Thinks that a single command and control provides for all the zombie lasers in regard to performance metrics
The old way of things was hey we think a tube will last X time depending on y usage rate. Well this is pretty much useless as what constitutes heavy or light usage. It is purely subjective. If I run my laser at 15% for 1000 hours and someone else runs it at 80% for 1000 hours take a guess whose tube has more miles on it.
So we need a better metrics some that aligns closer to consumption of a depleting resource
⢠Time of usage â nope the laser tube doesnât have an hours count down if it did then âdepending on your usageâ would not be the disclaimer when providing life cycle
⢠Photons emitted (micro moles) â would be darn cool but I donât know if this is realistic, quantum sensors are in the 1,000$ range. When I was setting up my hydroponics I rented one for about 150$ for a week. Canât imagine that setting up a licor would be cost effective
⢠Watts sent to the tube â as with everything complex there could be a way to just keep a running tally of what watts have been sent to the tube (not the system as we donât care to measure fan or stepper usage). This is raw consumption X watts have been sent to the tube with a capacity of Y (infinity because we donât know yet) and do the math and you can see yea Iâm getting close the end of this tube can shoot out.
⢠Power level at X duration â if measuring watts is not possible then measure duration spent firing at x power level certainty is possible. I ran for 1 minute at 100% power = the same as 2 minutes at 50% power (this assumes that tube burn down is linear (we know that power vs ablation is not linear but we are not measuring cut performance just consumption)). So once you have that you can normalize the data to get power used thus far similar to doing a wattâs measurement
Now for the power of the command and control to model how things are going to last or how to adjust for drift
⢠What is being gathered
o The material being cut (be it proof grade or user entered) be it specific or just general classifications this just depends on how you want your group by clause and what you want to provide as user enterable fields
o The thickness
o The power level used
o The speed used
o Maybe a check box that says is this a test/sample cut for calibration
⢠You can then start to create histograms of what settings are used for what X material. This can give you stuff like for X we find that 70% of people fall inside of here and 10 on this side for more power or more speed or less of each etcâŚ
⢠As the tube burns down uses some life etc⌠the users will re calibrate the settings to compensate something that you used to be able to cut at 20% now needs 30% power or needs 10% less speed etcâŚ
o Now that we know the consumed life on the tube (above) we can then take the histograms of just each material and where people typically fall for settings and then create bins for well these people have this y consumption and these people have y*.5 consumption and then look at what the normal range of operation is for that material between the different bins(as many as is reasonable)
o Splitting out by consumption and then by whatâs normally operations for settings for each material you can then start to project where things may start drifting to. This also has the potently for auto calibration of settings based on consumed life of the tube.
⢠The power of HUGE amounts of user data coming in that patterns are formed and association this wasnât possible on any of the previously deployed lasers. It was more instinctual knowledge or cheat sheets
or Iâm a nut and donât know what Iâm talking about
my daily database for what I do at work is about 9TB thatâs my configuration database
my performance metrics analytics database is 144 cpus, 432GB ram, 4.5 TB DB, currently doing analytics on 8.87 million metrics
This brings up a good point regarding proofgrade materials; the settings needed will be different for a brand new tube than for one nearing the end of its useful life. This leads me to the conclusion that they will actually need to be building this kind of per machine watts used data collection and bulk analysis in, and using that info to keep the proofgrade materials working correctly in the future on both brand-new and older forges. And since the tubes will be replaceable, they will need to be able to uniquely identify the tubes themselves, rather than just the machine, so they donât treat the new tube like itâs still the old and weak one that got replaced - they would really need to maintain the watts total per tube, not per machine Mac address.
Iâm a big fan of data and it might be useful for GF to collect such data. However, just reading up on tube life expectations and anecdotal information, experiences vary widely. Some tubes fall off slowly, some fall off sharply. More can go wrong with a tube than just the gases depleting inside - and anecdotally, tubes largely die not from gas depletion but other internal factors.
I feel like providing that info to the end user might set up some false expectations when everything looks good but a customers tube dies.
I think regarding proof grade, maybe setting up a software controlled fudge factor would be cool. Run a periodic calibration cutting and engraving test on your specific material and input those numbers for the material and it will adjust the cut as needed.
true but there has never been a way to aggregate all the usage together to find out if its really that wide or not. but yes there are other factors but at a least you can idendify the % of outlayers falling out side the norm etcâŚ
If they add such things I want 100miler badge for when your laser head as moved over 100 miles, and a monthish badge for when the laser has run 30 days total time.
Excellent post with lots of things to consider. The metrics possible with such a data set will be of enormous benefit to Glowforge at least in learning what to expect as time goes by. Sometimes, when I see these posts, it concerns me that the Glowforge printing process wonât be quite as easy as sticking paper into my HP laser printer. Proofgrade materials will be so helpful, but if I am true to my propensity to scrounge materials for projects, Iâm going to have to do a lot of testing. The community should be very helpful for sharing settings for whatever use, at least to get some starting point.
indeed community would be huge, but rather than 50000 posts of I cut x with abc settings would you rather a everyoneâs settings all auto report and pivoted and then you can select a class of material and see a bell curve of where most ppl fall out out layers would certainly help get you ball park to where to start without having to reference sheets. and who knows maybe we find out from 15,000 forges avging 3 hours of cutting a week that we find out that settings arenât so out of whack for everything after all from device to device from material x to material x variant
great thought ! that slipped my mind how to flag after maintenance tubes are not a data item so maybe a âreset tube life/ new tubeâ button so that the cloud knows to not 'reset the counters but create a new object for that forge with a new set of tracking. that way you save the bleed down from the first tube as it aged and how settings needed to be changed and then still get the metrics for the new tubes
From what Iâve read (little actual experience) tube wear is likely a baseline number plus some function of power that increases sharply at or near rated watts. But yeah, since Glowforge the company is going to be sending the power/time data down to our machines to instruct them to make cuts, it would be a piece of cake for them to hold onto some of that data and use it to build profiles.
I donât know if theyâre planning this, but if they keep a bunch of accelerated-life machines at HQ to do mixed real cuts and calibration cuts on proofgrade materials, the proof-grade adjustments could be semi-automagic: the cloud software knows what the material is, it knows how many watt-hours (or whatever) are on your machine, so it can tell the machine the optimal cutting speed and power settings without bothering you.
Ooh, I also wonder if there are any materials that have sharply visible cutting/engraving thresholds so that every now and then you could run a power/speed test pattern and then take a picture of the result that could be analyzed to tell you your current âtrueâ power settings.
would certainly help get you ball park to where to start without having to reference sheets. and who knows maybe we find out from 15,000 forges avging 3 hours of cutting a week that we find out that settings arenât so out of whack for everything after all from device to device from material x to material x variant