Need help with LM2703

[Xcandescent]

I'm new to the forum, and my electronics knowledge is extremely limited, so please -- be gentle.

I've been working on hacking (quite literally) an external lighting fixture for the Game Boy Advance. My intent is to power 2 white LED's (Vf=3.6V; I=20mA) from the Link Port, which provides either 3.3V or 5V depending on what type of game you're playing. I figured the best way to do this was to use a DC-DC boost converter, so I ordered some samples of National's LM2703.

The problem is I haven't enough knowledge to actually understand how to use it, or why certain circuit configurations work. I've noticed that most boost converters utilize some sort of feedback loop with 2 resistors to set output voltage -- but often replace the first resistor with the LEDs to be powered. I have no idea how/why this works, or how one would set the voltage properly in that case.

In the case of the LM2703, the two resistors used to set output voltage are expressed as functions of each other and Vout. Vout is known, but it's not clear to me whether either resistor value can be arbitrarily set (and then the second resistor solved for), or whether there's a secondary condition which must be met. If anyone knows, I could sure use the advice!

(Yes, to someone who knows what they're doing, I'm sure that sounds like a complete newbie question. I TOLD you my electronics knowledge is limited ...)

-XCN-

 

[yclo]

How about something like this?
It's called a worm light. A quick search will give you lots of info.
It should be less than $10.

YC

 

[Xcandescent]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by yclo:
How about something like this?
It's called a worm light. A quick search will give you lots of info.
It should be less than $10.<HR></BLOCKQUOTE>

In a word, no.

I didn't want to have to go into the history of why I'm persuing this project with tons of supposed "solutions" on the shelf, but to be brief:

The vast majority of GBA external lights, LED or not, flat-out suck. They're almost as bad as not having a light at all.

The reason they suck is twofold: NO product ships with a decent LED, and none of them power it properly. AFAIK, if they actually went to the trouble of using a decent LED *and* powering it properly, their product would cost considerably more than $10 (which seems to be their "magic" price point). Hence, quality gets stuffed so the shelves can be packed with useless substitutes.

To give you an idea of how bad your average off-the-shelf product is, most of them (with the exception of the two I mentioned above) can be outshone by a light shield modded with a Radio Shack 276-320 (see The LED Museum's White LED section for a picture).

I can think of two exceptions (Gamester Flood Light and TAM Arm Light Advance), but that's it. Of those two, the former is heavy, has glare issues, and sucks batteries like there's no tomorrow (it uses a CCFL and 3 AA's), and the latter has uneven screen coverage (even with 4 LEDs).

Since it's obvious that a better external solution is not coming, the only option left is to try to make one. Since this forum is packed with people who know more about selecting and powering bright white LEDs than anyone else I can think of, I figured this would be a good place to go for suggestions.

As I said before, I do have a boost converter IC (National's LM2703), but I haven't the faintest idea how to use it properly -- hence this thread.

-XCN-

P.S. Before anyone else mentions it, I'm aware of the existence of the Afterburner. In a nutshell, even if the Afterburner had turned out to be a perfect product (so far it's been hit-or-miss), there are enough issues with the installation difficulty, color wash-out, and possible incompatibility with newer GBA screens that persuing a better external solution is still worthwhile.

 

[jeff1500]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Xcandescent:


In the case of the LM2703, the two resistors used to set output voltage are expressed as functions of each other and Vout. Vout is known, but it's not clear to me whether either resistor value can be arbitrarily set (and then the second resistor solved for), or whether there's a secondary condition which must be met. If anyone knows, I could sure use the advice!
<HR></BLOCKQUOTE>

I've done it with Maxim chips. The resistor values just have to satisfy the Vout formula. Any way you do that is fine. The Maxim chips have a maximum value for one of the resistors. It's just a ratio thing. That way you have more freedom to use standard resistor values.

It's not the values of R1 and R2 that matter, it's the ratio of values that counts.

I don't have a clue about substituting leds for R1 or R2.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR> My intent is to power 2 white LED's (Vf=3.6V; I=20mA) from the Link Port, which provides either 3.3V or 5V depending on what type of game you're playing. I figured the best way to do this was to use a DC-DC boost converter, so I ordered some samples of National's LM2703.
<HR></BLOCKQUOTE>

If you've got 3.3 volts available that'll probably light up a white led but just not as bright. Maybe use 3 or 4 leds at 3.3 volts instead of 2 at 3.6 volts.

And if you grind the dome off the top of the leds and polish them with toothpaste, they become flood lights instead of spot lights.

 

[Xcandescent]

Thanks for the advice regarding the Maxim IC. I'm still curious as to what the deal is with replacing the resistor leading to the feedback pin with the LEDs you want to light, simply because I've seen this done in example circuits for various boost converters intended to drive LEDs.

