From the monthly archives: "March 2012"

 I try to stay away from UE commentary since there’s so much of my day job that overlaps with this, however it looks like I need to revisit the only device I have commented on for a 4th time. Hopefully this is it. Maybe I should take this as a lesson.

OK here are the previous comments firstly, if you want to catch up to my previous comments. 

No Surprise: New LTE iPad…Shock and Awe: Still 2 Models!! WTF???

Updated, Even More Digging on ‘The New iPad’ Regarding Carrier Lock Down

Follow Up Thinking on ‘The New iPad’ with LTE

In summary, I erroneously assumed Apple would have designed The New iPad with a WTR1605, or equivalent such as the new Fujitsu (see very bottom for specs) RF components. They did not, instead using the RTR8600 (Magellan) and therefore can claim to have spectrum issues to blame on not having a universal iPad that supports all carriers. This was a mistake on my part.

I just want to take this moment to examine why Apple has recently begun issuing refunds to Australian LTE customers on the basis they advertised it to the customers as 4G.

So the logical conclusion is the iPad doesn’t support their spectrum. Telstra has deployed LTE at 1800MHz, which is gaining a lot of worldwide popularity as an LTE band. No, The existing part, RTR8600 DOES have the capability of LTE on this band. So the design issue is not the transceivers it’s the amplifiers (RX and TX.) Here’s iFixit.com‘s picture of the board space for cellular RF and (large)    TechRepublic’s zoom in on the components… (large) 

There’s a TriQuint TQM7M5013 Quad-Band GSM / GPRS / EDGE-Linear Power Amplifier Module…and Skyworks 77460 LIPA  Module for WCDMA / HSPA+ Band IV (1710–1755 MHz)…so the ATT version is in these pictures. This is why Apple has separate versions and why they are having difficulty with Telstra’s 1800MHz LTE with Next G (850 WCDMA) deployment. There’s very limited space plus they’re trying to keep heat dissipation/power consumption down and these constraints together are tough to find parts for.

Regarding Australia, it is possible to cover the bands in the same spaces they have today so I’m only guessing logistics is the issue there. Until then, they will need to back off the 4G claims. 

Here is the confusing amount of possible bands….

I had assumed that there was more space allocated for more silicon so I was wrong about a universal iPad. The way forward for Apple and others is fairly straightforward, to make a universal version, they can:

  1. Create a modular RF section that can be user changed, perhaps embedded with the SIM holder???
  2. Shrink something like battery or use next generation MDM9625/WTR1605  to save some space to allocate for more chips. (Still, doubt you will get all PA/LNA combinations plus you need duplexers etc…)
  3. Go SDR!!! (Software Defined Radio..) Screw this approach, use their existing Cortex type of cores, license some IP from the myriad of A/D and LNA/PA guys out there create a solution that covers all bands.  
#3 is most difficult from a size and power consumption point of view but exactly the type of problem Apple is suited to attack given their resources, desire to create better solutions and end user device pricing. Perhaps one more process shrink of that hardware, like 28um at TMSC or IBM would get them to a ‘close enough for horseshoes and hand grenades so let’s go do it’ ability, I’m not sure. 
Here’s an article on a Wideband Power Amp for SDR that could be a model for the end device.
Lastly, an SDR project would benefit the end users greatly, but I don’t think Samsung or STEricsson would take the initial risk based on their previous performance. On the other hand, for Apple this doesn’t seem risky at all. They are moving something like at least 20M units of mobile devices annually, so developing this module would make sense for the upstream partners that would help make it happen. Apple could organize the effort on a project or tiger team basis with the expectation that the industry will adopt the approach once they have pioneered it. How much of a head start would this give Apple in the marketplace? Guessing about 12months or so at least. Not only that, platforms allows flexibility to really improve RF like through better interference cancellations and so on… Gets me excited to build one just thinking about it.

That’s it folks….

 Oh yeah, Fujitsu transceiver launched the other day and would work here too…

MB86L13A LTE-Optimized Transceiver

 There is an interesting flow of comments coming out of the Barclays Capital High Yield Bond and Syndicated Loan Conference. Firstly there was some interesting comments by MetroPCS regarding LTE, these are from Keith Terreri VP, Finance & Treasurer, MetroPCS Wireless, Inc.

