Pinephone LED Fresnel lens - Commentary

[mouffa]

Pinephone's case covers the LED with a Fresnel lens probably to diffuse the light.

The problem is that the lens analyses the spectrum of the white light and produces circular rings of different wavelengths(colours), so we start from a warm yellowish colour in the middle and we reach blue on the outter periphery.

There are various such things that I wish to modify on Pinephone even by replacing ICs if this is possible.

My general observation is that Pinephone seems to be purposefully designed to be of inferior performance comparing with phones of a similar class, everything somehow "isn't working so well" for no specific reason though we have already been many years in the age of smartphones and we also have many years of experience with Ubuntu Touch.

The same observations apply to the audio quality though I have not still researched whether I can somehow optimise it except for putting the Pinephone in a box to gain some bass.

[MtnSk8]

... Pinephone seems to be purposefully designed to be of inferior performance comparing with phones of a similar class ...

Of the $200 phones that run mainline linux kernels pinephone is the only one I know of. Cheap and BLOB free is the goal here not performance.  There are no phones in pinephones "class". The closest is Librem5 but it's $800.  The other "linux phone" projects are building proprietary OS walled gardens.

Fully open hardware! Fully open software!  That's what is going on here  

You can replace the open firmware IC's with Mystery BLOB IC's but what's the point?

[mouffa]

"same class" means similar hardware capabilities

and exactly you have a completely open source OS which counts 30 years of development with as much open source hardware as possible and a SoC from 2014 and you still can't make it to "just work" even on basic things

for example, after inspecting the i-phone LED lens I noticed that it has thinner and denser rings plus some more rings in the centre, it is just a cheap plastic, is it so much more difficult to put a similar one for proper light quality ?

and yes, I would replace a chip to get some things done like energy efficiency or audio quality, even the cheapest garbage has better audio quality than Pinephone, why?

[Trbl]

"same class" means similar hardware capabilities

and exactly you have a completely open source OS which counts 30 years of development with as much open source hardware as possible and a SoC from 2014 and you still can't make it to "just work" even on basic things

for example, after inspecting the i-phone LED lens I noticed that it has thinner and denser rings plus some more rings in the centre, it is just a cheap plastic, is it so much more difficult to put a similar one for proper light quality ?

and yes, I would replace a chip to get some things done like energy efficiency or audio quality, even the cheapest garbage has better audio quality than Pinephone, why?

Software support is up to the community - that means you. Make it work better instead of complaining. This device is being sold as development hardware. It's 'class' is development hardware.

[mouffa]

I will start micro-etching plastics soon, for now I am going to open the A64 SoC and fix the transistors of the Audio Module. At least with Manjaro Phosh you can have a working phone that will ring if someone calls.

[vortex]

Pinephone's case covers the LED with a Fresnel lens probably to diffuse the light.

The problem is that the lens analyses the spectrum of the white light and produces circular rings of different wavelengths(colours), so we start from a warm yellowish colour in the middle and we reach blue on the outter periphery.

Just tested the light and it is definitely rather yellow in the center and a bluish white around the edge.

I don't think it is chromatic aberration in the lens. I did some looking and I cannot find achromatic fresnel lenses, even big lenses from the folks who make lenses for instruments and scientists (Edmund, Thorlabs, Newport, etc.). There are more than a few papers on the topic though. Of course, it could be done and perhaps there are ones available that other phone manufacturers use (e.g. go to a custom optics manufacturer and have thousands or millions built at a time), but then most likely there would be at least someone selling them in smaller batches. So I think most likely no one is using achromatic fresnel lenses. Also, this is pretty strong for chromatic aberration. So I don't think it is the lens.

I think it is more likely that it is a problem with the LED's phosphors. Most likely there is too much in the middle converting blue to yellow. Means just the LED has to be switched for another one, for better (better selection of LEDs than lenses) or worse (soldering if doing oneself, possibly need to recertify if done as a revision in manufacturing).

My general observation is that Pinephone seems to be purposefully designed to be of inferior performance comparing with phones of a similar class, everything somehow "isn't working so well" for no specific reason though we have already been many years in the age of smartphones and we also have many years of experience with Ubuntu Touch.

I doubt that bad performance was the intent. However, it is definitely true that the chosen hardware has poor performance in many ways (including power). A lot of choices were made to get certain things at the price of performance.

The CPU and its internal GPU are quite puny for example, but were chosen because the pinefolks have a lot of experience with the processor and its relatives (helps a lot in designing the PCB and getting things like RAM to actually interface with it)  and it can be run with few blobs (which also helps with supporting the processor for a long time).

Separate discrete components as opposed to an all in one SoC means more power consumption and also worse performance (a SoC where the wifi and modem can just read and write to RAM at will will obviously be faster than ones that must go through the motions of a USB connection), but help with security since there is a verifiable barrier (a USB modem definitely can't access the RAM at will unless there is a bug in the USB controller in the CPU) and it is possible to for sure be able to power down the devices and know they can't possibly turn back on (hardware kill switches).

