NewBie
*Retired*
Okay, so what makes this sucker tick, and how does it actually perform?
First, the drive waveform on the pins of the LED, from the switcher that is built in:
I mounted the drop-in module into a 2D-Cell light, so that cell voltage drop would have much less of an effect on measurements.
Next we have preparation of the MagLite 2C LED module, where I have drilled into the side of the Luxeon, to access the Luxeon slug, backfilled with thermal grease for temperature measurements:
Showing the K-type thermocouple wires stuck into the hole, notice how fine of a guage they are:
Here is a picture of part of the setup after the 30 minute test was over:
The interesting thing here is to note just how hot the LED slug gets, and how it's light output drops as it heats up. We are looking at nearly a 60% drop in output, due to the LED getting hot, from a severe lack of good heatsinking:
At the beginning, I measured an input voltage of 3.109 V from the batteries, pulling 1,320 mA. This works out to 4.10388 Watts input. I did not have time to modify the output so that I could measure current to the Luxeon.
At time zero, I got 3.986 V across the pins on the LED in the module. This is the voltage applied to the LED.
The voltage to the LED rapidly drops, stabilizing over time, as the module heats way up.
Towards the end of the 30 minute runtime, the current dropped to 410mA at 2.95V on the input side, with only 3.115V on the Luxeon pins, after the MagLite drop-in heated up to 72.7 degrees C (162.9 F) on the LED slug.
As the module heats up, it looks like the circuitry severely kicks back the power to the LED, resulting in a major drop in output.
It would be interesting to see how it would perform with better heatsinking.
If I have time, I will do a better set of measurements, monitoring additional parameters.
--------------------------------------------------------------------------
Okay, well I froze the light body (without the batteries). I installed the room temperature D cells into the flashlight. Then I used M.G. Chemicals Super Cold 134 Plus (which can cool to -51C) Catalog #403A-285g to freeze the head section, to include the pillar/post and LED assembly. As soon as the LED slug rose to -40C, I started the test.
You will note a few quirks in the MagLite 2C LED 3W module as it is warming up.
FYI, the ambient temperature, in the room where I ran this test this morning is unheated, was 7.8 C (46F). It was about 10C cooler in the room than the test I ran yesterday.
If the ambient is warmer than 7.8C (46 F), then the LED would heat up more, and the light output would drop even further. It would also drop sooner.
--------------------------------------------------------------------------
Okay, I hit another Walmart and scored some stuff:
As it turns out, I opened this Maglite 2D 3W LED, really hoping to see some innovative heatsinking, and all I found was just a standard Maglite with the LED module in it:
One of the funniest things I found, is they still include the incandescent spare in the base of the MagLite LED 2D 3W.
Yes, it dims just like the LED module does when placed in a regular MagLite.
--------------------------------------------------------------------------
Yes, the current drops *alot* from cold to hot (72.7C or 161 F), in a cold ambient (I've been running in 45-55F ambient). Even the 10C difference makes a surprisingly big impact. I would expect these MagLite LED modules and MagLite LED flashlights to not perform as well in a warmer area of the country. When the module is hot, soaking them with alot of with cold spray causes the light output to jump back to when you first turned it on, and the current goes right back up at the same time.
In case anyone needs the SKU's to look around Walmarts in their area:
--------------------------------------------------------------------------
From the packaging of the Maglite 3W LED Upgrade, for those who want to know:
--------------------------------------------------------------------------
Here is the MagLite 4D 3W LED module, you can see the switcher running (ripple), at about 200KHz:
The ones that "buck" are not just resistored like some have said.
--------------------------------------------------------------------------
I've just got a waveform off the MagLite 3D LED 3W module. Notice I had to crank up the vertical scale to see the switcher ripple, so instead of being 500mV per division, it is at 10mV per division, so it has roughly 30mV of ripple (much less than the others), and it is switching around 333KHz:
--------------------------------------------------------------------------
I know everyone likes a beamshot layout.