  • Firstly they have accumulated 500K subscribers so far out of between 9M-10M total subscribers. 
  • They are looking for more spectrum and are high on LTE, with over $1B allocated to spectrum allocation.
  • Metro is working with Asian suppliers to introduce ~$200 LTE/Android UEs. Remember, Metro has deployed primarily in 3GPP Band 4 (AWS) with a few markets in (PCS) Band 2. 
  • They are pretty constant in EBITDA although there’s a severe market environment that puts lots of pressure on their prepaid subscribers, and competitive pressure on they voice minutes. Thus it’s probably fair to say LTE is helping the margins. 
Joseph Eutenuer, CFO of Sprint Nextel Corporation presented their network vision. Some tidbits:
  • Steve Elfman’s group doing well getting construction permits for their 6 markets. (Means it’s verrry early in their deployment!) Sprint is deploying  LTE in 3GPP Band 25, turning off iDEN (push to talk), and consolidating site infrastructure…
  • Sprint sort of preparing for LTE iPhone. Have a similar contract with Apple to Verizon Wireless and ATT so they will be able to offer it as well.
  • They feel confident they will sell LTE iPhone even without extensive LTE coverage initially.
 
Slides regarding Sprint’s Network Vision plan…  OET and PSHSB

 Here are some thoughts about the questions I get around SVLTE.

Notice the pic and remember this day? This was Verizon and Apple answering questions about the iPhone 4. The biggest technical tidbit of the day was that the 3G iPhone 4 would not allow a simultaneous voice and data session. Of course this was dictated by the 1X CDMA network architecture, not the phone.

Here is a good graphic sort of illustrating that difference.  


Yes, The same issue exists for LTE and CDMA voice!

See, there is a confusion based matrix of items to enable the use of data during a voice call. There is some technical detail to get into to understand what the problem is.

Some of the blame goes to the phone OS, some the phone HW and some is on the network.

HSPA/WCDMA 3GPP networks like ATT and Tmobile support this feature and so the remainder is inside the mobile device. 3GPP2 based Verizon and Sprint on the other hand don’t have this capability inherently in the network. 

 

UE (Mobile Device) UI Capability/function

The device User Interface (UI) manages the integration of voice and data services for us. As you know, for example, the iPhone visually displays a dial pad for voice calls and has other apps for data like browsers, email, FaceTime/Skype etc… Not so obviously, the multitasking nature of the UI and Operating System  that supports multiple execution threads etc, facilitate and encourage data use such as it is so convenient. The interface makes it easy to switch between these tasks and actually encourage parallel data/voice simultaneous use. So Smartphone = Attention Deficit Disorder (ADD) device.

UE (Mobile Device) Hardware (HW)

Beyond the user interface, the device needs to be enabled with the capability to transmit data and voice simultaneously (or apparently so.)

Many USB based devices are data only (much like the old flip phones were primarily voice only) and only have HW to facilitate a data transmission. No voice HW present, but no need either. Result is simple and cheaper device. Smartphones sort of beg dual use, and often have HW to support data transmission and voice encoding/transmission. Now for the caveat (confusing part), of course it’s possible to have a data application for voice like Skype and that would technically eliminate the need for separate HW however the first deployments in the US are only partial data overlays over extensive voice networks, thus most devices base their connectivity on 3G voice and need to have voice compatible HW onboard for those non 4G or 3G data covered areas. Secondly, the quality of voice in voice networks is currently more assured, or higher priority than the existing non assured data services. So in the case of Skype over 4G, your voice packet gets the same priority (currently by default) as my web browsing packet. (With the PCRF this doesn’t have to be true but operators have not really unleashed this tiered type of service over LTE yet.)

So hardware wise, UEs utilize (baseband) processors capable of 3G to in addition to LTE to be capable of communicating with both networks. The first devices had 2 chipsets, 1 for 3G processing and a second for LTE processing. Only recently has there been multi generational (and multi protocol) capable chipsets that are able to process both 4G data and 3G in one chipset. 

To be clear, a device that has a single RF/baseband processing path may only be communicating on one network at a time.

Here is a typical data + voice capable device architecture showing Qualcomm components. 

So if the processing is in place, the next thing in the chain is the Radio Frequency (RF) chips. You can’t physically interface to the network without the capabilities here so it’s essential for the device to be able to interface to the voice and data spectrum bands and modulate/demodulate the signals correctly. This whole subject is so deep, more in a follow up post on this.

 Qualcomm is arguably the market leader in UE components and architecture. The pictures are the previous MSM8960 components which were 2 pieces of silicon for 3G and 4G processing and others for RF. The MSM9600 chipsets that are shipping now are essentially 1 piece of silicon for 3G and 4G processing with a second for the RF interface.

UE Network Communication Protocol

In the US, MetroPCS, Sprint, US Cellular, Verizon are networks that utilize CDMA 1X technology. 1XRTT (IS-95)/CDMA2000 or 3GPP2 based voice is a circuit switched voice protocol, with all the bells and whistles of 1.25MHz channels, RCs, Walsh codes, PNs etc… For high rate data, these networks utilize 1xEVDO (Evolution Data Only) a code and time division protocol and have slots, 1.25MHz channels, DRCs and so on. In summary, CDMA2000 uses 2 separate radio channels with 2 protocols for data and voice communications. ATT, T-Mobile and others have deployed 3GPP based WCDMA networks with 5MHz channels for voice and HSxPA with 5MHz channels for high rate data. 

Just to be doubly confusing, functions and G’s aren’t necessarily the same. There is 3G high rate data and 4G high rate data, and there is 4G voice and 3G voice. 

The communications protocols for voice on 3G on CDMA networks uses Qualcomm’s EVRC (Enhanced Variable Rate Coder) and on WCDMA networks is AMR (Adaptive Multi Rate) coding. The coding converts sound to data bits. Just to stay confusing, 4G voice, under the Voice over LTE (VoLTE) framework uses AMR also, but the rest of the voice protocol is very different than 3G. 

Data story is similar with 3G high rate data on 1X CDMA 3Gpp2 based networks is EVDO (now Rev A is most common) and on WCDMA 3GPP based networks is HSPA (now HSPA+ is becoming the most common.) For 4G we will only focus on LTE which happens to be 3GPP based.

There’s a lot of complexity in communicating the bits back and forth from the network when you compare 3G and 4G and CDMA 1xRTT with WCDMA/HSPA. Due to technical and cost constraints, most devices up to date have been either CDMA 3GPP2 capable or WCDMA 3Gpp capable. As an example, Qualcomm Gobi based devices have support for just about every situation in one component whereas Marvell only supports 3G WCDMA/HSPA and LTE.

A strong desire exists for every device to support every scenario, however the complexity and risks sort of dictated this approach of multiple networks, separate devices. See this seamless mobility chart, key point is you need both protocols, 3G and 4G for now.

 

Wireless mobile networks have had this multilayered architecture to support 3G circuit switched voice and high rate packet data at the same time for a while now. Adding in LTE is another layer of complexity that manifests itself as a battery and size constraint for today’s UE designs (more HW + more communicating layers in the devices), and in fact Steve Jobs specifically mentioned it as a barrier to LTE implementation in the iPhones.

UE Parallelism/simultaneous use

So now we are clear a device that has a single baseband processing path may only be on one network at a time. That was very long windup and I apologize. Back to the problem of the Verizon iPhone vs the ATT iPhone and voice and data simultaneously…

A mobile device that only supports a single network (voice or data) device will have to pause communicating with the data network if a voice call is to be made or received and vice-versa.  

The first generation of 3G/LTE capable devices have 2 physical components, a 3G path processor and a LTE path processor.  Qualcomm, TI, Marvell and others are beginning to ship components to enable the 2nd generation of LTE devices, those with 2 paths and shared silicon (fewer components.) The first devices are hitting the shelves. The benefits will be improvements to battery life and form factor, however neither will be as best they can be yet. Have a look at a slide from Qualcomm presentation comparing various architectures based on heat (energy waste) output. So it takes 2 radios to have a 3G and LTE flow from the device at the same time in order to support true simultaneous data and voice use.

 

 

 

 

 

 

 

 

 

 

 

The next step in device design will either be VoLTE and/or a software defined radio approach that allows more effective conservation of resources but that’s a topic for another day. Back to voice+data.

 

 

OK so the ability to make 3G voice calls and use data at the same time can be broken into smaller pieces.

  • For 3G 1X or 3GPP2 based networks like Sprint, Verizon, MetroPCS etc…, 1X (voice) + EVDO (data) is called SVDO. Some devices support this but most networks do not…As an example, the Verizon iPhone either communicates data or voice at one time.
  • For 3G WCDMA or 3GPP based networks, the voice and data bearers can be controlled from a single control point, (before HSPA it was on the same channel) therefore the voice and data can flow effectively in parallel. I won’t go deeper but this is the end effect although there are some finer points.
  • 4G LTE based networks need to have Simultaneous Voice and LTE (SVLTE) capability on both network and devices to operate in parallel.  It’s a specific function that allows better control of the voice and data over the networks. 

Today’s 3G + LTE networks carry voice over 3G networks as circuit switched voice. In a circuit switched paradigm, voice is carried in a point to point stream and requires real time connection between the two points, much like 2 tin cans with string, or a wired phone. Voice over packet, such as voice over IP networks use a connectionless paradigm, no open line is required. The world is starting to prefer it because this method also allows the use of the time in between the sound pulses to carry other data. Skype is and example of voice over packet (sound is quantized and sent over packets on the internet.) 

So the next iPhone launch with Verizon (3GPP2 based) will require specific upgraded network functionality like enhanced High Rate Packet Data (eHRPD) to be deployed as well as a multitasking OS, and multiple connectivity chipsets to enable simultaneous use whereas ATT (3GPP based) will have an easier time and only require the UE capabilities.

As of right now, the Qualcomm MSM9600 chipsets appear to support the SVLTE functionality and based on their use in The New iPad it appears Apple is headed down this path so it will be up to Verizon and others to deploy the necessary network upgrades.

Hope this helps.

Interesting slides on VoLTE….

Download (PDF, Unknown)

Voice over LTE

I like this Aviat diagram. 

So have you seen all the press reports about The New iPad?  In the beginning it was about features etc…now the chorus of boos is about the LTE data plans from VZW and ATT. People are looking at the size of files that can be displayed on the tablet’s high resolution ‘retina’ display (2048×1536 pixels) and then looking at all of the application download sizes and coming to the conclusion that $5/10GB is a little high. I agree. It’s skyway robbery.

 

 

 

 

 

 

 

 

 

On the other hand, remember this chart? 

 Anyway the operators are banking on improving margins with you/we early adopters so this should not be shocking. They essentially need us to pay for the service but not really use it. Unfortunately this is not going to be the case with high speed mobile wireless. 

Unfortunately the wireless operator networks are organized for voice and sort of inefficiently designed for data, thus the costs for the operators will always be high and they will always be hesitant to give us a reasonable deal.

There is a way around this problem though. Operators could encourage the deployment of small cells (you are penalized now by having to buy one or complain to get one) and offer us free usage on small cells like femto cells. Traffic from a small cell is much less costly for them than over the macro network and while you are in coverage of a small cell you are going to be getting a very good signal, so high throughput is likely, thus this becomes a win-win solution.

So a lot of things on the operator’s side have to change to make this happen, but they are no different than the level of commitment it takes to launch LTE in the first place. It’s time for small cells to save the day. Let’s go get this done.

Links: valuebasedmanagement.net, Aviat

 Verizon must of had some test routes in or something…As of today their latency is much more reasonable. (30ms average reduction) The end user experience improvement is pretty good when testing with Skype. Their LTE service is getting to be pretty good now. I reside at the cell edge and 700MHz delivers where 1900MHz does not (where they have EVDO A.) I am getting 8-17Mbps down and 0.15-3Mbps at this location. It’s a good testament to the 700MHz spectrum and OFDMA approach f LTE in general. Well done!

 I’ve been playing around with The New iPad since Friday now, and the thing that’s most annoying at the moment is not the device at all. It’s Verizon. I’ve been of course testing LTE speeds and comparing to my Verizon 3G speeds and I must say, the latest iPad has a good receiver performance for a user device. I can burst at peak rates on Verizon’s network when I’m in good SNR conditions. I get lots of peaks with available high order modulations over the air, ready to travel but unfortunately the buffers are mostly empty when returning from the network. There is also some pauses in transmission but not due to link faults or retransmissions. Expectedly, the performance has been worst in public places like airports and close to business districts but the air performance is really not bad at all, it’s really really good according to my tools. Simplistically, this chart sums up what I (we all) expect from our LTE connection:

It’s not that different on ATT, only slightly more annoying.

This all points to one of two culprits.

  1. The Scheduler
  2. The Backhaul

I highly doubt it’s the scheduler because I’ve surely been on both vendor’s eNBs by now, and like I said, ATT and VZW have similar over the air performance. There is one smoking gun though. The latency. The packet latency is really pretty high on both networks! In SNRs of 30, I regularly get 50-65ms one way trips. Working backwards, I suspect that these guys are taking the packets on an odyssey from site, transport with various aggregators and ultimately to a distant EPC. I complained about the fact the iPad didn’t allow FaceTime on the LTE network… Perhaps that’s because of the fact that (A) the latency is crazy high and (B) the backhaul seems to be inadequate already.

Just note, same device tests 50% less latency on 3G networks than LTE. To me I suspect there is a ton of aggregation in the backhaul then that probably gets concentrated in to a few EPCs nationwide that don’t have your packet’s best path at heart. In the DFW airport area, best results I get for latency on LTE was communicating to servers in Level 3 facilities at the Dallas Internet POP, and they were still about 50ms> than 3G. Anyway, VZW take note, this is not cool but I am realistic and realize than I’m an early adopter.

BTW, NGMN has some great guidelines on requirements for optimized backhaul:

NGMN_Optimised_Backhaul_Requirements

Download (PDF, 1.5MB)

And provisioning backhaul:

NGMN_Whitepaper_Guideline_for_LTE_Backhaul_Traffic_Estimation

Download (PDF, 3.55MB)


I’ll stop complaining here. Have fun!

 Happy St.Patrick’s Day, yep.

Have fun and uh, be careful with your LTE mobile device.

 So now we can see a couple of things have been revealed about The New iPad. Firstly, I was wrong about the guess on the components. Obviously timeline were tight for Apple’s engineering team and they chose devices they had access to at the time vs the more advanced components available now. Firstly they use the MDM9600 instead of the MDM9615. The advantages of the MDM9600 are increased power thriftiness due to the process shrink to 28nm from 45nm  is capable of LTE R9 and is VoLTE enabled. So strictly speaking the VoLTE not important for iPad but the power savings would have resulted in either slightly less battery required or more battery life. 

The other major component difference is the use of the RTR8600 vs the WTR1605. There are more bands supported with the WTR but the RTR is no slouch either, oh yeah there are some power and size advantages to this component too. Oh so maybe that’s where this VZW vs ATT iPad thingy comes from, but I highly doubt it.

 

 

 

 

 

So a very ugly fact, the operators are not allowing FaceTime in LTE mode. It’s stupid. It would be a major selling point for using their LTE services and it’s only a drop in the bucket throughput hit. People are measuring their iPad LTE throughput at 20MBps down and 14Mbps up and Verizon and ATT are worried about a 64-128kbps stream? WTF??

 

Seriously, this is ANOTHER case of the consumer getting shanked. They could charge more or throttle more in the future (PCRF!) why not feed the addiction now?? 

Update: I forgot about the Over The Top (OTT) apps like Skype. Looks like only consumers that like FaceTime are penalized.

Net Net, The New iPad is a major step forward but there are some less than all out pieces that the competitors will now pile on (since they are fast followers) and leave us wishing for an update…

Happy Friday!

 

I noticed this AirWalk infographic on small cells and thought I would share. First, with cool Ubiquisys enterprise small cells info graphic, next the AirWalk.

 Total coincidence but the other day I was just trying to get in touch with AirWalk when I noticed their management had changed. (it had been a while) So imagine my surprise (not) when I see that Ubee Interactive is purchasing their assets for $45M. AirWalk is a venture funded startup that has got a little traction with Sprint, they have an enterprise femto that supports 1X and EVDO. I remember when Jake Han left us at Samsung to found AirWalk so the arc is complete. The first product line was macro base stations for CDMA, then that quickly changed to micro, then a change to small cells a few years ago. Anyway, Ubee seems to be a better runway for small cells strategy in general as they seem to have lots of funding, some success with their cable modems (complimentary product lines) and a strong Asian connection. Some of the team in Richardson, TX will likely have to move to Denver, Co. That’s an upgrade too!

VoLTE

VoLTE

Voice over LTE (VoLTE) is the next big thing. In fact, in 24 months, virtually all LTE enabled smart phones will support it. Curious?
Small Cells

Small Cells

Small Cells, previously known as 'femto' or 'pico' cells are possibly a savior to network operators. They offer capacity and coverage to the end user and are inexpensive for the network operator. Why aren't they everywhere?
Public Safety

Public Safety

LTE is and ideal technology for Public Safety use. See Why.