Alonefire H45 "45W" dissection and analysis

[ulf]

Please also fixate the position of the spectrometer.

 

[colinbm]

AloneFire H42 light, five minute test repeated with the exact same setup & Optimum SRI 2000UV spectroscope.
The four batteries were discharged & recharged on a Opus BT C3100 v2.2 four cell charger.

The setup, the H42 light front lens was 19" above the entrance pupil of the spectroscope......

 

Spectrometer report 1 at turn on......

 

Spectrometer report at five minutes......

[ulf]

Now the test setup looks OK, but it looks like the intensity is down to 40% of the start intensity after 5 minutes.

I am afraid your torch is also fading.

 

Do you know anything about the internal resistance of the batteries?

They are definitely a part of the primitive current "control", together with

1. the dynamic internal resistance of the LEDs
2. the resistance of tracks on the PCBs
3. the resistance of via holes on the controller PCB
4. the contact points between batteries and the controller PCB
5. the bottom contact springs against the batteries
6. the cables towards the LED board.
7. the ground path in the housing to the controller PCB

If the losses over the sum of all above generate enough voltage so that the voltage left for driving the LEDs are close to the forward voltage of the LEDs

you will see an intensity drop, just as I did when analysing the Nemo and reached 3.8V.

In the case of the Nemo the constant current drivers needed a few 100mV.

 

Without any constant current drivers, the Alonefire can draw current from the batteries with lower voltages, but then with constantly decreasing levels.

 

Here is the resistance chain doing the same.

 

You can check if the LED module is tightened by unscrewing the front ring and carefully remove the filter and silicone ring.

If then the reflector feels wiggly there is the same thermal connection problem as I had.

Also look for any brownish LED lenses.

 

My defect LED might have been caused by that the different LEDs had quite different characteristics, so that on of them got more than 1/3 of the current.

That is not possible to find out now when the damage is done to my LED.

[colinbm]

The Opus BT C3100 v2.2 four cell charger, has an ability to measure the, dynamic internal battery resistance, I will try this when I awake, it is now 2022 here so Happy New Year.

Quote from the manual....
" Quick Test mode: The charger will analyze the dynamic internal battery resistance by applying a load and the current reading is referred to the voltage drop detected on the battery. Within 10s, the tested battery resistance will be displayed in the unit of milliohm. For good quality batteries, the internal resistance is very low: in the range of 20 ~ 80 milliohms. If battery internal resistance is over 500milliohms, then these batteries are not good for supplying power to high current loads, such as digital camera etc. But they can still be used for low energy loads, such as clock, remote controllers etc. Always use batteries with close internal resistance range when they are used in serial to achieve maximum battery life. Alkaline and any other 1.5v batteries can be tested on this charger as well. If a completely empty battery is to be tested, it is not possible to give any correct reading. Please be noted that since the internal battery resistance can be very small, and contact resistance can be a major influence factor, thus same battery tested in different slot or even at the same slot with different contact condition, the reading can be varied for 10% to 20%. This is normal. Test twice if needed to assure a correct measurement result. Due to the construction reason, there is around 30milliohm contact resistance from the charger itself, and this resistance is calculated together with battery internal resistance. It should be deducted from the reading to get a more accurate battery resistance reading. When a high impedance battery( eg. Over 2500milli ohm) is loaded onto the charger, due to its high internal impedance, real displayed charging current can be much less than your pre-selected charging current. When this happened, it doesn’t means the charger is defective. Eventually it needs longer charging time until fully charged. "

[Andy Perrin]

Happy new year! 
 

colin, can you measure the current and voltage simultaneously (with two different multimeters) across the batteries as as a function of time? So make a chart that goes:

 

time Amps Volts 

 

[dabateman]

Happy New year Colin

Seems like your doing better than the torch. 

 

[colinbm]

The last two years have been interesting, it looks like this year will be more interesting for all the wrong reasons, but Hey we are still here, so hang in there.
 

Yes I have a power meter that will measure the Volts & Amps, but I will have to measure at time intervals manually.

I am still asleep, this is just a piss stop, I still have a few hours more sleep to get.

 

[colinbm]

Internal Resistance measured....

[colinbm]

Some figures for the Boys & Girls to play with....
I have set-up the AloneFire H42 with a power meter between the batteries & light & recorded the readings at one minute intervals.
So what do I have please ?

 

0min at start up......

 

1min......

 

2min.....

 

3min....

 

4min....

 

5min....

 

6min....

[ulf]

Hi Colin,

 

The numbers looks really strange. The current values are so low.

It is almost as you are running the LED-Head on just one of the batteries.

 

It looks like the insertion of the instrument has changed the circuit to much to give valid readings or possible is it not connected in a way that show what we are looking for.

Difficult to tell without any schematic drawing of how you connected your power meter gadget.

 

It might be difficult or almost impossible to do what you are trying here by inserting a current measuring device and any extra resistance added by cables conducting the current will change the circuit too much.

