News from all fronts


EvilDragon

Administrator
Staff member
Joined
Mar 4, 2003
Messages
29,986
Age
46
Location
Ingolstadt
Here I am, sitting in my plane for Greece, listening to some Arcade soundtracks and writing up a newspost.
It's way overdue, I wanted to write one last week, but well... let's just say I'm all up for a 30-hour day :)
I still will try to keep it short :)

Anyways, let's skip a long introduction and start right away.

1. The Case

As you can imagine, the reason I'm heading for Greece is to discuss the final tweaks of the case.
I've prepared a nice printed documentation (over 10 pages!) with all issues that still remain.
Don't worry, most of them are just tiny issues, like decreasing or increasing the size of some holes.
Or adding some peg to fix the MicroUSB port so it can't break off the PCB (how many smartphones have died because of this...?)

We should have finished cases for the GamesCom in August!

Monday, we'll work together on the physical aspects, on Tuesday we're discussing about colors (also for the aluminium metal frame).



2. The status of the PCBs

Nikolaus is currently finishing the design files and BOM for the (hopefully!) final board revision.
The one for the CPU board is already finished, the Display Board is almost finished and the Mainboard will come next.

We have also put the new 4GB RAM chips on one of our existing CPU PCBs. Once the new timings have been implemented into the bootloader, we can test these as well.

There have been many many small changes since the last revision. Nikolaus has almost finished all 30 topics he had on his to-do list in January.
Those were trivial things (like testing the sensors, etc.), optimization things (like improving the position of the LEDs for the keyboard backlight) and bug fixes (like only one of the two Space contact pads working).

Two a bit annoying things remain: The speakers are overdriven (could be a simple fix like adding a capacitor) and some boards frezee when running at 1,5GHz with 200% CPU Power for more than 10 minutes.

We've pinpointed that down pretty much already: It's most probably a bad soldering of the OMAP5, probably small micro-breaks that happen when they get too hot.
Such micro-breaks cannnot be found when X-Raying and they usually happen when the whole production process is not fully optimized yet.
So very unlikely to be a hardware design issue, but more tests will be done later.

We'll also make a second batch of prototypes, these will then also be the one that will be delivered to the prototype orders and most developers.
These should be bugfree in Hardware.

That one is planned to happen in August.


3. Part ordering!

As we're approaching the end, we started ordering parts with long leadtimes (like the OMAP5 or Palmas).
This is to speed up the production. Once we have all the parts in stock, we can start the production anytime after the prototypes have been confirmed to be fully working.

So yep, we're still hoping to deliver all preordered units within this year.



4. The new keymats

The improved keymats will arrive within July, so you can test them yourself at the GamesCom this year!


5. The new touchscreens

We'll receive three more different touchscreens (matte and clear) in order to improve the touchscreen quality.
Samples are expected before the GamesCom as well.
The clear ones will also have anti-glare filters applied, so they shouldn't act as a mirror.
We still hope to be able to use a matted one though.


6. Some software news

With the final hardware coming closer, it seems more and more developers show interest in the Pyra!
Some interesting developments have happened:

One devs working for ImgTech (the manufacturer of the 3D driver) has started to help us to get the 3D driver to work.
This is no official support from ImgTech, and of course he can't use or leak any internal sourcecode and details, but as he knows the driver, he should be a huge help!
So maybe we have working 3D at the GamesCom as well - fingers crossed.

Then another dev popped up who has worked with the TILER (the internal 2D rotator of the OMAP) and was wondering why we need the rotator chip and don't use TILER.
He hasn't used TILER on the OMAP5 yet, but on various other TI processors.
He ran a direct benchmark (without using Linux), and he thinks it should be possible to use TILER to rotate the image globally and transparent to Linux.
He did a quick check and has the impression that the Linux driver for TILER has some issues and bugs - and instead of fixing them, workarounds have been coded, that cause the huge slowdowns we were encountering.

This does mean: We MIGHT be able to rotate the screen using the internal OMAP5 hardware.

No one knows if that's true or possible, but it would make things a lot more easy: We wouldn't have to code some VSync into the SSD driver as the LCD timings would be enough, etc.

He will receive an OMAP5 EVM (devboard) on Monday and tries to work on it whenever he finds the time.
This would be awesome, and of course I'll let you know as soon as we have more details about it!

So, we're moving ahead, slow but steady.


7. The GamesCom 2016

As usual, we'll have a booth at the GamesCom 2016 in Cologne.
Everyone here is very welcome to visit us, have a chat (or coffee) with me and of course try out the Pyra.

Hopefully, we'll find the time to make a daily video from there.
Hope to meet you!
 
EvilDragon, regarding colors, have you decided what you are going to use, or is this still to be decided at a later date. I ask because you will be once again discussing case colors with them again. Dark grey/black and dark red seem to be the most popular with us here, but that may not reflect in future sales (you know that side of things better than most of us).


