Hacker News new | comments | show | ask | jobs | submit login
Inside a $34 smartphone (projectgus.com)
169 points by zdw 738 days ago | hide | past | web | 94 comments | favorite

I have actually done a lot of development on a very similar device, based on the Spreadtrum 6821 (the famed $25 firefox phone). Its a pretty neat device - 128MB RAM, 256 MB Flash, support for upto32GB microsd card, dual SIM (EDGE only), Bluetooth, FM radio, replaceable battery, 320x480 lcd touch screen, a seperate touch "strip" which you can configure as capacitive buttons, three hardware buttons (V+,V-, power).

I did a full kernel build, wrote the device driver for the LCD screen, am currently writing the driver for the SIM. I have a debian rootfs running on it - its a nifty little linux gadget. I am currently using a bunch of these as part of a linux kernel hacking tutorial for college kids.

I believe there is a lot of potential for these as alternatives to RPi and its brethren for certain applications. If you try to add all the above peripherals to RPi we quickly cross $500 (of course, you would probably not do so in reality).

If any one is interested in more info, message me.

Shameless Plug: I am also looking for consulting ops on such and similar (I have a LOT of embedded and electronic design experience - from vague user spec to full design).

"Eval" boards and systems like this have been available in the era before the raspberry-pi revolution, but rPi got a lot of hype for itself by actually shipping in large numbers.

These little systems are quite capable indeed! I remember compiling the Linux kernel on a 1ghz machine with 32Megs of RAM, back in the day .. and here we are with these little machines in our pockets. Fantastic stuff to be teaching kids to use ..

Pi also aimed at a different market. Most eval boards was either direct order or via B2B distributors. Pi aimed at kids and education, and went into the general consumer market.

There were eval and computing-platforms aimed at kids before rPi. None had the momentum of the nascent crowd-funding movement quite so potently as the rPi.

We're hiring contract and full time system software engineers at Skully: http://www.skully.com/careers/

Lots of things seem up your alley, like the need for a per pixel alpha driver for the Kopin display, etc..

do you mind sharing the curriculum and course material ? so we can benefit from it as-well......

hi ghoul2, i'd love to have more info on your linux kernel hacking tutorial.

Removable batteries are still the norm in the low-end Chinese Android market. There's a lot of commonality and standard sizes between them too, because these are all based on the same reference designs and the manufacturers don't see much point in changing things like the dimensions of the battery since that requires effort and would make parts sourcing harder - they're not after exclusivity or vendor-lock-in.

I remember the time when Mediatek was considered ultra-low-end and relatively unknown, and sellers would be selling MTK phones faked to show Qualcomm and other more "prestigious" brands of SoC. Faking MTK would be unheard of, because there wasn't anything below them (there were still fakes between the different SoCs models from them, and that still continues today.) Now that MTK has grown past that point and others have appeared below them, they've become a lot more desirable. I'd consider the fact that someone wants to fake some other brand as MTK evidence of that.

The interesting thing about the MTK platform is that it's "unofficially open" - various people have leaked detailed technical information on the SoCs, which doesn't seem to bother MTK; maybe it's somewhat deliberate. Many devices are also rooted by default, and it's really easy to root one (there's no bootloader locking, and being able to read/write the entire internal storage in raw format via the official SpFlashTool means they're practically unbrickable.) This means they're quite hackable, and the low price also helps. MTK themselves won't provide any support, but a whole community has formed in its place.

Perhaps the same might happen to Spreadtrum in the future, but the last time I looked there was not nearly as much information available for SC devices and rooting was not easy.

The "Time Switch Machine" is scheduled power on/off. http://en.wikipedia.org/wiki/Time_switch

I wrote an editorial on Mediatek[0] that you might be interested in. I disagree with Mediatek being "open" though, they're so closed source that they basically violate GPL.

The rooting is definitely to do with company culture though, the Jiayu S3 I recently played with[1] (and has become my daily driver) actually has a root toggle in the settings menu. Great stuff.

[0] http://www.neowin.net/news/the-qualcomm-reign-on-smartphones...

[1] http://www.neowin.net/news/review-of-the-jiayu-s3-hands-down...

A few years back i came across a video of someone that had picked up a 7" tablet in a Chinese store, and it was powered by a two Nokia form factor batteries in a hot swap configuration.

I also recently noticed a kickstarter that is trying to pitch a device that can act as a emergency charger or powerbank using phone batteries.

Mediatek is a company whose engineers attending Linaro conference used iphones - that makes you think.

