Destructive Incan Bulb Tests - Updated 8/27/2010 (Newer Info Added)

[jimjones3630]

Lux just read what I posted in all my above posts if you want to understand what I'm saying. The statement below reflects a part, small part of what I posted.

One thing we both understand is the lux measurement is the main factor determining the lumens of a bulb. So how a bulb with less lux rating 64430 "Top Chinese Bulb" can get a higher lumen output than the Tungsram 56580 is....

I don't think any of us are understanding what you are saying that reducing a bulb's life span increases a bulb's lumens, but oh well.
.

[pertinax]

Is he perhaps misapplying the fact that higer bulb temps mean more efficiency, and shorter life?

[jimjones3630]

As a bulb becomes more efficient by increading vbulb with subsequent higher lumen/watt ratio it's bulb life inversly declines.

Whatever causes a bulb to burn efficiently and subsequently brighter is measured with a lux meter. What I did not know is if the Hot Rater formula considered any factors other than lux measurements. It could have a bias factored in the formula. And still have a user bias by tinkering with the numbers.

The Hot Rater is in an Excel program and Excel does not generate formulas nor fix errors. I did have a circular error in my copy and spent the afternoon yesterday debugging it resulting in no change to lumen ratting of any bulb.
jim

[Timaxe]

I "ported" the Hot Rater to TK Solver during class, and it only relies on default lumens, and a ratio of stock/overdriven voltages. Specificially:

Overdrive Lumens = Lumens Spec * (Overdrive Voltage / Spec Voltage)^3.5

The rest of the spreadsheet works out voltage drop through the body of the light by ways of calculating the amperage drawn by the bulb at the overdrive voltage and using a given resistance value to find the voltage drop.

The formula for finding amperage appears to account for higher bulb resistances at higher temperatures (ie, overdrives) but I didn't bother to see if it correlates to the resistance of metals at elevated temperatures. (and it has no way to figure out what the temp of the bulb actually is).

The rest just provides interesting statistics that let you judge the validity of the numbers, such as comparing estimated CCT to the melting temperature of tungsten metal, calculating bulb life, efficiency, etc. Overall it is a very tight spreadsheet that has good insights into how the overdriving process generally works.

[Timaxe]

I'm interested in the 3.5 exponent used in the hot rater. It appears to be arbitrary, and warrants some investigation as to where it came from...so I had a very quick investigation into this...

Background:
Lumens and Lux are assumed to be related by a projected area (based on their definition - output & intensity). For a bulb, this area isn't expected to change during overdrive since the projected area depends on the filament shape.

Hypothesis:
The current equation for lumens overdriving from the hotrater spreadsheet (see previous post) should be applicable to lux.

Test Plan:
See how well Lux Luthor's data on lux matches the equation for lumens overdriving. We will look at the Osram 64625 & 62138 bulbs, since they are both 12v bulbs that are supposed to draw about 8.3 amps at default levels and are both bulbs used in the USL mod.

First, there will be a small modification to the equation to use lux values instead of lumens values. Our equation will be:
Od = Sd * (Ov / Sv ) ^ 3.5

Secondly, we will see what exponent actually fits the data collected, since this exponent is of interest to myself. With actual data we will solve the following equation for "x':
Od = Sd * (Ov / Sv ) ^ x

Results Part 1:
We will compare predicted lux with measured lux. The predicted lux comes from the formula: Odp = Sd * (Ov / Sv ) ^ 3.5

Part 1 w/ OSRAM 64625

Ov    Od   Odp    x
12    395 395    3.5
12.4  417 443.04 3.5
12.6  436 468.55 3.5
12.8  456 495.11 3.5
13    475 522.71 3.5
13.2  494 551.41 3.5
13.4  519 581.21 3.5
13.6  537 612.14 3.5
13.8  557 644.23 3.5
14    579 677.50 3.5

Sv = 12
Sd = 395

Note: Odp = Overdrive lux Predicted.  Od = Overdrive lux measured.

