Yeah, the results of the tests are not surprising to me, just interesting to finally have to back to back tests to back up what I already knew.

I never bothered to actually log the results of such a test because I had seen the results anecdotally many times.

Not to mention that when you think about it logically the results I got make perfect sense and are what I have been saying for some time.

I should be posting the results shortly, most likely the last thing I will be posting for a few weeks at least as internet assess is going to be very limited.

I’ve long wondered (ok, long = the short while I’ve had this flashlight obsession) how much clamping force the typical reflector provides when you screw down the bezel and it compresses the MCPCB. There are often people (myself included) sanding down centering rings/insulation gaskets, and I wonder if it is negatively affecting the clamping force the reflector provides.

Unfortunately I do not have any strain gauges at home to test any of this. Maybe I should get some, it might be a worthwhile data point.

Yep, I just ran a test earlier to put all of this to the test. Basically as long as you have ok clamping force, a thin layer of thermal paste will mostly make up for the lack of clamping force.

Excessive thermal paste though is far worse then no paste at all and good clamping force.

This also plays into what I was getting at with the reflector clamping to the mcpcb. That screw in the center is going to want to pull the center of the mcpcb off the shelf, possibly opening a gap there is someone got extreme with cranking that screw down. While the 2 screws holding the mcpcb could also cause unevenness.

I don’t think it is a make it or break it deal at all but I would like to for example see the screws that are holding the mcpcb instead go from the bottom of the shelf, through the mcpcb and then screw into the reflector since there is no shelf in the center for the screw to mount to.

There should be enough meat between the reflectors for this to work and should not be a hard change but it is also not a must. They just need to be careful to not crank that center screw down too hard.

For anyone interested here is the test I ran on thermal paste vs clamping force effectiveness: Just how important is thermal paste? The results may surprise some people... Tested by Texas_Ace

I go to great lengths on many of my lights to increase clamping force. Sometimes adding thicker o-rings, shimming the reflector or whatever else I can think of. I like quite a bit of force on mine, more then most.

That said as the tests above show, as long as you have a thin layer of thermal paste and there is at least some clamping force (you can noticeably feel the resistance when screw on the bezel), you should be ok.

Even in the worst case scenario it only lost about 9%, although I would worry about the LED life in that case since it was obviously much hotter.

Your data suggests something else. The output with thermal paste and clamping was consistently higher than without the thermal paste unless we are misinterpreting your graphs:

Think those lines are overlaying each other. The tests are certainly good data, but based on lumens alone, makes it tough to compare. Differences only show at high amps. Not sure how it would look with higher powered CREE's, or multiple LED MCPCB's for example. Think the patterns would be the same, but maybe more visible in differences.

Despite this thermal input I fear we (or I) will not ask for a shelf, ledboard AND reflector adaptation over it
We have a big ledboard and pretty good cooling as is, a change on all those parts would take months and sorry I cannot bring myself to it, not now we are relatively close and proto v2 does ~6K lumens after 30 seconds where we spec 4K.
It just doesn’t seem worth the hasle and time.
But maybe I am just getting a bit tired and anxious to hold a Q8 so in dire need of BLF input and push :wink:

The Miller I know you’re getting weary. We do appreciate your effort in this, and by the other folks involved also. Yeah, it’s been a long haul.

But will it work as is? Then OK. But if it screws over the modders it needs to be fixed and made right since modding was one of the design parameters. It needs to work for everyone based on the original design.

BTW, I’m not a modder.

I was going to mention this. Then I saw the responses. Then I figured I would toss it in anyway. Regarding another community with thermal compound issues.

People putting heatsinks onto various computer chips are commonly not knowledgeable about thermal compound. They slop it on and can’t figure out why their cpu temperatures are so high.

One suggested approach is to put a little dot of TC on the chip header, and then smear it around with something like the (non-scratching) edge of a credit card. Final operation after there is goo everywhere (including fingers, the can of beer, and the dog [if present]), is to make a single final pass with the edge of the edge of the credit card to remove excess compound. Then you add the heatsink. Which will, of course, now have goo on the outside of it. In the case of cpu heatsinks, they usually have an adjustable contact force from the heatsink to the chip. Doing that wrong can cost hundreds of dollars; so they tend to be very aware of applied force.

Sort of the equivalent of light modders and their assorted parts, the absolute best heatsink and thermal compound is the subject of countless debates and tests.