The values for the resistors in the LM2703 example circuits are rather high (510k, 33k, etc.) ... not sure why that's the case if the only thing that matters is the ratio ...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by jeff1500:
If you've got 3.3 volts available that'll probably light up a white led but just not as bright. Maybe use 3 or 4 leds at 3.3 volts instead of 2 at 3.6 volts.<HR></BLOCKQUOTE>

True, but this is coming from the Link Port -- which puts out either 3.3 OR 5V depending on what type of game you're playing. 3.3 will light a white LED; 5 will crispy-fry it.

(This is, BTW, one of the primary reasons current GBA lights powered off the Link Port are so dim. Rather than utilize a boost converter, they simply throw in enough resistors to make sure the LED survives at 5V -- which in turn makes it really weak at 3.3V. This is done, oddly enough, with both LED and incandescent light shields -- no idea what the deal is with the latter.)

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>And if you grind the dome off the top of the leds and polish them with toothpaste, they become flood lights instead of spot lights.<HR></BLOCKQUOTE>

I'm not nearly brave enough to try that, but thanks for the tip!

-XCN-

 

[MrAl]

Hi there Xcand,

As Jeff pointed out, the set point is
related to the ratio of the two resistors,
and the only reason for using somewhat
higher value resistors is because the
higher you can get away with the less
efficiency you loose.

If you want to substitute an LED for R1,
that's probably going to work ok, but
you will loose about 34 percent efficiency
by doing that. If efficiency isnt an
important issue, then all is well.
You substitute the LED for R1 and compute
R2 according to:
R2=1.2/I(Led)
For example,
if the Led current is to be 20ma, then
R2=1.2/0.020
which equals
60 ohms.

On the other hand, if efficiency is
an issue, then you would be better off
using resistors as the data sheet indicates
to create your basic voltage regulator
and use maybe a 1 to 10 ohm resistor in
series with the LED(s) to limit current.
That's the basic idea for using voltage
regulators to drive LEDs.

There are better chips out there too.
Try doing a parametric search on the
National site.

If you still want to use this chip, then
i recommend setting it up as a regular
voltage regulator and using series resistors
for the LEDs.


Good luck with your LED circuits,
Al

 

[Jonathan]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Xcandescent:
The problem is I haven't enough knowledge to actually understand how to use it, or why certain circuit configurations work. I've noticed that most boost converters utilize some sort of feedback loop with 2 resistors to set output voltage -- but often replace the first resistor with the LEDs to be powered. I have no idea how/why this works, or how one would set the voltage properly in that case.

-XCN-<HR></BLOCKQUOTE>

The above is a cute trick that uses the boost converter as a _current_ regulated converter rather than as a _voltage_ regulated converter.

The output regulation works to fix the voltage across one of the resistors. To use this for _voltage_ regulation, you build a resistive divider which produces the reference voltage when the output voltage is at the desired level. If the reference voltage is 1.25V, and you want an output of 5V, then you use a resistive divider with a 3:1 ratio...3.75V across the 'big' resistor, 1.25V across the 'small' resistor, 5V total, with 1.25V at the reference/feedback pin.

In general, with voltage feedback, you want to make the resistors as large as possible, so that there is minimal loss current through the resistors. On the other hand, they cannot be too large, otherwise the feedback circuit will be too noise sensitive and will not function properly.

If you place a constant voltage across a constant resistance, then you get a constant current. So if you use the _load_ and a resistor as the feedback divider, then the output voltage will adjust until the _current_ through the load (and resistor) is such that the _voltage_ across the resistor is the reference voltage.

If you use a couple of LEDs in series, followed by a 50 ohm resistor, and connect this as your divider, then the regulator will adjust the output voltage until there is 1.25V across the resistor...which happens at 25mA. The net result is that you get a precise 25mA through the LEDs, rather than having a precise voltage applied to the LEDs.

The benefit of this is that LEDs are ideally driven at a controlled current, but the voltage needed to supply this current will change quite a bit with small changes in temperature, production lot, age, etc. The use of current regulated output is a big win for LEDs. The cost is that you have the full LED current going through the sampling resistor, at the full reference voltage...which means considerable power loss.

There are various regulator chips which are specifically designed for current regulation. These chips use a much lower reference voltage. For example, the LT1937 uses a 0.095V current sense reference voltage, and is specifically designed to run 2-4 white LEDs from 3.3V.

There have been several discussions on boost converter chips for running LEDs, eg:
Best Chip...5W Luxeon

Good Luck!

-Jon

 

[Xcandescent]

Thanks for the help -- you guys rock!!!

