If the question of thermal step down for the irregular dips of the turbo behavior (the ‘yoyo’ effect), creating a larger interval between the perceived ambient temp and the upper limit (70ºC) has no bearing with my unit.
Also tried dissipating any heat from the FET or some of the 7135s:
Copper pill inserted between the FET and the charging port:
But I did stumble on something, with a partially discharged cell, much better results:
So I explored this avenue and had some interesting results.
Using a 20700 Molicel:
And then with a Sofirn 5500mAh / 26650 cell:
My interpretation is the PID behavior of Anduril when the ‘gate’ is fully opened there is an inrush current that prematurely steps down the current, the light then tries to climb back up but the current is again too high so it re-throttles. As was mentioned, Mateminco did not consult TK and there seems to be no tweeking for this particular emitter, or they had tested with a >50 mOhm cell.
It is not a workaround as rather a compromise. There is small loss of the turbo (~10%) but with a 64 mOhm cell the dip is but once.
Nice job, but sorry, I don't understand a couple things on those #'s. What is the calibrated max temp? What is the calibrated room temp?
The oscillation will of course be reduced with less amps because less amps is less heat. It would be more informational to publish the measured amps at turn on.
I really have no use for continuous operation at the max output setting of a flashlight. If I wanted to do that I would not set this kind of light to max/turbo, or some combo of using a higher resistance battery. To me if you wanted to do that with any of our over-powered FET based lights, I would be tweaking the top of the ramp to find a stable level. Most people would be greatly disappointed with what the light can do at a sustained level, but to me that's more valuable than getting a better oscillation pattern.
For me, I just want 30-60 seconds or so of max output. I'm more interested in short bursts. So if i can crank up Anduril's max temp to 90C (as I did), and set the max calibrated temp to 85C, then that is good for me with most of these lights.
Ohh - in looking at your results, it appears the heatsink you added made some huge differences - it boosted the start output by 23% (from 1462 to 1803), and not only that, it kept that higher output for longer. I can't explain why the turn on output is higher, but the longer high output makes sense if the heatsink is doing it's thing keeping the heat at the MCU more stable, less susceptible to spikes in temp.
You are right about tweaking the PID - I doubt Astrolux did anything - no tweaks, no testing.
Firstly, my intent of the tests was to reduce the ‘yoyo’ effect when at turbo, or for a max ceiling (I use stepped). So in this regards I’m testing at max output to determine if the erratic behavior of dipping then climbing up to then re-dip can be minimized.
So I had set the ambient temp at 5ºC and the max at 70ºC giving a range of ~90ºC (my room temp is more at 25ºC).
Taking the tail amps is rather problematic as the the light fluctuates enormously - anywhere between 22.5 to 20 amps and quickly drops. Again my goal was to find a middle ground for sustained output - say about 2 to 3 minutes. Sustained meaning it doesn’t fluctuate but of course will drop steadily.
I did do some tweaking to the max ceiling and the lowest floor on the last test. This after a number of tries to find an acceptable middleground. I didn’t post my notes as to not delve into the details.
Glad you noticed the higher initial output with the copper pill insert. I disregarded that as my objective is stability.
My use of the FT03S is in building inspection (outdoors - urban). Having a diminishing output, to regain brightness, and then repeat, is not acceptable. The throw is very good and the working hotspot is ideal.
I had initially posted on some of my findings in hopes of gaining some momentum of the issue. This would of perhaps get the technical team at Mateminco to look into it. I understand that there is a challenge to produce a light with this newer SBT90.2 and that some corners were cut. With some feedback from the community, perhaps the second batch could be tweaked.
I don’t have a Shockli 5500 mAh but a PLB 5000 which has 30 mOhms. The iJoy is about the same IR. So I didn’t test that one cell.
As for the copper pill discrepancy, I attribute more to the time I commenced the timer. The light drops very fast, so I may have gotten better at clicking the start. Next test (Molicel) supports this. As Tom E stated, the On Start (O/S) values should not have changed much as initially no heat (yet) thru FET. I had checked for continuity thru any exposed components*; wouldn’t want to short something out.
*The driver has a charging PCB above rendering component visuals and testing near impossible.
As my motive for posting this is to get some members to comment and maybe catch the attention of Astrolux. Then Mateminco may wish to further test and correct the firmware on later batch. But as the thread is rather lengthy, many would be turned off to investigate.
I had tinkered with various Inrush Current Limiters and some low valued resistances, I had gotten some varying results but the easiest compromise is a high capacity cell with 60 mOhms.
Again, my needs are for a sustainable (2 to 3 minutes) high lumen thrower. Having the ceiling set at 2600’ish lumens is underperforming for this emitter.
