14

What are the link layer protocols for cellular networks?

There are several platforms for high speed cellular networks. From GSM, to GPRS, W-CDMA, HSPA, LTE, etc. I (mostly) know about the physical details of radio transmission from what I studied at grad school (BPSK, QPSK, MIMO, Fading, etc.), and also the software details at and above IP. But I have no idea, and I can't seem to find information on what are the specific data link protocols in cellular networks. (Instead of Ethernet, 802.11 and the like)

I'm specifically interested in knowing if and how packet loss due to fading and other artifacts is managed at the link layer, or if they simply don't care and let the transport layer to care about this.

I am asking this because I want to evaluate whether or not I can manage to do some countermeasures against packet loss at the transport layer (UDP) specifically tailored to cellular networks.

My question is: What are the data link protocols for cellular networks? This is merely the starting point to solve a much larger problem, which is outside of the scope of this particular question.

Looking at some japanese documents, there seem to be some protocols such as DPCCH/DPDCH for W-CDMA, but frankly, the information on these is very scarce (even the wikipedia pages are orphans at this moment). Are there any good resources to get started?

Notice that I am a complete newbie on this site. Please forgive me if this question is wrongly worded, is not appropriate or if it has problems. I want to learn and become a good member of this community, so please consider giving me constructive criticism if you're considering closing or downvoting this question. I take criticism very well, and it won't be a waste of your time.

Panda Pajama
  • 369
  • 1
  • 2
  • 11
  • 3
    Cellular networks are typically not within the scope of IT professionals. The protocol stack is rather different from what "we" know. But there *is* student material for cellular networks which is publicly available, this for example: http://www.roggeweck.net/uploads/media/Student_-_GPRS_Architecture.pdf – the-wabbit Dec 18 '12 at 13:07
  • That's a rather good document. I'll read it in detail. However, it's unfortunately for GPRS which is pretty much obsolete. Where do you think I could have a better chance to get this question (and further subsequent questions as I dig deeper) answered? By the way, where did you find that document? – Panda Pajama Dec 18 '12 at 13:55
  • @syneticon-dj There's obviously differences at the transport layer but generally I'd have thought it was within our scope above that. A IP packet is an IP packet, whether transported by radio wave, fibre optic cable or carrier pigeon, after all. – Rob Moir Dec 18 '12 at 15:33
  • Yes, and I'm asking precisely about one layer below IP. – Panda Pajama Dec 18 '12 at 15:48
  • 2
    @PandaPajama Your best starting point may be [this Wikipedia article](http://en.wikipedia.org/wiki/Cellular_data_communication_protocol) -- Unfortunately the telecom Stack Exchange is defunct. We absorbed much of it, but I don't believe Server Fault has much expertise in this realm: We typically hand off IP packets to the network, and don't much care how they get to/from the handsets as long as it's not [RFC 2549](http://tools.ietf.org/html/rfc2549) or similar insanity. – voretaq7 Dec 18 '12 at 17:53
  • Maybe this is because cellular protocols are mostly industry driven and not very open? – Panda Pajama Dec 18 '12 at 18:14
  • 1
    Disclaimer: **am not a comms engineer**. I dimly remember there was a useful if a bit dated book: Radio Network Planning and Optimization for UMTS, 2nd ed. Laiho, Wacker, Novosad (eds.) Wiley, 2006. ISBN 0-470-01575-6. Unfortunately, while it is possible to derive BER (and hence, packet loss rate) from a given modulation and gaussian noise with a given bandwidth, I recall that errors due to multipath propagation are effectively unpredictable for a general case of urban terrain, and running experiments/simulations is highly advisable. Just my $0.02... – Deer Hunter Dec 18 '12 at 19:32
  • 1
    @PandaPajama I remember seeing a number of training documents from Nokia a while ago which I considered worth reading when the time comes. The GPRS document was yielded by a Google search of "nokia training document gprs". As the protocol set used in UMTS has [evolved from GPRS](http://www.mcl.hu/micromob/download/resources/unsorted/wcmc.pdf), they share a lot of terminology and many architectural features, so even if GPRS is obsolete, you likely would need its basics to work on to the UMTS protocol stack. – the-wabbit Dec 18 '12 at 21:45
  • Please clarify in which part, generally, major components are: BSS-BSC part - related to radio, MSC-HLR-VLR part - related to call routing and switching and subscriber state/DB, "IN" part - related to external services like SMS, MMS, Billing, Billing support (also eventually called BSS, but not related to Base Stations), GPRS core - related to IP packet transmission through mobile networks. What do you want to know? – GioMac Dec 24 '12 at 09:32
  • @GioMac I'm interested in knowing where you put your IP datagrams when sending user data (not voice or billing or whatever). As I mentioned, I want to be able to estimate the probability of packet loss at a given time, so I can probably send each IP datagram multiple times and get a slightly better probability of my data actually reaching its destination. I also want to know if it works the same way downstream and upstream. What communications standard do those protocols or technologies are related to? – Panda Pajama Dec 24 '12 at 14:29
  • 1
    First thing you should be aware of is that there is no guarantee about data channel on radio side - if all voice channels will be busy on the radio part, then your GPRS connection might be terminated, because usually voice calls have priority over data transmission - this also depends on BSC configuration of service provider. OK, so you want to know, what are your chances for reliable delivery (like checksums, auto retries etc) of your IP datagrams in case when mobile network is available and not interrupted and if yes - what happens, right? If so, I can get some real life stats/specs for you. – GioMac Dec 24 '12 at 22:54
  • @GioMac Thank you, I would really like that. But I'm more interested in Mobile WiMAX, LTE and HSPA than GPRS, so if you have some technical specs about their network stacks, I would really appreciate them. – Panda Pajama Dec 25 '12 at 01:24
  • 1
    For WiMax and LTE is IP, with some parts of old 3GPP, HSPA is just an radio part improvement of UMTS/3G which is based on 3GPP standards and protocols. read more at http://www.3gpp.org – GioMac Jan 15 '13 at 14:22
  • One of the few reliable sources I've found for information on cellular networks is the USPTO (Patent Office). Many of the patents for LTE, for instance, contain detailed implementation specs. You can search patents at http://patft.uspto.gov/netahtml/PTO/search-adv.htm. – dartonw Apr 05 '13 at 04:15

1 Answers1

3

The 3GPP specifications are free. In fact, a lot of the telecommunications specifications are free. They are often patent encumbered, but they are free for download. Some specifications, unfortunately, cost a lot of money. Anything from Telcordia is generally priced in hundreds of USD :( Alas, a lot of similar information is available elsewhere for free. For example, ISDN CPE recommendation formerly known as Bellcore SR-NWT-002343, the basis for implementation of libpri, covers the same things as ITU-T Q.9 family of recommendations. BTW, the current document would be SR-4994, weighing in at a "mere" $450.

You probably should start looking at 3GPP Technical Specification (TS) Series 24. The specification numbering reference comes handy. The specific protocols you're looking for fall under umbrella designation of IP Multimedia Subsystem. I'm not familiar enough with those to point you to just one document that addresses the specific layer you need.

The major problem you will encounter is that telecommunications are heavily layered. It's a problem when trying to comprehend things -- it's a godsend when you are implementing things, though. When developing and adapting protocols, they really do not want to reinvent the wheel. So each time some new functionality is needed, only this new functionality is added and the old protocols are usually retained and reused. For example, when GSM came around, a lot of SS7 (IIRC) signaling for things that landline terminals did was reused simply not to reinvent the wheel. They only added new signaling for things that are mobile-specific -- identification of terminals, security, handoffs, roaming, etc.

What you'll find in 3GPP specifications is a ton of references to other standards. It'll be rather heavy reading until you get a picture in your head as to how things interface. And they are not easy on acronyms. It'd be worse without acronyms -- the documents would be probably twice as long.

The 3GPP has good reference pages with "further reading" for most common technologies, such as LTE, LTE-Advanced, HSPA.

I believe it'd be incorrect to call 3gpp "old" -- it's a partnership project that links multiple standard development organizations and they are constantly producing new so-called Releases. Those encompass GPRS-EDGE, W-CDMA, HSDPA, HSUSPA, HSPA+, LTE and LTE-A. I think it's almost too good to be true that all of those technical standards are available under one roof, for free -- you don't even need to register on their site.

When you see references to other protocols, current ITU-T and all ITU-R recommendations are free downloads, as are 3GPP TS documents.

Community
  • 1
Kuba hasn't forgotten Monica
  • 237
  • 1
  • 13