Prototype mainboards coming soon

Black Sliver

Member
Joined
Apr 5, 2016
Messages
53
What if a trace or power plane catches the 1.5Ghz RF signal like an antenna?
I've managed to fry an µC with RF imission (30V/m (3x industrial limit)) once. That moment when you hold a measuring tape with λ/2 instead of cm or inch :D
But stuff like that would depend on location (external RF source) and not really on load/core frequency.
Quarter-wave of 1.5Ghz is 5cm. The space between the CPU board connectors looks close to that size.
If GND "swings" at an offset to PE, that should not be a problem for the CPU itself, since other traces should then just have the same offset to PE (on multi-layer). Assuming the DC/DC blocks noise from supply lines, the relevant distance is from the DC/DC converter to the CPU (you can apply noise between VCC and GND only there).
The CPU might be blasting it's 1.5Ghz clock signal into the main board
This would be more like cross-talk, which normally only affects high-impedance non-burried traces. So you'd probably see random SD/SPI/I²C errors before any cpu crash.
resonate with something, like an accidental LC circuit
(Assuming we are still talking about VCC,) the caps are way to big to get a (high-Q) resonator at GHz (or MHz). Besides that, the caps were already changed which would have shifted the frequency.
 
Last edited:

WizardStan

Mega GP Mania
Joined
May 24, 2008
Messages
16,733
Is there a remote possibility that it's the OMAP5 themselves? Like they're lower quality that simply can't function at the advertised rate because they're "pre-production" units or whatever you'd call them when they ship you parts to use in prototypes?
 

rv6502-2

Still Fresh
Joined
Jan 17, 2016
Messages
22
Location
Volcano Base
Website
rv6502.com
The board works perfectly fine with 1,5GHz, if you only use ONE CPU.
Using only one CPU core would roughly half the RF interference caused by CPU activity.


(Assuming we are still talking about VCC,) the caps are way to big to get a (high-Q) resonator at GHz (or MHz). Besides that, the caps were already changed which would have shifted the frequency.
True, so that's eliminated.

This would be more like cross-talk, which normally only affects high-impedance non-burried traces. So you'd probably see random SD/SPI/I²C errors before any cpu crash.
Probably, except that SD/SPI/I²C isn't running at 1.5Ghz so it can tolerate a bit more.


Anyway, just a wild theory. At those frequencies lots of annoying things happen.

It's not like weirder thing have never happened, like the Raspberry Pi 2's accidental "solar panel" crashing it by shutting down the power regulator.
https://www.neowin.net/news/a-camera-flash-will-make-the-raspberry-pi-2-freeze-and-reboot
[doublepost=1482081002,1482080496][/doublepost]
Is there a remote possibility that it's the OMAP5 themselves? Like they're lower quality that simply can't function at the advertised rate because they're "pre-production" units or whatever you'd call them when they ship you parts to use in prototypes?
Usually parts sent for prototyping/sampling are the better ones so you don't have an issue and think their chips are worthless.
"1st impressions" and all that.

There was issues with a certain game console and some games because a few of the consumer units had slightly slower memories than the dev-kits and test-kits that had "hand picked" better ram chip so the devs never saw the issue during testing.
The devs couldn't test on consumer units so they only got the surprise after launch.
 

EvilDragon

Administrator
Staff member
Joined
Mar 4, 2003
Messages
29,245
Age
42
Location
Ingolstadt
Usually parts sent for prototyping/sampling are the better ones so you don't have an issue and think their chips are worthless.
"1st impressions" and all that.
No, often samples are being taken from the first production run, which is often a prototype revision.
That's what's on the Classic Pandoras as well :D

However, our current boards have the latest production revision, so it's not that.

Im sure that I missed this, but when will the new CPU boards happen?
Probably soon. One production run costs about 10k EUR and takes 6 weeks, so we should do anything but RUSH it.
That's why we're doing multiple tests and are improving the general layout (adding more vias, etc.), to make sure to improve it as good as possible.

Adding more vias reduces the noise in the ground plane, so that could VERY much cause exact such issue.
But there are some other ideas worth checking. We're hurrying, but not rushing :)
 

Kippykip

BFG 9000
Joined
Sep 6, 2016
Messages
499
Age
21
Location
'STRAYA
Website
kippykip.com
No, often samples are being taken from the first production run, which is often a prototype revision.
That's what's on the Classic Pandoras as well :D

However, our current boards have the latest production revision, so it's not that.



Probably soon. One production run costs about 10k EUR and takes 6 weeks, so we should do anything but RUSH it.
That's why we're doing multiple tests and are improving the general layout (adding more vias, etc.), to make sure to improve it as good as possible.

Adding more vias reduces the noise in the ground plane, so that could VERY much cause exact such issue.
But there are some other ideas worth checking. We're hurrying, but not rushing :)
Not sure if I've missed something, but has the 1.5ghz issue not been fixed? Or does need new PCBs/cpuboards to test?
Sorry lads I just don't understand anything about power and circuits etc lmao
 

CapcomSuicide

Member
Joined
Mar 30, 2010
Messages
250
Is it possible to guess at a realistic date for the prototype pre-orders and then main pre-order queue? I must admit my pessimistic view was that the prototypes would be done in Dec, and the 900 main orders would happen in April 2017. That looks a bit optimistic now. Fingers crossed the latest boards work as expected.
 

ThinkPad

Very Active Member
Joined
Dec 21, 2012
Messages
423
As far as I can see,
If all the issues on the CPU board are finally fixed with the upcoming CPU board, prototype run,
June 2017 should be a reasonable ETA for the first 1000 units.


If not, well... Christmas 2017... Who knows... ?!

6 weeks are needed in between every failed prototype run, and this practice of trail and error will continue until all the issues are resolved.
And the upcoming run will be the 3rd or the 4th one in a row...

Even If everything is okay with this run, the preparation for mass production itself will take 2-3 Months
Also the end price for a single unit will increase with at least with 30-40 Euro if its not already.
 

CapcomSuicide

Member
Joined
Mar 30, 2010
Messages
250
Yes, I was thinking around June myself. Great that the CPU board is forwards upgradable, so the next dev cycle will be that bit easier.

I assume if the funds run out, we will be asked to settle the preorders in advance.
 
Top