4g lte will still work though no ?
It’s not huge impact if compared to something big, but the point of all corporate sustainability efforts are to aim for the best / most suitable solution. They add up.
It’s good to hear that there is something available. My hope is, that there’d be effort, and purpose to support this more - not just wishes to community. I’m trying to dangle the business aspect as bait, as benefits are not just for us.
And while 5G (and maybe Wifi ac) would have been nice futureproofing, we’ll have plenty for several years. I can already imagine scratching my head in 2040 at the nurcinghome, where I’ve hidden my repurposed mini-webserver…
Given that 5G won’t function outside most cities due to the high cost of implementing it, I think that we can count on 4G LTE being used for the next two decades. A number of companies are currently investing in advanced forms of LTE, that frankly makes a lot more sense than 5G in my opinion.
Considering that 2G GSM is just now being retired and it was first introduced in 1991 in Finland, I don’t think that we have anything to worry about.
About 80% of the total greenhouse gas emissions for a phone lie in its initial fabrication and shipping to point of sale. If the Librem 5 lasts 5+ years, it can be a more environmental phone than the typical Android phone that only lasts 2-3 years, if you keep using it or resell it when you no longer want it. Given the Librem 5’s unique features and its rarity, I expect that it will have a better resale value than even iPhones and Galaxy S’s and the market for used smartphones is rapidly expanding, so a long lifespan is likely.
However, the Librem 5 probably won’t be a very environmental phone, because it will take more energy and resources to create it than a normal phone. It will have separate chips for its cellular baseband, GNSS, Wi-Fi/Bluetooth, conversion to HDMI/DisplayPort Alt Mode (maybe?), digital signal processing for the camera (maybe?), and fast charging, which a normal phone has in a single SoC whose die size measures about 70 mm2 in a mid-range phone and 90mm2 in a flagship. In contrast, the Librem 5 will probably need 250-300 mm2 of silicon to implement the same functionality in different chips, plus it needs a big heat spreader in the cellular modem, an M.2 slot and a smartcard reader. All of these factors will increase the environmental costs of producing the Librem 5. On the other hand, the smaller amount of DRAM and Flash in the Librem 5 will decrease that cost to some degree, but not that much.
In addition, the Librem 5 will require more frequent charging because its SoC isn’t that energy efficient and having lots of separate chips consumes more energy. However, the operational energy usually isn’t that big of a factor.
On the other hand, the Librem 5 won’t be constantly connecting to Google servers like the typical Android phone (and even the typical iPhone) and consuming as much energy in the network. The energy to operate an Android phone is about 1/5th of the energy to operate the network and servers that feed a phone its data, so not constantly sending information to Google and other online servers is very important environmentally.
The question is how to weigh all these factors. I sat down and tried to guesstimate them:
If my assumptions are correct, then the Librem 5 has a higher carbon footprint than other types of phones if it only lasts 3 years. If it lasts 5 years then its footprint is between a mid-range and a flagship Android phone. If it lasts 7 years, then it is the most environmental phone.
@amosbatto Very nice!
The spreadsheet idoes make great assumptions - which can be either way - and I hope at some point we/someone can go deeper into them. Actual measurements as basis of extrapolated values after we get the phone would be nice. Logistics and production chain info would be a nice addition after release (as probably not a concernt right now) as they make a big chunk of the whole. Etc.
I’m very interested in your methodology.
How did you get to CO2-e? Some website?
Is the “lifespan (years)” for how long first user uses the phone or how long the phone is in use (second hand)? Because the average lifetime seems to be around 5 years (https://www.cta.tech/News/Blog/Articles/2014/September/The-Life-Expectancy-of-Electronics.aspx https://www.sciencedaily.com/releases/2018/10/181016142434.htm) and that would be a good reference point to show with androids too.
