Wireless N USB Adapter for GNU / Linux (TPE-N150USB) dropped packets

16 risposte [Ultimo contenuto]
yrk
yrk

I am a member!

Offline
Iscritto: 08/28/2012

hello all,

Ever since I've had it the thinkpenguin TPE-N150USB has been dropping between 3%-10% of wireless network packets under Trisquel (using the firmware from jxself).

I've done the troubleshooting to make sure that this isn't special to my network (happens on any network) or my router (changed router). I even (temporarily) compiled and ran the Intel wireless which usually sits dormant in my laptop and that indeed drops 0% consistently.

I'd rather have freedom than a perfect connection, so I'm living with it. But if I can have freedom _and_ a perfect connection that would be even better! Any help appreciated.

lembas
Offline
Iscritto: 05/13/2010

Which kernel are you on? Latest might be most useful or perhaps the latest backported saucy kernel.

One silly idea is to try different USB ports as they might give different amount of power and interference.

I found this interesting on zero packet loss, not that I claim to fully understand the issue. http://www.bufferbloat.net/projects/bloat/wiki/Introduction

yrk
yrk

I am a member!

Offline
Iscritto: 08/28/2012

I've tried this on every kernel release so far with the same results. So if it is a kernel issue then I haven't hit on one without the problem.

Different USB ports have been tried, as did connecting through the same USB to a different wireless device and seeing 0 packet loss.

Buffer-bloat is an interesting topic indeed! But it happens primarily in the router (and I've changed that).

mYself
Offline
Iscritto: 01/18/2012

You mean that the connection is periodically dropping, then reconnecting?

yrk
yrk

I am a member!

Offline
Iscritto: 08/28/2012

No, the wirelss connection remains throughout. You can tell that there is an issue since I can't stream video without it stuttering, for example from TWiT.tv "mplayer -aspect 16:9 http://bglive-a.bitgravity.com/twit/live/high".

And if I run "ping -c 100 something.org" I'll get proof of this with a 3%-10% packet loss report at the end.

mYself
Offline
Iscritto: 01/18/2012

Aha. I asked because I own this adapter and it newer worked quite well for me to the extent, that it caused kernel panics/system crashes. Sorry, but I cannot help you with this. But what I do recommend is to buy an internal wifi card instead if you can. This solved all my problems that I've got through years of using this wifi adapter.

yrk
yrk

I am a member!

Offline
Iscritto: 08/28/2012

Cool, do you have a specific recommendation for an internal card?

I don't mind keeping the USB one as a backup option. And while money is an issue, it is not _the_ issue. I'm just really fed-up with ethernet cables.

mYself
Offline
Iscritto: 01/18/2012

Yes, a "AR5B22" based card from eBay is the top pick nowadays. It should cost you something around 6 bucks (note that this card has bluetooth integrated, but it will not work since ath3k requires non-free firmware blobs - just ignore that). If that's too much for you, you can choose a lower-end "AR5B125" based card, that should cost around 2 bucks. The first card is 2x2:2 MIMO, while the second one just 1x1:1.

But I warn you, do not try to switch the internal card without guidance! Every laptop model have this card located in different parts, so you may found it directly accessible from under the keyboard (e.g. ASUS EeePC 1001PX), from the backside (e.g. Toshiba Satellite L40-14F), and sometimes you need to disassebly the top, or the bottom side completely in order to get an access to it (e.g. my Sony VAIO VPCZ13Z9E/X). And there are other things to pay attention to, so if you want to make the process smooth right from the first time, you will come back to this forum once you get your shiny new card to get help. You can then PM me through my profile contact form.

trisq

I am a member!

Offline
Iscritto: 09/03/2013

yrk,

I have that exact USB and running your 100 packet test to fsf.org ( ping -c 100 fsf.org ) had zero packet loss with only 2 bars of signal strength.

Chris

I am a member!

Offline
Iscritto: 04/23/2011

The AR5B22 is not a free software friendly card and I'd highly discourage it. There is no reason to get this card when there are other free software cards on the market. The main thing in getting one of these cards is verifying it is strictly an atheros 802.11N wifi card and not also including a bluetooth component. If your not willing to say no to proprietary software what makes you think your going to continue to have access to free software friendly hardware? Atheros is already reluctant to release the ath10k source code and nobody else is even thinking about it (this is for the newer 802.11ac standard).

We have lots of people streaming 1080p video with the TPE-N150USB. It's definitely not an issue with the adapter/firmware/etc. It may be a defect in the particular adapter you received or a similar type of issue related to the network configuration or USB ports/or USB controller. A newer kernel may resolve the issue or it may not. The issue may be partial support for the USB ports. Apparently there isn't full support for some USB controllers and that's causing problems with a lot of devices so despite things seeming to work with some devices other devices won't work or won't work properly.

I did a quick test a few days ago in fact as I was investigating various chipsets and for comparison I ran a quick test with the TPE-N150USB. I can say at that time I did not see any packet loss between the router we're using in the office and the adapter. This was tested from several rooms away too. I was testing against a new embedded Libre distribution we are working on and a free software friendly router.

mYself
Offline
Iscritto: 01/18/2012

Before I go further to comment on your statement, can you elaborate on what "a free software friendly card" means for you?

tct
tct

I am a member!

Offline
Iscritto: 10/23/2011

În 2014-03-18 13:45, name at domain a scris:
> Before I go further to comment on your statement, can you elaborate on
> what "a free software friendly card" means for you?

http://www.fsf.org/resources/hw/endorsement/criteria

--
Tiberiu C. Turbureanu
Președinte, Fundația Ceata
Telefon: +40-761-810-100
GPG: 8B51 53CB 354E 3049 FAE9 3260 F033 8452 4154 1967

Susții libertatea artelor și tehnologiilor?
Înscrie-te ca membru! (http://ceata.org/%C3%AEnscrieri)

mYself
Offline
Iscritto: 01/18/2012

So a hardware component working in a fully-free environment, that has free driver/firmware released under a liberal _free_software_ license is not enough to say that it's "free software friendly". Therefore beware Trisquel users, because many ThinkPenguin products does not gone through the Respects Your Freedom (RYF) certification process, thus "not free software friendly".

Also, if the non-free bluetooth part is a problem, then why the FSF is certifying such hardware as RYF?

Michał Masłowski

I am a member!

I am a translator!

Offline
Iscritto: 05/15/2010

I think "free software friendly" means not having optional features
requiring nonfree firmware. ThinkPenguin products don't offer Bluetooth
requiring nonfree Atheros firmware to be provided by the distro
regardless of them being certified or not.

Any specific source for RYF-certified hardware using nonfree firmware?

mYself
Offline
Iscritto: 01/18/2012

Sure. As I previously stated, the Gluglug X60 uses the AR5B195 wireless card, which is technically the AR5B95 w. an additional bluetooth chip. While in this aspect the seller of the laptop had the option to install a wireless card without the part requiring a non-free firmware blob (the AR5B22 doesn't have such counterpart), he have chosen no to do it. Also both AR5B95 & AR5B195 based cards are pretty much the first wireless cards by Atheros that fully supports the 802.11n standard (i.e. non-draft), but they're are now outdated. That's why I'm recommending newer models instead, all supported by the ath9k kernel module.

By my knowledge, none AR5B95 based wireless card is endorsed by FSF, then (following the above logic) why it's "free software friendly"?

Update: According to the Gluglug site, the laptop now ships with an AR5B95 based card, so my aforementioned statement is no longer valid.

Chris

I am a member!

Offline
Iscritto: 04/23/2011

I would make the argument that ThinkPenguin's following the Free Software Foundation guidelines to the extent possible. If your going to try and argue failures in our product line up I'm going to come back and argue the Free Software Foundation's not living up to its own standards. The reality is we're all living in an imperfect situation with goals significantly greater than what is achievable. What that means is we should all try and live up to those ideals to the best of each owns ability. I think Trisquel, the Free Software Foundation, and ThinkPenguin have all done that. I think other 'free software' projects have not despite that many have advanced free software adoption. Some have settled somewhere in-between and I'm significantly less critical of them (though the FSF is rightly not so, but then that is the FSF's job, it's why I contribute, and I think most others contribute to the FSF).

That said I believe the majority of the products in our catalog are certifiable under the current certification rules. Just because we have not taken the time or they are not yet certified does not make them less free software friendly. If something in our catalog is not technically certifiable then the Free Software Foundation is probably not in compliance with its own rules that sets out the standards in regards to hardware certification.

The certification process is time consuming and whether or not it's worth while is up in the air. We certify not because its profitable, but because we support the program. I firmly believe we should be promoting hardware certification over h-node and similar databases. I've been using GNU/Linux since GNU/Linux distributions have been around and worked in the consumer desktop GNU/Linux space since 2005. Even before graduating from university that is. I've been doing this for about 10 years and have seen the real world results of poor decisions in this regards. Individual perspectives on this give a warped perception of reality in most cases. Either because users are very technical or because ones world is too small. I can probably name only a handful of people that I've ever met who've truly understood the problem and as far I'm concerned there are only two efforts under way to fix it. One of them I founded in 2008 and the other is the Respect Your Freedom Certification. I might also add Gluglug to the list although my thoughts on that are a bit varied and reserved. I like the idea and applaud those involved although I'm not sure it's helping move things forward. However I don't have all the details so I'll refrain from going further. There are at least a handful of people here who I feel fully get what we're trying to do, and fully understand the hardware problems.

What I will say beyond this is that regardless of the Free Software Foundation's position on unsupported components my position on it is that we, as a community that is, should avoid them. I'd go even further and say we should pool resources in order to gain influence. Unfortunately I think that message has not gotten out there fully despite numerous advocates here. Personally I don't want to have to write out thousands of times the same thing. It's both annoying and self-serving, both in a good way, and potentially in a bad way. What I do do sometimes is try and correct people when they spout off really poor advice or otherwise make critical errors (it's not hard to do, even those very knowledgeable here who I highly respect do it occasionally). Undoubtedly most people are not as knowledgeable and do not have access to tons of hardware.

Divided as we are we will fail as a community to gain access to the source code under desirable free software licenses if we continue on this path. A vibrant set of free software friendly hardware and solutions will continue to be lacking and/or completely unavailable in many cases. The bluetooth minipcie wifi combo cards are a perfect example of this problem. There is no combination of bluetooth/wifi that actually works with 100% free software. Given that there are options and most people have the resources I'd suggest a USB bluetooth adapter /w CSR chipset and a mini pcie card that is strictly 802.11n atheros chipset based.

That said despite the sad reality this is not rant about giving up. ThinkPenguin's going to continue moving boulders and I'm going to continue advocating for free software. What this is is a rant about getting others to do the same. I don't expect people to work for nothing even if many purely free software developers and contributors currently are.

Unfortunately the projects which are most at the forefront get an unfair advantage. A discussion I just had today. How many people have contributed to a project that was not at the forefront? That is something you use every day, but don't see. Core utilities you might be using that have a GUI written by somebody else for example. Trisquel is dependent on a lot of other projects. Many of which are not in a position to request donations because there users never see there web site.

Mariakopp
Offline
Iscritto: 03/18/2014

Now-a-days there are many network connections available and some of them are used with the help of software. And therefore the use of wires have been reduced to some extent.
http://platinumplatypus.com