I assume you did not change anything on the ramp configuration and batteries were fully charged, right?
Can you please ramp up your both K9.3 on CH2 to CEILING (120/150) and then carry out a doubleclick to go to full turbo (150/150)? Best would be to direct the beam to a white wall in a dark room. Please observe the brightness change when going to turbo level. As reported before, my K9.3 gets noticeably brighter for a blink of an eye and immediately reverts back to CEILING (120/150) again. Since loneoceans reported a similar behavior on the repaired K9.3 with SST20-DR, it would be interesting to know if your lights act differently or the same way.
IIRC, the Flashoholic reported in his video review that his SST20 2700K 95CRI on CH2 stepped down after 15 seconds. They maybe have a higher Vf than SST20-DR 660nm and cause less excess heat with the linear driver?
In any case, I am a bit surprised how quickly CH2 steps down even if the thermal threshold is set to 65°C or higher. I wonder if Andúril 2 needs some more tweaks (e.g. thermal config separately for CH1+CH2) or if everything can be justified with the basic principle how linear drivers work with low voltage emitters.
Hey Lux, yep mine is factory default. I just tossed a fresh 30Q and did your test and I think you’re right. It does seem to go down to the ceiling max after a short time on turbo. It’s rather subtle and a little hard to judge it because my eyes are adjusting to the burst of red light in the dark room. I need to buy a nice device for measuring flashlight brightness
Yeah, judging is indeed a bit harder with deep red light. It felt a bit weird to my eyes to look at this intense spot going from ceiling to turbo. It appeared like a short flash.
Interesting side effect of looking at deep red light: Color perception will be altered, i.e. standard red (620nm) will look like amber/orange but not like red anymore. Fortunately, the effect is not permanent.
I have a Sofirn flashlight (with AA battery), for which I replaced the original XP-G2 emitter to a cheap board fitted XP-E far red. I had the chance to compare the color perception to SST-20 deep red, and I found, that even SST-20 deep red (~660nm) looks a little bit pale compared to the ~730nm far red - but the difference is very subtle without putting them side to side. Difference to ~620-625nm red is much more noticeable. Still, it would be lovely to use a multi-emitter far red flashlight (that comes only in SST-10).
Would it be more reasonable to use a LiFePO₄ battery with linear drivers and low-voltage emitters like SST20-DR or Osram W1/W2? The excess voltage should theoretically be lower as the battery‘s input voltage will not exceed 3.6V, thus less heat to dissipate from Mosfet and MCU.
I shared the same thought as you and did some experiments when the Emisar D4 and D4S were out. FET, 7135, and linear regulation drivers all wasted too much energy and turned that into too much excessive heat unnecessarily. However the results of IFR’s on D4/D4S were not quite impressive, because the LiFePO (IFR/LFP) batteries have not much energy.
NOTE: That A123 cell is actually manufactured by LISHEN, not A123 itself. The quality of the Soshine IFR cell varies noticeably from batch to batch, the one used was from the best batch I got.
NOTE: That K2 Energy cell is actually manufactured by PLB, not K2 Energy itself. The PLB 35A cell actually has more capacity and higher voltage under the same load. You may consider the K2 one as a degraded PLB 35A.
But I do prefer using an IFR in my Noctigon KR1’s (CSLNM1/CSLPM1) as long as the runtime is not a real concern, because the Vf of these emitters are so low that the wasted energy and excessive heat are not acceptable at all, though it’s a good pocket thrower. An IFR 18650 can effectively reduce the heat, though the max brightness and runtime are reduced. I’m looking forward to the Fireflies T1R (using the same 6A buck driver as E12R too) to replace the KR1.
I intentionally off-calibrated the temperature sensor of the tested lights so that no thermal step-down could happen. The huge decrease in brightness with the 3.7V li-ion was because of heat (and LVP in the end, of course). For the IFR cells, the drop of brightness in the very beginning was expected, simply their discharge characteristics.
Also I would like to share with you my runtime measurements with my K9.3. In most of the experiments, I intentionally off-calibrated the temperature sensor of the tested lights so that no thermal step-down could happen. But I did calibrate the temperature sensor correctly before conducting experiments for the purple curves (with temperature limit = 50°C) to see the stable brightness range under my room temperature condition (~22°C).
Don’t take the lumen numbers seriously, they are simply rough estimations. I cannot estimate the lumen numbers for the deep-red/yellow emitters since I have no corresponding reference lights, so I only show the relative output here.
The efficiency of the linear driver is not quite impressive
The red curve (Turbo) cannot be regulated, meaning that the Vf of the emitter is simply too high.
It’s interesting to see that the red curve (Turbo) performs noticeably worse than the orange curve (Step 7) here. The extra current doesn’t give any visible increase in output, simply more heat. That’s why the purple curve looks even better than the red curve in the first 2 mins, because it steps down and decreases the current.
