Concept: Humanly Phone II, the hardware sheet

In January we sketched the Humanly Phone, a 5.4 mm minimal phone with a black AMOLED screen, eight tools and no app store. We promised to come back with the hardware sheet: the screen decision, how you type, what it is made of and what would run on it. This is that sheet. We call it Phone II because the drawing moved on enough to deserve a revision number, but it is the same device, the same 62 by 124 by 5.4 mm body and the same short list of tools.

It is still a concept. We are a software studio, nobody is building this, and nothing here is for sale. Every number marked as an estimate comes from public teardowns and datasheets, not from a prototype. The point of the exercise is the same as before: hardware makes you pay for every idea in millimetres and milliamp hours, and that is a good discipline for software people.

The screen: AMOLED, and why

We went back and forth on this more than anything else, and the long version is coming later this month in e-ink vs AMOLED. The short version is that we kept the 4.3 inch AMOLED panel at 1080 by 2160.

E-ink is lovely for reading and sips power when nothing changes. The trouble is that our eight tools are the ones where things change. A camera viewfinder needs a live image. Turn-by-turn directions need a map that moves under your thumb. A call screen needs to wake instantly when someone rings. E-ink can do most of this with partial refreshes, but it ghosts, it is slow in the cold, and colour e-ink is still muted. We did not want to design around the screen every time we drew a new view.

AMOLED lets us have both halves of what we liked about e-ink. Black pixels are off, so an interface that is mostly black costs little to show. Panels with LTPO backplanes can drop their refresh rate to around 1 Hz for a static screen, which makes an always-on clock cheap. The phone looks off until it has something to say, and it can still show a viewfinder when you need one.

Typing without a keyboard

A 4.3 inch screen at a 2:1 ratio is about 49 mm wide. Spread a QWERTY row of ten keys across that and each key is under 5 mm, which is fine for short words and miserable for anything longer. We looked at three ways around it, and we would ship all three rather than pick one.

FIG. 01: three ways to reply. A predictive keypad, dictation that runs on the phone, and a short list of replies you set up yourself.

A T9-style keypad. Twelve keys, each about 16 mm wide on this screen, with predictive text. People who grew up on it still type surprisingly fast, and it works with one thumb while walking. The prediction list sits above the keys so you can see when GATE and HATE collide, which they do, because both are 4283.

Dictation on the device. Small speech models now run well on phone chips, and Android has offered an on-device recognition API since Android 12. We want dictation that never leaves the phone, so it works in a tunnel and nobody else hears your messages. Hold 0 on the keypad and talk.

Quick replies. Most texts need one of about six answers. A short list that you edit yourself ("on my way", "call you later") covers a large share of replies with a single tap. It is the least clever option and probably the one people would use most.

QWERTY is still there as a setting. We just would not make it the default.

Titanium frame, glass or ceramic back

The frame is titanium grade 5, the alloy Ti-6Al-4V. It is stiffer and much stronger than the aluminium alloys most phones use, which matters when the whole device is 5.4 mm thick and lives in a back pocket. It also costs more and is slower to machine, and it is poor at spreading heat, with a thermal conductivity many times lower than aluminium. Phones that use titanium often bond it to an aluminium structure inside for exactly that reason. In our case the processor draws so little that a graphite sheet under the board should be enough, but that is a guess we would want to test.

FIG. 02: candidate materials with typical densities, and an edge section through the five layers. The titanium rail wraps the stack and doubles as the antenna.

For the back we still have two options and we have not settled.

Matte glass is light, cheap and easy to source. Etched glass hides fingerprints and feels warm enough. It cracks like glass does.
Zirconia ceramic is much harder to scratch, feels dense and cool, and looks the same after years. It is also heavier (around 6 g/cm³ against roughly 2.4 for glass) and costs more to make in thin, flat sheets.

Both let radio signals through, which is the real constraint. A titanium back would look great and would block every antenna behind it.

Radios in a metal phone

A metal frame is a problem for radios, because metal around an antenna reflects and absorbs the signal. The standard answer, used by most metal phones for over a decade, is to turn the frame itself into the antennas. You cut the rail into segments, fill the cuts with a polymer so the edge stays smooth, and feed each segment as a separate antenna.

FIG. 03: top-down with the back removed. The board sits beside the cell, and each end of the rail is an antenna, with keep-out zones where nothing metal may go.

We draw four splits, which gives two antenna segments at the top and two at the bottom. That is probably too few. 4G and 5G want several antennas for MIMO, and Wi-Fi, Bluetooth and GPS need their own. Real phones usually hide extra antennas on flex circuits near the ends as well. Our honest count is four to six, and the drawing will have to change once someone who does RF for a living looks at it.

Thinness makes this harder. Antennas want space away from the display, the battery and the board, all of which are large pieces of metal in RF terms. At 5.4 mm there is not much space in any direction, so the ends of the phone become keep-out zones: no board, no cell, just the speaker, the USB-C port and plastic. Your hand also detunes the antennas when you hold the phone, which is why phones switch between antennas on the fly. We would need that too.

Charging and durability

Charging is USB-C and wired only. A cell of about 2,000 mAh (estimate) does not need fast charging. Around 10 W would fill most of it in under an hour, and slow charging is gentler on a small cell that has to last years. We left out wireless charging. A coil and its ferrite shield add a few tenths of a millimetre, and we do not have them to spare. A ceramic back would allow it later if the stack ever gets thinner somewhere else.

The battery layer is 3.2 mm, but it is not all battery. Lithium cells swell a little over their life, and the design has to leave room for that or the back gets pushed off. So some of that layer is air on purpose.

For durability we would aim for:

Flat cover glass that sits a fraction of a millimetre below the titanium rail, so a face-down drop lands on metal first.
Water and dust sealing in the IP67 class. Thin gaskets are the hard part, and we do not know yet whether 5.4 mm leaves enough room for them.
A simple bumper in the box. We said that in January and we still mean it.

The software: Android base or our own

This is the part we know best, and it is the part with the most open questions. Writing an operating system from scratch is out. The work is not in the home screen. It is in the modem firmware, drivers, power management and years of security updates, and chip makers ship their support packages for Android and Linux, not for anything we would invent.

So the realistic base is Android, built from the open source project, with no app store and no browser. Our eight tools would be the only apps, written by us as plain Android apps. That gives us working radios, battery management and a patch process that follows the monthly Android security bulletins.

What we have not solved:

Messaging. SMS and MMS are easy. RCS on most Android phones runs through Google's Messages app, which a phone without Google services may not get. That brings back the group chat problem from the first post.
Maps. Offline maps from OpenStreetMap data are good for walking and driving. Transit timetables are patchy and depend on the city.
Updates. Someone has to keep a custom Android build patched for five years or more. For a small team that is a real cost, not a footnote.

We build mobile apps for a living, so the eight tools themselves feel like the easy part. The phone around them is not.

Spec sheet, revised

Size	62 × 124 × 5.4 mm
Screen	4.3 inch AMOLED, 1080 × 2160, LTPO, true-black interface
Frame	Titanium grade 5 rail with four polymer-filled antenna splits
Back	Matte glass or zirconia ceramic (undecided)
Input	T9-style keypad, on-device dictation, quick replies, QWERTY as an option
Battery	About 2,000 mAh, two to three days (estimate)
Charging	USB-C, wired, around 10 W (estimate)
Sealing	IP67 class target, not yet shown to fit
Software	Android open source base, eight tools, no app store, no browser

The next thing we would draw is the antenna layout with real band numbers on it, and we would rather not guess at that one. If you design RF for small devices and want to tell us where the drawing is wrong, the contact page reaches us directly.