WiFi mysteries resolved once and for all — Power saving is clearly the culprit!

[_wb_]

I'm getting ~385mA with power saving disabled and ~185mA with power saving enabled. My setup is the same as yours (also Firefox running to type this post), but one difference is probably that I disabled my "power on" led because it is useless to me (I know my Pandora is on, because I never turn it off). The orange wifi led is on though. Also my cpu is clocked at 900MHz, OPP5.

 

[MarioPandio]

The annoying thing is the chip in use is iirc the same model in nokia n900 phones and they do not have problems. OPT have simply had disastrous luck here, as with many other calamities.  Porgs data is useful hopefully to attack the power saving and improve on it somehow.  

 

[notaz]

Nevertheless I would like internal WiFi power saving to work. Is an improvement likely? Could my report and data help to debug?

The performance and most other stuff related to wifi is actually controlled by the binary blob that's running inside the chip itself. TI is not providing any documentation or help on how that blob works or how to configure it. They only provided that info to Nokia to write the Linux driver for N900 that we are also using now. There is also a driver written by TI themselves that was opensourced during one of first Android phone releases, however it doesn't support normal Linux interfaces and is unsuitable to use. These 2 drivers are the only source of information about how to talk to the binary blob, but that doesn't make things easy.

Is the WiFi chip&firmware better on the 1 Ghz Pandora?

No, they are exactly the same as rebirth.

I guess what we need is to reverse-engineer the crappy proprietary closed-source wifi driver binary blob so we can 1) make the Pandora even more Open Source, and 2) fix the overly aggressive power saving.

True, but you can't easily make sense out of 192K binary blob (yes I have looked at it's disassembly), so you can easily write that off as impossible.

 

[notaz]

It seems calibrated very badly!!!

Actually this might be the case. I've read that after manufacture, every N900 (which uses the same chip) is separately calibrated using special signal generator in factory and software running on N900 itself to produce calibration file, which is then stored in device filesystem to be loaded before every use of wifi.
For pandora, the the chips themselves are precalibrated by manufacturer with calibration data stored in small eeprom inside the wifi module. When the wifi driver starts on pandora, it instructs the wifi whip to read calibration data from eeprom. On N900, it's loaded from file on device filesystem instead. On early days when we couldn't get wifi working at all (and later when we had it barely working at maybe 1K/s due to hardware issues), I had modified driver to load calibration from a file, and also had code to make calibration dumps from pandora wifi modules. What we found was:

- every wifi module's calibration data dump is different

- loading calibration data from another device produces non-working wifi

- loading calibration data from the same device it was dumped from produces working wifi, the same way as using normal driver

If anyone wants to research the calibration dump format, I could provide special driver that can dump, upload calibration data from/to the chip.

 

[danielo515]

This thread is getting very technical, what I like. Because that, I feel a bit bad about making this question, but here it goes : Disabling power saving affects the overall baterry life? I mean, with the wifi off

 

[Binky]

This thread is getting very technical, what I like. Because that, I feel a bit bad about making this question, but here it goes : Disabling power saving affects the overall baterry life? I mean, with the wifi off

With wifi disabled, the power consumed by the wifi chip should be tiny.

With the wifi setup as normal, the chip will consume a considerable amount of power, but be near-useless

With wifi powersaving off, it will gobble power like a ravenous goat and work reasonably well

 

[danielo515]

 
With wifi disabled, the power consumed by the wifi chip should be tiny.
With the wifi setup as normal, the chip will consume a considerable amount of power, but be near-useles.
With wifi powersaving off, it will gobble power like a ravenous goat and work reasonably well

If my assumption that there is a point after useles is correct, you have been very clear. Thanks.

 

[Michoko]

Thanks porg for this complete write up. It must have taken some time!

Actually I had missed the power saving stuff, despite the fact that I'm browsing those forums on a daily basis. So thank you for the reminder.

 

[porg]

@steveway:

I never claimed that I found something "excitingly unknown" concerning WiFi issues. The circumstance that the WiFi chip on the Pandora has issues and that "sudo iwconfig wlan0 power off rate 54M" may help is indeed well known and can be easily found.

And yes, be assured, I know how to search through OP resources (boards, wiki, issues tracker), and intensively did so prior all this (~8 hours). But all reports I found were very brief and/or subjective, no causal conclusions could be made. Therefore I asked you steveway for a particular well explaining source!

The novelty of my attempt was to, for the first time, provide substantiation to those claims, by empirical research & sharing that details, so that a solution gets better chances than from rather vague reports.

 

My efforts:

08h research in OP resources (boards, wiki, issue tracker)

20h data collection

10h graph plotting, commenting them, cutting/exporting them properly

06h compiling my report.

In total: 44h --> A man week of work!

 

Luckily the most people appreciated that effort (@F_Slim, @EvilDragon, @Michoko, ...) and some are already even getting motivated solving or at least further researching it (@_wb_, @bismuthdrummer, @notaz), and some contributed further interesting facts (@mindlord "from my research: routers make a difference", @MarioPandio & @notaz "N900 has same chip but works better. Each N900 WiFi chip got its very individual calibration") or just participated in a normal tone.