Sorry Photon Master, I don’t mean to bug you, but did you really have to quote the whole picture heavy comment just to say thanks to him? (Anyway, thanks to him for sharing his measurements as well.)
Wow this is a very detail tests, thank you toobadorz for making all the experiments and charts! I saw that you did same test with Fireflies E12R on here:
From what I can see to compare, using same Samsung 50S battery:
Noctigon K9.3 with E21A 3500K D220 R9080
420 to 600 lumen for 2h32m (auto regulate)
190 lumen for 7h 30m (cooled)
Charger 2h 53m
Fireflies E12R with 219B 4500K D220 R9080
1500 lumen for 1h 08m
900-800 lumen for 2h 07m (auto regulate)
800 lumen for 2h 22m (cooled)
225 lumen for 7h 0m (cooled)
Charger 3h 01m
Efficiency
I do not know the efficiency different between 219B and E21A, but it look like the E12R driver is much more efficient. From 50C no cooling test for both (I think flashlight settle at thermal stability after a while), the K9.3 go down to 420 lumen first when it is hot, then as battery voltage drop, my guess is driver is more efficient and heat less, so the brightness goes to 600 lumen until it turns off at 2h32m. The E12R somehow can stabilize at 900 to 800 lumen, about 2x to 1.5x brighter, and last for 2h 07m, shorter than K9.3 but much brighter. If I assume current is propotional to lumen within small range, then K9.3 is 77520 lumen*min, and E12R = 107950 lumen*min, so E12R driver is almost 40% more efficient (assume led is similar efficiency)?! Maybe my maths is wrong?
Charging
First it look like K9.3 charge about 10m faster than E12, but I notice that while max charge current is higher on E12R, it starts with low charge current for first 27m, before going to 1.9A max. Compare to K9.3 which start at 1.4A at max at 1.7A. Why is the charger doing this? I made some research and I found this: https://www.digikey.com/en/articles/a-designer-guide-fast-lithium-ion-battery-charging. Maybe E12R is charging the battery at low current from 2.8V to 3.0V for battery health before fast charging. I think this means that it may be good for health of lithium battery, but if your battery is 2.8V and you want to charge fast, the first part will be slow.
As I recall, E21A is a fairly low Vf emitter, thus, if a linear driver is used (even PWM-less), Vbatt - Vf difference is fairly big, so even if E21A and 219B 9080 has similar efficiency, E21A shows better these differences in driver efficiency - showing the superiority of the buck driver used by FireFlies more spectacularly. With keeping these measurements in mind, K9.3 seems definitely overpriced for me, even it is a very complex and expensive design. I have big hopes for the FireFlies E07X. If I get those with SST-20 4000K and 2700K (or hopefully ~3500K), I will trade my two D4SV1.
Because lumen brightness is low (600lumen for K9.3 / 9 led and 900lumen for E12R/12), I use lowest value on the chart per led which add up to about the correct total brightness:
E21A is 2.63V at 150mA
219B is 2.72 at 150mA
so I multiply by 2.63/2.72 and I get 34.6% more efficient for buck driver. I think most likely this is very simple estimation and not accurate, but even if it is different by +-10%, the effcieincy for buck is clearly shown in this comparison. K9.3 with buck driver would be amazing!
Thanks for your comparison, clientequator, that’s exactly what I would like to show. The efficiency of a buck driver is much better especially when a low Vf emitter is used and driven with low current.
My lumen numbers may be inaccurate, especially when it comes to a light with multiple emitters like this. I’m using a “lumen tube” for easier measurement, but its result is not that good for such kind of lights. I just calibrated my E12R results with maukka’s standard light (219B) and updated my E12R charts accordingly, and these results are now a bit lower (~7%) than earlier. Probably still inaccurate though
Hats off and thank you so much toobadorz for this detailed analysis. I took the liberty to summarize your observations and link to this thread on „the other forum“. ;-) Good to see some actual efficiency comparison between linear and buck drivers. :THUMBS-UP:
I personally prefer the E12R charger, because it ends with a lower cut-off current, leaving the battery quite close to 4.20V. The trickle charge phase for taking care of a low voltage cell is also quite good. The total charging time actually varies and is related to the cell and its voltage before charging (I own 2 pcs of K9.3 and 6pcs of the new Fireflies lights, all tested), so simply regard both of them as ~3hrs charging time.
Since the K9.3 is a switching charger, you might want to see what it output not want it inputs, since it is essentially working as a buck charger from 5V input source, so that means input current will be lower than actual output current, would be useful some output measurements rather input measurements.
Also would have been nice to see the “guts” of the E12R not only of the K9.3 in such massive amount of detail.
I believe both K9.3 and E12R utilize a buck charger, because they both consumed 4050mAh~4150mAh from the USB power source to charge a 4800~5000mAh cell.
