4 luxeon headlamp, need driver circuit

[rayearth]

I'd like to make a 4 white luxeon headlamp to use in close quarters for spot illumination. Goal is to get about 2-3 hours run time in a entirely headband contained unit. Due to weight considerations, I'll probably go with lithium-ion as they have some of the best energy storage density for the weight. I can get some surplus 7.2v packs cheap and I have the appropriate charger to go with it.

The difficult part of the design is a current limited driver circuit. I think I'll use the Linear Technology LT1308B, unless anyone has a better suggestion. Since the power output is current limited, I would hook up the 4 luxeons in series. I need help for
1) where's the best source to get the parts
2) what values to use as Ned Swiki's page on the Luxeon flashlight mod that uses this chip does not provide much in the way of values
3) is there a way to insert a variable resistor to adjust light output?

 

[bikeNomad]

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

1) where's the best source to get the parts
2) what values to use as Ned Swiki's page on the Luxeon flashlight mod that uses this chip does not provide much in the way of values
3) is there a way to insert a variable resistor to adjust light output?<HR></BLOCKQUOTE>
I can give you a better list of parts on the Swiki. Basically, they all came from Digi-Key except the LT1308B which I got from Linear's web site.
Here's the bill of materials with Digi-Key part numbers:

1 3K3 R-US_R0603 R2 311-3.3KGCT-ND
1 3V0 ZENER-DIODESOT23 D2 BZX84C3V0ZXCT-ND
1 6R8 R-US_R0603 R5 311-6.8GCT-ND
2 10n C-EUC0402 C2, C5 PCC103BQCT-ND
1 10uH PM54 L1 M4100CT-ND
1 12K R-US_R0603 R1 311-12KGCT-ND
2 10u 1210 (ceram?) C3 PCC2169CT-ND
1 100K R-US_R0603 R3 311-100KHCT-ND
1 B240-13 B130LSMB D1 B240DICT-ND
1 LM1308B LT1308BS8 IC1
1 R68 R-US_R1210 R4 P.68SCT-ND

And I can put the board design up there too if you're interested.

As far as variable output, the 0.68 and 6.8 ohm resistor in parallel are the current sensors. The LBI input expects to see 200mV to regulate. So if you added another sense resistor (sized so that at the minimum current you wanted to run at there would be 200mV across both the sense resistors), you could put a pot across the added sense resistor and run the slider to the LBI input. This would allow you to go from some minimum to some maximum current.

That is:

R1+R2 set the minimum current, R2 sets the maximum. There will be 200mV at the LBI terminal when it's running.

 

[rayearth]

Thanks for the info! Now I understand the values on the schematic, I don't have much experience, so I was expecting literal "6.8 ohms" instead of 6R8. I see now, you used those two resistances to get 200mV drop off an equivalent of 0.6182 ohms to get 324 mA drive power. That's pretty clever! About the only thing left now is a means of tuning the circuit for further efficiency, although given the datasheet, it is already in the neighborhood of 80-90%.

 

[**DONOTDELETE**]

Note that you can also buy the LT1308B from Digi-Key, for $6.50.

Although you can get the same chip directly from Linear Technology for $3.95, I suspect Linear charges a lot more than $2.55 for shipping, and you can get shipping for free from Digi-Key if your order totals $25 or more and you send them a check.

I might point out it's a bit ambitious to drive four LS's. You'll need to almost double the input voltage, and I'm not confident the above parts will accomplish that.

 

[bikeNomad]

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

I might point out it's a bit ambitious to drive four LS's. You'll need to almost double the input voltage, and I'm not confident the above parts will accomplish that.<HR></BLOCKQUOTE>

I just simulated the circuit, and found that the components that I used would easily drive 4 LS's from a 7.2V source.

As you might expect, the efficiency was better at this voltage than driving 1 LS from a 3V source (the drops across the switch, the diode, and the sense resistor make less of a contribution to the inefficiency). In simulation I showed about 92% efficiency; in practice I suspect it would be about 85%.

The output capacitor should be ceramic, of course, as it needs low ESR and a 16V unit will be operating very near 16V.