 

And @quartercast also honestly apologized after others made the effort & novelty of this report clear to him/her.

 

Only one guy, he knows that I mean him, could do nothing better than quickly and cynically bash. I hope he realizes and changes his tone. Otherwise, man, do not talk to me again! Spare me this. Don't bully me away from the OP community. I don't come here for such negativity. I rather prefer joint effort & collaboration! Hope you got it.

 

To all: Thanks for the fruitful discussion, hope that we can find a solution!

 

[porg]

It would be nice if the thread was a wiki page though porg

http://pandorawiki.org/Wi-Fi#Connection_Speed now contains the paragraph:

Also check this detailed report, which concludes, that disabling power saving very likely improves the situation (on the Original, Rebirth and the 1Ghz Pandora), as of 2013-01 SuperZaxxon 1.52, the power saving algorithm in the binary blob from TI and/or its calibration is so over-aggressive, that it fails in many situations.

 

[freamon]

Hi Porg,

Thanks for all your work with this. It's kind of the nature of these boards that the first responses you get to anything aren't always the most positive or the most considered.

Although, if you do end up completely solving the internal wifi's problems, then you owe me £8 for the usb adaptor I bought.

 

[porg]

freamon: LOL 

 

[Askarus]

Nice report.

Always nice to have such stuff visible.

Maybe we can improve the current situation.

Hi Porg,

Thanks for all your work with this. It's kind of the nature of these boards that the first responses you get to anything aren't always the most positive or the most considered.

Although, if you do end up completely solving the internal wifi's problems, then you owe me £8 for the usb adaptor I bought.

Does a USB Wifi adaptor work better than Pandora Wifi without power save?

 

[hailrazer]

 Otherwise, man, do not talk to me again! Spare me this. Don't bully me away from the OP community. I don't come here for such negativity. I rather prefer joint effort & collaboration! Hope you got it.

/\  This :  

Come on everyone.  Can we at least try to be civil and positive ? 

This is a community and as such don't we want Members who try to find out the reason for problems and try to help fix them ?

If you disagree or don't like a post, simply avoid the thread or state your opinion in a positive constructive manner. 

Then the community and the OpenPandora Project will continue to grow AND thrive !

 

[vandyswa]

Porg, "sudo iwconfig wlan0 power off rate 54M" is all over these forums! Its even in Dredd's signature afaik! Edit: daniel type that into your terminal ^

FWIW, this thread is where I found out about it.  I guess I wasn't googling the right keywords...

 

[quartercast]

Cool, glad my inappropriate rant helped someone

Edit: maybe every new pandora should come with a little slip of paper with that very command! Its pretty hard to remember and seemingly hard to search for.

 

[Rockthesmurf]

@porg Thank you for all the time you have invested into this problem! You have clearly conducted more testing (and more advanced testing) than anyone else who has posted (with possible exception of Notaz who, as usual, shows incredible amounts of knowledge in all areas).

@anyone negative: Seriously?

@evil dragon: I certainly wouldn't except anything negative from you, being that porg is attempting to help resolve a problem that ultimately helps the Pandora project! I hasten to add your comments have not been particularly negative, although perhaps a tiny bit dismissive at the start. I can certainly believe perhaps you didn't have the time to read his entire first post in detail of course.

@self: Who do you think you are! Stop telling everyone how to behave!

@self: Who said that?

@self: It was me, trying to keep things light hearted to avoid any nasty replies!

@self: Good move!

 

[porg]

Did anyone use this findings of this report and improve WiFi on the Pandora?

Am curious, also if it's just something yet experimental.

 

[PowerGod]

Hey, I just found this interesting thread, detailed work as always Porg

The situation with my router is more related to disconnections (when other devices are connected too) than speed...

Anyway, I also found strange things related to the speed that seemed indipendent from the power state, but more related to the application using the connection...

or maybe the way that different applications use to send/receive data can trigger somehow how the power save should work ?! :huh:

http://boards.openpandora.org/index.php/topic/11261-wi-fi-speed/#entry209836

 

[PowerGod]

It's happening something I never seen before, TeraCopy is reporting 768 KB/s copying a file from Windows to a shared samba folder on the Pandora... but I just copied files in the same way even at lunch, and yesterday, but the maxed speed was 60 KB/s like always ... so what's changed today ? I maniacally turned off power save for the entire day...

I never noticed so much differences when using power off, so after some trials months ago I just leaved it as it is in every connection...

The strange thing is that I used that "sudo iwconfig wlan0 power off rate 54M" at the start of EVERY connection today, and it was always going at 60 KB/s (except when using Transmission, in that case has always been >400 KB/s, with or without power save)... 

I don't know in what this connection differs from the others I've done today... anyway it's really stable at that speed, copied a 699 Mb file in some minutes, instead of hours... :huh: