Many little steps


EvilDragon

Administrator
Staff member
Joined
Mar 4, 2003
Messages
29,986
Age
46
Location
Ingolstadt
Whew, another few steps of the Pyra finalizing have been taken.
I feel a bit exhausted, but there's no time to rest right now. The final few percent are always the hardest to achieve.
But let's see what has happened!

1. New display board

The new and probably final display board has arrived and is working fine!
Thanks to all the cutouts, it's now 10g lighter than before.

displayboard.jpg

The LEDs have now been positioned properly to light the logo and we also switched them with LEDs that glow sideways.
A first test of the effect by Nikolaus is amazing, you can do very smooth color gradients (with ANY color, as both are RGB).
This can us used to show the battery status, any other notification... or simply to annoy users with a boring glowing fruit ;))
You could also use it as ambient lighting when playing games in the dark.

Of course, you can simply switch it off if you like as well.

Right now, we still have the rotator chip on there which we'll keep until we have a setup which works using the OMAP5 for rotation. zmatt is still working on that and I hope I have something to test next weekend, as I'd love to present a Pyra at the GamesCom running the normal Pyra OS using TILER as rotation already.
Well, I already have a display board without rotator, so I'm ready for the test :)

If we won't use the rotator chip, the PCB design can be changed within a couple of minutes, and as the PCB is very simply, it can also be produced within a short time. So we can wait until the very end with that.

Here are some pictures which show how smooth the gradient of the logo is:

logo1.jpg logo2.jpg

These are ripped from this little video Nikolaus made:


Of course, once I get my hands on one of these, I'll make more pictures and videos of it in better quality :)


2. The Touchscreen

I received four different touchscreens to test: Two from a partner of BOE (one clear, one matte) and one from the company that made the Pandora touchscreen (also matte).

Can you believe it's so hard these days to get a proper resistive touchscreen?
I couldn't. But now I know it is. But I also know that we will get the one we need in the end.

Well, ALL of these touchscreens had the same colored grain effect we already know. Interestingly enough, the one that was supposed to be clear was the worst of them all...

However, the one I just received for the Pandora two months ago would work PERFECTLY!
There's absolutely NO visible color grain effect when I put that one on top of Pyras' display - so yes, I KNOW a touchscreen we need exists and is still in production.
Okay, you can see on the camera that there's a very slight degradation of the image quality - but that's not visible with your bare eye.

The area you can see on the picture is about 3cm wide in reality :)

Touchscreen-Comparison.jpg

From left to right:
Without a touchscreen, with Pandoras touchscreen, Matte Touchscreen, Clear Touchscreen

Well, I gave BOE a sample of it back with them and received the information today that they now found another touchscreen material which is as clear as the one from the Pandora.

My BOE contact MIGHT receive it just-in-time for the GamesCom and as he might visit there, I could test and replace it at the GamesCom.

I also asked the company that made Pandoras touchscreen whether they can use the exact same material for the Pyra touchscreen. No answer yet, but as soon as I get one, I'll let you know.

So, let's hope for the next version!


3. The keymat

It's a bit similar with the keymat. With each change, you can improve things and also make others worse.
That's not a big deal though, as you can combine them in the end to create the wanted result.

Okay, so the new keymats just arrived today. I tested them but didn't have time to take pictures yet, but I'll do that, promised!

What has been changed?

  • Slightly curved keys feel A LOT nicer.
  • Height of the keys is pretty good now
  • Keymat isn't as stiff as before, so another improvement here
  • DPad and ABXY-Buttons are worse than before (we tried moving the carbon pad 0,2mm closer to the PCB, but that didn't work as expected).
All in all, it's a huge improvement!
Don't worry about the buttons and DPad: That's something I wanted to try out but as it didn't work as planned, we'll reverse that change. Then it will be as good as on the Pandora.

The most important thing is that the height of the keys and buttons is okay now. Changing the stiffness can simply be achieved using a different material, no mould changes needed for that.

Right now, we're using Shore 30, which should already be pretty soft - but I my guess is that the black color actually makes the keymat more stiff than it should be.
It should be possible to not color the DPad and the areas of the buttons that are needed to push it down. This should improve the keymat even further.

I'll make pictures on Monday.


4. Speaker Quality

The speaker quality has already been vastly improved again, as I fixed a bit of the sealing.
Now that I got a transparent case, I can even see where the sealing is properly done and where not :)

Well, to make them perfect, I need the sealing to be properly plotted. Manually cutting it is just not possible without having some leak somewhere.
I'm already in contact with some companies, so hopefully, I can finish the sound setup within this month.

