Commentary 4Mobility - Getting It Right:
* 3GAmericas.org GSM Voice Capacity White Paper
* Two Different Ways (EV-DO)
By Andrew M. Seybold
23 July 2002
Last week, after a barrage of press releases (one a day for a solid week) the 3GAmericas
organization (read "TDMA/GSM/GPRS/EDGE/UMTS" organization) held a Web cast about a new
white paper it published (www.3gamericas.org) that is designed to prove that the "new and
improved" GSM technology beats the pants off of CDMA2000 1x when it comes to voice capacity.
Since the Web cast, I have received a number of calls and emails asking me to comment on both the Web cast and the white paper. My comments? Show me a system thats up and running using all of this new technology in 40 or more cell sites, and why wasnt AT&T Wireless involved in the call?
Instead of discussing my views here, I would like to refer you to a report prepared by the very
knowledgeable folks at Deutsche Bank (U.S.). I think they have it right and we have permission to
provide a copy of their article. Click on the following link to read what they had to say:
www.outlook4mobility.com/special/S2N072202.pdf
See "Alice in GSM Fantasyland" on page 3).
==========
Pushing The 3G Envelope
I received the following email and thread this past week. Since I was already working on a
commentary that presented my ideas about CDMA2000 1x EV-DO network deployment, I
decided, with permission, to include Scotts ideas.
==========
Email from: Scott Zion
Hello Andy, I wonder if you have any thoughts on an idea I've been contemplating. In short, the idea
is for U.S. CDMA operators to collaborate on a national 1xEV-DO network (e.g., "Project
Rainbow") for a limited time (maybe 3 years). To get things started, QCOM should provide large
vendor financing. In addition, pricing should follow your model for various devices. One last point,
the proposed network would not be sold as a replacement for broadband to the home...it's purpose
is to spark high-speed WAN mobile data. --Scott
==========
Email Response
Problem: Qualcomm and the rest of the CDMA community realize that 1xEV-DO offers an
excellent opportunity to deliver serious WAN mobile data services. However, in the current
economic environment, CDMA operators and their investment community are unwilling to accept
the risks associated with a new venture and an uncertain ROI. As a result, CDMA operators will
be hesitant to deploy 1xEV-DO.
Solution: It is time for Qualcomm to unite the CDMA operators in the nationwide deployment of
1xEV-DO (e.g., "Project Rainbow"). Deployment risks would be shared, which would allow for
easier access to capital.
To get things started, Qualcomm should agree to provide about $1B in vendor financing. CDMA
operators would provide access to existing infrastructure and provision frequency bands
(dependent on availability, size of operator, POPs, etc. to be worked out beforehand) such that
5 MHz (two carriers) would be available in each market for 1xEV-DO. Operators would also
agree on data roaming rates.
Advantages
* verify the superiority of CDMA
* strengthen CDMA operators while weakening their competitors
* provide an immediate market for 1xEV-DO and grow economies of scale
* etc. etc. (you get the idea)
One additional thought: The CDMA community could unite to purchase Nextwave spectrum for
a nationwide 1xEV-DO deployment. BTW, I'm not sure whether the DOJ and/or FCC would
allow this.
==========
My Take
I read Scotts email with great interest because I and a few others have had similar ideas.
Lets set the stage here:
1) CDMA 1xEV-DO is up and running in Korea and is, therefore, proven technology
2) Verizon and others are trialing EV-DO in several areas
3) The uptake on 2.5G and 3G wireless data services in the U.S. is not setting the
world on fire
4) Before ANY CDMA operators decide to roll out CDMA2000 1x EV-DO, they will need
to see a much higher user rate for their existing CDMA2000 1x packet-data services.
It is also my view that even when this uptake occurs, the majority of the CDMA
operators will deploy CDMA2000 1x EV-DO only in specific parts of their networks. One of
the reasons for this is that EV-DO devices will be backward compatible with 1x data
services. Operating a network that is part EV-DO (where there is a high demand for
data) and part 1x makes a lot of sense.
As demand for faster data increases, CDMA network operators will add EV-DO to more of
their networks. However, it will take many years for an entire CDMA network to be
EV-DO capable, and only if there is demand.
By the way, I believe that the GSM/GPRS/EDGE/UMTS community will do the same. I
have been told that in most of Europe, UMTS systems will only be built out in major
metro areas and that users will "fall back" into GSM/GPRS when they travel outside these
areas.
You cant fault the wireless operators or either technology for this. Building out either 3G
system will be costly and operators are less willing to take a risk ("build it and they will
come") than they were during the dot-com era when it was thought that everyone would
get rich by providing the wireless Internet.
The above comments are based on wireless operators having to foot the bill for the total
infrastructure, back-haul, billing and other costs associated with getting the systems up
and running. I would like to take Scotts idea and put in a few twists of my own.
Suppose for a moment that all of the CDMA operators in the U.S. got together with
Qualcomm throwing some money into the pot. Now suppose there were some new
partners added to the mix. New partners you ask? What kind of new partners?
How about content providers? Or Disney, MTV, AOL, MSN, Yahoo and others? What if
each of these companies kicked in some bucks and in return got to sell EV-DO services to
their own customers? What if this nationwide EV-DO network could also be used for data
connections in homes where cable or DSL isnt available? Perhaps we could get Earthlink to
buy into this as well as one or two other ISPs, even some regional ISPs?
Isnt this what Nextwave was planning to do? Wasnt it going to build out a network and
then "rent it out" to anyone who wanted to resell the service? As Scott pointed out, the
Nextwave folks might like to come to this party and bring their spectrum with them!
The costs and the network would be shared and multiple companies would sell services.
Think about what a powerful network this would be, how fast it could be built and how
many years it would be before GSM/GPRS/EDGE/UMTS folks could catch up. Such a
network would clearly put the U.S. back in the forefront of wireless!
Now its time to stop thinking and start doing!
Andrew M. Seybold <andy@outlook4mobility.com>
++++++++++
Alice in GSM Fantasyland (Deutsche Bank)
Last Thursday we participated in the 3G Americas conference call, along with representatives from GSM operators, the 3G Americas organization (a pro-GSM organization) and a consultant that had been commissioned by 3G Americas. The purpose of the one-hour conference was to demystify the topic of GSM voice capacity. Unfortunately, for us it had the opposite effect.
Shortly following the conference call the CDG (a pro-CDMA organization) published its press release on voice capacity in CDMA2000 networks in a not so subtle attempt to refute statements made earlier in the day. After listening to the call and reading the white paper, which is available at the 3G Americas web site, we could not help but enter the ring with our own opinion.
Before we launch into a myriad of points in which we disagree with the white paper, we would like to point out those areas in which we basically agree with the paper. 1) GSM subscribers far outnumber CDMA subscribers. 2) Technology improvements can increase GSM voice capacity. 3) CDMA-based technologies (e.g., WCDMA, TD-SCDMA and CDMA2000) feature increased voice capacity over TDMA-based technologies (e.g., ANSI-136, GSM and PDC). 4) Today, GSM handsets typically cost less than CDMA handsets to manufacture. 5) When CDMA was first launched, the CDMA community was probably unable to deliver the voice capacity it claimed it could achieve at the time. Now, we will look at those points in which we disagree.
The white paper immediately begins with several misleading statements that set the tone for the rest of the paper. GSM is called the "de facto standard" with "capabilities today that match or exceed competing technologies." We would argue that at least 50 CDMA networks, 120+ million CDMA subscribers and the ITU would have a differing opinion. VHS is the "de facto" standard for VCRs instead of Betamax. GSM is not a de facto cellular technology. However, we take even more exception with the statement about GSM’s current capabilities in comparison with other technologies.
Sometime over the Rainbow
The biggest fallacy with the paper is that it treats future GSM voice capacity enhancements as being already proven and commercially available today, while presenting results that were made under the most optimistic scenarios. Conversely, it presents CDMA data that is ultra-conservative, to the point of being inaccurate. It also assumes that CDMA2000 is a static technology without ever considering that it has its own technology path that will also introduce voice capacity enhancements in the future.
We have read about Adaptive Multi-rate (AMR) vocoders and dynamic frequency and channel allocation (DFCA), with it always being another year away from being commercially available. As we understand it, AMR, as a stand-alone technology, does not increase voice capacity. However, if the voice signal is encoded using a lower coding rate, additional overhead can be allocated to error control. This error control can then be used to maintain a voice call with a weaker signal. Additionally, AMR is essential to help mitigate some of the interference introduced with the proposed voice capacity enhancements (half-rate channel mode, tighter frequency reuse, DFCA).
Half-rate mode (assigning 2 users to one timeslot) reportedly increases GSM capacity by 150% if all of the active GSM handsets in the network are AMR-enabled. However, the capacity claim is based upon the premise that essentially all of the radio channels have a high signal quality (C/I) that can support the half-rate mode. When the half-rate mode is used, there is significantly less error correction along with a lower voice coding rate. Voice quality is also impacted for those subscribers that care about quality.
Like 1xRTT, the handset must support the respective technology as well in order to provide the capacity boost. We asked Cingular when it expects all of its newly sold handsets will be AMR-enabled, but the question was not answered (Cingular did say they will introduce its first AMR handsets next year). Verizon and Sprint have been selling 1X handsets for several months (PCS since last year).
The paper then proposes that DFCA, a technology that has been proposed since 1995, but never implemented, can be used in addition to the half-rate mode to increase capacity another 40-50%. With DFCA, the network continuously monitors and adjusts assigned radio channels to each subscriber to ensure the optimal channel is assigned, effectively allowing more active channels (subscribers). Reportedly, this occurs without introducing noticeable interference.
Another way to look at it is that DFCA introduces near continuous handoffs in a GSM network (a GSM timeslot is 4.6ms in duration). Since subscribers constantly move throughout the network and in relation to each other, not to mention new calls being made or existing calls being terminated, the mathematical algorithm and process required to notify each base station and active terminal of its newly assigned channel is also extremely complex.
DFCA also requires a gps receiver in each base station to ensure highly accurate time synchronization throughout the network. While its cost is relatively low in comparison to the cost of cell splitting, it cannot be ignored. Cingular is reportedly going to try DFCA in its network next year so perhaps DFCA’s time in the spotlight has finally arrived. Our chief concern with DFCA is that we are very skeptical about the claimed capacity improvements, especially since the paper assumes DFCA can be used along with the half-rate mode, with both technologies achieving their maximum possible results without degrading voice quality. AMR must be one heck of a vocoder.
Head to Head Comparisons with an Uneven Playing Field
Ultimately, the paper presents its comparison between GSM, CDMA2000 1x and UMTS, suggesting that GSM-AMR offers a 16x improvement in voice capacity over Analog, CDMA2000 1x offers a 17x improvement and UMTS offers a 20x improvement over Analog. Once DFCA is introduced, GSM capacity increases to 19x Analog, surpassing CDMA2000 1x.
First, we take issue with the CDMA2000 estimates -they are ultra-conservative to the extent that they are wrong. SK Telecom has used CDMA2000 1x in its networks since October 2000. Other operators, many with whom we have spoken, including SKT, have since deployed the technology and can offer real world data, not theoretical estimates that could be biased by someone that was pro-GSM. Our data suggests that today’s commercial networks with CDMA2000 offer a 26x improvement in capacity over Analog. The GSM-AMR estimate, or for that matter the UMTS/WCDMA estimate, used in the comparison is not based upon actual results since it is not commercially available.
CDMA also has its own evolutionary path that will introduce additional capacity gains, beyond those cited in the report, the first technology being SMV. The Selectable Mode Vocoder (SMV), as the name implies, is a vocoder that can use a lower vocoder rate, somewhat like AMR, to increase capacity (claimed to be 32x Analog). SMV is part of the MSM6000 chip series and will be available in handsets next year. The paper does mention SMV, but does not discuss the potential capacity gain in a CDMA network.
However, the paper does suggest that GSM could use a similar technology (specifics were not provided) to increase voice capacity another 20% beyond that claimed with the half-rate vocoder and DFCA combination. In our view, the author is double counting capacity gains since he has already included the contribution from AMR in his calculations. We also wonder why the report does not provide the capacity gain for SMV in CDMA networks since at least two U.S. operators have already tested it.
Finally, CDMA2000 can use antenna diversity to further boost capacity. Diversity does require new antennas in the base station and handsets which support it (handset antennas are very inexpensive). With diversity, CDMA2000 networks should be able to support 47x the capacity of Analog networks. Today’s chip sets also support diversity.
Evolution or Replacement?
The paper also tries to lump UMTS/WCDMA and GSM technology together, calling WCDMA an "evolution" from GSM and a "compelling upgrade from GSM." If we believe the paper, "UMTS does not involve a complete replacement of user equipment and infrastructure."
UMTS, in reality, is 100% different from GSM in the handset, the radio access network, and the core network. It is not a compelling upgrade, it is a completely new technology that most GSM operators decided to use in order to deliver 3G services (China Unicom is a notable exception). A more accurate statement would be that once a GSM operator installs the data elements into its core network (SGSN, GGSN, etc) to provide GPRS and EDGE services, the operator should then be able to reuse those core network components when it deploys an entirely new radio network. In other words the operator must eventually replace its entire network, but not all at once. The operator could also maintain GSM/GPRS/EDGE services for its 2G/2.5G subscribers after launching WCDMA since its core network can support both technologies.
Further, all of our research and discussions with industry contacts suggest that those operators who initially deploy Release ’99 of the 3GPP standard, including EDGE, will require an entirely new core network (perhaps a new RAN as well) if they want to offer Release 5 services (real-time multimedia services, VoIP). The implication is that U.S. operators who are deploying GPRS today and EDGE next year will have to replace most of their network to deliver real-time voice services on EDGE. We also point out that the 3GPPhas not yet frozen the Release 5 specification, let alone completed the Change Request process, design work or interoperability testing. Release 5 is still several years from being commercially available and is not yet a proven technology.
Some readers may be a bit confused about the some of the terms we have introduced so far. As we discuss in our 3G report, the 3GPP is the standards board that is developing the UMTS standard, which is actually comprised of multiple releases (Release ’99, Release 4, Release 5 and Release 6). The 3GPP2 is the CDMA2000 standards board. CDMA2000 is comprised of CDMA2000 1X, 1xEV-DO and 1xEV-DV. Read our report for a more detailed explanation.
With respect to EDGE, the paper does correctly point out that most of the new GSM/GPRS networks being deployed today are EDGE capable. What the paper fails to mention is that these networks are EDGE Release '99 capable and not most likely not EDGE Release 5 capable. The paper also does not give any indication when EDGE handsets will become available or how much processing power is required in the handset to support VoIP. We submitted the question, but it was not answered.
What about Data?
Actual data rates (not theoretical rates) were not mentioned in the white paper or discussed in the conference call. We submitted a question to the CTO of Cingular that asked the operator what kind of data rates it was measuring in its network, but the question was not answered. Readers of our June 17 th S2N may recall that we measured rates on Cingular's network that were as high as the mid-20kbps, but that we were frequently being dropped from the network due to congestion. Last time we checked, we were consistently measuring average rates above 60kbps and as high as 112kbps on Verizon’s 1X network, although we note that results from our initial trial when the operator first launched its network were rather dismal (sub-20kbps).
The paper claims that EDGE data rates should be 3x that of GPRS (we agree), or 80-130kbps (we are skeptical). The problem we have is that the GSM operators are not willing, or able, to dedicate 4 concurrent timeslots to one user as the report suggests. It is the implementation of the technology, not the technical specification, that determines actual data rates - just ask the CEO of Vodafone.
Finally, the paper compares data rates between UMTS and CDMA2000. However, the paper uses Release 5 of the 3GPP standard as the basis for WCDMA rates instead of Release ’99 of the standard. We note that our research suggests that the 3GPP may even decide to allow 16-QAM (the modulation technique needed to achieve the highest data rate) to become optional instead of mandatory.
For CDMA2000, the paper uses 1xEV-DO to compare with Release 5, while correctly pointing out that 1xEV-DO does not support voice (Release 5 does support voice and data). However, the paper completely ignores 1xEV-DV, which we believe is the more accurate comparison and which supports voice and data. The 1xEV-DV standard is also an approved ITU standard, unlike Release 5, which is still being developed. For the record, we were a bit surprised that the 3GPP2 recently published the 1xEV-DV standard since our research conducted earlier in the month suggested the standard would not be completed until this fall.
In our view, it would have been more accurate and fair to compare measured data rates from 1xEV-DO commercial networks in South Korea, data rates from NTT DoCoMo’s FOMA network (pre-Release ’99) and UMTS/WCDMA data rates from trial networks in Europe. Perhaps, the paper could even present measured data rates from Verizon’s Express Network and Cingular’s Internet Express network (we could supply the data). At the same time, the paper could point out that 1xEV-DO does not support voice, but that 1xEV-DV, which like Release 5 is still a few years away, will support voice and data while delivering voice capacity comparable to 1xRTT and data rates comparable to 1xEV-DO.
Conclusions
Last week, nearly all of the major OEMs lowered their estimates for wireless cap ex infrastructure spending and handset unit shipments. Since we had already lowered the bar in late May, we remain comfortable with our earlier estimates. 2002 is turning out to be another wash like 2001, and 2003, at least for capex spending, is not looking too encouraging either.
We also provided our review of the white paper that the 3G Americas commissioned and which is available on its Web site. We believe that the report presented a very biased view of GSM technology and an overly optimistic outlook about the potential of future technology enhancements that can increase voice capacity in a TDMA-based network. We believe the technologies may work, but not nearly to the degree suggested by the author.
In addition to presenting unrealistic GSM capacity claims based upon simulations and theoretical models, the paper understates the voice capacity of commercial CDMA2000 networks today and assumes that CDMA is a static technology that will not introduce its own improvements to further boost capacity and/or data rates. CDMA2000 1xEV-DV was not even mentioned while Release 6 of the UMTS/WCDMA standard was discussed at least two times. The 3GPP has barely even begun determining what functionality to include in this release.
Finally, the paper inaccurately lumps WCDMA and GSM together, suggesting that WCDMA is a natural evolution from GSM. WCDMA is not a natural evolution from GSM, it is an entirely new technology that may even require an additional swap out of equipment to support Release 5 of the standard.
After last week’s conference call and the CDG press release, followed by our S2N commentary this week, the debate is far from being concluded. Since network simulation modeling is not an exact science and is subject to a lot of interpretation and criticism, we have our own suggestion on how to resolve the debate.
We invite the GSM and CDMA operators in North America to open up their networks and share actual performance characteristics with the public (or with us). Network operators have access to this information already and if it is as impressive as they claim, we are sure that the public and the investment community would relish the news. Another option would be for us to go out and get 50-60 of our closest friends to try and place simultaneous calls within the same sector of a base station (sort of like trying to squeeze the next guy into the backseat of the VW). Stay tuned for more news from the wireless world… |