The “1/5” network energy is a good point. Is you estimate just the background network & surveilance/service or do you also include user activity with cloud, streaming and searches etc. - I am assuming that L5 might use less cloud intensive as well as more local apps (again, something that needs to be measured, profiled)? (https://www.forbes.com/sites/forbestechcouncil/2017/12/15/why-energy-is-a-big-and-rapidly-growing-problem-for-data-centers/ https://www.nature.com/articles/d41586-018-06610-y)
Could the operating energy surprise us, or possibly get better after optimization over time?
The estimate is for an entire production batch, where larger quantities benefit from in most issues. However, the lifespan here doesn’t seem to take into account that L5 differs from those others with the likelyhood of how many of the batch are useable after a time (system is updated, battery can be replaced etc.). If a larger portion of the batch is useable, that should be counted as benefit for L5 in general (individual phones may vary).
I tried visualizing your spreadsheet and the first is straight up usage but the next is more interesting, as it takes into account lifetime, or rather spreads the CO2-e over time [(production/year)+(operating+emission x year)].
It doesn’t even look that bad for L5 in a few years! And this is even before any tweaks or enhancements to data or phone. There is almost enough here to utter “linux phone is green(er) design”
(Sorry about the “assumptions”, but I didn’t want to give the wrong impression to anyone if these spread)
If I may ask, and you have the inclination, could you maybe share your calculations on a spreadsheet, where new data could (hopefully) be updated/tested later?
I think lithium secondary cells and charging circuitry are so ubiquitous and standardised that it should be easily possible to source a battery of a suitable physical size, chemistry and pack voltage for at least the next decade. The only hitch might be if the connector is non-standard, then you would have to scavenge the connector off of the old battery.
Worst case scenario: some kind of graphene supercapacitor nanotubular fuel bells and whistles hydromabob comes out and lithium cells become obsolete and gradually stop being available. You have to hack together an adapter board to make one of these newfangled power sources work inside your 30 year old Librem 5 v1. (By now it’s had two new modems and seven battery replacements, and its computing power pales in comparison to the Librem Strap v12 quantum smartband, but you’re nostalgic for 2019, when we could walk outside without breathing apparatus and pigeons were mostly harmless.)
Really cool graphs that you made. I don’t have my spreadsheet file on this PC, but I will upload it when I have a chance.
If you read the scientific literature, you will find huge variations in the estimates of GHG emissions. I think that most studies have underestimated the GHG emissions, because they don’t do a detailed analysis of the materials and processes of IC and LCD production. The best studies were the ones by Eric Williams on DRAM production and a typical desktop PC in 2000 with a Pentium III CPU and 17 inch CRT monitor. Williams looked at some of the chemicals and the processes to purify them for use in a fab, and then looked at economic activity to estimate the rest of the chemicals where he didn’t have information, but his work is very outdated and the energy, water and chemical consumption have fallen, since fabs have gotten a lot more efficient over time and there has been a switch to renewable energy by some of the IC and LCD fabs.
Apple and Fairphone are the only electronics manufacturers who release GHG emission estimates on their devices. The Fairphone estimates are frankly baloney, and clearly don’t account for very much in the manufacturing process. Apple is not transparent at all, but the estimates from Apple used to be reliable in my opinion until the iPhone 6, when Apple realized that its rising emissions would be a PR problem. As smartphones have gotten bigger in screen size and battery size and their cameras and SoC’s have gotten more advanced, their emissions have gotten much higher, but Apple didn’t want to admit this.
Then, Apple embarked on a PR campaign to say that it was switching to 100% renewable energy and it had cut its emissions in half, which was a lot of greenwashing. Most of the emissions for Apple devices happen at its suppliers and assemblers in Asia (Samsung, TSMC, Foxconn, etc.) where most of the energy comes from burning coal and that doesn’t change in a year or two like Apple claimed. However, Apple has been leaning on its suppliers to use more renewable energy, so Apple does deserve some credit, but I still think that Apple is seriously underestimating the emissions for PR reasons.
