Important color changes


EvilDragon

Administrator
Staff member
Joined
Mar 4, 2003
Messages
29,986
Age
46
Location
Ingolstadt
Well, it's christmas today, so I hope you're all enjoying a calm and peaceful evening with your family or beloved ones!
And those who aren't: Well, don't let that bring yourself down.
You're part of our community at least, you got friends here :)

Anyways, christmas is also a day you give and receive presents - and while I can't give you your Pyra yet, I've got some quick news for you!

And no, it doesn't have anything to do with case colors... but some very important color changes in simulation.


Soo... your personal christmas present is:

The redesigned CPU board passes all simulated tests!

Nikolaus found the time to finish his improvements on the CPU board early this week and Juliano already found the time to run the simulations - thanks a lot to both of them!

And thanks to Juliano who made a nice video, you can get another view behind the scenes and see the improvements here:

https://www.openpandora.org/downloads/Pyra_DCIR_20171221.mp4

Just a quick reminder what was the issue and what happened lately:
The current prototypes we are using did fail horribly with the DDR connections... huge voltage losses, which are most probably the reason they only work with 2GB.
The first improvements made by Nikolaus end of November did fix all the DDR traces so they fully pass and don't have any voltage losses, but the power supply still had some areas left which didn't pass the test.
This doesn't necessarily mean that this is a problem - it just means that TI doesn't guarantee all OMAPs will run without issues if they're not being used within the specs.
And while we didn't have any power supply issues with the OMAP with our currently used CPU boards, it could still be they will cause issues under some circumstances (for example, when the 3D Chip is heavily used together with NEON code and a high clock rate, etc).
So we wanted to see if we can fix those as well.

A few weeks after the DDR fixes, Nikolaus tried to optimize these areas as well - with great results.

All of the tests pass - it's highly unlikely the CPU boards will have any issues now!

Yes, now the mainboards AND the CPU boards should be done.
Another puzzle fully finished, it seems!

Time to order prototype PCBs!

What have we learned from all of this?
Simulate, simulate, simulate.
Seriously, while this wasn't necessary at all with older SoCs like the OMAP3, it seems impossible now to get a properly working design with modern SoCs without simulating it.
This knowledge will greatly help with future CPU board upgrades (but let's finish the OMAP5 one first :))

For anyone interested, here is the comparison table with the most important power supply stuff of the OMAP5 of the version before and after the fixes:

https://www.openpandora.org/downloads/Comparison_200mA.png
 
Hey, this is great news! Just what we needed. 2018 is going to be an interesting year :)
And those who aren't: Well, don't let that bring yourself down.
You're part of our community at least, you got friends here :)
Thanks Michael.
 
good news everyone!
i will break out another bio-organic riesling-champagne on this. :)
*and*.... i will eat some more of these cocos-cookies....
but first, *i wish you all a verry happy xmass 2017*, wherever you are
and when you're alone,
just search another one who's alone..., see? ;)
 
great stuff, really improves confidence no end, looks like i'm gonna have to scrape together some cash soon! :-o
 
I'm glad a forum member, Juliano, had the simulation software. Worries me that the software costs so much and is so necessary for future community projects though.

You can also let a company do that for you, you usually don't buy the software.
This is still cheaper than having multiple production runs of the PCBs :)
 
Depending on what you try to build, a common solution to avoid the simulation is just to use the reference design...
 
Depending on what you try to build, a common solution to avoid the simulation is just to use the reference design...

That won't usually work. The reference design only includes the layout and how to connect everything. However, you still need to create your own PCB.
Doing that is easy if you have more than enough space, but the smaller your PCB is, the more complex it gets. And that's what we had to do with the Pyra.
 
Yes, that is definitely not a possibility for something as custom as a pyra... I was thinking for simpler devices, with enough room...
 
How many different board revisions did we actulay had so far? ^^" And would have it been less, if you would had access to the special software from the beginning?
 
My Pandora runs perfectly rock solid at 1220MHz as determined using the notaz's stress tester PND. But playing DraStic at that speed showed me that when it's running a lot of NEON code, it can't be clocked that high reliably. I had various bugs at that speed, but the most common in Mario & Luigi 3 was when sprites were scaled down the background that replaced them was sometimes the old unscaled sprite, sometimes that thing backwards, and sometimes complete gibberish, but rarely what was supposed to be revealed behind the sprite. If I clock down to 1150MHz though, all of those problems go away. I wonder if this is the sort of problem that could be avoided with the new revision. If so, this new board is very good news indeed!

Thanks for the cool video and chart, and thanks as always for documenting what's been happening. Helps me maintain some confidence that things are still getting closer to release every time I update the weekly pre-order stats thread.
 
Back
Top