I think there is a market for honestly low-spec smartphones, as there are some interesting uses for them, for example, a IP camera (plenty of apps for that), a dedicated XMBC/Kodi remote, or a home automation controller. You can think of them as Raspberry Pis but less hacking friendly (usually the source code of these things is not easily available/buildable), that cost more but also get you two cameras, speakers and microphone, a screen and a (sometimes dual-SIM) GSM modem, which if you're lucky can do 3G.

By the way, the fact that the author of the blog could not get past 2.5G could be because the phone only supports 3G in one of the SIM slots and the only SIM was inserted in the wrong slot. This is the kind of thing that can be confirmed by looking up the datasheet of that SC6825C SoC, to see if it supports 3G at all. The main problem is, before buying you can never be really sure what SoC it has...

The "Timer Switch Machine" is probably a function to turn the phone on and off automatically at given times. Most Mediatek chips support this (as well as ringing alarms when off), so it'd make sense if their competitor's chips also did this.

The little bubble thing is a weird chinese quirk - http://www.businessinsider.com/chinese-use-assistivetouch-on... if you're interested on that

Assistive touch is also useful if one of your real buttons gets broken. I borrowed a friend's iPhone to dev on it, but the power/sleep button was broken. AT solves this by providing an on screen version.

I was using it to bypass a broken lock button. Only catch is that you can't turn it on easily - plugging it in is the only way to restart the phone.

My brother found this because his home button was broken and now my parents have picked up on it even if their home button is perfectly functional.

Reading about what assistive touch does on the iPhone, i find myself thinking about faucet design.

These days you find this dual direction (lift to open, left and right to adjust temperature mix) type everywhere. But as best i can tell it was originally designed with handicapped users in mind.

A faucet with an interface in terms of the two attributes you generally care about (pressure and temperature) is in theory much better for everyone than one in terms of two implementation-detail attributes (amount of hot/cold water flowing into the mixture).

Unfortunately many of the one-control faucets have terrible mechanical engineering under the interface, and as a result you often can’t adequately control either pressure or temperature, and the two aren’t actually separated orthogonally (i.e. as you move the control along one axis both attributes change, and to change just one attribute you must adjust the control along both axes)

A lot of tech started with disabled people in mind: telephone, remote control, even typewriter iirc.

This is called [universal design], as opposed to designing only for users with abilities in a limited range (usually the "average" user)

By the looks of it, the physical button is just more annoying to press because it's at the very bottom of the device. You can't easily reach it with your thumb.

It's always on top and you can drag it anywhere on the screen any time you want. Sometimes it gets in the way of keyboards and etc and it has to be pushed over to keep working.

    > chinese 
Everyone in Thailand seems to have it switched on too

Yep, that's also my observation.

more proof users aren't dummies. they're saying they want to move more quickly through the OS. mechanical keys and accompanying sw animations are too slow.

This device is only running EDGE and not being a 3G smartphone is a big distinction. In practice, the low memory also makes it a very difficult sell to consumers, so I’m not entirely surprised by the marketing deception here.

Shameless plug, my startup wholesales a 3G Android smartphone with 4GB memory running on a Spreadtrum chip that costs $30: https://www.voxsupplychain.com/shop/users/#/shop/product/and...

We have found that there are some natural dividing lines on low-end Android specs. If you cut more corners, it really impedes the user experience.

4 GB of storage. It only has 512 MB of memory.

That's pretty standard for phones of this size, unfortunately.

I have no problem with that. My issue is about misleadingly talking about memory when you mean storage.

I didn't find it misleading because in my mind memory means storage.

For example: I use the terms flash memory and flash storage interchangeably. The other way round, RAM can be thought of as temporary storage that is lost when the device is powered off.

Are you sure that 3G license fees are paid on this phone? From my experience sourcing cheap phones this price is not possible unless you avoid the licensing.

Lowest we could go for a SC7715 was about 35$.

Yes, it is all rolled up into the price of the device. This pricing is just for delivery to a port (CIP incoterm), so doesn't include customs clearance, duties, VAT or a retail markup.

Being China, it could be a TD-SCDMA variant (the stylized '4G' logo hints at being a China Mobile variant). That would escape the WCDMA related patents, but its still technically a UMTS/GSM system.

How are you able to ship this to Germany for $1?

We can ship 1,000 smartphones to Frankfurt's airport (FRA) for $1,104 USD, which is about a dollar per phone. Our minimum order is 1,000 smartphones, so that's not the same as being able to ship 1 device for $1.

Also, we work directly with carriers and get wholesale rates for air and sea freight to major ports. Delivery to Hamburg by sea is about $0.05 USD per device for 1,000 smartphones.

I missed the Quantity field. My bad.

Have you thought of making it a bit more hacker friendly ? Having some available GPIO's somehow exposed and documented. A small (optional ?) component to point the camera forward. Something to mount these phones onto a frame in a solid way. An optional way to provide power to the phone without occupying the USB slot.

These phones would make excellent autopilots and controllers for RC cars, quadcopters, even regular helicopters, but there's just a few things missing.

We've gotten a few requests for modifications that we are looking at. Would love to hear more about what you are thinking here.

My email is in my profile if you'd like to be in touch.

tl;dr: It runs Android 4.0.3/ICS, has 200MB ram, just two cores, possibly only 2G/2.5G, is clunky but works -- despite advertising 4.4/KitKat, 8 cores and 4G.

A hardware engineer told me in discussions on chip subversion that BS like this is common in SOC's, too. He said they're so expensive to develop that vendors often include the functionality of all their chips in one design. Then, they flip a switch on the chip to enable or disable specific functionality. This means they all cost the same to make but they can give you specific features at specific prices. Hard disk makers do it too where you bought a "120GB" drive that uses same eg 1TB platter as the rest but firmware restricts user to 120GB.

You can't be sure whats in it unless a company such as Chipworks tears it down piece by piece and wire by wire. Sneaky is the norm in hardware.

>Hard disk makers do it too where you bought a "120GB" drive that uses same eg 1TB platter as the rest but firmware restricts user to 120GB.

Why isn't some smart hacker buying these in bulk, reverse engineering the firmware to enable the full 1TB, and reselling at a 1tb price point? Would that be a good business?

Reformatting a hard drive to a higher capacity would be akin to overclocking a CPU. The heads and platter might operate fine at the higher capacity, or they might not. Running the risk that your HDD might be quietly losing your data is not something most people are interested in.

It's a moot point anyways because reformatting an HDD to a different capacity would be an impossibly complex endeavor. Performing a low-level format on a modern HDD requires a multi-step (and multi day!) process of burn-in and calibration. None of it is documented publicly of course. Doing this outside of the factory is far beyond the realm of possibility.

>The heads and platter might operate fine at the higher capacity, or they might not.

If it's literally the exact same hardware, but different firmware, then they most definitely should perform exactly the same.

Heads and media are like silicon chips in that no two units are "exactly" the same. Yield in terms of usable capacity varies from one unit to the next just like clock rate for CPUs. When you buy an HDD it has been validated to operate correctly at the rated capacity, there's no guarantee how it may handle higher densities. Besides, unless you work at an HDD factory you have no way of knowing just how similar the parts of a particular 250GB drive are to a 1TB drive. For all you know they may have gotten a particularly dodgy lot of heads or patters which definitely can't handle 1TB.

The firmware on the device is a black box that controls a more complex black box in an undocumented way and where errors can break it. You don't JUST swap the firmware out. Even so, he may have meant it in the context of the onboard software and modifiable hardware (eg switch they flip at factory to permanently set behavior). Sometimes a hardware engineer will call both firmware speaking shorthand. If it was the latter, you're unlikely to change it because they blow a physical fuse (antifuse?) inside the chip.

Some IP vendors are also offering encryption, tamper-sensing circuitry, and obfuscation to help chip-makers protect the systems better. They really want that extra few hundred dollars. ;)

Someone should get that guy who ran Linux on his hd on it.

Good thinking, you two! The link shows that the black magic (spindle control, etc) is in a dedicated chip controlled by another chip. The "black box" controlling the "black box." I thought they'd be on the same chip and software package. That would increase risk. I didn't anticipate them putting it on a dedicated chip. It makes sense along some lines.

So, that counts out at least one HD vendor from my claim. More could follow if they do something similar. Thanks for the article, too, as I enjoyed reading it.

Heads would be the same, but there's a chance that they're using the platters which have defects on them, making them unusable for higher capacities.

They could be serious defects too, i.e. surface abberations that would damage the heads if they ventured into that area.

It would be a business that requires people to trust their data to voided-warranty drives. I don't think that'd be popular.

Backblaze might be interested. If they're happy to shuck drives from external enclosures they're possibly happy to hack firmware.

But I'm guessing the drive manufacturers test the platters and use faulty platters for these firmware restricted drives, like CPU makers test the cores.

I'd buy a new 1tb HD that comes from a known manufacturer if it was much cheaper than any others I could get.

It wouldn't be much cheaper, though. A WD Blue 250GB on Newegg is $47. A WD Blue 1TB is $55.

Exactly how is a "warranty" going to protect the data on your drives? Drives fail. If you aren't mitigating that fact, e.g. by using RAID and doing backups, then a warranty isn't going to bail you out.

