YAPDMM — Yet Another P7 D Mag Mod (Pic Heavy)

I put this together a couple of weekends ago but due to a combination of being busy, limited 'net access and not being particularly original, have only just got round to posting. Actually, I wasn't going to bother but there has been a question about KD D Lion cells and my driver solution is a bit outside the usual AMC 7135.

• Silver/Pewter 2D Mag
• Litemania H/S
• KD reflector (SMO or MOP, currently undecided on a permanent choice, leaning towards smooth)
• Option of KD Aspherical Lens (Jury also still out)
• LM338T Variable voltage regulator configured for 2.5A constant current
• Two KD D Lion Cells

LM338T & resistor (both in T0220 cases) mounted to H/S with Arctic Alumina epoxy, wire is 22AWG (I think) silicone.






KD spring needs tip bending in and down to stop it drilling through the bottom of the cells






Two KD Li-ion cells fit reasonably with the short spring — My driver solution wastes a bit more voltage than the AMC7135 boards so one cell won't do.






Beamshots — These are a fortnight old and I've forgotten which ones are which. I'll do some more carefully if the weather is good enough this weekend.














Kev.

awesome mod!

I did not know there are resistors out there that comes in TO-220 cases, thats a pretty interesting method to mount them...definitely a plus to keep the ends of ceramic resistors from touching the Heatsink:thumbsup:

Very interesting and nicely done.
I would be interested in aprox distance of your beam shots.

Thanks
X/BillyD..

I'm not at home and don't have access to the full res photos so I've had to blow these up from the ones I've uploaded. I'll replace them with 100% crops when I get home.

[highlight]After looking at them, I won't bother as they are no sharper than the blown up ones[/highlight]

The white house lit up by the uppermost two dice is at ~660ft/200m measured with Google Earth, I am impressed.




The hedge is at ~390ft/119m, the last of the three is a slightly unfocused aspherical, it appears to illuminate better, possible because it is less distracting than the Georgian window effect. Someone should come out with a bat shaped die.








Kev.

Cheesy said:

• LM338T Variable voltage regulator configured for 2.5A constant current
• Two KD D Lion Cells

LM338T & resistor (both in T0220 cases) mounted to H/S with Arctic Alumina epoxy, wire is 22AWG (I think) silicone.

Click to expand...

My electronic knowledge is hopeless; does this provide constant (flat) current? If so, what sort of efficiency? I am curious because I am looking for an efficient constant current driver for a Cree aspheric.

Thanks in advance,
Kurni

Kurni, don't feel bad, I'm electronically handicapped in the worst way. I thought resistors dissipated electrical energy into heat? but if they are being used to regulate current then it must just slow the output of the current? does it still lose energy in the form of heat radiation?
sorry if this is the wrong post I'll try to move my question.

nice looking build at any rate! look forward to understanding it

This is the next generation firmware for reporting windows errors.
Everytime the light is turned on Bill Gates mysteriously appears, in drag, which is odd. *

*I'm just joking, please don't sue me.

Very nice light btw. Sorry for the silliness, I can't help it sometimes.

It's a constant current regulator (see here), consisting of one LM338T voltage regulator and a 0.5Ω resistor to give 2.5A.

It provides flat regulation down to ~6.8V (Vf (LED fwd voltage ~3.6V) + Vdo (regulator dropout voltage ~2V) + Vadj (regulator adj pin voltage 1.25V)).

It's simple and the regulation is good but if you want efficiency, you need to use a switching regulator.


Kev.

Meltdown said:

Kurni, don't feel bad, I'm electronically handicapped in the worst way. I thought resistors dissipated electrical energy into heat? but if they are being used to regulate current then it must just slow the output of the current? does it still lose energy in the form of heat radiation?
sorry if this is the wrong post I'll try to move my question.

nice looking build at any rate! look forward to understanding it

Click to expand...

The LM338 is actually a voltage regulator. The resistor is called a sense resistor. While the LED does not obey Ohms law, the sense resistor does. You wire it in series with the LED. By monitoring the voltage drop across the resistor, the voltage regulator can effectively provide constant current.

Yes you are right about the heat. I looked at LM338's from 5 different manufacturers and all of them only regulate down to 1.2 volts. You need a little head room so say you use a resistor that will draw 1.3 volts @ 2.5 amps in series with the LED. That's 3.25 watts of heat lost in the resistor. That also doesn't count whatever inefficiency exists in the IC itself.