Part 1 w/ OSRAM 62138

Ov    Od   Odp    x
12    270 270    3.5
13    334 357.30 3.5
13.2  346 376.91 3.5
13.3  354 387.00 3.5
13.4  359 397.28 3.5
13.6  371 418.42 3.5
13.7  379 429.29 3.5
13.8  386 440.36 3.5
13.9  393 451.63 3.5
14    400 463.10 3.5
14.1  407 474.78 3.5
14.2  416 486.67 3.5
14.3  425 498.78 3.5
14.4  436 511.09 3.5
14.5  439 523.62 3.5
14.6  450 536.37 3.5
14.7  460 549.34 3.5
14.8  466 562.53 3.5
14.9  473 575.95 3.5

Sv = 12
Sd = 270

Note: Odp = Overdrive lux Predicted.  Od = Overdrive lux measured.

Analysis of Part 1:
There is a difference between what was predicted and what was measured. If the lumens formula was correct AND if the lumens to lux relationship is the projected area of light (which does not change at higher overdrives - but may change for bulbs) the numbers should have been the same or close. This does not appear to be the case.

Results Part 2:
We will try to determine the exponent "x" using lux luthor's collected data in the formula:
Od = Sd * (Ov / Sv ) ^ x

Part 2 w/ OSRAM 64625

Ov   Od    x  
12.4 417 1.653
12.6 436 2.024
12.8 456 2.225
13.0 475 2.304
13.2 494 2.347
13.4 519 2.474
13.6 537 2.454
13.8 557 2.459
14.0 579 2.481

Sv = 12
Sd = 395

Part 2 w/ OSRAM 62138

Ov   Od    x  
13.0 334 2.658
13.2 346 2.602
13.3 354 2.634
13.4 359 2.582
13.6 371 2.539
13.7 379 2.560
13.8 386 2.557
13.9 393 2.554
14.0 400 2.550
14.1 407 2.545
14.2 416 2.568
14.3 425 2.587
14.4 436 2.628
14.5 439 2.569
14.6 450 2.605
14.7 460 2.625
14.8 466 2.602
14.9 473 2.590

Sv = 12
Sd = 270

Analysis of Part 2:
For these particular bulbs, the exponent appears to be closer to 2.5, but these values seem to vary a somewhat within a particular bulb at various overdrive levels. Sometimes it decreases as overdrive goes up. Sometimes the exponent increases with increasing overdrives. But the trends are inconclusive with n=2 samples, and may not apply to bulbs from other manufacturers or of different filament design.

Overall Analysis:
The lumens overdrive formula seems to be over-optimistic when applied to lux overdrive values. If the assumption that the projected light area of the bulb is constant at different overdrives holds (which is used to relate lux & lumens in my experiment and validate using the same formula), then this may mean that x=3.5 is too optimistic for these particular bulbs. That said, the overdrive equations may need to be modified to include other important variables, such as the change in resistance of a bulb as temperature goes up (another assumption from electrical resistance of metals at high temperatures). It is unknown if all bulbs will share similar exponents, and it is inconclusive if there are any trends in the exponent at different overdrive levels.

Ideas for Future Research:
Obviously we should compile more data. It would be nice to get a better insight into the exponent used in the current hotrater overdrive formula so that we can get some insights into what needs to be done to write a new formula or modify the new formula to better correlate measured lux readings to predicted lumens readings.

Future research could go into looking at overdrive POWER consumption instead of voltage, since it is accepted that metals at higher temperatures generally increase their electrical resistance - so a formula that doesn't account for this will likely be off

I didn't look at power consumption in these bulbs at different voltages as I didn't want to derive a new formula using it - my focus was on determining the validity of the exponent 3.5 in the current hotrater formula. My results place some doubts on the validity of the hotrater equations (which have come into question recently, inspiring LuxLuthor's data collection), with the current item of major concern being the validity of the exponent x=3.5. Changing the exponent in the hotrater formulas will have MAJOR effects on the *advertised* output of today's hotwire lights, so more research is needed before anything conclusive can be derived from Lux's data.