I trust that WD's warranties are generally formulated to last for the statistical lifetime of the drive, and operate with that in mind. (And my home NAS is in RAID6, so, sure, I agree with you.) I don't trust some dude's homebrew hard drive controller flash, and I don't trust that I have no recourse if it goes south. I definitely don't value that over eight dollars.

Because its not true :)

There might be some 750GB disks using 1TB platters, but the difference is never x10.

That's for a different reason. Let's say you have 4 models of a chip, some with more internal RAM & Flash some with less and some with a couple of peripherals and features, and some with less. It is far less expensive to design one model with everything, and then disable some features either by not bonding the chip pads it to package pins or by laser cutting fuses.

One of the most expensive aspects of chip manufacturing is the setup cost, i.e. the mask and "tooling". So it is much cheaper for them to it once and have one chip that rolls of the line.

Also, things like RAM and EEPROM have fabrication errors just like other parts of the chip. It isn't uncommon to have "self healing" RAM. Basically the RAM has a few more cells or address rows more than the actual capacity. As part of wafer testing certain paths are cut so you still end up with the amount of RAM you need. Although it takes up a bit more area (because you're actually manufacturing more RAM) you still end up with higher overall yield.

So as chips come off the line, some get binned into lower grades, or RAM size and the rest of the features are disabled via laser cutting fuses or vias.

Yeah, that's what he said. So, that makes two of you saying this is industry standard for cost-cutting and increasing yield. I'd probably do the same if I made chips. We do it in security engineering for high assurance stuff for similar reasons: the more flexible, certified design can be reused in many applications with minimal extra work & (hopefully) max ROI.

Right. That makes sense too!

That's been going on for years and years and years, companies like IBM were doing it in the 60s.

I'm often keen to buy these super cheap phones, but then I see the outright deception involved and it's off putting.

I hope Mozilla can encourage interesting but honest budget engineering.

Yeah, the listed specs are more in line with $200 phones. Then too you need to be careful.

E.g the first one I had in that range was a (physically) near perfect Samsung SIII replica, complete with the dandelion screen saver, and the place I bought it for hinted strongly that you could get it with the Samsung logo if I asked. I knew what I was buying (a much slower MTK based phone), but I saw plenty of places try to pass it off as a genuine SIII too, which at the time was substantially more expensive.

This is often the best way of finding (reasonably) honest sellers: Look for other sellers offering the same phone and assume the worst listed specs are most accurate.

The interesting thing is how they've gotten to the point of faking MTK cheap-ish based phones too, rather than just faking more expensive devices.

Bit unrelated, but I hope someone here can help me out - couple friends and I want to develop a 3g-connected IoT appliance and we're looking for a chipset with reference design - if possible, a chipset that could power a today's high-end smartphone. Does anyone have an idea? Mediatek didn't reply to us in two weeks :/

If you want to use Chinese you're going to have to learn their customs and possibly the language - MTK will just ignore you if you try asking them unless you're already a huge established company.

Here's some good background reading on the culture of these ecosystems:



Actually I don't care much about the exact CPU vendor, all I need is an Android-compatible CPU with 3G support, a datasheet and if possible a basic hardware reference design... too bad none in the team are Chinese :/

Sounds like a opportunity for someone to launch a Chinese consulting company.

Hire some Chinese folks?

> This Huami H3 smartphone cost me $34US (210RMB) in January. It’s sold with a 5 inch screen, Bluetooth, WiFi, 4G cell connectivity, Android 4.4 and an 8 core processor. It's very cheap but good. And actually the processor has 2 cores.

While the phone doesn't meet it's stated spec, it's still pretty amazing that tech that'd have been top of the line a few years back now costs less than a night out.

Actually it is bit uncanny how you can basically apply a variation of Moores law. The specs of Huami H3 seem to be somewhat comparable to Nexus One, which was released almost five and half years ago. That puts 3.5 18 month cycles between the two devices. $34*(2^3.5) = $400, which sounds about right for "high-end" no-name Chinese phone.

$400 on Aliexpress gets you into the range of branded phones from brands like Sony, Xiaomi, Lenovo, Huawei, ZTE, Oppo, Jiayu these days.

Almost none of the phones in that range are truly "no-name" (though beware the fakes) - even Oppo and Jiayu are fairly established brands by now, though obviously nowhere near as well known.

I didn't think of Jiayu as being in that range of the market, but apparently their reputation has been good enough that others have been faking their phones.