MrAl: I'm sure there are better chips out there -- I just happen to have samples of the LM2703 on hand. Though the damn thing is so small I may have to find something larger (more below ...)

Jonathan: Thanks for the explanation of how current-limiting boost converters work! I take it that, once you set the resistor for the current you want (with respect to Vref), that the circuit will automatically adjust to whatever voltage load you drop in -- up to its maximum rated voltage?

I saw the LT1937 -- looks like a good bet for what I'm doing ... except for the size of the IC. I'm used to working with larger components, and anything that's intended to be surface-mounted poses problems as:

a.) I've never etched, drawn, or otherwise done anything to create a circuit board,

b.) I've never surface-mounted ANYTHING.

A lot of the converters I'm looking at are extremely small SMT components. This is a real problem. How the heck am I supposed to work with this stuff when all I've got is a breadboard, wires, and a soldering iron? I've already learned the hard way that large wires + small IC leads don't mix.

-XCN-

 

[dat2zip]

Just my two cents here.

One thing that is really easy to do with SMT devices is to dead bug or ratsnest the circuit. it turns out that it is rather easy, requires no special tools.

I highly suggest trying the following. Start with a DIP8 or SOIC8 converter. Bend all but the ground pins out horizontal like an airplane. leave the GND pins pointing down.

Now, solder the GND pins of the IC to a solid copper clad board (posititioned in the middle of the board). Continue this method soldering the Caps, resistors that go to GND directly to the copper board in the location that is most appropriate for your design. You can usually get a better placement this way then an etched board can. The solid ground plane is super ideal for switching power supplies. The remaining components can be soldered to the free side of the lead, that are not soldered to the GND plane.

To help make a solid input terminal you can use high value resistors as standoffs or posts where needed. Just solder one end to GND and use the other side connected to the input capacitor or input/output component. This will help relieve the stress on the critical component.

I had to do this just recently to prove my MAX1674 switching noise problem was my layout and not the IC. Turned out after deadbugging I got the circuit to finally behave correctly. It then took me another two layout revisions to get it correct, but, I had the knowledge that the circuit worked if done correctly (from the deadbug board).

Hey, it's lying right here. Let me scan it in and show it.

This circuit as I stare at it has had the inductor removed (two black wires) and possibly more components. It once worked, believe me!

 

[Xcandescent]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
One thing that is really easy to do with SMT devices is to dead bug or ratsnest the circuit. it turns out that it is rather easy, requires no special tools.<HR></BLOCKQUOTE>
My original plan was to do something like this with a breadboard -- solder wires to IC, connect wires to breadboard, make circuit. Unforunately, big wires make small IC leads snap like twigs when soldered to them.

So I've concluded that, since SMT components WANT to be soldered to boards, I may as well bite the bullet and learn how to etch one. (Granted, with the leads so close together, getting the traces right will still be difficult, but I think it's easier than praying that the leads don't snap.)

A quick trip to Radio Shack (quickly becoming the Acme to my Wile E. Coyote existence), and I'm now the proud owner of a "PCB Etching Kit" ... which is basically a Sharpie, two bottles of odd chemicals, a Brillo pad, and a plastic case (doubles as an etching tray! woo!). Oh yeah -- and two copper-coated plates, which are scratched up enough to make me wonder if they're usable at all.

So I look at the bottles of odd chemcials, and quickly notice that something is missing from the directions -- namely, how to dispose of them once they've been used. All those who think dumping Ferric Chloride down the sink is a smart idea, raise your hand. Right, then.

A quick trip around the Web reveals that etching solution seems to work better at warmer temperatures ... suggestions of using fishtank heaters abound, as do suggestions of printing PCB layouts onto iron-on transparencies (laser printer not included). However, no site answers the obvious question -- how the heck do you "dispose" of used ferric chloride? Are there etching chemicals that can be reused or easily recycled? Why does building a simple boost converter have to be such a pain in the -- whoops, wrong question.

So, my newbie question of the day is: how does one etch boards in an environmentally-friendly manner?

-XCN-

 

[Xcandescent]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
Making your own PCBs requires some chemicals, that you will not be able to avoid. Minimizing it can help the environment.<HR></BLOCKQUOTE>
That's what I figured ... I can't even imagine how expensive the mechanical etchers are.

Just trying to find the least toxic, easily disposable, possibly reusable option. Ferric chloride seems to strike out on all three of those counts.

What's even more amazing is that the bottle of ferric chloride I have says to "discard used solution into toilet bowl and flush". From what I've been reading about this stuff, that is a really, REALLY bad idea.