At any rate, Apple estimates that its iPhone Xs emits 70 kg CO2-e for the 64GB model and 99 kg for the 512GB model (so 0.07 kg per extra GB of Flash memory) and 77 kg for the iPhone Xs Max with 64GB (so 0.38 kg per extra cm2 of screen). For the iPhone Xs 64GB, Apple estimates that 81% of GHG emissions (56.7kg) come from production of the phone, 3% (2.1 kg) from transport, 15% (10.5 kg) from usage for 3 years and 1% (0.7 kg) from recycling. I distrust all of Apple’s current estimates since Apple has turned environmentalism into a PR game, but I find how the emissions were rising up to the iPhone 6 very interesting:
Most of the LCA studies are like Ercan (2013)'s study of the Xperia T (4.55" screen, 139 g), which estimates 70% of GHG emissions (34.5 kg) for production, 10% (5.0 kg) for transportation, 18% (9.5 kg) for usage and 2% (1.0 kg) for end of life. Then, these studies are multiplied by 1.5 billion smartphones per year for annual global smartphone emissions.
I think that these studies are dramatically underestimating the production emissions, because as Williams shows they aren’t capturing all the activities that are needed to produce ICs, LCDs and batteries, so standard process-sum methodology in LCA studies doesn’t work very well for electronics. The amount of energy and resources to purify metals, chemicals, water and air for use in silicon and LCD fabs is enormous and a lot of the chemical purification happens outside the fab, so you can’t just look at the fab operations and you can’t just look at what it takes to produce standard copper to estimate energy and resources to produce the copper used in an IC, since it is much purer. You also have to keep in mind that the average phone screen has grown from 50-60 to 80-100 cm2 in size, and the battery has grown from 2000-2500 to 3000-4000 mAh over the last 5 years and the weight has increased from 120-150 to 170-210 grams. Fabricating components such as camera sensors and digital/image signal processors have a higher environmental impact than they did 5 years ago because the expectations for smartphone quality has risen over time.
For these reasons, I’m estimating 80 kg for a mid-range Android phone and 130 kg for a flagship Android phone, and then adjusting it to 180 kg for the Librem 5, since I guesstimate that it will have 2.5 times more silicon die area and 2 times more circuit board area than the typical phone, and those components form the majority of GHG emissions.
One thing that I know that I got wrong in my previous post is the emissions from operating a smartphone. Even if we assume that the Librem 5 will use 80% of its battery per day, here is what I calculate will be the GHG emissions to charge the Librem 5 for a year:
|GHG emissions to charge the Librem 5|
|Battery (Wh) (assume 3.8V)||13.3|
|80% battery consumed per day (Wh)||10.6|
|“ with 12% DC loss charging battery (Wh)||12.1|
|“ with 27% AC-DC adapter loss (Wh)||16.6|
|Energy when not charging (W)||0.02|
|22 hours per day when not charging (Wh)||0.44|
|Total per day (Wh)||17.0|
|“ with 5% transmission line loss (Wh)||17.9|
|Yearly AC power (kWh)||6.5|
|Yearly kg CO2-e (if 0.6 kg CO2-e / kWh)||3.9|
I’m assuming that the Librem 5 uses the same type of battery as the Galaxy On5 which is 3.8V. I was surprised how little energy it takes to charge smartphones, compared to running other things. The average Android phone is probably more energy efficient than the Librem 5 and only uses 60% of its battery per day.
I overestimated the charging energy because today’s fast charging, larger batteries and larger screens have upped the energy consumption. Fast charging is less energy-efficient that slow charging because more of the energy is lost in heat and fast charging generates more resistance in the battery. However, when I do the calculations, I find that the energy to charge smartphones still isn’t that much, even with today’s 15 and 18 watt chargers.
The network energy estimates vary a lot too. Belkhir and Elmeligi (2018) have much higher estimates than Malmodin and Lunden (2018). I tend to think that the real amount is between these two estimates. Malmodin and Lunden seem to consistently underestimate emissions, but Belkhir and Elmeligi don’t seem to be taking into account growing energy efficiency and the switch to renewables in their estimates of network energy.