This combined with the small batch size of the PinePhone mean that the hardware limits to its performance are quite abysmal compared to other smartphones for the same price. This is in exchange for more configurability and control, being able to run something closer to a mainline kernel (and quite likely an outright mainline kernel eventually), better hardware security properties, higher chance for a long time of software support, actually getting the PinePhone manufactured in a reasonable amount of time after the project started (experience from the other pine products carried over due to similarities), etc. The small batch size definitely doesn't help, but one has to start somewhere for a project like this. Have to actually have some phone hardware in people's hands to get the development ball really rolling, and at this stage not many people are interested in a phone like this and at this stage in development and thus the batches are small.

So the question is, is the price in performance worth what is gotten? For me, I think yes. For others, the answer is no.

The software end is interesting. The main problem is that while software on the PinePhone stands on the shoulders of decades of software development for GNU/Linux on other computers, that same software was not designed or tested with the form factor PinePhone in mind and a lot but not all not designed or optimized for the performance limitations (I say not all because people have designed and optimized software for the Raspberry Pi which also has puny hardware). Screen size, on screen keyboard use, portrait mode for a screen with this aspect ratio, performance issues (a program that performs just barely well enough on a typical desktop is going to really struggle on the PinePhone). Also, the PinePhone has a lot of things not exactly on most desktops and laptops (e.g. camera flash) or are used differently (modem used to make calls, not just to get internet). So it is taking a long time to get software up to par for it since existing software has to either be adapted (time consuming and sometimes run into limits due to design issues) or built from the ground up.

Android software would be hard to port because the android ecosystem works very very differently.

As for software from UB, doesn't UB have an android-like API? If so, it would be quite hard to port software meant for that over. Not impossible, but a lot of work. But such ports would definitely be well suited since it was designed for smart phones.

The software definitely has a long long long ways to go.

The same observations apply to the audio quality though I have not still researched whether I can somehow optimise it except for putting the Pinephone in a box to gain some bass.

and yes, I would replace a chip to get some things done like energy efficiency or audio quality, even the cheapest garbage has better audio quality than Pinephone, why?

I'm still early in testing audio output of the phone. Haven't tested the bass range. What testing I have done so far indicates that the PinePhone's best audio quality is obtained on low volume. Speaker is definitely powerful, but doesn't work as well across the whole frequency spectrum at high volumes.

I've mostly been having problems with sound recording. Currently working on narrowing down the cause and hoping there is a software solution.

As for energy efficiency, it was the price paid. That said, there is room to improve the power consumption some from the software side. The work on putting new firmware on the modem has already been able to reduce power consumption some and some kernel improvements are being done to reduce power consumption as well (see https://xnux.eu/log/#033). But there are definitely limits to what can be done.



As for the various issues that the phone has:

There are hardware choices that are part of the price paid to gain desired features like rather open hardware. Not much that can be done about it at the present time other than spend effort reverse engineering other chips to increase the number of choices for future phones. In the future, there may be more choices available. For now, for the particular goals of the project, there are sacrifices one has to make.

There are hardware choices that do not impact the goals like open hardware where there is a lot more leeway in the choices and things can potentially be optimized. The LED for the flash is a good example of a hardware issue that can be solved and improved upon without compromising the main goals of the phone. [EDIT - added something missing] The LED flash, speakers, and similar things like that are definitely fair game for criticism and things to possibly improve in future pinephones when there are problems.

And then there are the software issues. Those require time and effort to mitigate by developers, tinkerers, testing by users, etc. (note there is a lot of overlap between these). There is A LOT of work to do.

[EDIT] - added something missing two paragraphs up

[mouffa]

I was more or less aware of these dimensions but I didn't thought the case was so much developmental.

i-phone 6 follows a similar approach of everything being in a different IC, namely, Audio, GPS, Wi-fi, baseband modem, sensor controllers, power controllers, screen controllers etc and indeed the battery doesn't last long but you get through the day without heavy use.

I think the most value I get out of Pinephone is the dramatic advance on my knowledge on Linux and hardware due to the problems that I face.

I want to develop a program that can print a keyboard on the terminal framebuffer and senses the screen so that you can logout the GUI and have a working phone that the battery lasts long enough to go through the day without missing calls.

LED - I considered removing the case to check the spectrum of the LED but I don't want to destroy the sticker on the screw and "void the warranty" for the moment. On commercial phones the lens's grating is very fine and starts from the centre with a dot (degenerate circle), Pine's leaves a large area in the centre uncovered and has a small number of thick rings.

Audio - I think the "real" quality is through headphones and it is really bad, it reminds cheap auto-tuning radios of the early 2000s which would drain 2 AAA batteries in two hours.

