Is Everything On Track For Early April Production?

[Monk]

At the moment, we get about 45KB/s max.

ED, please use full words when describing data rate. You've now said both Kb and KB, which mean different things. Either KiloBits or KiloBytes please?

+1

I thought that the pandora team was certain that the wifi problem wasn't hardware related.

They were. Now they are less certain. They won't be as certain again, either way, until they think they have found the ACTUAL problem.

However, we have no money to pay a team to get WiFi working.

This is something I can't understand. You would pay 1000 * 10EURO (that is price at AMAZON I could find). That equals 10k Euro.

When buying in batches of the thousands, rather than single units, they mayget a better price, and through fewer intermediaries. Going for better price is how they managed to DRASTICALLY cut the cost of the cases, for example, so your 10k Euro might be a considerable overestimate, and they've had this WiFi chip for a considerable time now already, much of that time in pretty capable hands (yeah, I know - on the PCB would be a abetter place to put it...).

 

[Bosbeetle]

A dongle is fine for me

So I guess 1st of April assemby is off, nice I didnt like that date

 

[ste_167]

My thoughts...

• I hope I'm in the first 1000
• 45 KBytes isn't terrible and I'll gladly take up the dongle offer

• Although the dongle and motherboard replacement offer seems like a no-brainer of a question, it has to be asked as OP can't afford masses of returns on the basis of 'broken WiFi'
• This offer has to be limited to 1000 given the free motherboard replacements, or it'll be too expensive for the company

• Will the first 1000 be contacted individually?

• Are TI in anyway accountable for their Wifi chip recommendation?
• Can they help? (Isn't the Pandora an opportunity to showcase their hardware?)

 

[Chairman_Now]

Would this problem exist if the Wifi chip had been interfaced via SPI? Why was SDIO chosen over SPI?

 

[dyslexic]

A dongle is fine for me

So I guess 1st of April assemby is off, nice I didnt like that date

Did you really expect them to start production on the 1st of April.
I would have been extremely surprised if they would have had all the parts on ore before that date.
It will be extremely likely they will still miss some important parts when they think they are ready to go.

The statement about the new machine with the factory sound colorful to me.

Well the offer with Wifi is a nice quick fix(offer) hopefully they get it working one day.

PS 1 kilobyte = 8 kilobits Example 1024 kbit/s : 8 = 128 kByte/s.

 

[notaz]

At the moment, we get about 45KB/s max.

ED, please use full words when describing data rate. You've now said both Kb and KB, which mean different things. Either KiloBits or KiloBytes please?

bytes, peak.

10k Euro is about 1 months for 2 people working 40h a week full time on the problem.

You won't find such people easily. You might find someone claiming he can solve your problems, but in reality they'll just take your money and waste your time.

OTOH notaz (and others?) are working on this for longer then 1 month hmmm..).

..a few hours after the day job? Remember that all software work on this project is 100% volunteer work.

 

[aq21]

• Are TI in anyway accountable for their Wifi chip recommendation?
• Can they help? (Isn't the Pandora an opportunity to showcase their hardware?)

Pandora might have been that two years ago. Now there are so many devices working with OMAP3.

And there is no such thing as "Linux SDIO problem". The smartq7 has SDIO wifi that has no problem transferring at least 7.5Mbit/s.

 

[Monk]

PS 1 kilobyte = 8 kilobits Example 1024 kbit/s : 8 = 128 kByte/s.

Technically yup. When stored or transferred using various mediums you can however add parity bits, or start/stop bits, which can increase the effective "size" of a byte to ABOUT 10 bits (rarely more I'd have thought, often less unless you have high end hardware). I don't know how many extra bits are transferred when using WiFi but assuming that any number of kBytes required exactly that number of kbits * 8 seems optimistic. I guess it depends where the bits are measured.

 

[notaz]

And there is no such thing as "Linux SDIO problem".

True. People are just blaming SDIO here for no reason, where it has been supported in Linux since somewhere around 2.6.22 (2007).

 

[Monk]

At the moment, we get about 45KB/s max.

ED, please use full words when describing data rate. You've now said both Kb and KB, which mean different things. Either KiloBits or KiloBytes please?

bytes, peak.

Thanks notaz - that's better than at first it seemed (about 8 times better!). Though "peak" could be important when speeds are this low.

Any chance you could tell us why speeds seem to be so slow compared to the multi-megabyte speed target? You seem not to think it's SDIO-caused, but I'm reluctant to make too many guesses it'd be nice to just be told what's happening, if you can spare the time.

Ta

EDIT for PS :

And there is no such thing as "Linux SDIO problem".

True. People are just blaming SDIO here for no reason, where it has been supported in Linux since somewhere around 2.6.22 (2007).

LOL! On first glance I was SO trying to figure out how Linux had been supporting SDIO since the second of June 1922, or the 6th of February 1922! Got there in the end, though

 

[mhadf]

45 Kilobytes (KB) Max = 360 Kilobits (Kb) max... Sounds fine to me.

That speed in the UK is about 1/5th of your 2Mb ADSL line. What's the problem again?

I understand wanting everything absolutely perfect, but there comes a point where you must draw a line under the hardware changes and focus on other areas of development that can be continued with minimal disruption to the production process.... i.e Software, firmware, etc.