Other random thoughts:
Does the definition of a lumen at a certain wavelength change output? Does the color temp of a hotwire torch change enough for this to make a difference in calculated outputs?

[Northern Lights]

The hotrater came from Andrew Wynn. Is it copyrighted? Although CPF had a fall out with him over undelivered goods; maybe someone can ask him how he developed the equations. Last I heard they were based on empirical data. There were several up dates form him and many of us, including me have been cutting and pasting and adding new bulb specs. I have no idea which are his original specs in the sheet and which have been added.

[jimjones3630]

Timaxe,

Thanks for your work and analysis of formulas in the hot rater. Appreciate any help.
jim

I'm interested in the 3.5 exponent used in the hot rater. It appears to be arbitrary, and warrants some investigation as to where it came from...so I had a very quick investigation into this...

[LuxLuthor]

I'm interested in the 3.5 exponent used in the hot rater. It appears to be arbitrary, and warrants some investigation as to where it came from...so I had a very quick investigation into this...

A great project, which I have been following. I have had conversations with AWR about this very subject. I tracked back the formula using the 3.5 exponent and other estimates in his Hotrater Excel spreadsheet. Here is the story on the history of that number.

Someone noticed that the WA website had a built in predictor on links with their various bulbs that listed changed webpage link values in the URL. I.E. this link regarding the 1185 bulb has Volts/Amp values of V2=10.8&L2=3.36 that can be manually changed, and link refreshed to give new values.

http://www.walamp.com/lpd/webstore/rerate.tpl?action=rerate&V2=10.8&L2=3.36&adj=5&partnumber=01185-U

This then evolved to people trying to guess the formulas that gave the different values. According to AWR, someone either affiliated with WA, or in some research website where their re-rating formulas were being discussed, came accross the formula and posted about it here. It may have been CPF user Newbie or Genseng, or someone of that kind of background.

AWR then took a spreadsheet chart that Litho123 had made to keep track of the various WA bulbs that he was selling...and which had some common voltage matchups he obtained from the URL manual formula. AWR added this formula to a revised spreadsheet which he needed for figuring out various voltage/current/resistance setups with his Hotdriver development.

When I started discussing with him all of the assumptions behind his Hotrater spreadsheet, and this equation for lumen prediction...it became clear that no one had done any data testing to confirm the accuracy of the Hotrater predictions. Basically, people used the WA URL predictor to give false promotion of WA Hotwire packages....saying with xx volts, the WA xxxx bulb gives a bazillion Lumen rocket that you should buy. LOL!

Then people started using this same WA 3.5 exponent formula that someone came across in researching WA bulbs, and using bulb retail website default listings of Osram, Philips, GE, Tungsram, Generic, etc. brands....just dropped those other bulb default Voltage, Amps (by dividing watts by volts), Lumens, Life into the same spreadsheet...and again assumed the equation must apply to all bulbs.

This was all done innocently, and to try and tout "my light is bigger than your light." When I began to get more people looking at a quick post I made trying to compare some Hotwires I had....and seeing that they were relying on it, and people selling various light combinations using lumens from the AWR spreadsheet...it became obvious that this was a case akin to "The Emperor Has No Clothes" fable, since noone did any type of data testing to correlate predictions.

Obviously, my testbed setup is not as precisely accurate as an expensive industrial research set of tests in an integrating sphere, but I tried to be as objective as I could, and controlling for variables like reflectors, reflected light, variations in technique...for a hobbyist setup.

When I started seeing the results, and talking with AWR, we realized that the variations that you noticed were not following a linear proportional relationship to voltage/amps that would be resolved by just finding the correct exponent. The actual results have typical "real life" complexity that likely involves many unique factors...and that may not even be the same complex formula from one bulb manufacturer to another, or from one same bulb model lot number to another.