A 5 year lifespan for mobile phones is very different from the average of the reputable studies whose median value is a little over 2 years:
The Consumer Electronics Association survey (since changed their name to the CTA) is poor methodology because people often expect devices to last longer than they do. People don’t expect their phones to break, but a sizable portion get cracked screens, get dropped in liquid, have the battery die, get stolen, etc. or people switch carriers or move and their old phone isn’t compatible. The Yale study also doesn’t take into account that the vast majority of phones never get resold as used phones. In 2017, there were 140 million used smartphones sold in 2017, compared to 1536.5 million new smartphones, so only 9% of smartphones got resold as used (although that percentage is rising). In other words, the Yale study only looks at the 9% of phones that get sold as used, so it is ignoring all the phones that were damaged, were stored or became functionally obsolete. It is mainly focusing on high-end models, such as the iPhone and Galaxy S series, that are more likely to retain their value over time and be resold as used phones, but that isn’t a representative sample of all smartphones.
The most accurate way to measure average lifespans of smartphones in my opinion is the data from Kantar WorldPanel, because it compares annual smartphone sales to the total number of cellular phone subscriptions in each country to determine how often people upgrade their phone:
A sizable percentage of phones get damaged. The charge capacity of lithium ion batteries start degrading rapidly after 500 100% depth of discard cycles, and over 90% of current cell phone models do not have replaceable batteries. The majority of old cell phones get stored because the owner buys an newer phone or changes the cellular provider, but keeps the old phone as a backup.
9% of old smartphones will be resold and 11% of mobile phones get recycled, so the majority of old phones are stored and not reused. I have a smartphone from 2011 and a feature phone from 2008 sitting in my desk which still work, but I haven’t used them in years, so I don’t know how I would answer a survey that asked me how long I expect my mobile phone to last. I might say 8 years since my 2011 smartphone still works, but the reality is that I have gotten a new phone every 3 years, which is why we can’t trust the CTA survey results, in my opinion.
One more point that I forgot to add. There are three critical factors that will affect the lifespan of the Librem 5:
- Availability of protective cases
- Availability of replacement screens
- Availability of replacement batteries
Cracked screens represent the majority of repairs for smartphones. Replacement batteries are critical for extending the lifetime.
A second point to consider is that “green” electronics which eliminates hazardous substances usually have a higher environmental footprint in fabrication. Energy-efficient electronics also often has a higher environmental footprint in fabrication, so the only way to make environmental electronics is to make electronics that lasts a long time, meaning that it has to be easy to repair and upgrade over time, which is the opposite of modern smartphone design.
For example, plastic parts without PVC, plastic wire insulation without PVC and phthalates, lead-free solder and circuit boards without brominated flame retardants all have higher carbon footprints. SSD fabrication has a higher carbon footprint than HDD fabrication. Using LCDs instead of CRT monitors is more energy efficient and eliminates a lot of toxins found in CRTs, but LCD fabrication has a very high carbon footprint. Smaller line widths in silicon fabs make for more energy-efficient ICs, but a cutting-edge 5nm fab will cost around $15 billion to build. It isn’t clear to me whether it is more environmental to keep using an old 28nm fab and a planar process which requires fewer processing steps or build a new fab with a cutting-edge FinFET 5-7 nm process that uses quad patterning and requires a lot more resources, but contains over 100 million transistors per square mm.
@amosbatto Your data is impressive, very thorough and well thought out of. Good work! I take it you are in the field?
I’m not concerned about the charge emissions calculations, as that goes for all, not just L5. The faster pace of changing phones is the real culprit and makes long lasting phones the heroes. I’m not sure how to account in a simple way the amount of differing reused/-sold phones, but anyway, the faster pace of new phones just moves the dotted android lines left [edit: and L5 even more to right, in theory].
The charging / use of power and use of network are something that could be measured, so data on that could at least be updated. Too bad we may never get good comparison data on production (to which I’d also like to add equivalents from organization: marketing, buildings and other organizational activities etc.). And such a complex whole may never be totally mapped, as you pointed towards, but I’ll take your approximations. L5 has some disadvantages to bigger players, but over time it still seems better (especially when including lifecycle and next phones).