Pandora is now at that point, I understand that most of the parts have shipped to be assembled. So leave the physical changes for ver 2.like the professionals do ie. DS > DS Lite > DS XL.

You've all done really well so far, just don't go backwards at the last step. I don't think people could take it.


PS. What chip is used for the WIFI?

 

[gruso]

And there is no such thing as "Linux SDIO problem".

True. People are just blaming SDIO here for no reason, where it has been supported in Linux since somewhere around 2.6.22 (2007).

That misconception has come through official channels I believe, even if it wasn't explicitly blaming SDIO. Certainly I have had the impression that using SDIO rather than USB for the wifi has been a contributing factor. I'll try to dig up the sources and we'll see where we've all gone wrong.

 

[mali]

That misconception has come through official channels I believe, even if it wasn't explicitly blaming SDIO. Certainly I have had the impression that using SDIO rather than USB for the wifi has been a contributing factor. I'll try to dig up the sources and we'll see where we've all gone wrong.

Yep, that's what I seem to recall, too.

So, is MWeston already redesigning the boards for SPI?

 

[Monk]

45 Kilobytes (KB) Max = 360 Kilobits (Kb) max... Sounds fine to me.

That speed in the UK is about 1/5th of your 2Mb ADSL line. What's the problem again?

Well, to be fair I've got a 10Mb Cable connection here in the UK, so it's about one twenty-seventh of my speed, at peak. But I'm darned lucky to see actual speeds of 10Mbps - the peak is FAR above the average.

That misconception has come through official channels I believe, even if it wasn't explicitly blaming SDIO. Certainly I have had the impression that using SDIO rather than USB for the wifi has been a contributing factor. I'll try to dig up the sources and we'll see where we've all gone wrong.

Yep, that's what I seem to recall, too.

thirded, which is one reason I'd love some clarity from notaz, if he could.

So, is MWeston already redesigning the boards for SPI?

If they haven't found the exact problem yet, redesigning the board might be premature.

 

[Pleng]

Any chance you could tell us why speeds seem to be so slow compared to the multi-megabyte speed target? You seem not to think it's SDIO-caused, but I'm reluctant to make too many guesses it'd be nice to just be told what's happening, if you can spare the time.

I guess they don't really know why speeds are so slow. If they did they'd have a direction to go with fixing it.

 

[Monk]

I guess they don't really know why speeds are so slow. If they did they'd have a direction to go with fixing it.

True, but I'd hope they have some idea of what they are looking at - for example, if they have (or can) rule out SDIO, what else can be ruled out? What does that leave? If they really have no idea why it's not achieving full speed... that would not be a good place to be in

 

[mali]

If they haven't found the exact problem yet, redesigning the board might be premature.

Thats what I would do at least, being an excellent armchair expert
The problem might never be found, so you'd have to use different hardware and try again. Thinking ahead might save some two months cycles.

 

[Monk]

If they haven't found the exact problem yet, redesigning the board might be premature.

Thats what I would do at least, being an excellent armchair expert
The problem might never be found, so you'd have to use different hardware and try again. Thinking ahead might save some two months cycles.

And designing a revised board now doesn't mean that they have to produce it if the problem is deemed to be software after all - a pre-emptive approach rather than waiting till the last minute and THEN starting to work on a fix. Work the problem from both directions so that "the final solution" is more surely found at best speed. I'm not sure MWeston would care for the extra work and pressure, but I like your thinking! Sortof. I like your thinking IF they think there's a different hardware approach that would run more smoothly (it can be tricky fixing a problem if you don't know what the problem IS). Sage Armchair Expert advice!

It'd be nice to know, but that's probably what most of the OP team is thinking right now too. Well, those that are awake. Oh - how convenient a time machine would be!

 

[Xtreme2k]

Nope. You never know unless you can try it... I already ordered a few of those small sticks.

Did you take a look at this driver? It seems to do working WLAN with SDIO on a TI1271. http://www.jorjin.com.tw/Support_Download.htm

 

[Polossatik]

Frankly, if wifi works, aldo slowly, and you add a dongle that actually works fast then I see no reason to return it, even if the internal wifi is never "fixed".
The general tone of this thread seams to be in the same line, of course this are not the 1000 pre-orders as such.

Just a proposal:
I think it would be quite possible to send out an email (later when things like the dongle are known to work) to ask to the first 1000

a ) are you ok with getting a Pandora + dongle explicitly stating the internal wifi might never work at full speed (and maybe even no promises about an hardware swap for this - but that's the teams call), if yes fill email and order nr here., if not then your order will be moved up to the next 3000 batch which might have a clearer picture about if it's "fixable" or not and the conditions for the (if applicable) later swap of the hw.

b ) after 2 weeks (tm) check how many of the first 1000 did not reply or said "i'll wait" (let's say 300) send out the same email to the next 300 inline.

This gives 2 things:
* people who don't care about the wifi beeing slow can get their pandora and might help in finding a solution. They know it will be "as is" (for the wifi).
* of the 1000 people let's say 300 who do care don't get something shipped out they need to ship back and may even cancel if this is a real "dealbreaker" for them (if it would be clear it cannot be fixed in current hw config or if the redesign would take long - which I would suspect it will)
* things about the "Q" place, which seams a hot topic are solved seen you keep your order place but spread over 2 "mini batches".