BTW: We also tested passive speakers with the headset port: It works perfectly, no further amplifier needed :)
Once I find the time, I'll do quality comparisons with Pandoras headset out.


5. Repaired PCBs

Nikolaus picked up some reworked PCBs from Global Components today and hopefully with these we can finally build more prototypes.

They haven't been tested yet, but that's what Nikolaus is doing within the next few days.
We'll also receive some more repaired PCBs and we'll try to build as many prototypes as possible for the GamesCom.

That's it for today.

See you soon!
 
Awesome as always.
About the audio : is it cleaner or louder than the previous video ? Even though the mic didn't do it justice, I thought it was the best in frequency restitution. All it could lack would be a bit of warmth, but I'm only guessing from the video.
 
Working almost perfectly:
very-tiled-fbdev.jpg


At least TILER is clearly getting involved... you can see the tiles ;-)
I am curious, what exactly is going wrong here? Is it swapping the color and location bits?
 
I can already see someone making the logo LEDs glow in a sequence that makes it look somewhat like actual flames, you know? like, glowing smoothly from red to yellow, while also changing their intensity. It can be achieved also with blue and green and other cool combinations. The thing is to find the right sequence for them to behave like the real thing.
I just found this video that effect is done with 30 RGB LEDs, Do you think it could be done with only two? I do.
Great Progress! and awesome updates! Thanks!
 
Last edited:
I just found this video that effect is done with 30 RGB LEDs, Do you think it could be done with only two? I do.
Great Progress! and awesome updates! Thanks!

Not exactly like that, but similar, yes.
You can control the color and intensity for both LEDs. By changing the intensity, you can move the middle point of the gradient.
 
I like how the screen is forcing you to get the rotation to work, unlike the Pandora, where rotating the screen simply doesn't exist ;) At least once I was missing being able to rotate the screen on the Pandora to be able to have the pdf reader in portrait mode. So I'm just glad the hardware is forcing you to implement that feature instead of being lazy and ignoring it.
Will that TILER thing also be able to do the trick for the HDMI out? Rotating the screen is awesome for coding, it would be sad if that doesn't work.
Oh, and the cutouts from the display board would be the perfect space for hardware mods, I guess.
 
Last edited:
Very nice. Glad to see good process is being made.

Michael, don't push yourself too hard. Enjoy the weekend at the very least.
You can ignore exhaustion for a while, but it if you do that too long then it will come back to bite you.
 
Working almost perfectly:
very-tiled-fbdev.jpg


At least TILER is clearly getting involved... you can see the tiles ;-)
That looks like bytes per pixel I remember messing with some hardware where bpp was offset by 1 so you'd need a value of 2 for a bpp of 3, worse still this wasn't actually documented just assumed - but at any rate looks like you're very close now, so it will be a real buzz when you've struggled and finally cracked it!
[doublepost=1470474457,1470474337][/doublepost]oh and ED if you're feeling exhausted its time for a day or ideally two's rest - look at the alternative a slight day or so delay or months....
put yourself first believe it or not your health is more important than the Pyra....
 
Sweet update. How much is done percentage wise? Would you say it's 90% done?
 
I would like the LEDs to do two things:

Either show the status of the battery (green to yellow to red) or be some sort of graphical light show for music :)
 
I am curious, what exactly is going wrong here? Is it swapping the color and location bits?
No, all the colors are fine, the pixels are just being shuffled around.

This was the result of my first attempt to force the legacy fbdev framebuffer into tiled memory to allow it to be rotated, since despite being legacy it still seems to be a dominant way of doing non-GL fullscreen stuff. My hack actually worked, but I initially tested this by catting a raw image into /dev/fb0. Using file ops on the framebuffer (rather than mmap) invokes a special hacky mechanism whereby the kernel maps a particular page of the framebuffer into kernel memory to be able to access it. Unfortunately in this case the pages it's using are the underlying storage, entirely bypassing TILER. As a result you can directly see the data rearrangement performed by TILER.

Using mmap() on /dev/fb0 worked a lot better:

tiled-fbdev-mmap.jpg


Now it's just using the wrong physical stride as a result of assuming the framebuffer is physically contiguous when it isn't. As a result 3 out of 4 lines are being scribbled into unused tiler memory (which fortunately is all mapped onto the same dummy page by default).
 
Last edited:
Nice news.
Isn't there some danger for the ear to be able to drive too much power in the headphones ?
 
The risk will always be there. Some audiophile headphones require a much higher impedance to be driven to acceptable volumes. If the Pandora and its successor are able to drive them, that would mean much higher volumes for more common headsets. Probably not to deafening volumes, but may get to painful levels.
 
Back
Top