I may be barking at the wrong tree; people are content on instantaneous high output.
Ahh, ok. To sustain 2-3 mins, understandable. Who knows what 'Astrolux' is exactly. TA and Lexel maybe have contact with them, but don't think the communication is very reliable from what's been posted, and not sure how they listen to us. 2-3 mins all depends on amps in a light of this design. Certainly a better light design for shedding heat would get higher amps for that runtime.
Yes, I do wish they had TK (or someone else) properly tweak this light for Anduril's PID algorithm for thermal mgt. From past experience, they want to get to market quick and don't want to delay time sending out a production sample to have the firmware tweaked - requests like that have been denied or not responded to before by Astrolux. Semeone like Hank at IOS is probably way better to work with in regards to this.
From the 3rd data set, where I accidently tested with a partial iJoy cell (can’t remember the voltage - somewhere in my previous posts though), the light can sustain high output for the 2 minute duration - 5000’ish lumens and 56ºC at the head / switch area. My lumens are not calibrated but a relative value for my needs. 1380 Lux is a better data comparative as using a ceiling bounce test, again not an integration sphere but the technique is somewhat acceptable.
Yes, I have tried to get confirmation from Astrolux of the turbo mgt. issue. However all I get is BangGood responding that “the driver is broken”. They offered a new driver to which I never yet accepted. I was lucky in my bad luck that the light stays on (dimly) and after showing proof of the fault, BangGood front rep never acknowledged the ‘Yoyo’ effect nor wish to transfer to someone at Mateminco. Typical corporate attitude.
Would Hank at IOS be our Hank here on the forum and you would be referring to his work on IOS firmware, the one that preceded Anduril? (thought that was TK’s)
After the initial 5 second drop with a 40T/P42A I get 770 lux. I just tested it with my shockli 5500 at 4.07v (85?) and got about 660lux. Ran it for 3 mins and it didn’t step down. Had to move my hand back a but because the head was getting hot (50° max) but I still managed to hold it for 3 mins. There’s a small 10 step down after about 30 seconds which is probably normal. It ramps up back a bit after a while but this is less than 5%. I do remember running it at a similar brightness using the P42A and still having the step down issue. I think if the shockli was fully charged it would’ve stepped down but the voltage was just low enough for it not to. That’s what it seems like anyway. The cell measured 3.92v after the test.
Edit: I wanted to also add that my E07 steps down at 22 seconds which is similar to this light. The current it pulls is also similar. The temp is set at 65 but it never reaches that temperature at stepdown. Not too much of a problem on that light cause there isn’t much thermal mass anyway. This FT03S however can handle a lot more than 20 seconds heat wise.
My conclusion is the PID is inferring the current delivered versus the heat. Inferring meaning it is using some delta temperature for re-adjusting the current to the FET. Some calculus for predicting the curve. This makes for a foreseeable thermal behavior given certain parameters. I don’t believe Astrolux / Mateminco changed these parameters.
This LED can produce 60+ watts, quite a lot for a single emitter.
My little tests did not see much change in the mood swings with a copper pill inserted on top the FET (and some 3 x 7135s). And any increase of start output is probably due to my quickness of pushing the ‘start’ button on the timer.
FreeMe had tested the driver with some “Freeze-it” and a meter at tail. He observed a gain in amps. I don’t know what to make of his test. It does not address the instability of the turbo.
Edit;
With a higher impedance cell or partially used, there is a voltage ‘sag’ that limits the response time to the current drawn. The PID isn’t programmed for a steep current curve.
Hank from IOS is "Hank Wang" here on BLF. I was the one who wrote Narsil, and wrote the first version of the IOS firmware for Hank, and was paid by Hank to do so. I actually named it IOS firmware because that's how I kept track of it. That's what I mean - he works with us, and will even pay for services - not many over there do. Another example, you may see on Richards' site, MtnE, he'll charge an extra $1 or so for certain firmware - this is because he's paying a commission to the author - that was Richard's idea, not the author's.
Oh so sorry, didn’t know you were the author behind IOS (but I did know of Narsil being of your hand) - give credence to where it’s due!
The link to IOS returns ‘Apache HTTPS Server Test’ and some IT about the site either being down or registering website content to the directory. I suppose there are links within the forum to get in touch with Hank. I’ve read some of his posts.
Some news ?
I have sst40 and xhp50.2. Do I have to change the thermal conductive paste if I take sbt90.2? I can’t do that.
Is it worth it to buy it at more than double the price of sst40?
Do you have problems with heating?
I only have the black 26650 liitokala, am I ok for this flashlight?