Yesterday I ordered a clamp current meter that can handle DC-currents without breaking the circuit. It had a long delivery time so I have to be patient.

[colinbm]

Thanks Ulf

When I know the Vf & Cf of the LEDs I will be able to run the light head without batteries or the driver, then I can see what the out-put is again.

[Andy Perrin]

Thanks for the efforts Colin! 

[KhanhDam]

maybe if one used mineral oil and a gasket to keep the torch liquid tight woudl help with cooling. eternal dc charing jack could be added if needed. 
soo many chinese led products have inferior cooling!  or maybe a heat sink could be put over the torch's shaft to incrase heat disappation. 

[ulf]

The latest development for UV-torches with this mechanical housing style is by adding 3 more quad LED arrays to a total of 6.

They have gone from a marketing power of "45W" to "120W".

https://vi.aliexpress.com/item/1005006137265886.html

As there is no added externa cooling surfaces or fans so here cannot be an input power anywhere near 120W.

It might have different internal drive electronics for the LEDs.

 

 

Beside the one defect torch I analysed above I got a second one that is working well. It too is fading when heating up. 

Still it is my most powerful UV-torch and I like it even if it is fading.

When I use it for shorter periods, less than 3 minutes and then it is more powerful than the stable "Nemo" style torch.

 

I fell for the temptation and ordered one "120W torch, to measure and compare the emitted intensity with my "45W"

With some luck distributing the generated heat over six LED arrays the losses might be transferred more efficiently to the case making it a bit brighter for a longer operation period.

 

 

 

[Stefano]

Well, that's powerful. There's no way it can run at full power continuously, but it shouldn't overheat if ran for a few minutes.

[colinbm]

Some good news if you are looking for a brighter 365nm UVA LED & it comes from an Australian seller.
How is this for output, Fluoromins 365nm, 36mW/cm2 @ 380mm ?

[ulf]

3 hours ago, Stefano said:

Well, that's powerful. There's no way it can run at full power continuously, but it shouldn't overheat if ran for a few minutes.

I think the LED Chips would overheat rather soon and eventually die.

 

Well there are also other things that makes 120W very unlikely unless you have a very powerful and expensive buck-boost regulator to drive the LEDs.

I think that such a regulator would be quite expensive and too big to fit inside the torch. 

 

The problem with the simple unregulated design as in the "45W" model is the accumulated serial resistance in all steps between battery and LEDs and also the internal resistance in the batteries.

120W from the battery at say 4V would give 30A. The voltage drop from 30A will be rather big in the supply chain. Then there are not enough voltage to drive the LEDs at such a high power.

 

The main problem in driving UV-LEDs with those LiPO-batteries is that the battery voltage becomes too low to drive enough current through the LED when drawing much current from the battery.

That was what limited the power output with the first "45W" model.

Also the MOSFET transistors used to switch the torch on and off naturally became hot when conducting current.

Their internal resistance also increase with increasing temperature.

 

With an expensive drive electronics (that do not fit in this torch) you could definitely push 120W into those six LED arrays, but remember that when pushing around 20W into the LEDs as in the "45W" model the usable time before the drastic intensity drop started was around 3 minutes (180s)

If you assume that this is OK despite the cooling problems in the mechanical design, then if the thermal design is rather equal the drop would start after 30s. 

The actual problem for the LED chips would be much worse as the speed of energy travel though aluminium is quite slow. 

 

They would rather quickly transform and change from being LEDs and become SEDs (Smoke emitting diodes)   

There might also be some crackling sounds that you get when overloading smaller indicator LEDs 

That process is unfortunately irreversible. 

 

I am not aware of any UV-LEDs in that four chip configuration that are designed for an input power of 20W.

Normally they are designed for 10-15W input power and then assumed to be used in a design with proper thermal design.

They might be able to handle VERY short pulses of higher power, but then you exceed the specified Absolute Maximum Rating for the current of the component.

 

Search for CUN6HB4A at for a rather current LED, (but not one with all chips in parallel, a 2 + 2 configuration)

https://www.seoulviosys.com/en/

 

 

 

[Stefano]

At least the new torch could be better than the old one if ran at lower power. The same power is distributed among more LEDs, and LEDs tend to be more efficient (and maybe last longer) at lower power.

[ulf]

19 minutes ago, Stefano said:

At least the new torch could be better than the old one if ran at lower power. The same power is distributed among more LEDs, and LEDs tend to be more efficient (and maybe last longer) at lower power.

That is very true.

[ulf]

On 12/16/2023 at 3:44 PM, Stefano said:

Well, that's powerful. There's no way it can run at full power continuously, but it shouldn't overheat if ran for a few minutes.

Unfortunately the new 6-LED version is not an improvement over the Alonfire H42.

 

The optical output from the "120W" torch starts at a slightly higher level ca 10-15% above the "H42" torch, so what we see is just an even more inflated false marketing specification.

Both torches drop in intensity and reach a rather low steady state after ca 4 minutes. The "120W" six LED version drops faster and level out at a lower level rather much lower than a Nemo-Torch.