News in all fonts!
Impossible, there are too many. This one doesn't even look accurate, and my favorite, which isn't in the board software (and which this is supposedly based on) isn't even represented in your post. Where are the serifs? I suppose I can use this one, or this, but they aren't mono, and this one doesn't appear to have serifs (and isn't monospaced), despite the sample when selecting it.
 
Last edited:
Even if none of the LEDs are bright enough, the screen backlight will be bright enough in a darkened room, I've found with my Pandora.

Also, longest short news evar!
 
Thanks for the update! Sounds like a lot of good news.

I do have a couple of questions regarding this statement:

It's most probably a bad soldering of the OMAP5, probably small micro-breaks that happen when they get too hot.
Such micro-breaks cannnot be found when X-Raying and they usually happen when the whole production process is not fully optimized yet.
So very unlikely to be a hardware design issue, but more tests will be done later.

Is the production process expected to be fully optimized in time for the production runs? How can any risk associated with this issue be mitigated?

Thanks!
 
If TILER works for rotate then would we even need the rotator chip anymore? Would removing it save power or would it be better to keep it just in case a new CPU board needs it in the future?
 
If TILER works for rotate then would we even need the rotator chip anymore? Would removing it save power or would it be better to keep it just in case a new CPU board needs it in the future?
It can be shut off and thus not use any power, and it does more than just rotate, such as hardware scaling, plus a future CPU board may not have any sort of built in rotation that's worth using (as it was originally assumed of the OMAP5).
I would vote to leave it, even if it's not required and just predominantly sits powered off.
 
Afaik, you can't switch it off. Not without an additional bypass in external hardware. And it adds an additional €5-€10 to the BOM.
 
Thanks for the news, ED.
[doublepost=1467000370,1467000148][/doublepost]I'd be for keeping the rotator chip, too. Chaining them should result in more rotations per time. The Pyra could become the gadget with the fastest spinning display content of the world. If that doesn't sell, I'm out of hope.
 
Wasn't the argument against using internal rotating, that it would saturate the internal bus and thus lower performance a lot?
So, we have a 720x1280 screen. Lets say we are using 16bit color, stored as 2 byte. Those would need to get loaded and then stored again. At 30fps we would have 2*2*720*1280*30 bytes/sec ~ 111MBytes/sec. At 60fps and 3 bytes per pixel it would be ~332MBytes/sec. According to Wikipedia ;) the RAM should have transfer rates of at least 10GBytes/sec. So 1% or 3% of bandwidth usage still sounds reasonable to me.
Anyway, great update! Looking forward to the Pyra. My Pandora just got hit by a planet. It was able to recover, but I'm not sure how long it's going to last.
 
It's 533 MHz DDR-3 iirc, 32-bit wide * 2 channels * 2 (DDR -> data is latched twice per clock cycle) * 533 MHz ≈ 7.94 GiB/s maximum theoretical peak bandwidth. The maximum performance of anything in computing, is the performance the mfg assures you it will never reach, so it'll be lower in reality.
 
It can be shut off and thus not use any power, and it does more than just rotate, such as hardware scaling
If it's designed in it can't be shut off since the signal passes through it. Also the omap5 has scalers in the display subsystem (it can composite four layers, three of which have independent scalers).

I'd be for keeping the rotator chip, too. Chaining them should result in more rotations per time.
lol

Wasn't the argument against using internal rotating, that it would saturate the internal bus and thus lower performance a lot?
TILER doesn't cause an additional copy of the framebuffer (like doing the rotation in the gpu would), it is integrated in the memory controller and swizzles addressing and data as it passes through.
 
It's 533 MHz DDR-3 iirc, 32-bit wide * 2 channels * 2 (DDR -> data is latched twice per clock cycle) * 533 MHz ≈ 7.94 GiB/s maximum theoretical peak bandwidth. The maximum performance of anything in computing, is the performance the mfg assures you it will never reach, so it'll be lower in reality.
Don't forget about latency, though. Internally the RAM chips clock a lot slower, such high bandwidth can only be achieved if the access patterns perfectly fit the internal DDR prefetch mechanisms.
 
good news everyone, somehow i'm verry tempted to have a coffee at the gamescom this year! :)

cheers to greece!
 
We should have finished cases for the GamesCom in August!
Just out of curiosity, how long did they tinker onto these Molds all together then?

Then another dev popped up who has worked with the TILER (the internal 2D rotator of the OMAP) and was wondering why we need the rotator chip and don't use TILER.
I guess he's not the only one. ;)

This does mean: We MIGHT be able to rotate the screen using the internal OMAP5 hardware.
Sounds like all the rotator-chip-drama was more or less wasted time at the end?

All in all it feels a little bit like we are in a treatmill to me. You see progress and it goes forward but somehow this streches infinite without reaching the finish line. I hope there will be final results soon without further "tweaks" needed.
 
Back
Top