The efficiency will go down a bit if you add more sense resistor for variable brightness. For instance, if you want a 2:1 range, you'll have to have twice the sense resistance; this means that at full current it'll be dropping 400mV, for an extra waste of 70mW. Though this is only about 1.5% extra inefficiency considering the 4+W output.

 

[**DONOTDELETE**]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>I just simulated the circuit, and found that the components that I used would easily drive 4 LS's from a 7.2V source.<HR></BLOCKQUOTE>

That surprises me, because as I recall, you could only pull 310mA through a single LS with a 3-volt source at 80% efficiency.

I'm just suggesting that a different set of components would work much better.

Since you can do a simulation, see what happens if:

<UL TYPE=SQUARE><LI>You use a Zetex ZHCS2000 Schottky diode, as MrAl suggested
<LI>You use an inductor with a DCR of 0.01 ohm
<LI>You use a ceramic output cap of 22uF
[/list]

 

[bikeNomad]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Duggg:
That surprises me, because as I recall, you could only pull 310mA through a single LS with a 3-volt source at 80% efficiency.

I'm just suggesting that a different set of components would work much better.

Since you can do a simulation, see what happens if:

<UL TYPE=SQUARE><LI>You use a Zetex ZHCS2000 Schottky diode, as MrAl suggested
<LI>You use an inductor with a DCR of 0.01 ohm
<LI>You use a ceramic output cap of 22uF
[/list]<HR></BLOCKQUOTE>

I deliberately limited the output to 320mA or so; the circuit is easily able to handle higher currents.

The 80% efficiency was because of the low voltage. Of course, you'd expect a higher efficiency with a higher input voltage.

Do you have a SPICE model for the Zetex diode? In simulation, I'm only showing the diode and inductor as being responsible for wasting 3.5% of the total power input (diode 129mW L 35mW). Even if they were improved by a factor of 2 (unlikely because the bulk of this is Vf on the diode), it would only be a savings of 1.75%. At least, it's not worth it putting in a bigger inductor (the one I'm using has a 42milliohm resistance).

The simulation of the 3V->1 LS showed 83% and measured about 80% efficiency.

The simulation of the 7.2V->4 LS shows 92% efficiency; I suspect it would be actually in the 85%+ range.

 

[**DONOTDELETE**]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>
Do you have a SPICE model for the Zetex diode?

Couldn't find one on the Zetex site. But here's one for the ZHCS1000, which is a 1-amp model that should still work:

In simulation, I'm only showing the diode and inductor as being responsible for wasting 3.5% of the total power input (diode 129mW L 35mW).

Wait, output power was 1000mW, efficiency was 80%, so input power must be 1250mW. 129+35=164, which represents 13% of 1250, not 3.5%.

The diode and inductor are usually the lossiest components.

At least, it's not worth it putting in a bigger inductor (the one I'm using has a 42milliohm resistance).

The handwound inexpensive inductor we used in the Zetex circuit has a DCR of around 6 milliohms.
<HR></BLOCKQUOTE>

 

[bikeNomad]

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

Wait, output power was 1000mW, efficiency was 80%, so input power must be 1250mW. 129+35=164, which represents 13% of 1250, not 3.5%.
<HR></BLOCKQUOTE>

The 3.5% was in the 7.2V circuit:
output power 4.34W
input power 4.72W
L 34mW
D 129mW
Rsense 64mW
LT1308 104mW

Thanks for the D model...

 

[**DONOTDELETE**]

Ahh, I see! Well those are pretty impressive results then!

Am a bit surprised that the chip and Rsense use as much power as they do, but can't really complain if the efficiency's above 90%.

I still think the Zetex diode would be an improvement though.

 

[bikeNomad]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Duggg:
Ahh, I see! Well those are pretty impressive results then!

Am a bit surprised that the chip and Rsense use as much power as they do, but can't really complain if the efficiency's above 90%.

I still think the Zetex diode would be an improvement though.<HR></BLOCKQUOTE>

The chip uses that much power mostly because it's handling peak currents of about 1.4A and has a bipolar transistor switch (if I recall right).

The sense resistor dissipation is unavoidable: I'm using a hack that uses the LBI input comparator to do current regulation. And it's set at 200mV. So the resistor will always waste 60-70mW, depending on output current.