The faster drop rate indicates a less efficient thermal transfer of the losses.

The "120W" LEDs also have a peak closer to 370nm so more UV will be blocked by the ZWB2-filter that will be more heated. 

[Stefano]

I'm a bit surprised by the output power being only slightly higher than the H42 (I don't remember what the advertised power of the H42 was, maybe 45 W?), I didn't expect 120 W but at least a bit higher.

 

Also, I think LEDs are a bit more efficient when driven at lower power, so, considering that the power is distributed among more LEDs, the flashlight having even worse thermal management is another surprise.

 

There exist very powerful flashlights, with hundreds of watts of power, but they have cooling fans and can't sustain their full power for long.

 

For example, the IMALENT MS32 flashlight has 32 Cree XHP70.2 LEDs, which can run at 29 W. Assuming they are run at that power, the total input power is 928 W. Maybe the power is even higher, as the flashlight should produce only around 137,000 lumen instead of the advertised 200,000. It can only keep that power for 45 s, then it will step down.

 

We have to consider that modern high-end white LEDs are very efficient, perhaps around 80%, so there's less wasted heat.

 

Anyway, thanks for testing.

[ulf]

Here comes a graph from the measurement session:

 

I bought the "120W" with batteries and they were of a lower capacity 1800mAh-type with lower inner serial resistance. My charger reported 70mOhm.

I used these batteries fully re-charged during the measurements of both torches.

[ulf]

I took the "120W" appart in hope of improving the performance.

The application of thermal paste was indeed lousy and I had high hopes for an improvement doing that better:

 

To take the torch appart and get the LEDs and electronics free to apply the thermal paste you have to first unscrew the retaining ring at the battery side and pull out that board to allow it to rotate. Then the front assembly with LEDs and reflector can be unscrewed forward by gripping into the reflector funnels.

 

New in the design is a thin waved stainless steel washer between the driver-PCB and the retaining ring. 

The design of the PCB is also new with slightly different components. They have changed the charging circuit to a cheaper linear design.

The design with simple parallelled resistors is still there. I did not look very much at the design, but saw that they have removed some big via holes that help cooling the MOSFETs and addad lacquer coating (now green instead of red) on part of the rim of the board.

 

I applied good HY510 thermal paste between the LED-PCB and the aluminium block and also on the threads of the block before re-assembly.

Note the Driver board inside.

 

When the torch was put together again I measured the optical output.

The improvement was stunningly marginal, ca 1% and well within the measurement's accuracy.

 

Even if the LED-block was really badly assembled from a cooling point of view that was not the limiting factor.

 

Now I think the main problem in this design is the lack of sufficient cooling off the MOSFETs on the driver board.

 

All changes they made, the thin extra washer, with it's few contact points, (air and stainless steel have a really bad thermal conductivity) and the changed layout with omitted via holes made cooling the board really bad.

 

My next step will be to try to improve that. I have some ideas I want to try.

 

To be continued!

 

[colinbm]

I've got the pop corn & waiting Ulf.....

[ulf]

8 hours ago, colinbm said:

I've got the pop corn & waiting Ulf.....

If you want a success story you can put the pop-corn away again.

The new design and layout is so stupid and not curable by any patches.

The main problem is that they have replaced two MOSFETs in parallell by one single MOSFET with worse characteristics in a less suitable casing that makes cooling it very difficult.

 

I really did a massive attempt to improve cooling, but in vain. If I had read the data-sheets first I would not have tried that.

The new component's electrical characteristics is not suitable for this task. It is causing the fading and low final intensity.

The two MOSFETs , in the original primitive design were better suited for several reasons.

 

Without any modifications the single MOSFET have very little thermal coupling to the torch housing. 

I fixed that and improved the thermal transfer to the retaining ring and housing.

On the new component the only big thermal connection must be electrically insulated from ground so I had to create a bridging via a double folded 0.2mm thick copper sheet that connects to the PCB's edge via a thin (blue) thermal interface material. The sheetmetal is also soldered together after folding.

I had to create a new connection point for soldering the black cable as this part is covering the old solder position.

 

I'll just post a series of images I took during the modification. 

For now I give up. 

 

Board before modification

Note the unused position for a second (bad) MOSFET. I use that and the edge of the mounted MOSFET to thermally anchor the copper plate

 

New holes added and filled with thin thread to pass heat to the opposite side of the board, without solder and soldered and filed flat.

 

Testing of copper plate shape and position:

 

Dito soldered in place and with blue TIM

 

The TIM is of good quality, but naturally a direct connection that was possible in the first design is better.

 

In the far future I just might decide to design a board with a constant current Buck-Boost converter including an overheat control that can be mounted inside at the aluminium  LED mounting block opposite the LED PCB.

 

If I design such a driver board it will cost more than the torch itself. As the torch's thermal design is really marginal it might not be worth the trouble, but it might be fun to see if I succeed doing so.