The Reality Of Dead Pixels, Soft Or Inconsistently Registering Buttons


Arkay Firestar

Very Active Member
Joined
Apr 28, 2010
Messages
243
So, something I've seen addressed here and there but not really extensively discussed is the reality of hardware defects. Myself or my friends/coworkers have all had at least one electronics product from a major company fail or arrive with some defect, whether it's a dead or stuck pixel, mushy buttons or buttons that don't register properly, etc.

I'm sure that OPT will try their best to make sure that all their customers receive a defect-free product, but the reality is that proper quality assurance testing involves going through every aspect of every unit. People are throwing around casual estimates for assembling a unit, but in reality, proper testing of the screen, the wi-fi, the buttons, etc will likely take as long or longer than physically putting the things together.

I'm ok with buttons that are mushy here or there, so long as they register, and even a button that seems inconsistent can possibly be fixed by the user, but what scares me to death is the prospect of a dead pixel. Not a stuck one, but a hot or completely dead one, which are typically irreprable.

I think given how slow things will be with production that it is unrealistic, despite ED at some point saying they'll take care of dead pixels, to expect that a system with a single dead pixel should be exchanged. I mean, I'm ultra OCD about that kind of stuff and even I'm going to let one or two go, because frankly, I'm buying something that was put together in someone's garage (well, village hall, but close).

But what if there are more than just a couple? Or what if a unit arrives with serious hardware issues? Sure, something could happen during shipping, but what is OPT planning in terms of customer support should we get our systems and it turns out 20% are DOA, or have screen issues, or speakers that don't work, etc?

My personal thought is that OPT needs to test each unit extensively, maybe have one person spend 15-45minutes with each system testing the s**t out of every aspect. Sure, it will slow things down much more, but how devastating would it be for a project like this to have a majority of systems need to be shipped back? After a point, they wouldn't even be able to afford it. Better to take as long as possible to make sure every unit works as it is supposed to.

No need to rush if we're indeed this close to the finish line. I post this because all these folks trying to figure out just how fast OPT can put this together and pop them out has me nervous that important steps will be skipped or overlooked in the interests of not delaying any longer, when, in reality, if you're going to do something you should still do it right, even if people complain.
 
They'll be doing basic tests whilst flashing. I'm sure they've already thought about this. I'd be pretty gutted if I got even a single dead pixel on my unit. I remember how annoying one was on my laptop.
 
There will be no quality control, and nobody will verify assembly. They will more or less stack the parts loosely in the box, and ship it to you. Once the box is in your hands, it's your problem.

Seriously. They have been going through with this build spending most of there own savings, and taking out loans to fund it. You really for a second think they haven't made up a plan on how to test systems b4 shipping them of? I'm just speculating, of course, but i would assume it's tested right after the up to date software is blown in to them.
B!
 
It's called a burn in test. They gave 105 boards some serious (metaphorical) hammering. 103 survived.
Knowing that they've got such a low board failure rate, it's really enough to run the basic tests to make sure everything is working (no dead or stuck pixels, all keys respond, it clocks to the guaranteed rate and runs some program that hits every instruction, writes through memory, that sort of thing). As far as I know, that's the plan. They'll try to catch the 2% that fail, but if they don't, that's what warranty is for. There's no need to spend more than a minute or so looking at the screen and pushing the buttons to make sure they respond.
 
WizardStan said:
It's called a burn in test. They gave 105 boards some serious (metaphorical) hammering. 103 survived.
Knowing that they've got such a low board failure rate, it's really enough to run the basic tests to make sure everything is working (no dead or stuck pixels, all keys respond, it clocks to the guaranteed rate and runs some program that hits every instruction, writes through memory, that sort of thing). As far as I know, that's the plan. They'll try to catch the 2% that fail, but if they don't, that's what warranty is for. There's no need to spend more than a minute or so looking at the screen and pushing the buttons to make sure they respond.

WizardStan,

Note I didnt really question the board itself. That is the one thing where, as long as the system turns on, you know core things are working and can run a quick software test for things like wifi and bluetooth and whatnot. The LCD, the buttons, etc need to be tested by sight and touch. Look, I'm not intentionally trying to be a pessimist, but I thought I'd throw the worry out. Not because I'm expcting no one to consider quality assurance testing, but rather because of the reality of producing hundreds of portable electronics devices. Some of us will get defective hardware. It's fact. It's part of life. Every major hardware maker has it happen to them.

I guess my fears will probably be addressed when OPT starts assembling and possibly films/webcams the process so we can see units being tested and whatnot.

Again, not trying to be a pessimist (just happen to be one sometimes :) ).
 
Last edited by a moderator:
Arkay Firestar said:
The LCD, the buttons, etc need to be tested by sight and touch.
I know, that's why I suggested they'd be running the quick test. But I honestly don't see what you're hoping to gain by using it for upwards of 45 minutes. Either there are stuck pixels or there aren't, you can tell with a single glance. Either the buttons respond or they don't, just hit each one a couple of times in turn. If there's a mass failing like you fear, one that isn't caught with such simple testing, then testing them longer isn't going to prevent it, it'll just make it obvious that there's a fundamental problem with the system.
I hate to say this, but if some huge number of Pandoras are really failing within minutes of use then OPT has failed so very hard that the only logical course would be to stop shipping them until the problem is found, and personally I'd rather have my Pandora at risk of some squishy buttons.
 
Last edited by a moderator:
Arkay Firestar said:
There's really no need for this thread. A year ago, maybe, yes. Now that we're into the home stretch, then there's no need.
People are waiting and want their Pandora asap, NOT 2011!!!
 