I'm currently working on them, here is the first teaser:
--------------------------------------------------------------------------
Added a few lights to the array. Beamshots are after about 25 minutes runtime here. The LED MagLites were alot brighter at the beginning. All lights have new cells, and cost quite a handsome sum to go fresh. Note the Fenix L2P shown here is quite the cream of the crop...
--------------------------------------------------------------------------
Here we go, one of those for your eyes only, which reveals
the secrets and the shortcommings, and explains a great
many things...
Potting this puppy might help alot...but keep the potting out of the contacts.
I really cannot believe MagLite actually did this...
--------------------------------------------------------------------------
I have not had a chance to characterize the module yet, but I will try and get some curves on it over the next few days. I can tell you it still works at 1.45V in, but it will only pull 182mA out of the cell. As such, this is well below the 0.8V of 2 depleted D sized Alkaline cells in series.
--------------------------------------------------------------------------
Because I have two 2D Mag 3W LED drop-in bulbs. I fully expected to have to ruin one when revealing it's internal guts.
Brightness difference, in a side by side, I have seen folks that can just make out 20% when I test them.
I did notice the tint shifts as they get hot.
Okay, here are two of these 2D 3W MagLite LED drop-ins, side by side. At the start they are equal brightness, very slightly different tints. I just set one up here for about 15 minutes, and let it run, and then turned on the other, and got a picture within the first minute.
--------------------------------------------------------------------------
Measuring the MagLite circuit was a little squirrley.
The reason why is that it is not very efficient, and it produces lots of it's own heat all by itself (no LED).
I measured it with the circuit board in open air, with an LED that was remotely located, the LED was very agressively cooled.
The LED I used for the test with the Mag circuit has a little high of a Vf, I will be repeating the test later on with an LED of a lower Vf. I expect the circuit efficiency to increase a little with a lower Vf bin.
It was neat to watch the circuit crank up the output power when you lightly blow on it, it responds very quickly.
I would expect the circuit to deliver more power in the module, as it looks like the temperature measurement point would end up cooler in the beginning. But once the module heats up, it would crank the power down anyhow.
I may end up having to modify a MagLite, with the module mounted in it, to get more realisitic numbers.
--------------------------------------------------------------------------
Hitting the circuit with chill spray, causes the current to jump up, and as it heats up, it draws more and more current, then it rapidly kicks down.
At the peak (frosty cold), input current hits 1.952A, and output current hit 0.952A. Output power was 4.76W and input power is 6.04W. Efficiency in this state is 78.8%. This high power spot climbs from cold, hits a peak, and rapidly drops back down. I do believe it probably happens right at the same point that my chilled runtime plot I posted earlier peaks way up.
--------------------------------------------------------------------------
Okay, I took an uber low Vf G bin Luxeon I and re-ran the same test, under the same type of conditions as mentioned before:
It was interesting to see the temperature hysteresis and thermal time constant interactions. Without the circuit mounted in the normal PR housing, the thermal time constant is faster, and you can watch it oscillate a bit, until it stabilizes.
--------------------------------------------------------------------------
Some folks wanted a clearer picture of the custom LED in the MagLite drop-in, this should clear up the custom leads on the LED:
Here is the 4 Cell version of the drop-in:
--------------------------------------------------------------------------
The power side is nothing special, it is just a low efficiency non-synchronous buck, or low efficiency non-synchronous boost. Folks have been using circuits like this around here since 2002, but most have moved on to more efficient designs.
I'll not be getting into the proprietary circuit on the other side, I will leave that up to the end user to figure out on their own.
Now for modders, there are a few items in here that could be changed to improve the efficiency of the module, most notably is the schottky diode.
On the "power side" of the board, we have the 33uF Tantalum capacitor, 10uH inductor, and the 0.1 ohm sense resistor.
Okay, lets start identifying some of the semiconductors used on the power side:
4 Cell
S3G- Toshiba P-CH MOSFET part number TPC6107 (55 [email protected] Gate)
2 Cell
S2C- Toshiba N-CH MOSFET part number TPC6004 (32 [email protected] Gate)
2 Cell and 4 Cell
1L1 OnSemi Powermite Schottky Power Rectifier MBRM110L rated for 1A, 10V reverse voltage, 0.365 Vf @ 1A & 25C, Average power dissipation at 1A square wave 0.27W.
--------------------------------------------------------------------------
Another shot of the MagLite 3W LED module (3 cell shown here):
Internal board pictures:
--------------------------------------------------------------------------
The standard step-down controller, without the MagLite modifcations, that is used on the MagLite 3 cell and 4 cell internal boards, and goes for 0.55 ea can be found here:
http://focus.ti.com/docs/prod/folders/print/tps64201.html
Datasheet is here:
http://focus.ti.com/lit/ds/symlink/tps64201.pdf
--------------------------------------------------------------------------
Good Question. I've been working with solder for 36 years now. Normally, all the lead-free solders I seen, have a grainy appearance to them, and they are somewhat harder then normal solders.
This solder doesn't appear grainy (in some areas it does, but that may be the process and what floated off the leads, particularly the inductor).
As you can see, on a close-up of one of the leads of the uber tiny SC-70 package parts, one sees none of the grainy stuff, just a little oxide in places. Keep in mind, these leads are small, ~0.225mm wide (~0.00886 inches) :
Nor is the solder hard.
I lack a lead test kit for a definitive answer, but these ones I bought at WalMart, appear to be done on a lead process.
--------------------------------------------------------------------------
Okay, started a runtime on a second "bulb", with the following results:
This test is on going, the batteries are only down to 1.27V each at the end of 24 hours, so there is a very long way to go yet:
--------------------------------------------------------------------------
As promised, Moon Mode runtime plots for the 2 cell boost version, in a 2D light.
The 3D and 4D buck would run alot longer...
It was still emitting light in moon mode 184 hours later (over 7 days). The 4D would probably still be emitting light more than 15 days later...though I have not had a chance to do the runtime on it.
------------------------------------------------------------------
If you take a look over at Quickbeam's site, he just did a review, with final runtime graphs in process on July 16, 2006. You'll notice these puppies are owning the EverLED, and pretty close to the Diamond for Throw and Overall Output. Please note that the human eye needs about a 20% difference in output in a side-by-side comparison test to tell which one is brighter, and for most folks, even 20% doesn't result in a 100% ID of the brighter one...
Stock Maglite:
http://www.flashlightreviews.com/reviews/maglite_3d.htm
3-D
Throw 6185 (78.65)
Overall Output 2300 (23.00)
4-D
Throw 9720 (98.59)
Overall Output 3800 (38.00)
MagLite's LED module:
http://www.flashlightreviews.com/reviews/maglite_mag-led.htm
2-D
Throw 6800 (82.46)
Overall Output 3800 (38.00)
3-D
Throw 7500 (86.60)
Overall Output 4750 (47.50)
4-D
Throw 7000 (83.66)
Overall Output4300 (43.00)
EverLED:
http://www.flashlightreviews.com/reviews/everled.htm
3C
Overall Output 1740 (17.40 standardized)
3D
Overall Output 2350 (23.50 standardized)
4D
Overall Output of 2440 (24.40 standardized)
Diamond 3W
http://www.flashlightreviews.com/reviews/diamond_magbulbs3w.htm
3D
Throw 8400 (91.65)
Overall Output 4200 (42.00)
Most notable, you can see the Diamond drop in output in a bigtime hurry- to 50% output on the graphs. Well, the EverLED also drops in it's graphs. And if you look, you'll see even the Incandescent bulbs drop even worse. So the drop isn't unique to the MagLite LED module...
A link for the poor incandescent bulb performance:
http://www.flashlightreviews.com/reviews/maglite_3d.htm
Keep in mind, the human eye needs about a 20% difference to even see a difference in a simultaneous side-by-side comparision...
As usual, refer back to the first post in this thread for alot more details.
First, the drive waveform on the pins of the LED, from the switcher that is built in:
I mounted the drop-in module into a 2D-Cell light, so that cell voltage drop would have much less of an effect on measurements.
