RG1 1.8v regulator

Time keeps going by and I don't see any information being shared by Pete or anybody from RPF about the issues and what we can expect to get them fixed.

-J

Jamodio,

  right, Pete has gone silent, and JamesH says he can't say much.

Pete said he was in full-duplex communication with SMSC, but

apparently he's not eager to share.

People have asked for gerbers to see if there's a way to disconnect

1.8v by drilling a hole in a strategic place, but no response.

That doesn't mean progress isn't being made, such as people

experimenting with heatsinks.   Maybe SMSC could be asked

if the networking problems are a plausible result.  Maybe someone

will use sandpaper to reveal exactly how the power is connected.

I don't agree that everybody is blaming the 1.8V problem to keyboard repeats or anything else related to USB, at least I don't.

That's is simply a clear and flat error that needs to be fixed, again we can theorize and overanalyze the potential side effects of the error but it does not fix it and there are so many variables and other issues that make things not work for some folks and work for others.

I don't believe much that the USB problems are related to this one, I'd lean more in the direction that we are trying to squeeze a full featured USB host from a flaky USB OTG not very well known and coded on the drivers implementation.

-J

It is not clear who is telling who what, but what is pretty clear is that we are not being told anything !!

I strongly believe that at this instance the distributors should stop producing and selling the current revision of the board, which is clearly faulty.

-J

Flaky? I have the impression that the hardware is not flaky. Its just that the hardware is a bit minimal. It defers to software all that CAN be deferred to software. Together with the minimal documentation, and interactions with other Linux drivers, it turns out that it's not quite finished yet.

... stop selling ... faulty.....

They clearly should continue selling the boards, which "work for  me".

With a $1 heatsink on the lan chip you can (over) compensate for the extra heat if you want.

Part of "getting this software problem fixed" is in getting enough devices out in the field so that some brilliant linux-hacker gets one and says: Oh, this is a simple problem to fix, and fixes the driver-problem. The original plan was to get 10k units out to just Linux-hackers. But how do you restrict access to the people you want? Impossible. The only way is to flood the market, and hope the right people obtain one.

From a power supply design is both flaky and faulty, from a user perspective is flaky and unreliable.

I've a heathsink on both the SoC and SMSC and that does not fix the problems with USB.

When I found and detected the 1V8 problem I didn't have a R-Pi board in my hands, the error was just clearly visible in the schematics.

We already know that the "original plan" was badly exectued by the RPF folks, they didn't listen before, during and after the launch, and they continue on that behavior.

Flooding the market with a faulty product is not a reasonable and measurable approach to fix problems and improve hardware/firmware, putting the boards on the right hands would have been more effective, hope is not a very reliable tool of engineering.

I said it many times already, I'm not a fanboi but neither a detractor of the R-Pi, I still consider it a good idea, something worth to have, but it needs to get FIXED !!

-J

Let me add another one ...

Performance seems to be an issue for some folks, so what you can do ? "over clock it,", but that will void your warranty and it is somehting

we (RPF) can't recommend officialy ...

That's flaky++

-J

If performance is an issue for you: BUY SOMETHING ELSE!

When you overclock, chips get hotter than what they (and/or their cooling system) were designed for. If you overclock your desktop, and burn the CPU, that's your own fault. Due to errors in the system, you might not get caught if you bring it back to the shop and claim it was a manufacturing error and want it exchanged.

Roger Wolff wrote:

 

They clearly should continue selling the boards, which "work for  me".

Let's pin that one up on the Logical Fallacy of the Day board and move onto something more interesting,.

Part of "getting this software problem fixed" is in getting enough devices out in the field so that some brilliant linux-hacker gets one and says: Oh, this is a simple problem to fix, and fixes the driver-problem. The original plan was to get 10k units out to just Linux-hackers. But how do you restrict access to the people you want? Impossible. The only way is to flood the market, and hope the right people obtain one.

That paragraph makes no sense, because at the time that the 10K were envisaged, the Foundation didn't know about these faults so they couldn't possibly have even envisaged that there was anything to fix.

