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


Edit 2:  If it's an interference issue than this is really a goof-up.  I think it was Mr. Weston's initial idea to tape it to the back of the lcd screen in the first place (?) and this was presumably to enhance the signal and not hinder it...BUT as we all know now, it took the OP Team years to realize that the (magnetic?) tape was masking the signal as well.   :unsure:
I kinda understand the OP team. Wireless signals can be so mysterious at times, almost like black magic. ;)

I mean, even Apple got it wrong with iPhone 4.
Good link and good point there, Yoyobuae.  I remember hearing about Apple's Iphone 4 problems...and this is coming from a billion dollar company with more R&D money that you can shake a stick at.  Plus they make their products and Texas Instruments wouldn't even give the OP Team all of the driver details for our wifi-drivers (calibration?) and such...so what can ya do (if you're ED, Craig, Notaz, Skeezix, etc)???  Still, any improvements and tinkering with what we CAN solve on our own would be great IMO.
 
The antenna, in this case, is a track on a PCB, not just a length of cable.  The tape holds the PCB in place.

- Neelix
Well I've seen the PCB (in pictures) - see Ingoreis' post (#181 on this thread), so I sort of get that the start of the trace is coming off of the wifi chip...but I'm not sure why taping the antenna to the back of the lcd screen is holding the pcb in place...I must be missing something.  Maybe you're talking about a different piece of tape?
 
There's a wire which runs from the Pandora PCB, through the hinge, up to another PCB behind the LCD. That PCB has the antenna on it. It is held in place with some metalic tape.
 
There's a wire which runs from the Pandora PCB, through the hinge, up to another PCB behind the LCD. That PCB has the antenna on it. It is held in place with some metalic tape.
Okay, thanks for that clarification as well, WizardStan.  Man, I really should have watched that video again of ED (or Link?) putting that part of the Pandora together...referring from memory is failing me here.  Duhhhhh...   <_<  For some reason I thought the antenna came right off of the main PCB (through the hinge and then was strung and taped to the back of the lcd.  So that is where a smaller pcb is taped...huhhh...


Still, it'd be interesting to see the exact same power draws with LCD on and the smaller pcb and antenna in it's normal place versus having it strung away from the screen.  Has anyone tried this approach yet I wonder?

Ha ha!  Thanks for that - that's the thread I remember...  Time for me to reread it I suppose.   ;)

Edit 1:  Okay, (after re-reading that thread) it looks like Skeezix tested just that theory - having the antenna and small pcb outside of the case and no noticeable improvements were found.  Just might be a power draw and subsequent wifi speed "down-throttling" at work here after all... sheeesh.  Since my Pandora arrived from Link this year and after ED had instructed GC to properly affix the aluminum tape (last Oct/Nov), then I probably can't improve any interference issues.  Must look at power usages then... Well, at least looking back has convinced me that I really need _wb_'s System Monitor program.  So that's something!   :)  


>>>-------Time for some more testing and analysis------->
 
Last edited by a moderator:
1) a ) I tested again, ensured that the Pandora is the only device on my wireless network.

The results were bad as usual. 30-60 KB/s, either staying that disappointing, or getting to a stall somewhen.

b ) Btw, the WiFi channel distribution in my urban neighborhood:

  • ch1 (3 networks, -90 - -65)
  • ch6 (4 networks, -95 - -75)
  • ch8 (1, only my network, -45, my amateurish thought: "Using the gap" within channel 6 and 11, preferring "lighter collisions" with 8 networks' partially overlapping frequencies, instead of heavier collisions with less networks (3-4) but fully overlapping frequencies)
  • ch11 (4 networks, -92 - -82)
porg-wifi-neighbors.png

2) Meanwhile my http://webserver-wired-100mbit.local/bigfile -O /dev/null gives me steady 2.4 MB/s.

d ) wget http://webserver-on-same-wireless-54mbit-network.local/bigfile -O /dev/null gives me stable ~1 MB/s.