They have some... interesting models like this one, which is far cheaper than $400:


I have the quadcore (MT6589) version and have gotten quite a few "is that an iPhone running Android?" questions, which somewhat adds to the experience of owning one. It's slightly bigger than an iPhone though.

This falls squarely within Moore's Law in the classic definition of transistor density.

This because transistor density allows for two things, but Intel is only pushing one of them.

First it allows for more compute power for a certain price.

Second, it allows for a set compute power at a lower price.

The second comes because you can stack more chips on a single wafer, and so get a bigger yield out of each production run.

I suspect Intel avoids that one because it will inevitably lead to a race to the bottom.

In case you haven't noticed, Intel has been consistently been pushing towards smaller dies for some time now. Atoms are prime example, but it also happening in their Core series. See e.g. this http://www.kitguru.net/components/cpu/anton-shilov/intel-10n... article for "Weighted Average Die Size" graph. Or this graphic: http://729675461.r.lightningbase-cdn.com/wp-content/uploads/...

Wow, only 200 MB RAM built in while Android still mostly working.

The Galaxy S shipped with only 256 megs and a single core. It was a capable device at the time, although it can barely run any modern apps. I know this because I replaced mine only a month ago. As recently as 6 months ago it would run Spotify and Runkeeper simultaneously, though with recent versions of Spotify I lost the ability to multitask while not on Wifi. On Uber the only things preventing me using it were the custom text fields, which wouldn't allow me to key data in properly. The biggest memory hogs were recent versions of Google Play at 40 megabytes of ram, so I would keep the Google services downgraded unless I needed to install an app.

The base Android system will run with surprisingly little RAM.

Reminds me when I installed Win XP on a 64MB/Duron box for my mother. The thing was running almost as fast as my Athlon/1024.

On a side note, i find myself reminded of a TED talk.

Mostly because it held a story about the lady out buying a handbag, and using her detailed knowledge of brands to get the seller to bring out ever better copies of a certain brand from storage.

I wonder how much this falls within a culture of haggling.

(Now lets see if i can find the right one...)

We already have $10 smartphones here (US) though they have 3.5 inch screens are run 4.0 not 4.4.

You can get higher spec better screen phones for $75 or less if you send in rebates. In the US we have hit smart phones that are affordable to everyone and can do "most stuff" already.

The phone from the article fakes 4.4, it's actually 4.0.

I meant that 4.4 is latest version that most higher end phones use (most people have not upgraded to 5.0)

I did read the article and saw the phone in question fakes most of the specs.

Examples? The cheapest I can find is the ALCATEL onetouch Evolve 2 for $19 at MetroPCS[1].

[1]: https://www.metropcs.com/phones/details/alcatel-onetouch-evo...

Alcatel C1 was $9.99 (att prepaid; earlier this month), runs 4.4.2

ZTE Whirl 2 is $9-19 and runs 4.3

Moto G is $29 (<!!!, as of last week - probably sold out by now, but max 2 per person on bestbuy.com; boost mobile moto g's standard price is $39 either way though: this is probably the best price:performance ratio in terms of not being a complete POS)


I pick up a few oddball phones for android testing/development.

These are locked to a specific carrier every time I checked - they are subsidized. I've never seen an unlocked Android for less then $39=~$40 (But I don't look often, so I might have missed some)

Probably. I don't activate these, I just use them to test my apps on small screens and incredibly shitty or old OS or other stuff.

Some of these have like.. 300x480 screens or something equally ridiculous, which is a nice way to test for "see how bad your app can get". It helps that they are all "legitimate" phones, i.e., have the full Play Store and ecosystem (supporting push, google accounts, etc).

Hmmm, can't seem to find any links for the prices you mentioned. Citations?

Alcatel C1 is over now, it was $9.99 (also https://www.att.com/deviceunlock eligible, IIRC) - bestbuy

ZTE Whirl 2 is $10 still (active, in store only) - http://www.familydollar.com/pages/hotitems.aspx

Moto G is over now, it was $29 on bestbuy, carrier locked to your choice of Boost, Verizon, or a few others. Not rootable

The Alcatel C1 is MT6572-based so should be easily rootable and unlockable.

True some are carrier locked but unlock code can be purchased for cheap (for some phones).

Plus there are lots of used phones smartphones for cheap.

Keep in mind that the price given was without carrier subsidy.

Heh, i recall using compressed ram as swap during the waning days of my N800 usage.

Regarding 'rootability', it'll almost certainly be vulnerable to towelroot: https://towelroot.com/

Unfortunately, my tablet can't be rooted by any method I've tried, including towelroot. DAMN YOU HUAWEI!

What model is it? (and what kernel?)

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact