Pandora Wi-Fi On 350kilobytes/s

[paddy]

Tom has a very good point in that it looks like it IS the hardware to fault and yet OP are doing a quick
dirty fix to at least the first 1000 units ,and that quick fix might also be bottlenecking other operations
that OP may or may not retest.In light of the hardware issue i don't think it's a good idea to ship any
Pandoras untill everything has been tested and wifi has been optimised,even with this apparent fix for
the Pandora,if the hardware is to fault then the other 3000 units will also be faulty minus a resistor
unless everything hardware wise has been checked properly and wifi has been optimised in good time ,will
OP replace 4000 boards ? i can understand getting 1000 and replace when the time is right but 4000 is bad.

Zotty also has a very good point ,if in doubt about net speeds ,then use
home network speeds ,transfer and receive a large file over a home wirless
network and you will get better numbers.

 

[SiegfriedCroes]

Oke didn't read the whole topic but just a question: When you say 350KB/s speed, do you mean when downloading something you get that speed or what does it mean? Where can I see how fast my internet is?

 

[renwils]

As others have said, if this is without much packet loss, it should even be fine for streaming video locally. When I transcode video on the fly to my PS3, I minimize the bandwith to 450 kB/s for 720p video without much visible compression. For 480p, 350kB/s should be fine.

 

[fusion_power]

One tiny resistor caused all the mess? That's it? :blink:
So the Software-work that notaz did was just fine. It was indeed the hardware that caused the trouble. Hm, I'm sure, MWeston did exaclty the specs of the Chip-Company, could it be, that they gave him a wrong spec for the transistor? So at the end it WAS the fault of the Chip-Manufacturer?
And does this also explain the issues, other Companies have with this WiFi Unit?

Then we need the cases - and assembly will finally start.

Uh, please ship the cases as soon as possible, it sounds like you will wait again (for the boards), before the Chinese Company ship all the cases. BAD idea, please learn from the past.

 

[WizardStan]

Oke didn't read the whole topic but just a question: When you say 350KB/s speed, do you mean when downloading something you get that speed or what does it mean? Where can I see how fast my internet is?

From EvilDragon just a few posts above yours:

Well, Michael did test a download from a normal internet site, not internal network.
I don't know how fast his internet connection is, but most probably not 54Mbps.

So the 350KB/s was MWeston actually downloading a file from somewhere off the internet. In other words, that's a real world figure and likely limited by internet traffic, not the chip. This means that local transfers MAY be faster. I'm guessing.

 

[Caine]

Then we need the cases - and assembly will finally start.

They STILL haven't shipped those? What's the hold up this time?

 

[Exophase]

Not mentioning that if you need to decrease the value you can simply solder one "on top" of the existing one - no desoldering needed and damn fast to do (under one minute even counting taking the resistor and board and putting it back) by someone who is used to soldering smd's.

Even desoldering the existing resistor first will only add a little more time if you use a decent smd workstation.

Manually changing one resistor should be timewise faster then the assembly process, so ONE person soldering instead of assembling the first 2 day's or so.

Yes of course in my post it was under the assumption that he was replacing the resistor. I don't know what kind of tools he has available. But I think that it might take longer to do them when you're doing 500 straight than if you were doing just a few. My figure was for about 10 minutes.. an assumption of less than 1 minute each and doing that for over 8 hours straight is a bit much, especially when that's supposed to be the extreme end in time! These people are not machines...

Yeah it's faster than the overall process but it's still an extra overhead, which is why I asked if more than MWeston is going to be doing it (because that's not what was implied). ED says a few minutes, going to assume it's not a big deal with the rest of the assembly process but unless they have more people it'll still add a little time overhead.

 

[LuvClub]

Why is everyone talking about good enough for internet-surfing? Hopefully we will still get the full 54mbps so that one day we could use it for other things such as say emus being able to load the larger zipped roms at a reasoable speed from a network drive. Of course I congratulate the OP team on the current progress with the wifi, but hope that they still intend to aim for the full theorertical speed less overheads.

 

[mali]

For everyone too lazy to read properly, just check the quotes thread

 

[craigix]

Not mentioning that if you need to decrease the value you can simply solder one "on top" of the existing one - no desoldering needed and damn fast to do (under one minute even counting taking the resistor and board and putting it back) by someone who is used to soldering smd's.

Even desoldering the existing resistor first will only add a little more time if you use a decent smd workstation.

Manually changing one resistor should be timewise faster then the assembly process, so ONE person soldering instead of assembling the first 2 day's or so.

