Development effort on Pi's USB problems

Ugh.. depressing reading.  Light at the end of the tunnel is a good thing, but it's amazing the number of problems in external devices this is exposing - like cheap hubs with only one TT messing up audio.

The only thing for certain is that I'll never look at USB (in general, not specific to the Pi) in quite the same way again. 'Universal' takes on quite a different meaning...

selsinork wrote:.

 

The only thing for certain is that I'll never look at USB (in general, not specific to the Pi) in quite the same way again. 'Universal' takes on quite a different meaning...

USB is a black box that must forever remain black.  If it ever stops being black, the hordes of Mordor emerge through the opened portal and lay waste to humanity (and hobbits) while searching for the One True USB Frame.

Unfortunately in the Pi, the box isn't black.

The term "black box" is a bit misleading and I find it gets bandied about a bit too freely.

There may be some opacity in the encapsulation, but given a bright enough light (of the right frequency) then all will become visible. Whether or not you can actually touch the insides is the only issue - and it only really matters when the contents of the box don't work as they should anyway...

Sometimes people leave the box open because they know you'll need to get inside and poke about with a soldering iron because they've been too lazy and dim to finish the job properly themselves. And what's the practical difference between "closed" and "a million combinations of undocumented command line switches that only seem to make it worse"?

The Pi's USB thinger has been particularly troublesome, but it was predictable given the provenance of the SOC and the limited USB functionality of the box that originally used it. I maintain that if any blame is to be apportioned then then it is with the tech "journalists" (their term, not mine) that originally hyped it before it was even released and to a lesser extent with the Foundation for letting the hype get out of hand.

I guess that the thing with single board computers is that they're delightfully cheap, but only if everything works. At least with modular x86 if there was some fundamental disagreement between (and I'm showing my age now) an nforce2 motherboard's USB and your new piece of hardware then a £5 pci USB card with a VIA chip would doubtless sort it. There is no such recourse with credit card sized computers and you have to hope that the manufacturer sorts it in software before your throwaway computer reaches obsolescence..

Or if it's all open then you can spend a year of your life fixing it yourself while the people that sold it to you are sipping cocktails somewhere...

I guess it all comes down to buying the most suitable (in engineering terms) part in the first place.

I guess it all comes down to buying the most suitable (in engineering terms) part in the first place.

You are correct, however the end user may not be aware of those issues at purchase time.

Myself and a collegue both have RPi and use a wireless keyboard and mouse.

We thought we had faulty Logitech devices due to the random stuck keys, wrong keys, etc, but they work elsewhere, and are very strange.

It turns out the USB might be the problem.

It won't change our minds about buying/using RPi but isn't something the average user would necessarily know about, until they got sick of replacing hardware.

(it sounds like one of our systems where the optical mice kept getting blamed)

Thanks for the update Morgaine

I did note how the overlord decided the minion's form of punishment wasn't enough. ...scary

Mark

Mark Beckett wrote:

 

I guess it all comes down to buying the most suitable (in engineering terms) part in the first place.

You are correct, however the end user may not be aware of those issues at purchase time.

 

Which is why it's up to all of us to make sure that e.g. the lazy and feckless technical press start writing about the here and now rather than just dreaming about some imaginary point in the future. When was the last time you saw a roadtest that went past a "deboxing" video, for instance? And whatever happened to "it's thirty quid, so it's quite possibly going to suck really badly" style pragmatism?

Hope you get your issues sorted out btw.

rather than just dreaming about some imaginary point in the future

Jonathon

You are quite right.

I've done the odd review of Arduino books for the magazine I send words to (the editor makes them into an article), and try to give a reasonable view on who it suits, would I buy it, etc.

Unless you're like the UK Top Gear team where absolute truth is a marketing advantage, many times the end result of reviews isn't as honest as it could be.

For products it may also take a lot more in depth probing that can take far longer than the deadlines available ...however that only lasts for the intial launch.

Whether telling people the problems afterwards sells copy is another thing.

Publishers tend to like repeat business, and the advertising executives tend to see the glass half full, so honesty probably isn't the highest on the list.

When was the last time you saw a roadtest that went past a "deboxing" video, for instance?

Funny I have something sent to me to do an unboxing video.

Personally they don't interest me in deciding to buy the product or how useful it is.

If I was a company looking to stock/puchase them then I might be slightly interested, but for reasons other than its function.

Hope you get your issues sorted out btw.

Thanks for the kindness.

Its not a huge issue, and is more of a nuisance that we can live with.

Mark

Jonathan Garrish wrote:

 

And whatever happened to "it's thirty quid, so it's quite possibly going to suck really badly" style pragmatism?

Hopefully it died a death in well-deserved ignomy.  There is no price range in which badly sucking product is acceptable.

Morgaine Dinova wrote:

 

Hopefully it died a death in well-deserved ignomy.  There is no price range in which badly sucking product is acceptable.

While I don't disagree, there's something to be said for the level of pragmatism that realises that if you buy cheap crap you really do get what you pay for.  Or equally that if you buy from certain suppliers you are agreeing to be locked into a gilded cage.

Mark Beckett wrote:

 

You are correct, however the end user may not be aware of those issues at purchase time.

The other side is that the manufacturer may not be aware of those issues either, and if they are do they care or do they just want your money ?

Whether we like the idea or not, it's not about the product, it's about parting consumers from their cash. In general terms, companies will do the absolute minimum required in order to accomplist that goal.

Jonathan Garrish wrote:

 

There is no such recourse with credit card sized computers and you have to hope that the manufacturer sorts it in software before your throwaway computer reaches obsolescence..

Or you throw it away in disgust.

There is some light in the tunnel. A couple of the boards have PCIe available, so it could be possible to add that cheap replacement card. It's probably not actually worth the trouble doing it though..

selsinork wrote:

 

Morgaine Dinova wrote:

 

Hopefully it died a death in well-deserved ignomy.  There is no price range in which badly sucking product is acceptable.

While I don't disagree, there's something to be said for the level of pragmatism that realises that if you buy cheap crap you really do get what you pay for.

But there's a difference between "sucking badly" and being manufactured with an eye to extreme low cost.  "Cheap crap" is different to just plain "cheap".  Many things that are extremely cheap are not crap at all, and low cost should never be used as an excuse for "sucking badly".

Generalities aside, in our specific area of ARM technology, it is completely unacceptable to try to justify failure to operate correctly on the basis of "Well it's cheap, what do you expect?", particularly when the faulty feature in question is listed as adhering to a commercial standard such as USB.  A USB device that drops USB data isn't "a cheap USB device", it's a faulty USB device.  (Think about USB devices in general, not Pi specifically.)

In analogue electronics it is very common and normal for cheaper devices to have poor specifications, but even the cheapest are required to work as specified.  In the mostly yes/no world of digital electronics, working in accordance with a specification is even easier to distinguish.  For both kinds, there is no price range at all in which faulty operation is acceptable.

Addendum:  "You get what you pay for" doesn't extend to faulty operation.  Cheap stuff may be manufactured poorly or use low quality materials, but it will still work or it won't be of merchantable quality, at least in this corner of the world.

selsinork wrote:

 

Morgaine Dinova wrote:

 

Hopefully it died a death in well-deserved ignomy.  There is no price range in which badly sucking product is acceptable.

While I don't disagree, there's something to be said for the level of pragmatism that realises that if you buy cheap crap you really do get what you pay for.

But there's a difference between "sucking badly" and being manufactured with an eye to extreme low cost.  "Cheap crap" is different to just plain "cheap".  Many things that are extremely cheap are not crap at all, and low cost should never be used as an excuse for "sucking badly".

Generalities aside, in our specific area of ARM technology, it is completely unacceptable to try to justify failure to operate correctly on the basis of "Well it's cheap, what do you expect?", particularly when the faulty feature in question is listed as adhering to a commercial standard such as USB.  A USB device that drops USB data isn't "a cheap USB device", it's a faulty USB device.  (Think about USB devices in general, not Pi specifically.)

In analogue electronics it is very common and normal for cheaper devices to have poor specifications, but even the cheapest are required to work as specified.  In the mostly yes/no world of digital electronics, working in accordance with a specification is even easier to distinguish.  For both kinds, there is no price range at all in which faulty operation is acceptable.

Addendum:  "You get what you pay for" doesn't extend to faulty operation.  Cheap stuff may be manufactured poorly or use low quality materials, but it will still work or it won't be of merchantable quality, at least in this corner of the world.

Morgaine Dinova wrote:

 

 

But there's a difference between "sucking badly" and being manufactured with an eye to extreme low cost.  "Cheap crap" is different to just plain "cheap".  Many things that are extremely cheap are not crap at all, and low cost should never be used as an excuse for "sucking badly".

I'm not suggesting it's an excuse, just observing that in reality cheap often does equal crap.  However much we wish it wasn't the case, that doesn't change the reality.  The only way to change the reality is to hurt the profits of the companies who produce this stuff by not buying their products.

selsinork wrote:

 

The only way to change the reality is to hurt the profits of the companies who produce this stuff by not buying their products.

Indeed, and when stuff doesn't work, I'm certain that happens.

And in these days of  Internet forums and blogs and social networking sites, the negative feedback resulting from a poor experience with a product is much greater than just the linear effect of voting with your money.

Neither form of feedback has anything to do with cost though.  Costly and cheap items are both expected to work, or their manufacturers will suffer the wrath of the masses.

Morgaine Dinova wrote:

 

selsinork wrote:

 

The only way to change the reality is to hurt the profits of the companies who produce this stuff by not buying their products.

 

Indeed, and when stuff doesn't work, I'm certain that happens.

 

And in these days of  Internet forums and blogs and social networking sites, the negative feedback resulting from a poor experience with a product is much greater than just the linear effect of voting with your money.

 

Neither form of feedback has anything to do with cost though.  Costly and cheap items are both expected to work, or their manufacturers will suffer the wrath of the masses.

I'll toss another factor into the conversation.  Sometimes you have cheap items that don't comply with a standard, but most users won't notice so the seller can get away with it.  My favorite example is cheap HDMI cables that don't implement the individual ground wires, but instead use a foil shield for all grounds which is connected to the connector shells at both ends of the cable.  Such a cable will work fine for most HDMI uses, especially if it's a short cable.  This is because most HDMI products short together all the grounds and the shield, and since HDMI uses differential twisted pairs for the fast signals they really don't need individual grounds.

However, some devices do need the individual grounds.  For example, the Motorola Atrix Lapdock uses one of the individual grounds lines to detect that an Atrix phone (or Rasperry Pi) is plugged in.  The cheap cable won't work, and have fun figuring out why not.

The nasty part is that you usually can't find out whether the item is compliant or not without ordering one -- often from China.  And you don't know when you order the next one whether it will be compilant either, since the seller may have changed manufacturers.

Very good point, John!

That reminded me of our extensive discussions about USB compliance in powered hubs that backfeed power to the upstream host.  Yes indeed, standards compliance is a very thorny issue, and it's clear that compliance checking is not given much importance in some parts of the world.

It looks like Raspbian was updated on July 26.

AFAIK there aren't any release notes, but presumably it must fix something.

I see a July 20 post by jdb on the FIQ thread regarding USB audio DAC issues that says:

These aren't going to be quick or simple fixes.

So apparently there's still more known work to do.