Wifi instability on RPi 0.3-alpha

[Stwert]

Okay, I think I may have found the root of the problem.
It was niggling in the back of my mind that something wasn't right, and I did a quick google search and confirmed that the Red power light should not be flickering. On my pi it blinks off and on frequently. This seems to indicate a drop in voltage.

I tried switching the power supply to an official Raspberry Pi PSU (I was using a Samsing charger) but it's still doing it, particularly when the Edimax dongle is plugged in, I think. I will try to do some more experiments with different cables, but could this be a fault in the Pi itself? I'd appreciate any further thoughts.

[Stwert]

Okay, I seem to have found a more stable power supply using an Apple Charger, at least the red power light does not flicker or blink.

And I seem to be able to get an IP for my wireless dongle more reliably when the ethernet is plugged in. But, whenever I turn off the pi, unplug the ethernet and power back up, I can no longer connect, even after waiting several minutes for an IP to be assigned, since I've read it can take a while (though I'm using a static IP, so I don't see why it should).

Thinking I might just wait for the beta and see if anything is different. This is quite frustrating.

[cmh714]

try to test with an ethernet loopback adapter and you may find it work better. that was my experience with Volumio when I was having issues. AND I still use it with Rune and never have that issue.

YMMV

[bleu34]

Thanks for the trick with the Ethernet loopback adaptor (I built quickly one by cutting an ethernet cable and making all connections).
It improved the wifi stability on my B+

[Stwert]

So I had to look up what an ethernet loopback adapter was... apparently "a loopback adapter is useful for tricking a network card into believing it is plugged into a network."

I can definitely try making one with a spare ethernet cable, but I'm wondering why this might be necessary? Surely the necessity of faking a plugged-in cable means there's a software bug somewhere?

[cmh714]

Well I cant speak for a bug....all I know is many, many years ago some Gear I was working with, I think it was Cisco or maybe BlueCoat had one in the box and I read that it is sometimes needed so I just gave it a try when I first started playing with Volumio and it worked.

And yes, it basically cheats the network by short circuiting the send/receive so it appears like its actually plugged in.

Another thing I do, is after 5 minutes I disable the ethernet altogether, not sure it helps me much, but I try everything to make the music better. The 5 minutes allows me to use the ethernet if for some reason my wifi wont come up. I have that added to my rc.local script.

[Stwert]

Did some more playing around this weekend and I'm finding more strange behaviour. With both wireless and ethernet plugged in, Rune UI shows that I have an IP for both, green check marks. If I ssh into the pi through the eth0 IP, sure enough ifconfig shows that I have an IP for my wireless adapter. However, if I try to access the Pi through the wireless IP, I can't connect at all, with ethernet plugged in or not.
I quickly did a test with a Pi B and with a different wireless adapter (same SD card and OS) and I had pretty much the same problem. So I'm feeling really stuck. I don't know what error logs to be looking through to focus in on the problem.

I appreciate the help so far and future help anyone can offer. I'm hopeful that 0.3-beta will somehow fix the problem, but it doesn't hurt to keep pursuing the issue; I learn new things all the time when I'm troubleshooting.

PS I did construct an ethernet loopback adaptor, but I can't really utilize it as I don't have a spare keyboard and monitor nor any HDMI cables to plug directly into the pi, and without ethernet plugged in, I can't access it at all. I can't access through the wireless with real ethernet or "fake" ethernet plugged in.

[johnbanks]

@Stwert

Are you really sure it’s not a power problem?

My working RPiB with Wolfson and Edimax adapter draws 540mA from the power supply and 430mA with the Edimax unplugged.

That’s quite a current draw thru’ a USB cable that may look hefty but typically is a lot of plastic with not a lot of copper. The resistance of the lead causes a volt drop which leaves the RPI short.

If you can borrow a voltmeter then carry out the test – see link above in this thread.

Or you could try a separate powered USB hub and plug the Edimax into that – avoid the cheap ones as they don’t work as I know to my cost.

My approach is probably OTT. I use a USB A female to Micro USB B male adaptor similar to It saves wear & tear on the RPi power socket and is easier to grab hold of. And I also make my own USB power leads with lightweight speaker flex with plenty of copper. Wikipedia shows the connections for USB - only ground/earth and the 5 volt lines are needed.

The power supply I use is currently out of stock However you may find the product description is informative.

Hope this helps

[Stwert]

So I finally got a hold of a multimeter and checked the voltage across PP1 and PP3 on the B+ (that seems to be where to do it?). The voltage read between 4.88 and 4.93 when it was running. Do you think that is too low?

[Stwert]

Okay, I'm at least getting audio out of my IQaudIO Pi-Dac+, thanks to a kernal update that some helpful people were sharing. So I'm listening to some nice music through a veeerrry long headphone cable into my receiver. But I'd really like to get the wifi working so I can connect the line outs to my receiver directly.

The wifi still doesn't seem to connect when the pi boots up, but I did notice that after hitting the Refresh Interfaces button in the UI (while still connected with ethernet), then the wifi seemed to connect. Obviously I couldn't do that through the UI with the ethernet unplugged, so I'm wondering if I could add a bootup script that would do the "refresh interfaces" command (whatever that is) after a bit of a time delay? Any ideas about that?