Yes of course in my post it was under the assumption that he was replacing the resistor. I don't know what kind of tools he has available. But I think that it might take longer to do them when you're doing 500 straight than if you were doing just a few. My figure was for about 10 minutes.. an assumption of less than 1 minute each and doing that for over 8 hours straight is a bit much, especially when that's supposed to be the extreme end in time! These people are not machines...

Yeah it's faster than the overall process but it's still an extra overhead, which is why I asked if more than MWeston is going to be doing it (because that's not what was implied). ED says a few minutes, going to assume it's not a big deal with the rest of the assembly process but unless they have more people it'll still add a little time overhead.

But a lot more preferable to paying for 4000 new boards and 1000 wifi dongles!

 

[TrashyMG]

Ohm my god -/\/\/\- fail!


yeah bad..

 

[EvilDragon]

Sorry but I can't help wondering; after all the 'bitching' recently, why on earth test speed with something as unstable as an internet connection? While the new speed is good news, please please please do some SOLID STABLE PREDICTABLE TESTING!

Most probably because Michael doesn't have a local server at home and he is the one working on the hardware

You can't assume everybody got a local server at home.

And yes, this is just the beginning. Michael did discover that a few hours ago (he just did send the eMail to us before he went to sleep at 4am).
He will continue to work on it tomorrow and try to tweak it more.

Tom has a very good point in that it looks like it IS the hardware to fault and yet OP are doing a quick
dirty fix to at least the first 1000 units ,and that quick fix might also be bottlenecking other operations
that OP may or may not retest.

Definately not, as this one resistor is only there for the module (it only affects the curve of the 26MHz clock AFAIK).
It doesn't have anything to do with RAM, the OMAP, SDIO, etc.
And all those had already been tested since we got them working for a while.

The only bottleneck we seemed to have appeared after we got the WiFi driver to work, the rest was working on our boards since months.

 

[scottm]

This is great news! Can't believe some people are still complaining! Ship all 4000 now!

 

[Dead1nside]

Yes it's great news. However what the limiting factor is now, is still unknown. 66KB/s stable is clearly not anywhere near the realistic expectations for the chip when performing properly, so that's my concern. Just have to wait for MWeston to explain what's happened in his testing.

 

[skeezix]

Also to note .. I can't count how many gadgets I have that have these little last minbute changes on them; its almost the norm for first-batches, just you usually don't know about it

jeff

 

[paddy]

Ed ,what i am talking about is retesting all parts after the removal of the resistor ,you really ,really
do not know for sure if it will cause a problem else where ,and what do you mean ,not everyone has a
home server ,does MWeston have a pc,laptop or any other device capable of being on a network ?
i am pretty dam sure MWeston has a pc ,otherwise why have an wifi internet connection ? a simple
file transfer from and to the Pandora would give much better numbers and if it does only max out
at 350 then i see this is a major problem unless the driver again is to fault as it may not be
fully optimised seeing as there was no current testing done on the driver so far.

 

[droneb]

XD, i cant believe people saying it takes loong time to make such a replacement.

i've seen people solder wiikey modules in under 2 minutes (not to say he did 100 in one shift), and you say instaling a resistor takes more than 1 min?

you just get the smd rework to heat the resist pull up and put the new one and that should be in less than maybe 10 secs top ?? the more you take to heat it up the more you are closer to damage/unsolder neighborhood components.

 

[EvilDragon]

Ed ,what i am talking about is retesting all parts after the removal of the resistor ,you really ,really
do not know for sure if it will cause a problem else where

We really really really do know for sure. Michael designed the whole board and does certainly know what part does what.
There's a crystal that gives the module a 26MHz clock, and the curve is optimized by some parts.
Those parts have ABSOLUTELY nothing to do with any other busses on the Pandora.
On the old boards without WiFi they didn't even exist.

 

[WizardStan]

Ed ,what i am talking about is retesting all parts after the removal of the resistor ,you really ,really
do not know for sure if it will cause a problem else where ,and what do you mean ,not everyone has a
home server ,does MWeston have a pc,laptop or any other device capable of being on a network ?
i am pretty dam sure MWeston has a pc ,otherwise why have an wifi internet connection ? a simple
file transfer from and to the Pandora would give much better numbers and if it does only max out
at 350 then i see this is a major problem unless the driver again is to fault as it may not be
fully optimised seeing as there was no current testing done on the driver so far.

I think you're asking a little much right now. ED just said it was 4am when MWeston finally figured it out and did a quick test. I won't fault him for not doing a local test. Maybe he hasn't setup his PC for file sharing or anything. Let the fix get out there first, give him some time to sleep and figure it out later.

 

[TrashyMG]

From a rough calculations 350k/s is around 2.8mbps, I may be wrong.. However if I'm right, this is speedier than my internet connection. Also the Nintendo DS can only do 2mbps as well.. so I'm fine with it.