I thought of this a little while ago when I plotted the Convoy C8 data. If it was the meter being swamped somehow, then the actual light being used should not matter. However, in my data the initial sag is only being recorded on certain lights (in my tests, the EA4W and the M1). My thread will be posted shortly with the results.
Quick preview for those interested; M1 has sag issues, C8 is rockin’ it.
tickle me pink, i would not have expected sag issues on the M1
I should have said that relative outputs were in throw on that first graph. Here’s the same thing in lux which makes the percentages larger but the overall picture is the same.
I’m using an Extech SDL400 which I set to record every two seconds and which auto-ranges. Each light began at room temperature and I’d start recording in the dark and turn the light on. Since the meter’s already running, initial readings were recorded somewhere within in a two second window. But so what? IMO initial “actual max” is irrelevant compared to three minutes of output.
Big time sag issues. As can be seen in the above graph and longer term here:
As cheap saggy lights go, the M1 actually does not sag too bad… It’s basically just not regulated.
You are potentially missing the most important period of lumen sag; the first two seconds. IMO, ignoring this is unfortunate and invalidates your data. The variability in your readings alone causes problems. What if the M1 happened to get sampled right when it turned on, but the EA4 happened to get sampled right around the two second mark? On the EA4, the difference from time zero to the two second mark for me was about 45lm. This does not totally explain how your data shows about 4% sag during the first three minutes and mine shows almost 15% sag during the same time period. Something is still amiss. Did I read somewhere that you also applied forced air to the light under test?
Even my sample rate is limited by the meter; it is documented at 400 milliseconds in the manual.
Agreed, the M1 is not as bad as some. I still need to investigate what’s going on, I think the star contact may be limited.
For those interested, my thermal testing post is up, here with more lights tested. Future test results will be added to that thread.
I think initial testing temperature is very important. You must let your flashlight get cold before starting doing test. Maybe let your flashlight sit there for 2 hours before starting test. Else sagging percentage might not be accurate.
Hi OL. I dont think anyone is listening to you anymore. There all on there own mission now. I’ll see you back at the build thread. 
Yep, that’s exactly what I was going to say.
Surely any sag in the first two seconds couldn’t be due to poor contact from the heatsink to the body, but must be due to the efficacy of heat extraction from the LED itself?
Why do you think the first two seconds are so important that they “invalidate” the following three minutes? I have five luxmeters and can easily capture the initial max but IMO it’s irrelevant and won’t bother unless someone gives me a good reason. Who cares about when in the two second window the sampling started? There’s no “variability” that “causes problems” as anyone can see by looking at the graphs. The curves are relatively smooth for three minutes with no spikes or drops. To claim that almost 100 data points on each of five lights are invalid because of the “reasons” you’ve given seems more to me like grasping at straws because you don’t like the results. And no, there was no forced air applied. What meter are you using BTW?
Why is sag in the first two seconds at all relevant? If someone gives me a good reason why it’s important, I can capture it and graph it for each of the five lights. But it seems to me that obsessing about it is pole vaulting over mouse turds and missing the bigger picture 
The output at zero seconds is the actual capability of the emitter. I care about that number. Having good heatsinking allows me to get closer to that number for longer.
That still doesn’t invalidate other baselines… and as Calvin pointed out, there’s the question of meter accuracy when initially flooded and based on how far off some of your lumens readings look in your thermal performance thread, I wouldn’t be so sure of your measurements.
You’re also misinterpreting the data based on some of your comments.
What luxmeter are you using BTW?
Isn't ANSI spec 30 seconds, to stabilize the light? I would think any readings before 30 seconds would be erratic and not viable. Isn't that why there is an ANSi spec? I know you are checking temps, but that should be erratic at first turn on too.
Yes, between 30 seconds and two minutes, great point. I was thinking the exact same thing which is why I made this graph of lux sag over the entire ANSI/NEMA FL1 testing window.
I’m not too concerned about the meter, as long as the same one is used for the entire time period. I’m looking for relative numbers only. I responded to Calvin’s comment; I guess you missed that. If it is the meter being swamped by the initial turn on, I would think 1100+ lumens from the HD2010 should have really messed with it; but my graph shows otherwise.
Clearly the initial sag doesn’t matter to you. ANSI cares about it. They do not want a manufacturer to test a light within that period and publish the result. Why? Well, because the initial sag on most lights is significant.
As I said, I only care about the initial output so I know what the potential of flashlight is. From there I can decide if some heat sink mods are required.
Clearly, you do not care about this and I’m happy to let you continue down that path.
I thought this was a EA8 thread….nice chart BUT where is the EA8? I think this belongs in a different thread unless you add the EA8 to the mix.
I already started another thread, linked in a previous comment. So far, not many questions like this over there. This is my last non-EA8 tear down post in this thread.
In case anyone runs into real Loctite, here's what the Mfg says;
red
It is only
removable once cured by heating up parts to 500°F (260°C).
blue
For disassembly, shear with standard hand tools and remove with methylene chloride. In rare instances where hand tools do not work because of excessive engagement length, apply localized heat to nut or bolt to approximately 482°F (250°C). Disassemble while hot.
green
For disassembly, shear with standard hand tools and remove with methylene chloride. In rare instances where hand tools do not work because of excessive engagement length, apply localized heat to nut or bolt to approximately 482°F (250°C). Disassemble while hot.
The red thread locker on the Nitecore is NOT Loctite brand, as it gave up before the Ano discolored. I imagine it's an overseas budget type locker.