A whole other level of complexity was introduced when I started doing "ideal bulb voltage" correlation of the predicted lifespan part of the Hotrater spreadsheet...which started again with default life values. I was amazed with some bulbs, that at a particular voltage they started out at an expected Lux, but then the output dropped precipitously after a relatively short time of being overdriven.

For example, one bulb that had a 12 hr predicted life at xx overdrive voltage/amps....began dropping Lux with run time...down to 30% of starting (peak) Lux after 4-5 hours...and burned out at about 9 hours. Equally perplexing is seeing that other bulb brands/models did not have the same dropoff using the voltage/amps that predicted the same 12 hr. life duration. Obviously, this has now become a whole other crucial aspect to using a particular bulb.

It is very hard to objectively appreciate with our eyes the drop over hours of intermittent use with a particular flashlight, in various nighttime ambient light environments over time. Suffice it to say that having two identical lights (identical batteries with same charge, same reflector, same internal resistance) but with a new and moderately used bulb shining side by side, a difference will be appreciated, that was not seen with used bulb alone.

We would tend to adjust to a decreasing Lux/Lumen output as bulb ages in our flashlight setup, thinking it is close to the same as when bulb started new....or chalk it up to waning battery remaining charge if not using a regulated light....or just not remembering objectively.

Of course, even controlling for voltage/amperage by using my test platform (vs. draining batteries and/or varrying ambient light environments), this drop in Lux/Lumen output over bulb lifetime varies from bulb to bulb, brand to brand, model to model, and with degree of overdrive. I have not yet published any of the dropoff Lux results with run time....but the projected life on the Hotrater is also capriciously unreliable.

[LuxLuthor]

Also, the Hotrater spreadsheet that a few of us have been using for the last 5-6 months has a much more complex formula for estimating lumens....the 3.5 exponent got dropped a while back. It is now using an "S" curve function.

[LuxLuthor]

Big thanks to Pertinax for sending me two 64610 bulbs and Ictorana for two Energizer HPR-71 (6V/10W) bulbs for me to test. I'll post testing results when I get to them in next few days.

I was kind of stalling when I got a note that Litho123 was sending me a bunch of new bulbs to test, which I just got yesterday...so now I will get these all caught up and tested soon.

[Firemedic262]

And the great work cont.

[lctorana]

I was kind of stalling when I got a note that Litho123 was sending me a bunch of new bulbs to test, which I just got yesterday...so now I will get these all caught up and tested soon.

Understood, Lux; it makes more sense to settle doen to do a decent test session in one hit.

[LuxLuthor]

OK, got first round of new bulbs tested last night...but I like to repeat another round with second bulb on a separate night to make sure that my test bed setup is being reproduced the same way, so I take it all down and setup fresh each time...which I believe helps eliminate any human testing method errors. I even check the battery level of the Light Meter each time.

I will be posting results of these bulbs I got from pertinax, lctorana, & Litho123:

Osram 64610 12V
Energizer HPR71 6V 10W (made in Germany)
WA 1331 - Frosted
WA 1111 - Frosted
Carley 809
Carley 1057

[LuxLuthor]

I have received a number of questions on the still listed Lumens on my spreadsheets, which are mostly not accurate. Here is a reply to a recent PM about the 64625 bulb for my advice on Lumens, and which will be added to spreadsheets on post #1 as a new column calculation:

Don't use the Lumens that appear on my spreadsheets, as they are only determined by the AWR Hotrater spreadsheet, which is not very accurate. I think a better way to estimate lumens in most cases, is using the manufacturer default voltage/lumens and show an increase in that starting lumen value--in direct proportion to my actual Lux measurements. This is an answer I gave to another member which I think gives the answer. I will eventually setup this calculation as another column when I post latest bulb tests shortly.

I think the most accurate way to figure out lumens is to look at the destructive test thread I did. In particular the chart for the 64625 bulb:

I know my Lux measurements were accurate, as I repeated them at least twice for all the bulbs on this thread. I also know that projections for Osram Lumens using AWR's Hotrater spreadsheet were overblown, so you cannot use those higher values.

