So basically the TN31 is just a larger TK35 with a third battery.
Thanks, you just saved me some money.
No, it isn't. Direct output comparisons in my lightbox and ceiling bounce show the TN31 has ~40-45% more output than my TK35 at 30 secs into the run (although that drops down to <35% after the intial step-down of the TN31). See my
TN31 review for output summary values of both lights.
You need to compare relative output measures within individual testing methodologies, not across reviewers (see comments below).
In the case of the TK35, I imagine that difference has dropped slightly with the new TK35 recently released (with improved output bin). But keep in mind, the average successive Cree output bin increase is only ~7% more output (and normal variance could make that anywhere from 1-14% for any given two samples).
EDIT: Also, throw is hugely different - my TN31 clocks in at nearly ~90K lux @1m, vs my TK35 at a little over ~20K lux @1m. That is over twice the relative beam distance (i.e., overall "throw")
Could he be that far out in his testing.
I am sorry if I have heaped rubbish on the TN31 for no reason if his test light was faulty or the batteries not being fully charged.
I can't comment on any other reviewer's methodology, but I suspect the issue is simply one of lumen estimate variation (i.e., it has nothing to do with thermal drop-down, just different baseline estimates).
None of us have permanent access to true NIST-calibrated and maintained integrating spheres of appropriate size for all lights. And even if we did, there are still plenty of factors that can lead to spurious results. Plus, you must understand the natural variation between samples, runs, etc. Manufacturers are not required to report these variations (i.e., ANSI FL-1 just requires reporting of the average of 3 samples - no variance measures reported, and 3 is not large number to start with). Given all the adjustment errors involved, no one should infer any sort of true accuracy to the implied level of precision some people seem to like to report for their own boxes.
Keep in mind, precision is based only on your ability to get reliable repeated measures of the same light in the same box, over and over again - it tells you nothing about how representative that one sample is, or how accurate the box is (i.e., it can only be used for relative internal comparisons).
My own lumen estimates were developed based on a comparison of my ceiling bounce readings for specific lights where the same model was measured in true integrating sphere. I make no claim to the accuracy of my numbers (as suggested by my relatively low precision in reporting - 2 sig figs with a half digit variance on the second digit). But here is a table comparing my estimates to manufacturer specs for a number of high-output lights:
The main anomaly is the SR90, but that is presumably explained by the very early testing sample I received (note the dates). I am quite confident that currently shipping SR90s meet or exceed the current ANSI FL-1 rated spec from the manufacturer, due to improved output bins on the SST-90.
The point here is that my values are directly based on the relative performance of my sample lights in my ceiling bounce room, linearly adjusted to estimated lumens by a conversion multiplication factor. It is quite possible that my values are not accurate - but the relative proportionate readings are surprisingly consistent with manufacturer specs.
Put it another way, if my TN31 were
really 800, or 900, or 1000 ANSI FL-1 lumens, etc., then you would have to adjust all my other lumen estimates down by the same percentage as they are based on a linear conversion of direct ceiling bounce readings (i.e., the lights are all proportionately relative to each other). Of course, it is possible that someone else has a 800 lumen TM31 - I have no idea of what normal variation is, given my single sample. But is also quite likely that their "lumen estimate" calibration differs from mine.
You have to compare any one individual's lumen values against only the other lights they have tested. You cannot compare across reviewers and samples unless you do a detail correlation analysis of all figures they report for common lights to see how much they personally differ (as I did
here for my basic lightbox lumens).