3) Power usage of OpenPandora's internal WiFi versus USB WiFi adpapter:

  • While active transmissions, the internal WiFi (966 mW) and USB WiFi (900 mW) consume about the same power.
  • In idle mode the internal WiFi (30 mW) consumes considerably less, as for external WiFi, both the idle USB-host (160 mW) and the idle USB-WiFi (240 mW) consume more, giving a total of 400 mW for idle USB WiFi.
  • While idle operation on the USB WiFi, occasionally there is an increase of 300 mW lasting only very shortly (data link layer, I guess).
Note: How did I get these numbers? I ran sysinfo. Assessed my Pandora's base consumption — 700 mW from using: CPU @ 800 Mhz OPP5, BT/USB/WiFi all off, screen on low brightness, power-LED with brightness 32/256, 2 SDs inside, XFCE window manager, Terminal window (for wget tests), sysinfo showing power tab) — and then created the other situations (enabling internal-WiFi/USB-host, plugging adapter in/out, running WiFi idle/active), and calculated the difference to the previous situation, and considered it as the consumption for this device or activity.

4) Practically, the issue is solved for me. But for my hacker, tinkering & community morale, it is yet not solved! Our current status:

  • I think a mis-calibration is unlikely to be the root for the problem, as alternative calibration profile loading from users porg, klapse & Galaxis brought no significant improvements. Anyone still believing, the issue could be here?
  • Power problems of the Pandora (hi-speed WiFi drains quite some battery, almost 1W), giving the WiFi chip not enough power, seem logical at first, but not so much at the second glance, as USB WiFi drains about the same power (900 mW) but works at high speed without problems! Could only be true if the power line to the WiFi chip delivers less current (I remember something about wrong chosen resistors in very early Pandoras…) than the one going to the USB-host-port. Ideas?
  • If people experience good network speeds with the lid closed, they could misinterpret the WiFi issue as a chip-power-shortage issue (thought: "LCD backlight drains a lot of power, which if off, is then available to the "power-hungry" WiFi chip), which in fact could be rather an interference issue — LCD signals interfering with the mini-PCB, or the connection cable between the main Pandora PCB and the mini PCB.
  • More realistic to me seems a mix of an antenna/interference issue (loose contact, radio-blocking metal tape, interference with the LCD) in combination with the power-management to cut down aggressively, due to the bad reception, being the fault of the antenna/interference issues, and possibly slightly responsible: an unoptimized WiFi chip calibration.
These are my last guesses. I am running out of ideas. Any thoughts?
 
Last edited by a moderator:
I guess the Pandora has the worst WiFi Antenna of the Galaxy. :D

However, my Pandora wifi has good and bad days but since the last Firmware Update, it seems to only has very bad days. 3-20 kb/s right in front of my WiFi Router (!!!) is a "little" bit disappointing. I've tested various Wlan channels over my FritzBox but nothing helped, not even Pandora Wifi powersave off.

Something must be really, really wrong with the Wifi of the Pandora, may it be the Firmware, the drivers, theCip or like mentioned just the Antenna, however, I hope that if there will ever be a "Pandora 2", the Devs pay more attention to good WiFi, even if they have to put an Broomstick of an Antenna onto that Thing!  -_-
 
I think that you've done precise and exhaustive research there, Porg.  Deductive reasoning that Sherlock Holmes would be proud of.  Maybe it is as Fusion_Power says...in the end, the Pandora may have the worst wifi antenna (reception) in the Galaxy.  I can't fault any of your steps.  I've done minor tests, but mostly I've bounced ideas off of other people and tried to figure things out, but I haven't busted open my case yet - I don't want to do that until something else is wrong...(knock on chinese plastic)....

I'm like you (I think), I like to know things about the devices that I own and I enjoy tinkering with them and getting the most out of them.  It's a labor of love to improve the things that we hold dear.  Also, I just like mysteries and problem-solving...one day we'll get to the bottom of this wifi dilemma.  Until then?  I might just pick up a wifi usb stick.  I'll "pull it out" to save battery life when I'm not needing wifi access.  Your findings seem to suggest that under heavy usage, the wifi stick you use consumes about the same power as the internal wifi anyway, but with a much better signal and range.  I guess that I'll have to go down that road.