Next we have preparation of the MagLite 2C LED module, where I have drilled into the side of the Luxeon, to access the Luxeon slug, backfilled with thermal grease for temperature measurements:
Showing the K-type thermocouple wires stuck into the hole, notice how fine of a guage they are:
Here is a picture of part of the setup after the 30 minute test was over:
The interesting thing here is to note just how hot the LED slug gets, and how it's light output drops as it heats up. We are looking at nearly a 60% drop in output, due to the LED getting hot, from a severe lack of good heatsinking:
At the beginning, I measured an input voltage of 3.109 V from the batteries, pulling 1,320 mA. This works out to 4.10388 Watts input. I did not have time to modify the output so that I could measure current to the Luxeon.
At time zero, I got 3.986 V across the pins on the LED in the module. This is the voltage applied to the LED.
The voltage to the LED rapidly drops, stabilizing over time, as the module heats way up.
Towards the end of the 30 minute runtime, the current dropped to 410mA at 2.95V on the input side, with only 3.115V on the Luxeon pins, after the MagLite drop-in heated up to 72.7 degrees C (162.9 F) on the LED slug.
As the module heats up, it looks like the circuitry severely kicks back the power to the LED, resulting in a major drop in output.
It would be interesting to see how it would perform with better heatsinking.
If I have time, I will do a better set of measurements, monitoring additional parameters.
--------------------------------------------------------------------------
Okay, well I froze the light body (without the batteries). I installed the room temperature D cells into the flashlight. Then I used M.G. Chemicals Super Cold 134 Plus (which can cool to -51C) Catalog #403A-285g to freeze the head section, to include the pillar/post and LED assembly. As soon as the LED slug rose to -40C, I started the test.
You will note a few quirks in the MagLite 2C LED 3W module as it is warming up.
FYI, the ambient temperature, in the room where I ran this test this morning is unheated, was 7.8 C (46F). It was about 10C cooler in the room than the test I ran yesterday.
If the ambient is warmer than 7.8C (46 F), then the LED would heat up more, and the light output would drop even further. It would also drop sooner.
--------------------------------------------------------------------------
Okay, I hit another Walmart and scored some stuff:
As it turns out, I opened this Maglite 2D 3W LED, really hoping to see some innovative heatsinking, and all I found was just a standard Maglite with the LED module in it:
One of the funniest things I found, is they still include the incandescent spare in the base of the MagLite LED 2D 3W.
Yes, it dims just like the LED module does when placed in a regular MagLite.
--------------------------------------------------------------------------
wwglen said:Even dimmed it will probably be bright enough for extended run times.
Question:
Does the battery current drop as the protection kicks in?
Thanks.
wwglen
Yes, the current drops *alot* from cold to hot (72.7C or 161 F), in a cold ambient (I've been running in 45-55F ambient). Even the 10C difference makes a surprisingly big impact. I would expect these MagLite LED modules and MagLite LED flashlights to not perform as well in a warmer area of the country. When the module is hot, soaking them with alot of with cold spray causes the light output to jump back to when you first turned it on, and the current goes right back up at the same time.
In case anyone needs the SKU's to look around Walmarts in their area:
--------------------------------------------------------------------------
From the packaging of the Maglite 3W LED Upgrade, for those who want to know:
--------------------------------------------------------------------------
Here is the MagLite 4D 3W LED module, you can see the switcher running (ripple), at about 200KHz:
The ones that "buck" are not just resistored like some have said.
--------------------------------------------------------------------------
I've just got a waveform off the MagLite 3D LED 3W module. Notice I had to crank up the vertical scale to see the switcher ripple, so instead of being 500mV per division, it is at 10mV per division, so it has roughly 30mV of ripple (much less than the others), and it is switching around 333KHz:
--------------------------------------------------------------------------
I know everyone likes a beamshot layout.
I'm currently working on them, here is the first teaser:
--------------------------------------------------------------------------
Added a few lights to the array. Beamshots are after about 25 minutes runtime here. The LED MagLites were alot brighter at the beginning. All lights have new cells, and cost quite a handsome sum to go fresh. Note the Fenix L2P shown here is quite the cream of the crop...
--------------------------------------------------------------------------
Here we go, one of those for your eyes only, which reveals
the secrets and the shortcommings, and explains a great
many things...
Potting this puppy might help alot...but keep the potting out of the contacts.
I really cannot believe MagLite actually did this...
--------------------------------------------------------------------------
I have not had a chance to characterize the module yet, but I will try and get some curves on it over the next few days. I can tell you it still works at 1.45V in, but it will only pull 182mA out of the cell. As such, this is well below the 0.8V of 2 depleted D sized Alkaline cells in series.
--------------------------------------------------------------------------
martytoo said:Newbie,
Why do you have two beam shots for Mag 2D LED Bulb (to the right of the Inova beam) ??
And, isn't it true that even though the measured beam is 50% of maximum when hot, this is just about the limit of what we can discern in brightness difference when using a light. That is, we will notice that the light is a bit less bright once it heats up, but only a barely perceptible difference will be seen.
Has anyone been able to make beam shots of the Mags cold versus hot? I think that would be interesting. ??
Because I have two 2D Mag 3W LED drop-in bulbs. I fully expected to have to ruin one when revealing it's internal guts.
Brightness difference, in a side by side, I have seen folks that can just make out 20% when I test them.
I did notice the tint shifts as they get hot.
Okay, here are two of these 2D 3W MagLite LED drop-ins, side by side. At the start they are equal brightness, very slightly different tints. I just set one up here for about 15 minutes, and let it run, and then turned on the other, and got a picture within the first minute.
--------------------------------------------------------------------------
Measuring the MagLite circuit was a little squirrley.
The reason why is that it is not very efficient, and it produces lots of it's own heat all by itself (no LED).
I measured it with the circuit board in open air, with an LED that was remotely located, the LED was very agressively cooled.
The LED I used for the test with the Mag circuit has a little high of a Vf, I will be repeating the test later on with an LED of a lower Vf. I expect the circuit efficiency to increase a little with a lower Vf bin.
It was neat to watch the circuit crank up the output power when you lightly blow on it, it responds very quickly.
I would expect the circuit to deliver more power in the module, as it looks like the temperature measurement point would end up cooler in the beginning. But once the module heats up, it would crank the power down anyhow.
I may end up having to modify a MagLite, with the module mounted in it, to get more realisitic numbers.
--------------------------------------------------------------------------
Hitting the circuit with chill spray, causes the current to jump up, and as it heats up, it draws more and more current, then it rapidly kicks down.
At the peak (frosty cold), input current hits 1.952A, and output current hit 0.952A. Output power was 4.76W and input power is 6.04W. Efficiency in this state is 78.8%. This high power spot climbs from cold, hits a peak, and rapidly drops back down. I do believe it probably happens right at the same point that my chilled runtime plot I posted earlier peaks way up.
--------------------------------------------------------------------------
Okay, I took an uber low Vf G bin Luxeon I and re-ran the same test, under the same type of conditions as mentioned before:
It was interesting to see the temperature hysteresis and thermal time constant interactions. Without the circuit mounted in the normal PR housing, the thermal time constant is faster, and you can watch it oscillate a bit, until it stabilizes.
--------------------------------------------------------------------------
Some folks wanted a clearer picture of the custom LED in the MagLite drop-in, this should clear up the custom leads on the LED:
Here is the 4 Cell version of the drop-in:
--------------------------------------------------------------------------
The power side is nothing special, it is just a low efficiency non-synchronous buck, or low efficiency non-synchronous boost. Folks have been using circuits like this around here since 2002, but most have moved on to more efficient designs.
I'll not be getting into the proprietary circuit on the other side, I will leave that up to the end user to figure out on their own.
Now for modders, there are a few items in here that could be changed to improve the efficiency of the module, most notably is the schottky diode.
On the "power side" of the board, we have the 33uF Tantalum capacitor, 10uH inductor, and the 0.1 ohm sense resistor.
Okay, lets start identifying some of the semiconductors used on the power side:
4 Cell
S3G- Toshiba P-CH MOSFET part number TPC6107 (55 [email protected] Gate)
2 Cell
S2C- Toshiba N-CH MOSFET part number TPC6004 (32 [email protected] Gate)
2 Cell and 4 Cell
1L1 OnSemi Powermite Schottky Power Rectifier MBRM110L rated for 1A, 10V reverse voltage, 0.365 Vf @ 1A & 25C, Average power dissipation at 1A square wave 0.27W.
