Why voice calls on 4g may suck at first | Hacker News Reader

Why voice calls on 4g may suck at first

reghardware.com

58 pointsAndrewDucker51 comments

I wonder how efficient VoLTE will be. I love the idea that my carrier is just a dumb pipe that I use to make VoIP calls with whatever service I desire but from what I have heard that may be a poor engineering decision. I remember a Qualcomm (its their baseband chip in the iPhone 4S and 5) engineer state that there have been decades of research into getting the highest # of calls and best call quality per Hz of OTA bandwidth. The engineer went on to say that operating at a packet level is too high of an abstraction if you want to squeeze the most out of the cell network. He stated that you must drop down into lower, more niche/optimized layers to really match the efficiency that they already have achieved on GSM/CDMA voice.

Your're probably right, but maybe it doesn't matter? Voice traffic is a diminishing percentage of overall cellular traffic. LTE optimizes for the increasing majority of traffic - data.

There are extra advantages to LTE -- for example, T-Mobile is using it to make it so that you can seamlessly handover phone calls between cell towers and wifi access points (using IMS, not VoLTE). Previously this was only possible via complex technology embedded in the phone's baseband (UMA).

More calls routed over wifi means less cell utilization.

More details needed. (Ex Qualcomm engineer here, who finds the argument you were told tenuous. CDMA and GSM use digital vocoders, some adjust the bitrate for the characteristics of what is being encoded, but I'm just curious about the details of the argument.)

Packets have more overhead than TDM circuits just because of headers, but LTE does contain several optimizations for VoIP like header compression.

> Network operators might not care about that if voice didn't contribute so much to the money they make off each user. There may well come a time when their revenue will come solely from data traffic. At that point, the networks shouldn't care whether voice is handled by their app, Skype's or another - it's all just data, after all.

I would actually be quite happy if this came to be true. Both mobile and at home the main use of my service is data, but both come bundled with a voice plan that I don't use. Heck we don't even have the phone plugged in at home as it isn't used. I would happily get a data only plan if one was available at a good rate.

My prepaid €20 3G SIM with 15 GB gives me a responsive network connection with low variance latency. Know what that means? I just open up Skype and use it for every call back home. Works better than Skype over any public shared 802.11 network.

YMMVUYAIET (your mileage may very unless you are in europe too)

Actually network operators are extremely worried of becoming just a dumb bit pipe. They have seen what happened to fixed telephony and don't want to get the same route.

I mean how else would they make users pay 20c for 140 bytes :)

For example, many global operators tried some variant of a walled garden (apps, music, ringtones, etc). Another example: afaik there are no technological limitations of having something called "homezone" - i.e. when at home all your mobile traffic goes via WiFi and not through operator's network.

Luckily for us, users, innovative competition (like free.fr - it's in French) will help break the giants and lower the rates.

Disclaimer: I work in Telecom sector, not for an operator though

If you live in the US it might be a good idea to have a telephone plugged into your landline anyways. This way if there's a combined electrical failure and cell phone failure, you'll still be able to make 911 calls. There is also a reverse 911 conference call system that works with landline phones.

At least in California, if you're renting, your landlord is required to provide you with a working telephone line. You and them don't even need to pay anything, just make sure there's a working dial tone.

Interesting... I'm actually in Europe, and the phone line is on the back of the cable modem, so if there is a combined power failure we are out of luck...

My cable modem has an internal battery backup (Comcast Business in California); if you unplug your modem does the phone still work?

I love the pervasive myth that you don't need electricity to use a landline phone in 2012. You're more likely to have a working cell phone (generator @ the tower) than a working phone line in an area with FTTH (say, Verizon FIOS deployments, or cable co. VOIP deployments). Sure, the PON and gear has a battery backup, but it doesn't last anywhere near as long as the generator at the cell tower, or the battery in my cellphone.

In the Bay Area, I think most homes still have POTS lines. I wasn't aware most lines weren't already. Where are POTS lines no longer available?

By law, they're available everywhere if you specifically request them.

Verizon, when they install FIOS, will rip out the copper, by default. If you switch to another CLEC at a later date, you (or they) will have to pay to have copper put back in. You can request Verizon not remove the copper though, but you have to be aware of the need beforehand.

The main concern seems to be

Will they accept a lesser experience than they're used to?

And quite frankly, of course they will. Thats the beauty in having a monopoly on cell phone coverage.

I'm not sure I understand what you mean. Even in the US there are multiple carriers active in every market, and I assume that 3G is competing against LTE as products for the same provider, in the same market. That should mean that LTE subscribers might turn off their LTE to get better call-connect speed, or simply choose a 3G subscription if that functions adequately enough for their general data needs.

That might mean that the big problem will be to convince people to jump to LTE (and the higher price) with slightly worse end-user voice experience. Some customers will, but perhaps many won't. Artificial quality degradation of 3G might be the answer for that, or limiting the 3G-only contract data limits to encourage LTE adoption.

In Germany (the largest economy in the EU, and a population of 80M people) there is exactly ONE (1) carrier whose LTE works with iPhone 5.

There are zero (0) that work with the LTE chip/frequencies in the new iPad.

Which carrier is it? Which band do they use?

Deutsche Telekom. They use 850MHz AFAIK. o2 and Vodafone use 800MHz (and I think maybe another band).

http://www.apple.com/iphone/LTE/

The European iPhone 5 does 2100/1800/850.

Deutsche Telekom does not offer LTE on prepaid plans. Their minimum contract length is 24 months, and auto-renews for another year at the end of the 24 months. It is not cancellable within the last 90 days of the 24 month period, so if you don't remember to cancel at or before 21 months, you are on the hook for 36 months.

It's also something like 70€/mo for 2GB of LTE. :(

Being the first mover is not the same as being a monopoly.

There are two other LTE networks here. iPhone 5 only supports one of them due to the frequencies in use. :(

They will because voice calls are on major decline. It's all about data and SMS now, sadly.

VoLTE doesn't even make sense to use from a network provider's perspective. 3g architecture already exists and isn't going anywhere soon while people still have 3G phones. Calls generally aren't made over 3G to this day, they're made over 2G. Since frequency spectrum is so expensive and limited, packet sizes need to be as small as possible...so unless they can fit more people on an LTE voice channel compared to a 3G voice channel, the networks aren't going to be jumping to use VoLTE. LTE should be focused on data for right now, and keep 3G for voice. When was the last time someone said, "Gee, I wish my call was faster!" anyways?

The point that keeping the phone in 3G for voice and LTE for data will drain your battery doesn't make sense to me. Keep the phone in a 3G state for SMS, voice, and push notification type data interactions, and when data is used, then it would utilize LTE. That would actually SAVE battery life, since as the article states, LTE is a battery hog.

It's not about the speed of the call (from the consumers point of view), it's about the audio quality. Higher bandwidth will enable vastly improved audio quality.

People may not say "Gee, I wish my call was faster!", but they sure say "Gee, can you speak clearer?"/"Gee, I can't understand you" thousands of times every day due to muddy/compressed call quality due to the compressed nature of our current cell phone calling infrastructure. Apple made a big deal about the so-called HD Audio capabilities of the iPhone 5, which is still below FM (and VoLTE quality).

And while LTE may be a battery hog right now since it's so new, don't forget that 3G was way worse when it was new. Battery consumption on LTE will improve drastically quite fast. In a year or two it probably won't use much more battery than 3G does.

It's always been vaguely surprising to me that Apple hasn't used the iPhone platform to force the carriers to upgrade the voice quality. It's exactly the kind of thing that would cement the iPhone brand as the high-quality, aspirational choice.

Either Apple decided it wasn't worth the effort, or the carriers flatly refused I guess.

It's not in Apple (or any phone maker) control today. With current technology you pipe 8 kHz voice to the wireless modem and the encoding is done there, following strictly defined cellular standards. So it's in the operator control.

Apple has extra leverage, though, because of its perceived importance in the market. Look at visual voicemail: Apple decided IVR voicemail sucked, and worked its way through carriers until it could find one that would support changing its voicemail architecture. That carrier then turned both the iPhone and that feature into a competitive advantage.

If Apple really wanted to, it seems like they could have strong-armed their way into some higher-quality audio standard -- earlier deployment of VoLTE if nothing else, since that's going to happen eventually anyway. That said, I'm not especially surprised that they didn't; Apple has historically actually been pretty timid with wireless technology, and the biggest bafflingly absent feature of the original iPhone was that it didn't include a 3G radio even though 3G was already widely deployed.

Exactly. They didn't even need VoLTE; higher quality voice has been possible on 3G networks for years. If you have an HD-voice capable phone, then I believe you should get HD calls if you use Orange (now Everything Everywhere) in the UK for instance.

Lots of modern smartphones support it, but it's not always turned on apparently.

There are limits to Apple's leverage in the mobile marketplace. When the original iPhone was introduced, they got AT&T to implement Visual Voicemail and ship the phone without a carrier branding logo, but not much else. They couldn't even get AT&T to agree to ship the phones without a subsidy lock, despite the fact that there was no subsidy for the first few months (nor could you legitimately unlock an iPhone associated with AT&T ever until earlier this year).

At this point, analysts and investors expect Apple to ship over 5 million iPhones in the first weekend of availability. Resistance by a single carrier (there are 3 now, and that's just in the USA) will seriously impact that number. Moreover, carriers have other serious challengers at their disposal and are pushing them hard (e.g. Samsung Galaxy S3).

Asking a carrier to promote a phone is one thing. Demanding they overhaul their entire network is probably asking too much.

Apple controlled the choice of hardware in the iPhone and could easily have picked a modem that came with AMR-WB codec support. It's not as if the codec has onerous hardware requirements.

Yes, putting it in the original iPhone would probably have been a feature too far, but it could have been added to any iPhone from the 3G onwards as I understand things. It's exactly the kind of premium feature that you'd think Apple would have loved to have had associated with the iPhone.

Whether they could force the operators to upgrade their networks is another question though.

> Calls generally aren't made over 3G to this day, they're made over 2G

You must be talking about CDMA (EV-DO vs 1xRTT) only. Not true with GSM+UMTS networks.

As far as network operators are concerned, the more data (incl. voice channels) carriers that can be fit into their spectrum allocation and/or existing site layout, the better. Of course they want to use LTE for voice, but the VoLTE saga has been a long and drawn out one.

So if I understand this correctly, 4G, despite having higher bandwidth, results in either bad voice quality, or a 2-4 second delay? Voice quality over the phone has been good for a 100 years, and now they can't match it with a higher bandwidth? Back to the drawing board I say!

Voice quality has never been good on cellphones, it's actually far below even so called FM-quality, which is far below CD-quality. The microphones are capable of recording voice in a much better voice quality than what is currently transferred through the networks. Phone audio is normally extremely compressed. So called HD Audio, audio with a wider bandwidth, was actually touted as great new iPhone 5 feature, since we've all noticed that the audio quality of cell phones calls are horrible compared to the real voice or even audio recordings made by the same microphones.

VoLTE is a new technology and isn't totally ready yet, but when it is it'll enable vastly improved call quality.

http://www.engadget.com/2012/02/24/fraunhofer-full-hd-voice-...

Voice quality used to be very good on mobile. My first CDMA phone was indistinguishable from a landline, with some people surprised I was right outside their door when I called. And it was a big selling point in the early mobile analog to digital transition. My first cellphone was from Primeco PCS who made had no qualms about saying that TDMA (the competing standard at the time) sounded like shit compared to their CDMA phones.

As all good things eventually end Primeco was sliced up into Verizon Wireless and Sprint. With the only competing tech being TDMA from ATT Wireless they compressed the voice channel even more to accomodate more customers.

TL;DR: CDMA audio, currently, is almost indistinguishable from landlines but the carriers choose compression over quality

Quite true. On my old Verizon Motorola P8767 (a color flip phone -- it had four colors) I could enter "engineering mode" and increase the bitrate of voice calls over the network. It meant you were using more than your share of the voice pipe, but at least my voice sounded excellent.

Switching to a 2007 AT&T iPhone from Verizon was the equivalent of going back to two tin cans on a string. A string that a vengeful voice god would cut at random between 5 and 15 minutes every call.

"Indistinguishable from a a landline" is a pretty low standard. Landlines have terrible quality compared to nearly every VoIP system I've used.

Yes, was about to respond and say exactly this. The frequency response is something like 300-3500Hz (vs. 20Hz to 20kHz for normal human hearing), and the dynamic range is also pretty mediocre. Skype is much better, and many in-network cell calls that never touch POTS also have better call quality already than landline phones do.

Yes, now imagine a mobile phone system with quality that couldn't even match that. People can usually tell you are calling from a mobile based on the quality of the call.

Skype is great, until about 1800-2000h. Then the quality turns to crap as everyone comes home, does netflix, bittorrent, youtube, etc. and saturates the bandwidth. Unlike most offices, homes don't get a SLA.

You're absolutely right. They chose compression over quality to load as many concurrent calls per tower (CDMA, mind you) as tolerable. It used to be (maybe still is) that on the Motorola Razr v3m you could drop into a programming menu, and force the handset to use the higher bitrate vocoder, whereas the towers would request use of the lower bitrate vocoder. Around Chicago, in 2000, for example, Sprint CDMA was indistinguishable from a landline, likely because the towers were telling handsets to encode with the high bitrate.

"Enable" is indeed the key word. VoLTE does support AMR wideband, and there are other wideband codec evolutions in the pipe. But it also supports AMR narrowband.

So it'll be up to the operator to decide between higher voice quality or higher capacity. That will depend on how much spectrum is available, and load. It may even change during the day depending on a cell load.

I've always found Skypes audio quality to be far superior to phone or cellphone...that is when lag doesn't come into play. I wonder how much more or less bandwidth they use for audio vs traditional voice networks.

The billing system causes a 2-4 second delay when first connecting a call. Using Skype or other "data is data" VoIP app (that doesn't treat voice-data differently from data-data) would avoid that problem and have plenty of bandwidth.

I think this is really a case of someone looking at a emerging technology and saying well its not going to be good because it isn't good _now_. In America the time frame where we even start to think about VoLTE is middle to end of next year , and there are hundreds of engineers from a dozen companies working on making it better. From a technical standpoint there is no reason why it would be worse, in fact all the technical reasons point to it being better once we figure out the implementation. This is like saying PC gaming is going to suck because General Custards Revenge sucked.

I confirm the two to four seconds delay in experimental deployments. When I first saw the proposed architecture, I couldn't believe it - kludge was the first word to spring to my mind. All that while the deployment of IMS platforms is gathering steam... Why can't IMS become the voice carrier ? SIP everywhere and no more circuit switching... What's wrong with that ? NIH ?

Another "why can't they do this already" technology that impacts millions of people every day: seamless GSM to wifi (and back) handoff. They had specs, prototypes, and demos back in 2004. It never went anywhere. Is it just because "they" wanted to overcharge for microcell hardware and tack on monthly fees?

Imagine being on a call outside, walking inside, through an elevator, and up to your room/office/flat/apartment and not have the call disconnect. No GSM repeaters in the loop. Your call went from AT&T->Wifi VoIP automatically and it'll switch back to GSM when you leave wireless range again.

Handover between GSM/UMTS and Wifi works in the wild... But there is more to market acceptance than just having a perfectly fine technical solution : http://www.kineto.com/wi-ficellular-handover-the-state-of-th...

"VoLTE" seems to be an new term for some subset of IMS. There are three problems with end-to-end SIP as of now:

1) You want to make and receive calls where other end is in PSTN or non-AIN PLMN or otherwise unable to directly process RTP payload stream (you essentially can use SIP for circuit switched data paths, although terminating SIP sessions at adge of circuit switched network seems like better solution).

2) Handover LTE->GSM when leaving coverage area. It is possible to handover an active packet data session from LTE radio channel to almost anything else (including WiFi), but with active call the target transport has to be able to support required bitrate and latency, which is clearly impossible with plain GPRS, barely attainable with EDGE. On the other hand LTE allows transparent handover to just about anything that is IP (wired ethernet, anyone?). Cludgeness of the IP handover architecture is caused by two things: no real deployment of IPv6 when first GPRS networks were designed and desire for this architecture to support non-IP protocols (things like X.25 over EDGE are specified as possible, although probably non-existent).

3) QoS on backhaul network. While LTE All-IP backhauls are simpler to build (and in fact mostly cheaper, because designer has more freedom with network topology), they present different set of challenges than normal circuit switched networks when it comes to QoS. This tends to be complicated further by desire to tunnel PCM/ATM based interfaces of legacy network base station components of newly built sites through same backhaul Ethernet/IP link (there goes your topology freedom...).

Also there is slight problem of how to bill SIP/RTP calls spanning different networks, it is technically possible to capture data you need to do normal per-minute billing, but it is not clear whether that is good model for LTE. In my opinion best solution is to bill for data transferred across LTE network, for call-management operations (ie. per-call fee) and for possible costs of call termination through circuit switched network all at once, with possibly separate entities providing each of these services. This seems to me like fair pricing scheme, but on the other hand it is not understandable by common consumer and thus probably infeasible for carriers.

Not doing voice calls over LTE and relying on fallback to whatever legacy circuit switched technology is supported by network and MS is good way for carrier to workaround these issues. But if I correctly understand conditions of LTE spectrum allocations in Czech Republic, best way to meet these conditions is to actually roll-out VoLTE from the start (these conditions mostly require availability of voice service everywhere with expectation that missing coverage areas of GSM/UMTS will be filled with LTE, which seems like technological nonsense to me, but it is how it is).

Billing has seen much worse cases. QOS is just fine too, at least on the radio segment : LTE allows dedicated bearers, each with differentiated QOS, and no one in his right mind will try VOIP on GPRS or EDGE. Backhaul QOS is currently contentious - which is why we are scrambling like mad to lay fiber everywhere, so it will only be a problem for a while. Inter-protocol handover is hairy, but it has been done before - though it is indeed a significant technical hurdle. So handing over between SIP voice and CS voice seems to be the harder problem. Has anyone seen research in that direction ?

Nokia E-series handsets can be persuaded to support VoIP over EDGE and it sort of works (few years ago we did some testing of that with intention to use that as part of cab telematics), one would say that it works better (certainly it's more reliable) than VoIP over WiFi with same handsets.

Interprotocol handoff between two CS protocols is mostly equivalent problem to normal GSM handoff, handoff of packet switched transport is mostly transparent to higher level applications (ie. voice). Clear problem with IP->CS handoff is sudden change of QoS parameters of channel which I think cannot be overcame in any sane way and any solution involving IP->CS handofs will reintroduce audio artifacts of NMT. (hypothetical insane way: pump full RTP datagrams through UMTS radio channel instead of bare audio, which is impossible to do with GSM radio layer). CS->IP handoff (which seems to me to be more interesting in czech environment) has same problems and also lacks any support in existing CS protocols that are mostly set in stone.

On the other hand I think that LTE deployment will go reasonably fast so these issues can be mostly ignored by doing fallback to CS where possible and using only LTE when deployment is sufficient. So I don't think that anyone is going to seriously consider solving problems inherent in CS<->IP handoff.

So say, every Android and iPhone Devices, and may be Windows Phone 8 will come with a default hidden VoIP inside, if it is collected using LTE it will use that instead, the data would then be discounted over in the Mobile Network. This to act as a temp solution until VoLTE comes.

Like Jeremy Clarkson in Top Gear, How hard can it be?

The article also fails to mention why VoLTE is hard and takes so long. It said many different network operator has many different implementations. And even if that is true there is only a handful of network provider these days that network operator uses, Errisson, Nokia-Simens, ZTE, Huawei, and i dont know who else. ( Samsung?? ) They had to test again 3 others. Again how hard could that be?

Here's the punchline that most of the writers in the industry don't discuss.

IMS is an interoperability standard, but everyone implements their own version. The original intention of IMS was to provide a method of dealing with non-IP systems as IP (Particularly SS7 related functionality). The issue is that SS7 was the last pre-divestiture routing system (thus it was the last ubiquitous standard), and there can be no consensus in the industry anymore with respect to standardization.

So why no VoLTE? Because there isn't a standard that everyone is willing to implement, and there probably never will be. IMS is different everywhere you go.

The telco world tried to do deploy IMS already in the 3G era but couldn't get it working due to complexity, now they rely on it for voice in 4G .. resulting in no voice service?

Reply on news.ycombinator.com