2002
Messages
1,262
Location
Sunnyvale, CA
Mike, the soic packages (that aren't thermally enhanced) have limited power dissipation capability. The pins provide a direct path from the package to the board and of course the plastic provides some path as well. The problem is you want to get the heat off the semiconductor and that's buried inside the package.

The DFN packages can be more problematic to solder (without solderpaste and some experience/technique), but provide a much better thermal from the semiconductor to the PCB since the die substrate is directly attached to the thermal pad that is on the bottom of the IC package. Soldering the thermal pad to the PCB provides a direct heat path to the semiconductor die - hence a major reason for developing that type of package in the first place. Additionally it also minimizes connection paths from the die to the 'pins' and so reduces inductance and lead resistance etc.

With a single nimh cell, and a white LED you're really asking too much of that switcher IC. Even with zero resistance you still have 1.1V or less and 3.3V out and that means ~50% efficiency and a lot of heat to dissipate.

cheers,
george.
 

Mike S

Newly Enlightened
Joined
Apr 29, 2011
Messages
132
Thanks, I'll continue to search for a more efficient boost circuit and save the remaining 3490's for 2 cell lights. I may even try to improve the PCB and experiment with the DFN package a little more. Some claimed to have hand soldered this type of IC using the swiping method with a fine point tip. It sounds difficult, but is worth a shot after having such messy results with solder paste.

I spent a couple of hours finishing up the light. A 4000K XP-E was chosen for a tighter hot spot and it was reflow soldered to a 16 mm copper disk. Getting the LED centered was a bit of a challenge. It's still slightly off center which can be seen in the wall shot. A TIR optic might be the best solution.

xpeSoldered.jpg


lightComplete2.jpg


lightComplete3.jpg


beamPattern1.jpg
 
Top