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


Maybe we're onto something though....I've routinely seen electronics breakdowns (from various computer techy websites) and I've seen the shielding separating lcd screens from other components (within a very tight space)...like the Nexus 7 tablet.  They do that for legal reasons sure (FCC regs I guess), but it also reduces a lot of signal/interference traffic.  Maybe the OP Team could play around with it a bit.  I'm too afraid to risk it yet.  My baby Panda is too young to die....

Edit 1:  From ED: "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."

Now I'm firmly back into the "wifi interference" camp again.  I'm flip-flopping like a salmon over here....or a US politician....
 
Last edited by a moderator:
Might it be worth talking to the GTA04 guys about the issue? They would have experience with radios sharing board space with an SOC no?

- Neelix
 
The power savings trick does help, but for now for me the best fix was a ten dollar Edimax USB Wifi nub.  Greatly increased speeds.
 
The USB WiFi adapter works seamlessly, but in idle mode, it consumes considerably more power (example: Compo4all, while 3min game WiFi is idle, then only busy for 1-3 secs while uploading score, then idle again for a long time — during all that idle time lot more energy is wasted than with internal WiFi).

For a longer battery runtime it is still worth for me if the internal WiFi could get a fix.
 
Notaz: […] turning off LCD completely by closing lid […] turns off most signals on LCD cable […] is helping wifi.
Question 1: Is there a reliable way to run the Pandora, while having the LCD completely inactive*, and the WiFi active?

* Not merely the backlight off, but really completely off! Both the signal cable transmitting nothing to the LCD, and the LCD's circuits completely off.

Does lid-closing do this? Or can/must I toggle the LCD off and on with a certain keystroke or a commandline which I type in blindly?

This is a test I could do.

As the screen is off, I could either log with sysinfo and watch the results later as a log/plot (can writing to SD cause interferences?),

or watch at the speed values server-side (the least intrusive test method, the Pandora does only WiFi data transmission, nothing else).

Question 2: The latest suspicion is that the WiFi-chip and the System-On-a-Chip (SOC) are too close together.

In that case my "LCD-off + WiFi on" test won't be of much help, would it be? I may nevertheless do it. Not so much effort.
 
Last edited by a moderator:
porg if you have a TV-out cable and close the lid you can watch sysinfo (not only the log afterwards) and LCD should be turned off.

If you don't have one close the lid and check the logfil.

As far as I know closing the lid turns it off.
 
you can also use the Pandora with ssh if it can keep the connection up
 
I have three Pandora(s) CC, Rebirth and 1Ghz and never had much issues with the wifi disconnecting and such on any of them.. I do only have maybe only five WIfi Access Points around me that are not mine.
 
Last edited by a moderator:
Notaz: […] turning off LCD completely by closing lid […] turns off most signals on LCD cable […] is helping wifi.
 Question 1: Is there a reliable way to run the Pandora, while having the LCD completely inactive*, and the WiFi active?


* Not merely the backlight off, but really completely off! Both the signal cable transmitting nothing to the LCD, and the LCD's circuits completely off.


Does lid-closing do this? Or can/must I toggle the LCD off and on with a certain keystroke or a commandline which I type in blindly?
Just closing the lid fully turns off LCD too.

Question 2: The latest suspicion is that the WiFi-chip and the System-On-a-Chip (SOC) are too close together.


In that case my "LCD-off + WiFi on" test won't be of much help, would it be? I may nevertheless do it. Not so much effort.
When LCD is turned off good portion of SoC (the display controller) and it's clocks/PLLs are turned off too, so it makes a difference.
 
Last edited by a moderator:
If the display was enabled but the LCD cable was disconnected the display controller would still be running yes?

If so, that would suggest to me that trying to reproduce this behaviour without an LCD connected would show whether or not the display controller is actually responsible for the interference.

- Neelix
 
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 normally have a good and steady wifi connection. It is not very fast though.


I noticed, that some pnds, mainly stuff using gles, made the wifi disconnect.


Posted that earlier already.


One example is falling up.


Maybe the wifi really doesn't like what the graphics department in the Pandora is doing.
 
I only have problems when the connection drops below 50%. Then I can not send any posts one the forum for example. Turning off the power-saving  in the terminal, and I can send the post. So this is definately related to the power-saving, and not some kind of Wifi-Voodoo ... using a Fritz-Router by the way.

If I'm closer to the router and have a better connection everything is fine with the standard settings.

Apart from that I rarely had any problems, and I have been on several conferences, using all sorts of unknown Wifi routers at Universities, institutes and Hotels. Only if the connection is weak I have to turn off the power-saving.

I think it was the same with my Rebirth Pandora (upgraded to 1GHz), but back then I did not try the "sudo iwconfig wlan0 power off rate 54M"-trick when the connection was bad.

Anyway, a connection saying ~50% does not seem that bad to me, and yet it is bad enough to cause a lot of problems on the Pandora with the current settings.

I think it would be already a help to have a button to turn off the power-saving in the menu where you can toggle Wifi, USB etc.?
 
Ran some tests again, my device again being the only device on my WiFi network, in my urban environment surrounded by 10-15 neighboring networks.

At the beginning I had wonderful 500-600 KB/s for half an hour or so with powersaving off!

Thought, that's a typical experimental effect: When you want to test for a negative phenomena, it suddenly is absent, while the positive phenomena is surprisingly present.

But than after this ~0.5 hour, the usual happened. On every new connection attempt ~60 KB/s for a few seconds, then getting to a complete stall.

1) En/dis-abling the LCD (by openening/closing the lid or running the related command line) changed nothing! (Also not while I had that wonderfull 600 KB/s! My WAN's max speed is ~810 KB/s, the mirror server quite reliably always provides this downlink speed, it did so on my MacBook Pro minutes before I started my Pandora tests, and also afterwards). UPDATE: I also tried a delayed triggering of wget via sleep 10; # closing lid within the 10 sec pause; wget …  but it also did not help.

2) Disabling powersaving helped, again. It then ran around 500-700 KB/s. Also after reactivating powersaving it still worked fine. (Again the treshold theory: Once the treshold is overcome it works)

3) A new thought, which came to my mind: Closing the lid is quite a vibration ("clack"). Maybe this is changing the positioning/contacts of Pandora parts (cables, mini-PCB) each time, and thus varies the WiFi experience? Or does the WiFi experience variation rather come from the changing environment conditions, how many WiFi neighbors actively networking.

Besides this, I am quite at the end of my wisdom. The only test remaining for me is at a rural location with a web server connected to a WiFi access point. Dunno, when I'll have time for this.
 
Last edited by a moderator:
I've not looked at the PCB, but it may help to add some decoupling of the power (and other lines) going to the wifi module. A USB module will already have this built in. Ideally a small series resistance, and about 100-200pF 0805 capacitor between power/ground pins (without any extra wires) will act as near enough a short circuit at GHz frequencies.

A screening can over the wifi is probably a waste of time, I never found these had much of a reliable effect other than to isolate RF circuits from proximity effects of hands near the CB etc. Better to absorb the interfering signal with some resistance.
 
^ An interesting query.  I was basically wondering the same thing (back on post #217 of this thread).  Anyway, for now I'm at the limits on what I'm willing to do.  I've only had my Giga-Pandy for a month and I'm not brave enough to tear it down and experiment.  I'm just going to turn off power-saving mode and use a usb-wifi-stick when I need that extra signal strength.  My wifi-stick should arrive from ED shortly...and we at least know that these options help.
 
I found another interesting thing... I had always the issue of disconnection when also my PC was connected to wifi AND using a lot of bandwith, and I was "resolving" this using the PC with just an ethernet cable...

now something happened, I was in the other room with the Pandora, downloading at 300 kb/s (with power off) for about half an hour... ok, all normal... NO, WAIT !! My PC was connected through wifi this time, downloading with P2P clients !! So why the Pandora was still connected ?! ... and while thinking about this I was walking toward my pc, I gave a look at the Panda and was disconnected... while sitting near the PC I tried reconnecting it different times but it was disconnecting almost immediately...

...so I went to the other room and tried again connecting, and ... yeah, stable and fast connection  :blink: tried this dance for three times and now I'm almost sure that the problem is not with the router (because I was thinking it wasn't able to manage 2 devices using full bandwith) but with the Pandora wifi itself... seems it doesn't like interferences so much

At least now I know that I can actually use both devices with wifi on, but at a certain distance !!   ^_^
 
Last edited by a moderator:
To me the Pandora is also "the weak wifi kid on the block", which cannot "make itself heard" among a surrounding "loud" crowd (interference). A shy kid!

But with the help of a USB WiFi adapter, its ego can be really boosted. Sadly though, stronger egos require more energy (battery).  ^_^
 
Performed some more tests under my recent suspicion:

I have a totally new idea concerning my WiFi and charging issues. What has NOT changed throughout all the 3 Pandoras (2 originals, 1 rebirth) I have had is ……… the battery!!! Maybe this is causing both my issues with charging and potentially WiFi, as the WiFi chip is directly connected to the battery, according to notaz. It is likely that at the OpenPandora workshop the testing was done with their spare batteries, and therefore the issues were unnoticed? I will ask ED & Askarus directly, and if I get infos, follow up here.
Booted my Pandora without a battery, just connected to the original OpenPandora charger.

Result: WiFi has the same behavior (mostly faulty, sometimes amazingly speedy).

Details: Now since Sysinfo 0.6 having the ability to toggle WiFi, and to toggle its powersaving, giving you instant control & feedback, the correlation between powersaving and speed felt even more obvious.

Again: Not so obvious that without powersaving WiFi was working 100% and without powersaving not, but quite tendentially: With powersaving working 0-10% (0-60 KB/sec, stalling soon, but rarely if "critical treshhold" (see details in my previous posts) is reached sometimes keeping almost max WAN speed of 650 KB/sec steadily), without powersaving working 30-60% (not always maximum speed, but at least better than with powersaving at the very recent moment, i.e. getting 30 KB/sec instead of 0, or 60 KB/sec instead of <10 KB/sec, and often if it reaches high speed, it then keeps it, if switching to powersaving mode, sometimes keeps it up, sometimes drops down).

Conclusion / Follow-Up: I have almost nothing left than to believe that I have had enormous bad luck by getting 3 Pandoras all with HW faults (WiFi & charging).

(Except if the battery is needed in the circuitry to "smooth the current" (or something like that) before it reaches the WiFi chip, but I doubt that. Confirmation anyone?)
 
Last edited by a moderator:
Have you tried switching your Router to a different Wifi-channel already? If the Pandora is very sensitive to interference this could be your problem. I set my Router to a channel which is not occupied by another access-point. But probably also one of those phones without a cable could cause problems, so it's definately worth a try.

Also, what is the reported connection quality? I start to get problems only if it falls below 50%.
 
Back
Top