I generally test many different applications to make an assessment whether the problem is either on the software or the hardware side, for example with mpv or ffplay I can smoothly reproduce even 1080p videos but through the browser only up to 720p with a lot frame of dropping and only for adequately designed websites like youtube.

I had had a bad experience with A33 Android 5.0 tablets in the past but since Android had being ported to Allwinner and with the help of AOSP I would expect it to be much easier, I haven't yet tested GloDroid.

Now I am mostly excited with the modem and whether it can be hacked to a level that you can control it out of specifications and use it like a software defined radio to research the cellular network.

[mouffa]

here is an apparently abandoned project implementing much more than I had in mind, it can become "Manjaro fbUI"

GitHub - 8l/fbui: Framebuffer UI (fbui) in-kernel Linux ...

I am giving a good video link too for anyone who wants to follow the journey

https://youtu.be/x1oXByIJcHU

[vortex]

I was more or less aware of these dimensions but I didn't thought the case was so much developmental.

It was a bit of the opposite for me. I was expecting more trouble in some areas. Though, I've had some trouble in areas I didn't expect, like recording sound. But was mostly expecting more trouble.

i-phone 6 follows a similar approach of everything being in a different IC, namely, Audio, GPS, Wi-fi, baseband modem, sensor controllers, power controllers, screen controllers etc and indeed the battery doesn't last long but you get through the day without heavy use.

Didn't know that about the i-phone 6.

For the PinePhone, one of the issues presently, from what I understand, is not shutting off unused cores when it is away as well as powering other peripherals unnecessarily. Also, supposedly the alternative firmware for the modem helps a bit. So, there is some room for battery life improvement. But it might not be that much. I can get mine to go over 20 hours, but awake usage very quickly drains things.

I think the most value I get out of Pinephone is the dramatic advance on my knowledge on Linux and hardware due to the problems that I face.

It sure does that.

I want to develop a program that can print a keyboard on the terminal framebuffer and senses the screen so that you can logout the GUI and have a working phone that the battery lasts long enough to go through the day without missing calls.

That could be pretty neat. Would reduce the load on the GPU quite a bit in addition to the CPU, which would help. Might even be able to get away with running only one core most of the time when it is awake.

LED - I considered removing the case to check the spectrum of the LED but I don't want to destroy the sticker on the screw and "void the warranty" for the moment. On commercial phones the lens's grating is very fine and starts from the centre with a dot (degenerate circle), Pine's leaves a large area in the centre uncovered and has a small number of thick rings.

The large center are with no rings actually makes a decent amount of sense. The lens is needed to focus the light so it goes forward. Light already going forward from the LED doesn't need any steering since it is already going in the right direction. As one goes to the edges, the light needs more bending and the curvature of the lens increases. Then the rings begin to avoid the lens becoming ludicrously thick. Depending on how wide a beam is acceptable, one could even just leave the center region flat and it be good enough.

As far as investigating the LED further, it is probably easier to lookup exactly which LED is used, lookup the datasheet from the manufacturer and hopefully there will be a picture of the light it emits cast on a surface and the pattern will probably be noticeable. If not, one could buy the LED and a driver and hook them up and test it. The parts would be fairly cheap, though one would need some protoboard if one doesn't have some already and an SMD breakout if the LED is SMD (or solder wires to the pins and just have it stand up).

Now I am mostly excited with the modem and whether it can be hacked to a level that you can control it out of specifications and use it like a software defined radio to research the cellular network.

That would not be possible without a lot of reverse engineering. The radio is interfaced by Qualcomm blobs that would have to be reverse engineered before the radio could be driven by user designed software. And the hardware itself may limit what can be done. One would need a good faraday cage to do that reverse engineering testing in, and then a radio license out of the cage unless one can get it to operate in a band where a license is not required.

[mouffa]

For the time being it seems that the most stable OS is the Manjaro Plasma, though there is much trouble caused if you leave the screen Auto-Turn on, it consumes much power and makes the Plasma unresponsive for some time especially if you hold it upside down as it is the only phone that rotates the screen 180 degrees and this causes problems.

If you check the `/sys` the sensors work all the time though in the `net.hadess.SensorProxy` in D-Bus only the accelerometre returns the readings.

I remember that the Coolpad Torino (Android 5.0) had an energy saving mode that hibernated completely everything except the modem and you could only make calls, calculator and alarms.

I consider the Pinephone the reference device for a Frame Buffer UI, old 40nm power consuming hardware that needs to be stripped off of any complex modern software technology, everything should be used as needed and then be hibernated.

We have a relevant discussion for the modem here

https://forum.pine64.org/showthread.php?tid=14007

I have seen such big Fresnel lenses on the rear windshield of the buses, in the middle there is a lens and gratings from half a diametre to the edge for focusing the image from the total width of the street.