Bottomline is that we don't know why some people claim to have decent wifi and some don't.  It's probably subjective and up to personal usage(s) and maybe due to environmental conditions and thicker walls...and router channels and wifi traffic and, and, and...my head is spinning.  I wasn't about to ask Link to test numerous units during the PAX East conference...because one of those was probably mine... :D ...but maybe next time a bunch of Pandora users gets together in one place, it'd be a great socially Pandorific wifi test to see how each unit does under the same (environmental) conditions - provided the units are the same - similar models produced at GC, after last October (properly affixed aluminum tape)...  That would be an interesting experiment to observe.

Onwards and upwards.  This machine IS getting better though...and I realized it when Link had them in stock again, last December and the price was dropped (Christmas sale)...I had to finally pounce after I got my tax return back - if not now, when (I said)?  Plus, we need to sell a bunch of these units if there's going to be a successor to the P1...so I'm onboard.  Let's improve this thing as good as we can.   :)
 
I think that you've done precise and exhaustive research there, Porg.  Deductive reasoning that Sherlock Holmes would be proud of.  until something else is wrong...(knock on chinese plastic)....
Thanks for the flowers!

…until something else is wrong...(knock on chinese plastic)....
 

And after all that charming, you even make me *LOL*. Thanks! :)
 
No worries there, Porg.  It's just good to see people pulling together to try and solve something, even if it's unattainable in the end.  I've been reading these forums for years, and I'm almost bursting at the seems to properly express my enthusiasm now (that I finally have my Pandora).  The Pandora community...Gotta love it!  Beats the angry sports fans on Facebook any day.   :lol:   Good to be amongst intelligent, reasonable people...working towards a common goal.
 
I guess the Pandora has the worst WiFi Antenna of the Galaxy. :D
It's not the antenna.

We tried using a professional external antenna connected to the PCB, and it didn't change much.

However, we know that switching off the display (or even the content of the display) does affect Wifi, so it might be an interference to the Wifi Module...
 
I guess the Pandora has the worst WiFi Antenna of the Galaxy. :D
It's not the antenna.

We tried using a professional external antenna connected to the PCB, and it didn't change much.

However, we know that switching off the display (or even the content of the display) does affect Wifi, so it might be an interference to the Wifi Module...
Dooohhh, so many twists and turns...LOL.  I keep flip-flopping between interference and power (drainage)...Interesting timing too - I just ordered a wifi stick (and signed PCB from your shop) - you convinced me!   :)
 
  • Power problems of the Pandora (hi-speed WiFi drains quite some battery, almost 1W), giving the WiFi chip not enough power, seem logical at first, but not so much at the second glance, as USB WiFi drains about the same power (900 mW) but works at high speed without problems! Could only be true if the power line to the WiFi chip delivers less current (I remember something about wrong chosen resistors in very early Pandoras…) than the one going to the USB-host-port. Ideas?
The wifi module is connected directly to the battery, so if that is true it's the chip's design error.

  • If people experience good network speeds with the lid closed, they could misinterpret the WiFi issue as a chip-power-shortage issue (thought: "LCD backlight drains a lot of power, which if off, is then available to the "power-hungry" WiFi chip), which in fact could be rather an interference issue — LCD signals interfering with the mini-PCB, or the connection cable between the main Pandora PCB and the mini PCB.
If that was the case then changing backlight brightness would affect it (backlight uses a lot of power) and it doesn't, PowerGod confirmed that too.

What could cause interference is the fact that LCD cable is twisted around the wifi antenna (see ED's LCD cable replacement video), and as LCD cable carries 24 lines of data bits switching at (IIRC) 36MHz + some control signals, these could be causing some unwanted effect on radio magic inside of antenna cable.

Another possible issue is that wifi module is placed too close to OMAP on the board, if you look at phone teardowns, wifi modules are usually placed further away from the main SoC, sometimes on dedicated mini-boards and shield cages. The SoC has many high speed clock signals inside of it (for CPU/GPU/LCD/whatever), and clock signals are more or less a form of an alternating current, and where is alternating current there is always at least some radio waves produced. The LCD cable could be acting as an antenna to radiate those unwanted radio waves, and this is what is likely interfering with wifi, and that's why turning off LCD completely by closing lid (which turns off most signals on LCD cable) is helping wifi.
 