Here is a National link of their LM338. About half way down the page you can download a pdf on this regulator IC. At the top of page 5 in the pdf they have an application hint called load regulation. This shows how to hook it up.

I wonder what would happen if you used this IC in its normal voltage regulating capacity. The diagram to the left of the load regulation diagram shows a simple diagram of voltage regulation, although not labeled. LED's when they get hot tend to have their Vf drop. That could be disastrous if the regulator kept dumping more power to the LED as it overheated. But then again, look at all the folks running direct drive without incident. Also using an adjustable R2 as they suggest would give a variable output.

I have to admit here my knowledge of electronics is limited. I would be curious to know if the LM338 would work, if not great, at least better than direct drive as a voltage regulator in this application.

excellent, thanks very much for the detailed explanations

Ha, nice and simple

So by regulating the resistor voltage to constant 1.25V, the LED also receives constant voltage hence constant current.

Assuming Vled=3.6 and Vres=1.25, you're losing 26% of the power through the resistor alone; have I understood it correctly?

Thanks,
Kurni

The powerconsumption is as much as 2 P7 LEDs in series.

The lm338 converts the excess voltage into heat.

so overall consumption with full batteries would be 8.4V * 2.8A = 23.52W

kurni said:

So by regulating the resistor voltage to constant 1.25V, the LED also receives constant voltage hence constant current.

Click to expand...

Not exactly, the resistor voltage drop is a constant 1.25V and therefore the current is constant. In a series circuit the current is the same through all the components therefore the LED sees the same constant current regardless of how it's Vf varies with temperature.

kurni said:

Assuming Vled=3.6 and Vres=1.25, you're losing 26% of the power through the resistor alone; have I understood it correctly?

Click to expand...

Not knowing how you came up with the 26%, I think you are confusing voltage and power (P=IV). Efficiency calculations are power in versus power out.

By my reckoning, the driver losses (regulator & resistor) are around 40% with fully charged cells (~8V under load) down to 30% when it drops out of regulation at ~6.8V.


Kev.

kurni said:

Ha, nice and simple

So by regulating the resistor voltage to constant 1.25V, the LED also receives constant voltage hence constant current.

Assuming Vled=3.6 and Vres=1.25, you're losing 26% of the power through the resistor alone; have I understood it correctly?

Thanks,
Kurni

Click to expand...

I was about to say yes until I saw Der Wichtel's post. As I said, I have a limited knowledge on the subject myself. Perhaps Cheesy can chime in with his test results. I took it for granted he had his DMM wired in series with the LED when he set it up for 2.5 amp draw. Or was he reading 2.5 amps at the tailcap? He didn't mention anything about heat. But 21 watts must be almost too hot to touch the head after just a few minutes.

I already have a 3D MagLite. I used a 0.1 ohm 5 watt power resistor wired in series with a CSXPI-P7. I get exactly 2.8 amps with my 3 D NiMH's fresh off the charger. It only has a 10% power loss in the resistor. That's as good as any regulator in terms of efficiency. But by the time the batteries drop down to 1.2 volts per cell, I'm only getting 2.1 amps. I'm not that happy with the arrangement.

If I used 4C's in the same light with a modified tailcap spring and an LM338, there wouldn't be any extra voltage. Just enough for the sense resistor and the LED. The LM338 is cheap enough to test and discard if it doesn't work. I think it's worth a little experimenting. I can cobble it together on a power supply at work before I spring for the batteries.

Cheesy said:

Not exactly, the resistor voltage drop is a constant 1.25V and therefore the current is constant. In a series circuit the current is the same through all the components therefore the LED sees the same constant current regardless of how it's Vf varies with temperature.



Not knowing how you came up with the 26%, I think you are confusing voltage and power (P=IV). Efficiency calculations are power in versus power out.

By my reckoning, the driver losses (regulator & resistor) are around 40% with fully charged cells (~8V under load) down to 30% when it drops out of regulation at ~6.8V.


Kev.

Click to expand...

A 40% power loss doesn't sound too bad. It sounds a lot better than 21 watts. What resistor did you use? What current measurement do you get at the tailcap of your light? The LM338 spec sheet says estimating from a graph, there is a dropout voltage (I/O differential) of about 2.3 volts at 25°C with a 3 amp load. So 2.3 + LED + sense resistor = 6.8 volts? I guess this means my idea of 4*NiMH isn't going to work. But 6*AA eneloops in a 2D MagLite might.