Regarding your three points, I’d suggest adding to the hardware centric list also OS and software. They are a major component of keeping a personal information device operatinal and current. Secondly, I think over time it may be even possible to increase efficiency and extend battery life as well as minimize network usage (although, I also imagine networks will be used more heavily on apps). Also, an environmental conscious user should prefer using apps and data that reside in their device - which mostly aligns with security/privacy conscious approaches. One example is how screens go dark and save power. Nokia - once upon time - squeezed extra power saving from their lock screen clock display by changing the font: a hollow font lights up less white pixels (edit: could now only find clock) - even design choises add over time (although is minor compared what else can be done).
Although there might be difference between 28nm and 5-7nm, I’d also consider reliability of “cruder” (which sound funny to me as I’m talking about nm scale ) tech as an asset, considering reliability as well as probable longer lifespan.
For future… Should battery usege - and all usege for that matter - be differentiated to 3 to 5 profiles for better accuracy? One average does give enough of a picture, but if there is changing in use profiles over time, it would be interesting to keep an eye on. And people could see what difference changing habits makes. Although, at the same time, I recognize that might be a bit burdensome and there probably aren’t similar comparisons.
Thanks in advance for the spreadsheet. No hurry though, as shipments are a few months away, probably. I would like to see Purism make some sustainability report at some point (and perhaps recognize your contribution).
I helped write a report on reducing e-waste back in 2006-7, which got me interested in the topic. I’ve been reading the LCA studies for electronics and ICT since then, mainly because I work at a company that makes an open source web application, so I’m fascinated/horrified by the environmental impact of my job.
Good point. The fact that most Android phone manufacturers only guarantee 2 years of security updates is one of the main reasons why Android phones are junked so quickly. Google only mandates that Android licensees provide 4 security updates in the first year after launch and 1 update in the second year. Most low and mid-range Android devices only get one major OS upgrades and flagships generally only get two upgrades. Google, OnePlus and Android One phones are a little better, guaranteeing 3 years of security upgrades and 2 years of upgrades. The best is Apple which supports its phones for 5 years. Fairphone sold the Fairphone 2 for 39 months and aims to provide 6 years of software updates, but the upgrade to Android 7 cost Fairphone 500,000 Euros and I doubt it will ever upgrade again.
Due to Google’s requirements and compatibility tests for Android in the Open Handset Alliance and the licensing of Google Web Services, it is very difficult to keep providing Android upgrades as Fairphone has discovered. Linux and unauthorized mods like LineageOS are the only way to make a sustainable phone. If Purism and PINE64 can prove the viability of mobile Linux, it wouldn’t surprise me if Fairphone switches to Linux in the future.
I was very surprised at the sheer quantity of network traffic that Android and iOS causes. Douglas Schmidt at Vanderbilt studied Google Data Collection and found that an idling Android phone sends 40.2 requests per hour and sends 4.4 MB of data per day to Google servers. In comparison, an idling iPhone sends 0.73 requests/hour and 0.76 MB/day to Google servers and sends 4.2 requests/hour and 0.63 MB/day to Apple servers. Then, think of the difference in network traffic between downloading a map once from Open Street Maps and having to download map data every day from Google Maps when you want to consult a map. Also think of the amount of network traffic you reduce by not having to download endless propaganda and not sending your data to trackers.
I would love to see the difference in network traffic and phone operating energy between someone using the Librem 5 + Librem One services and someone using Android + Google services. Not having to constantly wake up and send data to Google servers is going to improve battery life.
Other parts of electronics fail far sooner than silicon gates, so I don’t expect that we will be able to notice any difference between a 7nm FinFET and a 28nm planar chip in terms of lifespan as long as moisture doesn’t get into the chip package. This post covers the ways that silicon can degrade over time, but I don’t think that this is a serious factor in the lifespan of a phone. Modern SSDs with wear leveling are designed to last for over 10 years even with heavy use.
The one component that really degrades is the battery, which is why we need a Battery Charge Limit app for the Librem 5, like exists for Android, because limiting the depth of discharge and the max charging capacity can prevent battery degradation.
In case you don’t know it, the L5 should be capable of limiting charging
Long story short in the Librem5 we will implement a smart charger chip that can be controlled by software
It’s great that Nicole Faerber has studied the problem and is thinking about it. None of the phone companies provides software by default to set a max charge limit. I hope that Purism will include this, but even if it doesn’t, it shouldn’t be hard to implement if there is hardware support for it.
I guess day one you can directly set it somewhere like
And then it should become a slider in the settings
Any idea of the difference on reacting to moisture on those or would it ultimately be about something else? Better survivability of accidental nautical incursion would be a positive side-effect
Wouldn’t thicker/wider mean better when mechanical forces (vibration, bending) are applied? Something that SSDs experience less than phones.
And the battery setting (with appropriate wisdom to use it right) needs to be added on someones todo-list (along with some other stuff from this thread).
vibration, bending, thermal extremes and moisture can cause the solder joints and thermal paste to crack and break circuits on the board, but usually the circuits and gates inside silicon chips are not damaged. When a board is bent, the solder joints to the chips usually break rather than the chips bending with the board. If you recall “Touch Disease” on the iPhone 6, which was caused by bending, the chip itself wasn’t damaged and could be resoldered to the board. If you pull apart old electronics from the 1960s and 1970s, the ICs usually still work.
I think this is important and on my Galaxy S3 I am using this:
@amosbatto I took another look at the graph and noticed, I’d made a mistake that made Flagship Android seem better than it was. I also updated the averages of getting a new phone, which were worse for Androids (and I used maybe a bit optimistic average for L5 users getting a new one, but the difference is, we should not have to, though). That also led me to rename the lines: the solids are in fact about the phones and then there is what users will (have to) do. It is partly about selection, what consumer choices one makes, but short lifespan phones definitely force towards a bad path.
(edit: and now that I look at it, I notice that all phones have a slight cumulative error as extra usage is added to user line, but that is on all and overall does not change - will come back to this when there is more actual data, someday)
If someone is interested, a simplified comparison: how many trees it would take [sources for calculations: https://projects.ncsu.edu/project/treesofstrength/treefact.htm http://www.carbonify.com/carbon-calculator.htm].
Tree can absorb about 22kg (48 pounds) of CO2 per year (we are talking good forest size “adult” trees here - multiply by about 20 if you think about planting a seedling). Production of an L5 would mean that roughly 9 trees need a year to offset that. For all phonetypes in graph, each individual users phones CO2-e during 5 years would need (if they don’t get a new one) about 2 trees to offset them during that time (production+use&network). But if a Mid-range Android user gets their third phone then, they should have about 4 trees - double - during that whole time plus have plans to plant more. Now multiply by number of users. Expecting to use more of the same tech, especially following the blue or orange dotted lines, users - or countries - need to plant bigger forests. And this only counts for a phone and CO2 equivalency. This is why a longer lasting phone (and tech in general) is good as reduction is better than just offsetting.
Simply capping full charge at 4.2 volts instead of 4.3-4.4 volts (which is overcharge territory for lithium ion) will greatly extend battery life. 4.1 volts as the full charge voltage will let the battery last a decade with good runtime. 4.1 volts is usually where electric vehicles top out their charge, but 4.2 is considered “fully charged” for lithium ion chemistries.
Are you talking about 18650 lithium ion (NMC, NCA or LCO) batteries?
Almost all the batteries found in phones are lithium polymer with LCO (lithium-cobalt-oxide) chemistry that have a max voltage of 3.8 or 3.7, but the same principal that you describe applies that limiting the charging capacity to 80% or 90% (i.e., limiting the max voltage) will greatly extend the number of cycles before degradation in capacity.
Given a name of “mrtsolar” maybe the reference is to the lithium chemistry found in PV system storage.