So if we assume that Osram's published lumen estimate at default is accurate, then we can cross compare that to my actual Lux measurements when overdriven, and set up a simple algebra equation, and solve for "x lumens"

If at default 12V, Osram says this bulb has 3,600 (bulb) lumens, and I measured 395 Lux. If we look at higher voltage of 13V it gives Lux measurement of 475, which we use to setup equation:

3600 Lumen / 395 Lux = X Lumen / 475 Lux

To solve for X, you 'cross multiply' to get:

3600 x 475 = X x 395

Solve for X:

1710000 / 395

= 4,329 (Bulb) Lumens

Now, that assumes the voltage at that one value, even though without regulation, the battery voltage starts high, and drops over run time.

[

[Raoul_Duke]

So going off that the Ozram 458 crunches out to

1800 x 157 / X x 723

8289 lumens (bulb lumens (65%) ~ 5388)

And the osram 623 =

2800 x 303 = X x 713

6589 lumens ( bulb lumens (65%) ~ 4282 )

[Raoul_Duke]

ooops double post;.

Oh well may aswell use this chance to say another thanks for all this quality testing

[LuxLuthor]

Except I would not use the values just before it flashed. The more you overdrive, the more inaccurate it all becomes.

I think I will try doing the 64655 & 64657 bulbs...not sure how my 30V 20A PS will hold up....but I have everything setup. What's the worst that can happen? Let's see...I do have my homeowner's insurance paid for the year....ok good to go!

[Raoul_Duke]

Excellent, I have wanting to see your results for the 64657 for a long time.

Looks promising on paper, and has a longer life than the 655, should be able to handle the overdrive better.

Any info on your real world eye comparison of these lamps.

I have all the bits and have been planning to stick 6 or 7 emoli up it for too long now, but no time to do it in.

[Fulgeo]

Wanted to open with saying thanks LuxLuthor for spending the time and effort to make these charts available. I find myself mulling over them alot. Fuel for creative dreams.

I was curious has anybody underdriven the Osram 62138 or 64625 at 11.1 volts? Is the beam too yellow? Anyone have a lumen estimate or reading? How does this effect bulb life? Is it detrimental? I have three Emoli and a 3D Mag waiting to go.

Anyone have a good source to purchase the 64447, bulb connection does not seem to carry them?

Thanks in advance. Perhaps I should change my handle to "Question Man"

[petrev]

Halostar 64447 IRC 65W 12V GY6,35

Svetila.com

[LuxLuthor]

I finally added the bulbs I got from pertinax & lctorana:

Osram 64610 12V
Philips HPR71 6V 10W (made in Germany)

(Note: the frosted WA 1331 & 1111 that I got from Litho123 have the same Lux & Life as non-frosted)

I have the test results done for these bulbs, just need to input in spreadsheet. Note: t still have the two Carley to post from Litho123 , and my own 250W bulbs which pushed my 30V 20A power supply !!!

Carley 809
Carley 1057
Osram 64657 (OMG this bulb got hot)
Osram 64655 (Even more OMG as this bulb got REALLY hot)

Also note that I added a new column (in yellow) with the Philips HPR71 & Osram 64610 that multiplies the ACTUAL measured percent increase Lux readings times the default manufacturer lumen rating, rather than using the inaccurate Hotrater spreadsheet. I will be adding this column to the other bulbs when I update the above 4.

I believe this is the most accurate, "real world" measured overdrive estimation using the default baseline Lumen reading we are given. The only baseline Lumen rating that is whacked is the Top Bulb Chinese Generic 64430 listed as 650 lumens at 6V, which should actually be a default rating of about 230 lumens at 6V.

[Lips]

Lux

I have two WA1160 bulbs that everyone uses in the MagChargers if you don't have those I can send.


Lips

[Jay T]

Anyone have a good source to purchase the 64447, bulb connection does not seem to carry them?

Thanks in advance. Perhaps I should change my handle to "Question Man"

http://www.servicelighting.com/catal...m?prod=QQ64447

[modamag]

Lux, this is AWSOME stuff. Although I'm an LED guy, but I surely can appreciate the amount of work and thought that goes in your testing process.