The resist ink solvent smells suspiciously like isopropyl rubbing alcohol -- which I already have an abundance of. This is unsurprising, since isopropyl alcohol is a base of many dry erase marker solvents.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>I'd suggest the toner transfer method or the blue poly. Both use a Laserjet Toner to be transfered to the bare copper clad board. No chemicals are required during this phase.<HR></BLOCKQUOTE>
Don't have a laser printer, but I've read that xeroxing to transparencies/special medium might work. Of course, this assumes that I have the software to create said layout -- which I don't. Guess I'll have to look around for possibilities ...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Coming back to the dead bug method, one can use a dremel to cut out large islands. Say, three islands. One ground, one VIN, and one VOUT. Carefully arranged, one can solder more pins/pads/components to the appropriate islands of copper. Especially handy is the Input and Output capacitors now can be soldered directly across the cut between the GND and VIN or the GND and VOUT.<HR></BLOCKQUOTE>
Interesting possibility for a rough prototype ... perhaps an X-Acto or utility knife might work as well? I'll have to think this over. Think I need to pick up some thinner/higher gauge wire first. Probably end up having to mail-order some, since Radio Shack doesn't seem to carry anything above AWG 22.

-XCN-

 

[MrAl]

Hi again Xcand,

You can use ferric chloride over quite a few
times before you have to discard it.
After you do an etching, dump the contents
back into the bottle slowly, untill you get
down to the final ounce or so. Throw that
last ounce out right away, because that
contains the more 'used' solution then
the rest.
Doing these tiny boards one bottle will
last a few years :-)
When you do have to dispose of it, you have to
remember that some pipes are made out of
copper, which it likes to eat though.
The key is to highly dilute it by
mixing a small amount into a large bucket of
water first. After you dump it, keep the
water running for several minutes to wash
any remaining amounts out of your pipes.
Cold water is best, because heat activates it
more.
If you can find a place to dispose of it
as the other posts suggest, that's even
better.

I would think that if you want to try
point to point wiring with these small
ics then you would want to use fine gauge
wire to attach to the leads.
Some very thin magnet wire would work great
i bet.

I was considering turning the package upside
down and glueing it to a fiber board, then
soldering individual #32 wires to the leads
and soldering these to a larger terminal
that i could connect to the rest of the circuit
just for testing. I'd like to test some
ic's i got a few weeks ago that have a lead
pitch of 0.0256 of an inch, and that's
pretty darn small.


Good luck with it,
Al

 

[dat2zip]

Most cities have some disposition methods available for the general public. Like recycling, chemical disposal is a major concern.

I'd check with the local city and see if there is a possible chemical waste drop off for your city. They usually take paint, thinners and other chemicals. The local Fire department might also be a good resource.

Making your own PCBs requires some chemicals, that you will not be able to avoid. Minimizing it can help the environment.

I'd suggest the toner transfer method or the blue poly. Both use a Laserjet Toner to be transfered to the bare copper clad board. No chemicals are required during this phase. You will still need the etchant. Amonium Persulfate is another etchant and ferric chloride is the primary used etchant. Of course if you don't have computer or laserjet that would be a serious investment.

I put my amonium persulfate in a heavy duty freezer zip lock. I then place the zip lock in a glass cake pan over a single element heater. The glass cake pan has water in it and i turn the heater dial to low low-med to get the water to luke warm or slightly warmer to help speed up the etching process.

The zip lock seals in the chemicals while etching the board.

Coming back to the dead bug method, one can use a dremel to cut out large islands. Say, three islands. One ground, one VIN, and one VOUT. Carefully arranged, one can solder more pins/pads/components to the appropriate islands of copper. Especially handy is the Input and Output capacitors now can be soldered directly across the cut between the GND and VIN or the GND and VOUT.

You can still use the sharpie pen to label the three islands before soldering any components down.

This method only requires a dremel and a dust mask.

When making protos I always order two or three sets of components. Make the first proto spacious but tidy. You want to verify the design works and possibly play with different component values, so, packing them too close makes life difficult to solder and unsolder components. Once you get a working proto, on to the final or next one. Keep the first one in case you run into new problems down the road. You can use the first board to help debug the problem.

Good luck

-WayneY

 

[Xcandescent]

Thanks for all the help, guys! Just wanted to drop a quick note to mention that I finished the "deadbug" prototype -- and it actually worked the first time I tried it. (YES!) Pics and more details in a GBACentral Forum thread here:

Pics and Stuff

Next step is to swap out the voltage-regulated boost converter for one that's current-regulated, and build another prototype. THEN the real fun with board etching, SMT soldering, etc. starts. Knowing nothing about any of this, I haven't decided yet whether selling a drop-in board would be a good idea, but I intend to at least make one for myself.

-XCN-