Sofirn SP36, impressions of a sample

Hammering the driver out with a pin through the switch hole appeared fruitless and only damaged the driver. So a new approach was needed, something drastic because the glue was tough. With the 80W beast I soldered a sizable strip of copper on top of the batt+ ring.

Chunk of iron was stuck through the loop, the head was placed on top of the partly opened vice, with the loop and chisel hanging loose in between. Holding the long end of the bar I started hammering carefully on the short end, each time a bit harder until the driver broke loose. (or the strip, or the batt+ring, luckily it was the driver :slight_smile: )

So here are some pics of what’s inside. As with the Q8 they used a nice copper board and the reflector is fixed in place with scews coming from the driver cavity, with the ledboard sandwiched in between reflector and shelf. The shelf is 5mm thick which is very adequate.
The reflector btw only clamps against the front lens, it makes no contact with the side of the bezel, there is a small gap to the side. But I can’t think of a problem in that.

Mind that this is a prototype so some improvisation is expected, I see that the driver is hand-soldered, I see an extra cap soldered on top somewhere and a little jump wire, I see that they insulated the ledwires from the reflector with some led-centering thing (the concept of that is :+1: because it removes a potential risk).


(these are transistors? I found these components on aliexpress marked X15V 3 K and those were transistors)


(stacking an extra cap. Is this the charging chip?. Stereopic)


(the SP36 board is smaller than the Q8 board, but the same type, they just put the led+ trace on the outside instead of inside, but there’s clearance everywhere it could touch the shell)

So does it look good from a heat-shedding standpoint?

Thanks for going the extra mile to get a look under the hood!

No complaints here. The heat path is in fact very similar to the Q8. Of course with the light being a bit smaller and slimmer the heat-soaking mass and the outside surface area is less but still I get 4.25 minutes of 4600 lumen average turbo before stepdown is needed (needed as in: too hot to hold) so the design works fine spreading the heat away from the leds to the complete shell.

But getting the heat path right is no rocket science, it is in fact a miracle that so many budget light manufacturers manage to screw that part up.

Now that I have the light open without (I think :cowboy_hat_face: ) demolishing it, I ordered some 90CRI 4000K LH351D leds from led4power to make it into a flashlight that I will actually use (please forgive me my high CRI addiction)

Thanks for your hardwork. It looks great from a heatsinking standpoint. 4600 lumens for 4.25 minutes makes this a very practical light relative to the norm. Looks like Sofirn has another winner. I’m just hoping they start putting in USB type-C recharge in their C8F 21700.

Nice improvisation on the driver removal djozz. I’ve not seen a cold chisel put to this use before. :beer:

Sounds like it could be a nice gift light for a responsible non-flashaholic. Christmas is coming soon when you factor in shipping time from China.

Exactly my thought. I have no definite information about the longterm safety of having three cells parallel permanently inside a flashlight (sources of information vary, but some say that having a fusable resistor inline with every individual cell should provide enough safety, which is exactly what the tail springs provide), but give it to a responsible person with the story to check the cells every now and then, and he/she should be fine.

I hope Sofirn is correct about their time frame. I was told about a couple weeks. (crosses fingers and plans to order a trio of 30Q cells from Mtn.)

I’m not sure, just before oktober Barry told me of some production delays because old models needed to be produced first for some sale. At least I did not have a long list of complaints about the SP36, I just asked for a driver retaining ring instead of glue, and for a neutral emitter choice. That would delay it some more but I can imagine that Sofirn will ignore my tips, and then it will still be a nice light.

Thanks djozz for sharing your preview for this flashlight.

Good to see more additions to this format, which seems to be a sweet spot for output vs weight since HaikeLite MT01/MT02 :slight_smile:

The worst thing I can see happening from putting three new, good-quality cells of the same model into a light like this and handing it to a “muggle” to use and recharge in-light is that they won’t think my own lights are as cool as they used to.

At high charge and discharge currents (much higher than anything in this light) the differences in internal resistance between cells might be enough to cause problems, eventually, but even then the outcome would be accelerated wear on the pack.

At the charge and discharge rates of this light, I wouldn’t even be too concerned about “a 2200mAh store brand protected cell […] put in the light alongside a high discharge 30Q and a 3500mA high capacity cell.” It’d be a stupid thing to do, and wear and tear would certainly be accelerated, but it doesn’t strike me as a dire situation. The 2200mAh’s protection might trip, leaving the 30Q and the high capacity cell to carry the load. More of the load would fall to the 30Q, which is capable of direct-driving 4x XP-L2 emitters on its own, if need be. During charging, the 30Q would draw more of the current, until it didn’t. It wouldn’t overcharge. None of them would.

Safety and caution are important. Making a much bigger deal of something than necessary isn’t cautious and doesn’t serve safety.

Hi, can we use flat top 30q for the flashlight? Or is this a button top cell flashlight?

This is a button top flashlight, but cells with only the slightest protruding top will work.

Thanks. Will need to get 2 sets of button tops for this.

I got a message about the SP36 running too hot with NarsilM, with a request to try to fix it. NarsilM is not maintained any more though, and has no real thermal regulation, so I sent this instead:

http://toykeeper.net/torches/fsm/anduril.2018-10-17.SOFIRN_SP36.hex

I don’t have a SP36 so I had to guess about some config options, but I’m hoping this will work. Is there any chance you could try it?

Thanks for taking the time to adapt Anduril for the SP36. I did not see the overheating problems on my SP36 sample (it is not that different from how a Q8 behaves) but I would not mind Anduril on the SP36 :slight_smile:

But concerning my flashing skills, they are on pre-historic level. For me to flash the MCU would require lots of time to figure out how to do that on my ancient Mac, reading Hoop’s tutorial, acquire clip and other stuff, and probably an extra brain.

The chance of succes would increase enormously if I sent you the SP36 sample (which would hurt quite a bit and cost me 20 dollars, but for the sake of progress would be ok), or even better if Sofirn/Barry would send you an extra SP36 sample.

The glue could be an issue, and I’m a little concerned about the wire going from a MCU pin to some other parts. Would have to remove that before flashing. But maybe I won’t have to?

It’ll be faster if we don’t have to wait for shipping and such, so I’m hoping Sofirn can test it.

To make my SP36 sample a bit less sample and a bit more useful, yesterday I changed the leds to Samsung LH351D 4000K 90CRI (bought them from led4power). I changed the led wires from 24 AWG to 18 AWG to compensate for the higher Vf of the LH351D compared to the XP-L2’s that came with the sample. I did not bypass the tailsprings.

The light is slightly less yellow than my BLF Q8 with the same leds (those were obtained from AEDe) but nevertheless very comparable: at low level a bit above the BBL, diving just under the BBL at max. Great colour rendering with a nice neutral tint but without the rosiness of the high CRI Nichia 219B.

The output on freshly charged 30Q’s is 4250 lumen at start, 4000 lumen at 30 seconds (corrected to maukka’s calibration). The temperature stepdown (that I had set previously at 60 degrees as measured on the outside of the head) kicked in at 4 minute 40 seconds, so comparable to the XP-L2 situation, the wattage seems to not have changed much.