Last edited by a moderator:
^^^ Oh that's right (I've seen that video)...in order to speed up the assembly process (I'm guessing), the lcd cable gets rolled and the the wifi cable/antenna gets threaded through the hinge and inside of the rolled lcd cable....interesting.  That also could potentially cause some interference.  Why wouldn't it?  The only weird thing is that Skeezix said (in the older wifi thread, about 7 months ago)  that he had the antenna and smaller pcb outside of the case (i.e. away from the screen) and he didn't notice any improvements in the wifi data flow.  I assumed that he didn't have the antenna wrapped inside of the lcd cable, but I wonder now...

The wifi chip being closer to the central part of the main pcb (and cpu) is also interesting.  Potential interference there as well, like you said.  That hadn't occurred to me, but looking at the photos I wonder why the trace wasn't drawn outwards towards the edge of the pcb and the antenna then drawn off of the edge and just looped around the "circumference" of the screen?  I'm guessing that it was Mr. Weston's idea and not TI's to draw it up like that?  My knowledge of the original (pcb) designs are a bit vague.

Edit 1:  What I mean to say is, why wouldn't the wifi chip itself be moved away from the other noisier components on the pcb, just drawing the trace out from the battery towards the chip?  (too late I guess for that) The chip and antenna should be isolated from any potential noise/signals (IMO).  Would drawing a direct current from the battery also affect the wifi chip?  Is there a fuse or capacitor in between the power supply?  Maybe we can't fix this generation of Pandoras but learning from past mistakes is a worthwhile goal as well....especially if us fans can convince you that a P2 is a necessity.   ;)  

Looks like Askarus (or anyone that routinely assembles Pandoras) could do some interesting testing anyway.  For the betterment of future Pandora 1 customers.
 
Last edited by a moderator:
^^^ Oh that's right (I've seen that video)...in order to speed up the assembly process (I'm guessing), the lcd cable gets rolled and the the wifi cable/antenna gets threaded through the hinge and inside of the rolled lcd cable....interesting.  That also could potentially cause some interference.  Why wouldn't it?  The only weird thing is that Skeezix said (in the older wifi thread, about 7 months ago)  that he had the antenna and smaller pcb outside of the case (i.e. away from the screen) and he didn't notice any improvements in the wifi data flow.  I assumed that he didn't have the antenna wrapped inside of the lcd cable, but I wonder now...
 I don't think that's the issue.

The Wifi cable is fully shielded, so no interference should go in there.

Also, as mentioned, and external antenna doesn't improve it, you can also use the same antenna and cable but NOT put it through the LCD cable (by running the Pandora with an opened case), and it also doesn't change anything.

The wifi chip being closer to the central part of the main pcb (and cpu) is also interesting.  Potential interference there as well, like you said.  That hadn't occurred to me, but looking at the photos I wonder why the trace wasn't drawn outwards towards the edge of the pcb and the antenna then drawn off of the edge and just looped around the "circumference" of the screen?  I'm guessing that it was Mr. Weston's idea and not TI's to draw it up like that?  My knowledge of the original (pcb) designs are a bit vague.
That's what we currently suspect. Both Michael and Jorjin were experimenting with Wifi, and AFAIK, when the module was connected external, it worked better.
 
That's interesting....

So is there any way a shield cage could be added around the wifi chip? Could the answer be as simple as a tin-foil hat? ;)

- Neelix
 
So is there any way a shield cage could be added around the wifi chip? Could the answer be as simple as a tin-foil hat? ;)
Hey yeah....like an RF shield of some sort...that'd be a great test to try!
 Not if it shorts something on the board and fries your unit...    you'd want to be *very* careful trying something like that.

- Neelix

Edit: added quote for clarity due to new page
 
Last edited by a moderator:
I tried that already, but it didn't change anything.

It's hard to put some shield around the Wifi chip down to the ground, so maybe I didn't get it fully down.
 
Back
Top