Be that as it may, it became even less plausible when RPF decided to advertise the device as a capable media player and to market it to all and sundry rather than only to developers.  That transformed your already questionable premise of "sell in the hope of a fix" into a completely unreasonable one.  You don't sell a product to hundreds of thousands of users in the hope that a problem which you don't realize exists will be fixed.  It may have the merit of being funny, but we're not aiming for comedy here..

The Pi is a board with 3 USB-related faults which have just recently been accepted by the Foundation:

• 1.8V board design fault which is acknowledged to be out of spec, and which we know from basic principles can result in excess power dissipation in the LAN9512 in an unknown proportion of boards.
• Extreme USB endpoint constraints which prevent Pi from being used like a normal USB host that works with a reasonable number of devices attached to hubs.  These constraints are like none ever experienced by ordinary computer users on any kind of host computer equipment whatsoever, and must be bewildering.  From any reasonable user perspective, this constitutes a major equipment fault.
• USB operation is acknowledged to lose data intermittently, because the driver requires realtime response from the Linux kernel when handling USB's split transactions. Since Linux is not a realtime O/S, those transactions can fail depending on random Linux scheduling, and so the Pi's USB system is plagued with intermittent USB data loss.  This is extremely serious, and is unprecedented in any well known Linux system.

A normal product on the market would almost certainly be withdrawn from sale if such a catalogue of faults came to light.  Also quite likely, the units it has already sold would be placed under voluntary recall, not for safety but out of sheer embarrassment.  Problems on this scale could destroy a well-known brand like Apple i* or Google Nexus.

The Pi is not a normal product of course, and the Foundation is not known for acting rationally which complicates the matter still further.  On top of this, one should add that their Apple-like Reality Distortion Field has made some unknown but vocal proportion of Pi owners consider faulty boards to be satisfactory, and also that the $35+shipping cost of Pi is low enough that more than a few owners don't really care that the board has a few problems.  Far outweighing these considerations though, the Pi is not a normal product because RPF are driven by image making and facial egg, not by a desire for good engineering and open honesty..

One can only guess what course this will take, but from past precedent my guess is that they will put their spin machine into high gear and declare that no important problems existed and that nobody significant was affected, while secretly trying to remedy these severe issues which affect vast numbers of people.

For our part, I expect we'll just have to live with this dysfunctional little board and work within its limitations until a better version appears.  We can then deposit this original on the mantlepiece as an engineering curiosity.

Morgaine.

Part of "getting this software problem fixed" is in getting enough devices out in the field so that some brilliant linux-hacker gets one and says: Oh, this is a simple problem to fix, and fixes the driver-problem. The original plan was to get 10k units out to just Linux-hackers. But how do you restrict access to the people you want? Impossible. The only way is to flood the market, and hope the right people obtain one.

 

Roger,

  Of course there is a way to restrict access to the people you want.  You declare the item

to be in beta test.  On the front page you talk about what issues are known, and what is

being done to resolve them.  Of course, that doesn't prevent over-eager buyers from

buying the product and then complaining when they are disappointed, but it gives you

a good answer that you made a good faith attempt to disclose.

   Instead, the front page highlights only the happy stories, showing pictures of

7-year olds, with no indication whatsoever that the product is not intended for

general use.  In fact, when asked if the product is in beta test, a firm "no" is the

response.  Bug fixing is happening, but it is happening in off-site github repositories,

and is "pushed" to end users through low-visibility mechanisms.

Hardware errors are in a different class than software errors, since it is possible

to fix software errors in the field by means of downloading new versions. 

Intentionally selling known faulty hardware that apparently can't be fixed in

the field doesn't strike me as something a responsible organization would do.

I suspect that Maplins signed on as a retailer without knowing of the 1.8v problem.

When they find out, it will be interesting to see what they do.  I also wonder if

they know about the claimed PCB problem that prevents residential FCC/CE

certification.  The foundation may not be worried about its reputation, but

I think Maplins would be.

jamodio wrote:

 

I don't agree that everybody is blaming the 1.8V problem to keyboard repeats or anything else related to USB, at least I don't.

 

That's is simply a clear and flat error that needs to be fixed, again we can theorize and overanalyze the potential side effects of the error but it does not fix it and there are so many variables and other issues that make things not work for some folks and work for others.

 

 

 

...

Jamodio,

  While this certainly is an error that needs to be fixed, I think it does help

to theorize and analyze.  When you first pointed out the schematic error,

that wasn't enough to get any traction.  But then when further analysis and

experimentation showed the connection to the hot lan chip, that was

motivation to contact SMSC.  And when they replied, that got the

attention of element14, and they got Pete's attention.

  At this point, the problem isn't considered serious enough for Maplins

to put a hold on the product, or for either of the distributors to make

a formal disclosure.  If further theorizing, analysis, experimentation,

and contact with SMSC can provide a link to USB and/or networking

misbehavior, that would provide more incentive to promptly fix the

problem, both for future customers and current customers.

jamodio wrote:

 

It is not clear who is telling who what, but what is pretty clear is that we are not being told anything !!

 

I strongly believe that at this instance the distributors should stop producing and selling the current revision of the board, which is clearly faulty.

 

-J

Jamodio,

  You are not the first to suggest this.

On the RPi.org forum, user lajos wrote:

   "I hope this next suggestion won't be taken the wrong way... but would it make sense to tell buyers that these are known hardware issues, or even stop manufacturing and sales until they are fixed?"

Amazingly, he hasn't been banned, belittled, or had the thread locked.

So maybe the RPF is actually considering that.

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=24&t=14478&start=10

jamodio wrote:

 

It is not clear who is telling who what, but what is pretty clear is that we are not being told anything !!

 

I strongly believe that at this instance the distributors should stop producing and selling the current revision of the board, which is clearly faulty.

 

-J

Jamodio,

  You are not the first to suggest this.

On the RPi.org forum, user lajos wrote:

   "I hope this next suggestion won't be taken the wrong way... but would it make sense to tell buyers that these are known hardware issues, or even stop manufacturing and sales until they are fixed?"

Amazingly, he hasn't been banned, belittled, or had the thread locked.

So maybe the RPF is actually considering that.

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=24&t=14478&start=10

coder27 wrote:

 

On the RPi.org forum, user lajos wrote:

 

   "I hope this next suggestion won't be taken the wrong way... but would it make sense to tell buyers that these are known hardware issues, or even stop manufacturing and sales until they are fixed?"

 

Amazingly, he hasn't been banned, belittled, or had the thread locked.

So maybe the RPF is actually considering that.

 

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=24&t=14478&start=10

Don't get your hopes up too high.  While it would be nice to think that RPF have finally woken up to reality and will start being open and honest, precedent doesn't support this possibility very strongly.  Rather more likely is that convincing spin is proving more difficult to manufacture as the evidence accumulates.

Morgaine Dinova wrote:

 

coder27 wrote:

 

...Amazingly, he hasn't been banned, belittled, or had the thread locked.

So maybe the RPF is actually considering that.

 

Don't get your hopes up too high.  While it would be nice to think that RPF have finally woken up to reality and will start being open and honest, precedent doesn't support this possibility very strongly.  Rather more likely is that convincing spin is proving more difficult to manufacture as the evidence accumulates.

Maybe manufacturing spin is taking too much time and effort, so they decided to outsource it to China.  But it'll take a few months to get the process working.

John Beetem wrote:

 

Maybe manufacturing spin is taking too much time and effort, so they decided to outsource it to China. 

Gosh no, we only outsource the productive parts of manufacturing that create jobs, grow native production capability, and increase knowledge of technology among the population.

None of that stuff is needed in the West where the biggest growth industry is litigation, where the USPTO's primary role is chilling innovation, where the media companies treat their audience as criminals, where states block science in education, and where governments' only interest in technology to keep their populations in observational prison.

Spin is a key competency right here at home, and a necessary process in making a dysfunctional board brilliant.