Did you try when you were experimenting to just regulate the voltage of the LED with the LM338. That is to say instead of regulating the voltage of the sense resistor, regulate the LED directly. OK, it would no longer be a constant current regulator. And there would be the danger of a thermal runaway in the LED. But it would allow the introduction of a simple potentiometer as a dimmer. And it would be more efficient in the bargain.

Actually isn't kurni correct? Since current is everywhere the same in a series circuit, voltage is equivalent to power. At least in terms of a ratio. A ratio of 1.25 volts at the resistor, vs a total of 4.85 volts would be a 25.77% loss just for the resistor. Plus whatever the regulator uses.

Cheesy said:

Not exactly, the resistor voltage drop is a constant 1.25V and therefore the current is constant. In a series circuit the current is the same through all the components therefore the LED sees the same constant current regardless of how it's Vf varies with temperature.

Click to expand...

Yup; exactly; the current in resistor = LED.

Cheesy said:

Not knowing how you came up with the 26%, I think you are confusing voltage and power (P=IV). Efficiency calculations are power in versus power out.

Click to expand...

Sorry, I took a shortcut.

Pres = Vres*Ires
Pled = Vled*Iled
Ires = Iled; let Ires = Iled = I
Loss = Pres / (Pres+Pled)
Loss = Vres*I / (Vres*I+Vled*I)
All I can ben ignored because they're in numerator & denominator; hence
Loss = Vres / (Vres+Vled)
Loss = 1.25 / (1.25+3.6)
Loss = 26%

This is just calculating what came out of LM338T; obviously LM338T consume some power as well hence the total loss would be greater than 26%. I didn't include LM338T in the calculation because I didn't know how (it's not just a resistor). Der Wichtel appears to suggest that total loss is around 50%, which sounds about right to me because 2 LiIon should be able to power 2 P7, but in this case only 1 P7.

Cheesy said:

By my reckoning, the driver losses (regulator & resistor) are around 40% with fully charged cells (~8V under load) down to 30% when it drops out of regulation at ~6.8V.

Click to expand...

40% @ 8V to 30% @ 6.8V makes sense; the higher the voltage the more power converted to heat.

This is a different subject, but relevant. I understand how this setup convert excess power to heat; but could anybody give me a very brief explanation how other drivers such as nFlex can have at least 85% efficiency (less than 15% power loss)?

Thanks in advance,
Kurni

kurni said:

This is a different subject, but relevant. I understand how this setup convert excess power to heat; but could anybody give me a very brief explanation how other drivers such as nFlex can have at least 85% efficiency (less than 15% power loss)?

Thanks in advance,
Kurni

Click to expand...

This is an interesting set of articles I found on National's site. Part-2 - Buck whenever possible, gives a good if brief description of buck constant current regulators. The high efficiency ones are switching regulators. They use a combination of inductors and capacitors to smooth the final waveform. They give a few actual examples though none is powerful enough to drive a P7. I guess you could say the high efficiency of constant current switching regulators is due to their being designed for that purpose. The LM338 was designed to be a voltage regulator.

I think I've made an error in my calculations, the real figures are probably closer to 50% than 40%. Let's just say it's not very good. It does, however, represent the best compromise for my purposes. I intend to build another with a D2DIM which should allow long term use without excessive heat build up.

The LM338 is a linear regulator, it dissipates excess energy as heat, irrespective of operating as a voltage or current regulator.

The nFlex is a switching regulator, they work by switching a fixed current either fully on or fully off, varying the duty cycle to alter the average current at the output. This greatly reduces the heat generated and doesn't waste battery by burning off what you don't need, hence the better efficiency.


Kev.

Wow; thanks Al Combs, Cheesy

It makes a perfect sense now; switching regulator trickles the power in a very rapid frequency. Inductors & capacitors smooth the trickle by absorbing and releasing jaggy output from the switching thingy. Instead of trickling the input, LM338T & resistor burn the excess power; hence the heat.

Thanks for the references; I'll read it when I'm no longer overwhelmed with electronics; these may be a simple concept but really new to me

Kind regards,
Kurni

There are some high current LDO's out there (Micrel, Linear) if you want to increase efficiency. If you're lucky to get a low Vf P7, you can parallel the D cells. That way you're not dumping all that power into the resistor and regulator. Another option would be to use one of those Point of Load switchers in constant voltage mode. This gives efficiencies in the 90+ %. However, you need to tweak the voltage carefully.