Last edited by a moderator:
ajum, even Dell doesnt change LCDs for a single dead pixel.

last time i checked Dell said they wont change the screen unless there are 6 or more VISIBLE Dead pixels (i guess they mean full rbg dead).

the machines IIRC will have an auto test script, that should cover most of everything either in assembly or for future reference.

i believe they will change products that really have issues on the quality, beyond that, dont expect beyond what large corps do for warranty.

EDIT.

also, whats all the panic about?
even the very best of the best QC you get a % of fails its impossible to narrow it to zero
(if humands could, planes wouldnt ever fail, space ships wouldnt go boom).
 
I can appreciate the sentiments, but I believe your fears are unfounded. Even if they weren't, there's literally nothing that can be done about it now. Shipping is way past due. A further delay could very well kill the project.
 
Every board is completely tested and given a 24-48 hour burn in test before being shipped to Craig. It's part of the reason why they aren't making them as fast as they first promised. They are trying to help us guarantee a 100% success rate other than damage that could happen in shipping or during assembly (hopefully not). It is costing us more money this way but hopefully it works out in the long run with as close to zero RMAs as possible.

The assembler plus one dedicated engineer at TI who stuck with us through the entire process worked to create a test jig and automation software to provide a simple test procedure requiring minimal human interaction. The boards lock into the jig and have a bed of nails setup to test many different points on the board while live. They are still working on reducing the human interaction which is also why boards are not being delivered as quickly as first expected.

Basically, all features of the board are tested before shipping, including the pads that the game buttons and keypad hits (via bed of nails). All we have to do is make sure the LCDs work and that the keymat buttons work, both of which are part of Notaz's flash loader software.

As for dead pixels, replacing a unit for a single dead pixel is unrealistic. Our position on the failure limitations must match that of the LCD manufacturer because any panels we have to replace need to go back to be replaced by them too. I don't know what the specific criteria is as I've never talked to that company directly and hopefully Craig has those details.
 
WizardStan said:
It's called a burn in test. They gave 105 boards some serious (metaphorical) hammering. 103 survived.

So who's the lucky one gonna be?
There will be 76 of youse, come out wherever you are!

*shame it couldn't be 66 of 4000 (118 of 120 tested), then we'd think its an evil sign and no coincidence :p
 
Last edited by a moderator:
Sphinxter said:
Confirmed; 1-2 days and 8 minutes to build one pandora.
youse forgot the
2 seconds
2 miliseconds and
2 plank times.
 
Last edited by a moderator:
MWeston posted on 16 September 2008 - 07:25 PM said:
I've not seen any dead pixels on the LCDs we have, but if you had them and they were bad (ie you could notice them) we would replace it.

Post by Craig at: http://www.gp32x.de/board/index.php?/topic/44169-dead-pixel-policy/page__view__findpost__p__648776

Highlighting by Monk.
 
Last edited by a moderator:
MWeston said:
Every board is completely tested and given a 24-48 hour burn in test before being shipped to Craig.
Wow, 24h per Board, thats are 4000 Days for 4000 Boards :lol: I really hope they test more than one Board per time. ;)
 
Last edited by a moderator:
This is REALLY easy. Take a look at this table:

http://en.wikipedia.org/wiki/ISO_13406-2

Keep in mind, this is for 1 million pixel, we have 800*480 = 384000.

So, all we have to know from MWeston is which class they ordered. If it is class I then we are save ;)
 
Back when I bought my Neovo LCD monitor, they used zones.
Zone 1 (center of the screen), if there was 1 defective pixel in the zone, you could ask for a replacement.
Zone 2 (the rest of the screen), it was 2 contiguous or 3 separated defective pixels for replacement.
It was like 5 years ago.
 
With the proportions of defective units that we would expect, it's probably much easier and much faster just to ship units that are potentially defective, and replace as necessary, at least for the first batch. Obviously some problems can easily be caught by a quick test, but otherwise it's probably not worth it.

It's not entirely professional of course, but I think at this stage most people would rather get their Pandora sooner, and risk having to return it.
Later on, more rigorous quality controls will be necessary, since people buying in online shops are going to be less sympathetic if it doesn't work when it arrives.
 
The whole dead pixel issue was a huge topic for manufacturers back when the PSP was launched. Sony's problem was relying on the class of part they ordered from the LCD maker to determine their failure rate. They chose to go for the cheap route to lower costs thinking that the faulty units would not be significant (afterall even Nintendo's godlike GBAs weren't immune as I had two SPs with a small number of stuck pixels) when the class of LCDs were found to be more of a problem then they expected, they switched classes for a lower failure rate. I think it's unrealistic to expect there to not be a single dead pixel in the whole bunch but assuming they picked a decent class of screen the number of instances should be fair.

And supposedly it's the same screen that Archos uses so the number of faulty screens shouldnt be any worse than what they've experienced (I'd say a check of consumer reports for Dead pixels on archose units with the same screen would give a early indicator of whether or not there's a problem with them.)
 
I guess i have been fantastically lucky. I have never owned an LCD device with a dead or stuck pixel that was noticeable. Even back in the day when the Turboexpress which was known for that due to the tech not being as good as today. I take that back. I might have had one Stuck red on one of the 3 Express systems (it was kinda hard to tell as the screen was really small and the res wasn't fine enough so it might have been stuck but not really noticeable) I had one new, 2 used over a few years. My mp3 players, DS's, PSP,GB,GBC,GBA, 3 laptops and one 720pTV etc. No noticeable dead/stuck pixels, knock on wood my dream device will allow me to keep that streak alive. I think things so many screens have been made for so many years, that while it still happens, it's probably much less a concern now than ever. Not that it doesn't still happen of course.
 
Back
Top