You're making it much harder for me to keep up with the incands crowds.

Great Job! :

[LuxLuthor]

Lux

I have two WA1160 bulbs that everyone uses in the MagChargers if you don't have those I can send.

Lips

Lips, that would be cool. I don't have those bulbs. PM sent with my address.

Modamag, thank you sir! Your work is legendary.

[lctorana]

Lux,

Given a linear relationship between lux and lumens...


May I draw any conclusions from the variance of the Lux : Lumens ratio from bulb to bulb at default voltage?

If I can, then that would seem to indicate that some bulbs are wildly optimistic in their claimed lumens, whilst others are actually rather conservative...

Or is there another factor at work?

[Timaxe]

Given the same & constant surface area for the bulb filament, the lux and lumens should have the same linear relationship. Otherwise it is possible for this line to have a different slope depending on surface area, the shape of the filament (blocking/diffracting the light), etc. While in some models it is possible to assume that all bulbs have the same filament sizes (due to how the light meter may work - see how most people working with LED flashlights assume it is a point source, which can be good in some general cases) I'm not sure how LuxLuthor's setup and the bulbs under test fare in this regard.

Most filaments are produced via powdered metallurgy, which results in a relatively high surface area to volume ratio. LuxLuthor has commented that over time the output of a bulb goes down. One possible explanation for this behavior is that over time at high temperatures the tungsten will go through some annealing processes (or perhaps also some of the halogen cycle) and reform into a lower energy shape - reducing surface area and thus lumens output.

It would be interesting to have detailed photos/models of these bulb filaments, but that's probably too much to ask out of LuxLuther who has done so much for us already. If we want to produce a 'complete' simulation/estimate for bulb outputs it will be a useful piece of data, but our estimates are often good enough and not too critical of +/- 100 lumens (or 10%).

[LuxLuthor]

I just finished adding the yellow column to every bulb listed in first post--had to add it in each spreadsheet manually, and fix up formatting and colors. Man, that was a lot of work. Still didn't add the last few tested bulbs yet.

I double checked on every bulb, so if you are not seeing the new yellow column with red type either on page one, or when clicking on larger images, hit your browser refresh.

I changed the default TopBulb 64430 Chinese clone default to 230L, instead of their 6v=650 lumens listed on their website here, because it is obviously a bogus value. I also extended its life up to 4,000 hrs from their spec of 2,000 for the same reason. These changes are based on reasonable comparison of my Lux readings with other brands, and you will see them in red text.

I additionally extrapolated the default 6v lumens for AW's Chinese generic that he sells ("6V 30W 2000Hr Chinese Generic"), again based upon Lux comparisons. This allows us to use the new yellow Lumen columns on all bulbs, based on measured Lux.

I believe this new yellow column is the most accurate assessment of Lumens, but it does assume the default lumen spec value is accurate...since the percent increase in measured Lux is multiplied times that value. Only other option is to start using standard of Lux measurements from my tables.

Regarding the decreasing Lux measurement as an overdriven bulb ages, it is variable with the bulb model, and degree of overdrive beyond default. I almost think it is better to use the Ostrich "Head in the Sand" strategy on the fading output with degree of overdrive and life issue. LOL!

[DM51]

Lux, I've just had a look at a couple of those tables. I didn't think there was much room for improvement over what was there before, but you have managed it.

These tables are now absolutely superb - clear, easy to access information which we have never had before - we were floundering around 'in the dark', mostly just guessing at what the figures were. This is definitive, unarguable data - invaluable.

Every CPFer should be grateful to you for all the amazing work you have put into this project. This is a fantastic resource to have.

I take my hat off to you! Bravo!

[Action]

Any chance that these tables can be available in an excel format? I really would love to combine these into a table that is based either on lux or voltage or some lux per watt ratio vs. based on individual bulbs...

This is just awesome work Lux, a huge resource for the community!