Cut up an old credit card and you have a tool for dealing with flashlight innards.

Hey Miller, so what are we (you and Tom) going to ask ThorFire for?
To the knowledge so far the main issues seem to be wrong LED’s, LED adjustment as not yet perfect and Retaining ring for easy access. If so it seems very close already. Or am I missing something?

I don’t understand the OP’s statement about this either. :question:

IF I am reading the results correct, thermal paste with weak clamping had the most lumens .
And was only marginally less in the Amps department.

What am “I” missing
EDIT: Below is a Summary Table sac02 put together of T_A results. Thanks to both of you. :+1:

I'm calling the better Q8 proto #2, and the other proto #3. Better one (#2) has the easier switch and bit higher output. The one I took apart is #3 and found the screw securing down the reflector was pretty loose. Might explain lower lumens if the LED's weren't in as tight, and centered as they could be -- not sure though.

I'm not fond of the black alum switch trim. Barely touched it with needlenose and it removed some of the anodizing - this is before I applied any pressure at all to unscrew it.

I'll dig into both of them a bit more to see if #3 has any detriments in current flow, or extra resistance, or LED issues. LED's appear to be the same and in domes are in good shape but haven't checked their reflow quality yet.

From what I know proto 2 needs:
Correct XPL V6 3B LEDs
Little larger oring
Good finish of inside of tube
Latest Narsil
Maybe another FET, maybe

Things are better
Stainless steel bezel
A true Narsil driver

From what I know proto 2 needs:
Correct XPL V6 3B LEDs
Little larger oring
Good finish of inside of tube
Latest Narsil
Maybe another FET, maybe

Things are better
Stainless steel bezel
A true Narsil driver

I got some homework to do. I'll go thru our spec list this eve and see where we stand.

Just to update that list for fixes:

Correct XPL V6 3B 3D LEDs

Little larger oring at the lens - might be able to stretch it out, gotta see

Good finish of inside of tube

Latest Narsil

Maybe another FET, maybe

Things are better:

Stainless steel bezel

A true Narsil driver

double springs

lost the retaining ring but we do have screws

BLF Q8 labeling

Other things to note:

  • tube no longer LEGO's with other SRK's, but there were issues with that anyway
  • new one is 2 mm shorter (good)


(edit: Tom beat me to it)

To really test thermal compound, you need to put it between a heat source and a heatsink. Then, with controlled power in the source, measure the temperature at some relevant location on the heatsink for various power inputs. While current through a device does have a temperature dependence, I am not sure this is a preferred evaluation technique. Mapping the device curves for current could be useful. But how much of the observed effect is in reality a temperature feedback loop within the semiconductor device. And then again, I could just be perusing angels and pin landing sites.

For picking a compound, Google some of the PC overclocking sites or just general testing sites. Be aware, some compounds also conduct electricity.

Big Big thanks for that!!! I am searching for a FW with Ramping (for clicky lights) for a long time. Hope that it will be available soon! Even if i have to change the tiny13 with a tiny25 to get the best result.

@Tom E
Can you see which FET they used on the latest Proto Driver?
Is the polarity protection ring on top of the screws glued in? I have to say i do not like glue in a light! And i absolutly do not like countersunk screws to secure a driver!
It looks cheap and the slightest misalignment between the holes in the driver and the holes in the light could warp the driver. Also if you are using countersunk screws on a PCB you can easily crack the PCB if you are not really careful with tightening the screws. I really hope that they find another solution!!!
It might be ok for most of the people but it is not really mod friendly. Sure i can find a mod solution (glue in a new driver. . .) but i would only use glue as a last resort!

Yes, we know what FET it is and we will ask them for a change on it. It's the NXP PSNM3R0-30YLDX, so not the better one. Better one is: PSNM2R0-30YLDX.

Yes, the polarity protection was glued on, but not badly. I used the pry tool shown in post #5314 to get it loose with some minor damage. We are discussing options there with them.

Agree, not the best way to go w/screws there, but they seem committed to it, being that the battery tube and head were designed for it. For modding, a replacement driver could be press fit with no glue - this is what I usually do, but couple dabs of thermal epoxy could work also.

Oh boy -- driver diameter is 47.6 mm, slightly bigger than most SRK clones, and our BLF Q8 (OSHPark) driver at 45.86 mm. HQ's driver is 46.23 mm with wider tabs to sand down.