--------------------------------------------------------------------------
Another shot of the MagLite 3W LED module (3 cell shown here):
Internal board pictures:
--------------------------------------------------------------------------
The standard step-down controller, without the MagLite modifcations, that is used on the MagLite 3 cell and 4 cell internal boards, and goes for 0.55 ea can be found here:
http://focus.ti.com/docs/prod/folders/print/tps64201.html
Datasheet is here:
http://focus.ti.com/lit/ds/symlink/tps64201.pdf
--------------------------------------------------------------------------
Ray_of_Light said:Newbie, I have a question for you, since you have disassembled them: is the solder leadless? I have read somewhere that the new leadless solder has a limited number of thermal cycles, then it cracks. END EDIT
Anthony
Good Question. I've been working with solder for 36 years now. Normally, all the lead-free solders I seen, have a grainy appearance to them, and they are somewhat harder then normal solders.
This solder doesn't appear grainy (in some areas it does, but that may be the process and what floated off the leads, particularly the inductor).
As you can see, on a close-up of one of the leads of the uber tiny SC-70 package parts, one sees none of the grainy stuff, just a little oxide in places. Keep in mind, these leads are small, ~0.225mm wide (~0.00886 inches) :
Nor is the solder hard.
I lack a lead test kit for a definitive answer, but these ones I bought at WalMart, appear to be done on a lead process.
--------------------------------------------------------------------------
Okay, started a runtime on a second "bulb", with the following results:
This test is on going, the batteries are only down to 1.27V each at the end of 24 hours, so there is a very long way to go yet:
--------------------------------------------------------------------------
As promised, Moon Mode runtime plots for the 2 cell boost version, in a 2D light.
The 3D and 4D buck would run alot longer...
It was still emitting light in moon mode 184 hours later (over 7 days). The 4D would probably still be emitting light more than 15 days later...though I have not had a chance to do the runtime on it.
------------------------------------------------------------------
If you take a look over at Quickbeam's site, he just did a review, with final runtime graphs in process on July 16, 2006. You'll notice these puppies are owning the EverLED, and pretty close to the Diamond for Throw and Overall Output. Please note that the human eye needs about a 20% difference in output in a side-by-side comparison test to tell which one is brighter, and for most folks, even 20% doesn't result in a 100% ID of the brighter one...
Stock Maglite:
http://www.flashlightreviews.com/reviews/maglite_3d.htm
3-D
Throw 6185 (78.65)
Overall Output 2300 (23.00)
4-D
Throw 9720 (98.59)
Overall Output 3800 (38.00)
MagLite's LED module:
http://www.flashlightreviews.com/reviews/maglite_mag-led.htm
2-D
Throw 6800 (82.46)
Overall Output 3800 (38.00)
3-D
Throw 7500 (86.60)
Overall Output 4750 (47.50)
4-D
Throw 7000 (83.66)
Overall Output4300 (43.00)
EverLED:
http://www.flashlightreviews.com/reviews/everled.htm
3C
Overall Output 1740 (17.40 standardized)
3D
Overall Output 2350 (23.50 standardized)
4D
Overall Output of 2440 (24.40 standardized)
Diamond 3W
http://www.flashlightreviews.com/reviews/diamond_magbulbs3w.htm
3D
Throw 8400 (91.65)
Overall Output 4200 (42.00)
Most notable, you can see the Diamond drop in output in a bigtime hurry- to 50% output on the graphs. Well, the EverLED also drops in it's graphs. And if you look, you'll see even the Incandescent bulbs drop even worse. So the drop isn't unique to the MagLite LED module...
A link for the poor incandescent bulb performance:
http://www.flashlightreviews.com/reviews/maglite_3d.htm
Keep in mind, the human eye needs about a 20% difference to even see a difference in a simultaneous side-by-side comparision...
As usual, refer back to the first post in this thread for alot more details.
Last edited: