Parts, Boards and the Rotator


EvilDragon

Administrator
Staff member
Joined
Mar 4, 2003
Messages
29,986
Age
46
Location
Ingolstadt
Well, while we didn't have as many updates as the PGS team had on Kickstarter, I guess our updates make a lot more sense :)
So here's what's been happening lately.

1. The case

The team is fully working on finishing the molds. Most of the fixes I discussed with them when I visited them have already been fixed and only a few remain.
The first molds are already being hardened (which you do after they are finished, so that they're more robust and the plastics doesn't have these ugly discolorations anymore).
A lot of companies in Greece have a one month holiday in August, and FormAction is no exception here.
They're doing their best to send us final cases BEFORE August. We can then test those at the Gamescom (that's a hardcore test...) and shortly after that we can prepare the mass production of the cases.

We told them I'll have a GPD Win prototype with me at the GamesCom, and that their case is finished already... which certainly boosted their working speed, so I'm positive that it'll work out.


2. Keymat

Nothing breathtaking to report here.
The new samples are being produced and should be finished early August (so also in time for the GamesCom).


3. New touchscreens

BOE will have the new touchscreens finished around July 21st.
Then they'll be shipped to Germany and my BOE contact here will visit me so we check the new screens together.
One will be matte (as this is my preferation) but shouldn't have the rainbow-grain effect.
And to be on the safe side, the other one will be non-matte with an anti-glare filter.

If the matte works without a quality decrease in the picture, we'll probably use that one.
But I will let you know once we got it and will take pictures of it.


4. The PCBs

The layout for the CPU PCBs has been submitted to Global Components, so that production of those can now start.
We plan to produce 20 more prototypes, as we were really unlucky quality-wise with the first 18 we produced.
It seems that some of the PCBs weren't the best quality, so we'll also tell GC to keep an eye on that and maybe use a different PCB manufacturer.
It's not our problem for the mass production (as we have a maximum failure rate in our contract, the rest would be covered by GC), but I guess we all want to prevent production issues, right?
That's why you do prototype runs. Having quality issues with a mass production run would be horrible.

The production of the CPU boards will take approximately 6 weeks.
The other PCBs can be done a lot faster, so we haven't ordered those yet.
Especially as we want to test different SD Card slots (as the SD Cards right now are being inserted upside down) and optimized the position of the LEDs for the keyboard backlight using a dummy PCB Nikolaus is designing right now.

As mentioned, ALL parts with a long leadtime that we know we need for sure have been ordered already for the mass production already. So these won't delay the mass production :)



5. The rotator chip

Well, zmatt here from the boards has worked hard with the OMAP5 devboard I sent him to find out while TILER (part of the OMAP5 hardware that can be used to rotate a picture) is so damn slow.
And - what should I say? He's very close to have fixed it.
It looks like the driver had workarounds for bugs instead of fixes which worked but slowed them down a lot.

So yes, as it looks like right now, we can probably get rid of the rotator chip and use the internal OMAP5 hardware to achieve the same!

He already made a few patches and benchmarks show that it's fast enough to rotate without any speed loss.
The next step is to test that with actual programs and implement that into the kernel, so that our screen is always rotated.

What would that mean?
Well, the Displayboard would be 2-layer instead of 4-layer, power usage would go down, we wouldn't have to fix the SSD driver and implement VSync, etc.

It would be a huge relieve - and you can be sure I'll tell you once it's working!

So, thanks a lot to zmatt for that help!


6. Preparing for Gamescom

As the Gamescom is approaching fast (only 4 weeks to go!), Nikolaus is working hard to fix as many Prototype PCBs as possible.
Quite a few have shortcircuits somewhere, and repairing them is not trivial.
Unfortunately, the next prototype run is AFTER the Gamescom, so we don't have any other choice.

Everyone is very welcome to visit us at the Gamescom, test the Pyra and have a chat or coffee with me.

Thanks a lot to Nikolaus for the hard work here!


7. Investments

Just a pre-information:
My biggest investor (with 100K EUR) is bailing out at the end of the year, so in case your interested in investing into the Pyra, you can do that at the end of the year.

Just contact me for more information in case you're interested (EvilDragon@openpandora.org).

That's it for today, back to work :)
 
Its great to see solid, steady progress and best of all that you're keeping people in the picture...

About the touch screen would it be possible to take photos of an angled reflection of a strong light source? (think of a triangle with the pyra at the top point and the camera and light source at the remaining points) if the pyra were a mirror you should be able to see the light source...

I promise if I win the euro lottery I will invest!!! (so don't hold your breath!)
 
Did you manage to rescue enough of those 18 CPU board prototypes to build into prototype Pyras and send them out to coders? Or did you only get enough for your own and Niklaus' use?
 
Great news about the rotation.
Well, the Displayboard would be 2-layer instead of 4-layer
Does that mean you'll remove the rotator chip from the displaybaord?
I tough you was going to keep it in case it's needed with a new cpu.

The production of the CPU boards will take approximately 6 weeks.
The other PCBs can be done a lot faster, so we haven't ordered those yet.
Especially as we want to test different SD Card slots (as the SD Cards right now are being inserted upside down)
Seems like it's still possible to change some components. I made a thread about it, I'm pretty sure that it would be better to switch to usb-c instead of the micro usb 3.0. What do you think about that?
 
Great news about the rotation.

Does that mean you'll remove the rotator chip from the displaybaord?
I tough you was going to keep it in case it's needed with a new cpu.

I doubt any other SoC would need the rotator chip. They all have features like that included.

Seems like it's still possible to change some components. I made a thread about it, I'm pretty sure that it would be better to switch to usb-c instead of the micro usb 3.0. What do you think about that?

It wouldn't fit in the case.
 
It wouldn't fit in the case.
Damn, that was my biggest concern, too bad, I'll be happy with micro-b, nevertheless.

@EvilDragon Edit: I'm not sure if you were talking about the space in the case or the hole for the port in the case. If it's the hole, I think it worth to change the mold a bit.
 
Last edited:
Nice to hear about the rotator Chip,
But im a bit confused about the CPU Board and the case:
You said, CPU Boards will take 6 Weeks, so 6 Weeks until you get Test Samples, then test these and then it will take another 6 Week for the Mass Produktion of the CPU Boards??

And when there is a 1 Months Holyday in Greece, this means whe will see needed amount of cases in Septembre..

So 2 Months to goo ??

On reason more im happy to get my Pandora back..
 
Good luck with the TILER thing. Will that also mean you get a little more profit, as in boards are cheaper and there's less components?
 
Still curious how TILER's gonna do that with 0 performance impact. I'm all in for miracles but it doesn't want to sound plausible to me. Wouldn't it at the very least either block framebuffer access during rotation or need it's own copy of it?
And how does it impact applications wanting to use TILER?

It'll probably just work out some way and no one will even think about it anymore once it runs, but I'm curious.
 
2. Keymat

Nothing breathtaking to report here.
The new samples are being produced and should be finished early August (so also in time for the GamesCom).

I forgot what was wrong with the last ones ?Ah, thickness and stiffness ? Is that optimization you talked about in June, or did I misss a version of the keymat in between ?

4. The PCBs

The layout for the CPU PCBs has been submitted to Global Components, so that production of those can now start.
We plan to produce 20 more prototypes, as we were really unlucky quality-wise with the first 18 we produced.
It seems that some of the PCBs weren't the best quality, so we'll also tell GC to keep an eye on that and maybe use a different PCB manufacturer.

The production of the CPU boards will take approximately 6 weeks.

So the 4GB memory is already completely tested, right ? Timings and idle power and all...
I guess some of the new 20 prototypes will have 2 GB and others 4 GB to make
sure the GC process mounts both RAM options right.

The other PCBs can be done a lot faster, so we haven't ordered those yet.
Especially as we want to test different SD Card slots (as the SD Cards right now are being inserted upside down) and optimized the position of the LEDs for the keyboard backlight using a dummy PCB Nikolaus is designing right now.

Thank you. I think I understand about the keyboard LEDs (had a flash of people spraying purpurine over the board to reflect the led light up to the keys :p).
But I don't really understand what's wrong with a SD slot upside down. Is there any problem with contact or being somehow loose
or something or it is simply unusual to have to remember to turn them upside down ? They won't fit in the wrong orientation anyway, will they ?
I for one wouldn't care which way the SD is up. I have one camera where I always forget what's the orientation and keep having to reinsert the other way around. No big deal.

As mentioned, ALL parts with a long leadtime that we know we need for sure have been ordered already for the mass production already. So these won't delay the mass production :)

Great! Written like this sounds like some part with long leadtime and uncertain need may be unordered and might eventually delay mass production if finally needed.
Don't worry, make sure you like every decision before you order and feel free to risk any delay. Since I haven't been around so much time as
others I may contribute a little patience. :)

5. The rotator chip
[...]
He already made a few patches and benchmarks show that it's fast enough to rotate without any speed loss.
The next step is to test that with actual programs and implement that into the kernel, so that our screen is always rotated.

Great!. I guess there's always a risk that the technique that works standalone may conflict with some kernel convention,
and some difficulties implementing it into the kernel, but I'm confident it will be finally solved and not only the Pyra
will become cheaper for you to develop and produce, less power hungry, simpler (so less likely to have bugs) and maybe faster,
but also other devices running linux on OMAP 5432 may benefit from this work.
It may also give example of good practices in open design methodology.

Does that mean you'll remove the rotator chip from the displaybaord?
I tough you was going to keep it in case it's needed with a new cpu.

I have no idea of the available features in SOCs, but at least for any SOC used in mobile phones it seems logical to have some
similar functionality, since people rotate their mobile phones quite often and expect video to work roughly as fast in portrait as in landscape.
SOCs tend to integrate many components, so if they already have CPU, GPU, VPU and MMU, it would be strange to expect
the rotator to be discrete. But on the other hand if there are discrete rotators in the market it may be that they are sometimes needed...

Seems like it's still possible to change some components. I made a thread about it, I'm pretty sure that it would be better to switch to usb-c instead of the micro usb 3.0. What do you think about that?

Pity if it doesn't fit.

So, thanks a lot to zmatt for that help!
Yes, thank you zmatt, and good luck with the work ahead.

6. Preparing for Gamescom

As the Gamescom is approaching fast (only 4 weeks to go!), Nikolaus is working hard to fix as many Prototype PCBs as possible.
Quite a few have shortcircuits somewhere, and repairing them is not trivial.
Mmmm... hopefully the info gathered helps GC improve quality ? (maybe the work falls on the wrong shoulders due to time constraints
but at least it can serve to improve things).

Unfortunately, the next prototype run is AFTER the Gamescom, so we don't have any other choice.

Everyone is very welcome to visit us at the Gamescom, test the Pyra and have a chat or coffee with me.

So this gamecom thing seems a big event. Well, I'm sure there'll be smaller events (game related or not) where
prototypes will progressively shouw up as the first ones get to coders and more people start working on Pyras.

Thanks a lot to Nikolaus for the hard work here!
Yes, thank you.

7. Investments

Just a pre-information:
My biggest investor (with 100K EUR) is bailing out at the end of the year, so in case your interested in investing into the Pyra, you can do that at the end of the year.

Just contact me for more information in case you're interested (EvilDragon@openpandora.org).

Not sure I understand the economics, should I ? I won't be able to invest anyway (because I don't play lottery, so where would I get the money from?), so I don't really care much.
I thought by the end of the year the plan was the preorders and sales would already be sufficient for ongoing production and
no further investment was needed because the product would self finance. Maybe you're trying to cover the eventuality
it isn't ready this year ?

Is there a list of small tasks for non-specialists without hardware to help ?

That's it for today, back to work :)

Thank you very much, take care and remember to rest sometimes.
[doublepost=1468781591,1468780521][/doublepost]
Still curious how TILER's gonna do that with 0 performance impact. I'm all in for miracles but it doesn't want to sound plausible to me. Wouldn't it at the very least either block framebuffer access during rotation or need it's own copy of it?
And how does it impact applications wanting to use TILER?

It'll probably just work out some way and no one will even think about it anymore once it runs, but I'm curious.

Yes. I don't understand it either, but the doubts I have are so obvious that it must have been solved when designing the SOC or TILER wouldn't be there.
zmatt or someone explained it's about the MMU changing memory addressing, I guess that while other parts think they are accessing a framebuffer by rows you
can tell TILER to make them access the same framebuffer by columns, or something like that. It's not that the framebuffed gets rewritten in place with a rotated image or
that a rotated copy is made in another buffer.
I can understand page tables and so on changing the addresses sent to RAM so that a virtual address maps to different physical addresses, or so, because
those are continguous chunks of memory being mapped. So I can imagine some different TLB changing low order bits instead of high order bits (or more complicated
but fast arithmetic) and get adresses issued to RAM in jumps of 1 row (n columns) instead of jumps of 1 pixel.
But for individual bytes or words to be accessed contiguously or with accesses spread across the range, and
speed not be affected, I don't understand how they get it, at least without a framebuffer cache in the SOC. It is somehow counterintuitive with the idea
of locality of access being more or less always assumed to speed memory access...

But what I intuitively imagine is that if a discrete chip can (de/)compress and rotate fast enough, some logic in the SOC itself should be able to do it faster.
 
We told them I'll have a GPD Win prototype with me at the GamesCom, and that their case is finished already... which certainly boosted their working speed, so I'm positive that it'll work out.
Lol, nice. :D Tell them how fast the chinese made the GPD case all together from scratch and we may get nice looking Pyra molds and cases within a day. ;)

So yes, as it looks like right now, we can probably get rid of the rotator chip and use the internal OMAP5 hardware to achieve the same!
...like often suggested right from the beginning. ;) Would have saved alot of time and money for sure. The Idea was nice but my gut feelings always told me that this damn chip will make more trouble than do any good. I would like to sue the company that made this useless piece of silicone actualy. Or make them proof it works at least.
 
What else is on the display PCB besides the RGB LED and the rotation chip? I'd like to mod my Pyra to remove the whole PCB if possible and print a slimmer case using a Pandora battery.
 
What else is on the display PCB besides the RGB LED and the rotation chip? I'd like to mod my Pyra to remove the whole PCB if possible and print a slimmer case using a Pandora battery.
Isn't the Pyra battery wider not thinner?
 
ooh, or you could stick another battery in the display part (and 3d print a new display case), for super long battery life.

it might be a little top heavy...
 
Damn, that was my biggest concern, too bad, I'll be happy with micro-b, nevertheless.

@EvilDragon Edit: I'm not sure if you were talking about the space in the case or the hole for the port in the case. If it's the hole, I think it worth to change the mold a bit.

Would probably add 3 months delay (changing the motherboard, another test-run, then changing the case, etc.) and add 10.000 EUR to the costs.
Naaaah. I could offer that later as an upgrade, but money is tight and time as well :)
 
Thanks for the update! It's such a contrast from another company not worth mentioning. Jealous of anyone in the EU next month going to Gamescon. Take lots of pictures!
 
Back
Top