3Q 2006 Wireless Carrier Results

Bill wrote:
> You also have to add in receiver senseivity/antenna gain on each band.
> Your test is still good because it is testing the raw coverage for that
> phone model to a single tower over two bands. It may be that the
> phone is optimized for 1900 Mhz and will give equal results on both
> bands.

Bottom line is the *ad hoc* test showed similar efficacy of
800 MHz and 1900 MHz when used is a rural environment.

Both Nokia phones were dual-band models.

---

› See More: 3Q 2006 Wireless Carrier Results

SMS wrote:

>
> Thanks for all the interest. The number of hits on this page has been
> extremely high, as a compilation of this data is not available anywhere
> else.
>
> "http://nordicgroup.us/marketshare/3Q2006/"

How about profit per carrier if that is available? While the data is pretty good at showing the
metrics, one can't tell if a carrier is sacrificing profits at the expense of market share or any
other metric.

Thanks,
-Jason

> Bill wrote:
>> You also have to add in receiver senseivity/antenna gain on each band.
>> Your test is still good because it is testing the raw coverage for that
>> phone model to a single tower over two bands. It may be that the
>> phone is optimized for 1900 Mhz and will give equal results on both
>> bands.

Actually you can ignore receiver sensitivity as long as the phone is
well designed and limited by the system noise - which is what I was
suggesting. I also mentioned antenna gains, as part of ERP on transmit
and as part of effective aperture in the handset (receiving).

decaturtxcowboy wrote:
> Bottom line is the *ad hoc* test showed similar efficacy of
> 800 MHz and 1900 MHz when used is a rural environment.
>
> Both Nokia phones were dual-band models.

Yes, that definitiely shows they had the same performance but it doesn't
say too much about the reason or whether there is was a frequency
dependency differentially affecting the transmission losses. For that
you need to do significantly more, as described.

For more detail on frequency dependent attenuation, look for a paper by
Brown & Currie. A microwave-millimeter model from them:

> ke = 0.5*f^(0.75)
>
> where f is in GHz, and ke is attenuation in dB/meter for 1 way
> propagation in the summer for coniferous and deciduous foliage.
> In the winter, ke is about a factor of 2 smaller.

which gives about .44 dB per meter at 850 MHz and .8 dB/meter at 1.9
GHz. If the path from the tower has significant foliage, there will be a
significant increase in attenuation, due only to the foliage, at 1.9 GHz.

g

[email protected]lid wrote:
> SMS wrote:
>
>>
>> Thanks for all the interest. The number of hits on this page has been
>> extremely high, as a compilation of this data is not available
>> anywhere else.
>>
>> "http://nordicgroup.us/marketshare/3Q2006/"
>
> How about profit per carrier if that is available? While the data is
> pretty good at showing the metrics, one can't tell if a carrier is
> sacrificing profits at the expense of market share or any other metric.
>
> Thanks,
> -Jason

Yes, I could add the profit and the margins, quarter after quarter.

However it really doesn't tell you about sacrificing profits for market
share. AT&T and Cingular had low margins at the height of the TDMA to
GSM conversion because of the high expenses of the network conversion,
compared to Verizon which has enjoyed very high margins. Now Cingular's
margins are increasing because the network conversion is complete.

g wrote:
> Yes, that definitiely shows they had the same performance but it doesn't
> say too much about the reason or whether there is was a frequency
> dependency differentially affecting the transmission losses. For that
> you need to do significantly more, as described.

As far as the typical user-facing experience, results were similar.

"decaturtxcowboy" <[email protected]> wrote in message news:[email protected]...
>g wrote:
>> That's a tough comparison to make accurately both because of the
>> absolute accuracy of the RSSI built into the phones and because of the
>> variability due to propagation. Still, an interesting one to attempt!
>
> Actually, reading the dB scale wasn't necessary. Most importantly we
> did a side by side comparison from same cell tower location and similar
> preforming phones. So the only significant variable would be the
> propagation difference of the two bands (assuming similar radio
> performance).

Variations in the nominally comparable RF performance of the phones
could easily mask the 7 dB difference to be expected from the
operating frequency difference. These types of studies are normally
carried out with spectrum analyzers costing thousands of dollars.

--
John Richards

On 2006-11-12, SMS <[email protected]> wrote:
> Perhaps, but this is demonstrably untrue in suburban areas, where the
> tower placement is to cover a geographic area, and the capacity is not
> the issue.
>
> T-Mobile is great how their web site lets you go down to a specific
> address, and you can clearly see the gaps caused by insufficient towers
> in many areas.

I wish Verizon had a map like that. I'm not sure it would have fewer
gaps in suburban areas (like where I live), though there is no way
to tell currently.

> I don't think that anyone argues that 1900 MHz has as much range or as
> much penetration as 800 MHz. Not even Navas would claim something like
> that. The rule of thumb has always been 2x the distance, mathematically
> it's more than 2x, but their are other factors (geologic features,
> buildings, etc.) that make the increase in range less than ideal.

I think this is a bit bogus. I believe the "math" you are considering
is the path loss alone which, as another post here pointed out, is about
7 db greater at 1900 MHz than at 850 MHz. A 6 db difference in the
overall system budget would indeed be expected to double your
line-of-sight distance, so the decrease in frequency might double your
distance if nothing else in the system changed with frequency. It is
the latter assumption which is highly dubious; other things change
with frequency too.

As an example of what, suppose we were talking not about cell phones
but rather a point-to-point microwave circuit that we were picking a
frequency for. Suppose the transmitter power and parabolic dish sizes
were fixed. In this case, I think a 1900 MHz link would give you twice
the distance of an 850 MHz link. While 1900 MHz would increase the
path *loss* by 7 db, it would also increase the antenna *gain* (for
same size antennas) by 7 db each, and since there are two antennas
involved the net improvement at 1900 MHz would be 7 db. Path loss
is not the only thing which changes with frequency, and some of the
other changes can be to the benefit of the higher frequency. In
particular, while path loss increases with increasing frequency,
antenna performance, given realistic constraints, tends to improve.

The same situation exists with cell phones, it is just that since
I don't know the particular numbers (I don't think you do either)
I can only guess at how this might come out. Since the output
power on either frequency is constrained to be the same (by the
handset) we're down to a comparison of likely antenna performance.
One thing we know about this for sure is that if the antennas at
the two frequencies are the same physical size, the 1900 MHz
antenna is very likely to have higher gain. This is exactly
the situation of a dual band handset where the same antenna is used
for both frequencies. Worse, a quarter wavelength at 850 MHz is
a highly unstylish 3.5 inches, but at 1900 MHz is is a svelt 1.5
inches, so 850 MHz performance on a modern dual band handset is
increasingly likely to be compromised with an electrically short
antenna. I hence wouldn't be surprised if the handset antenna
gain gave 1900 MHz a 2-5 db advantage (look at 802.11a/g dual band
access points, which sometimes also share antennas at two frequencies;
if they list antenna gains for each frequency then you'll generally find
802.11a has an advantage of several db, even without the antenna
size constraints that are important for a handset).

At the base station the same issue applies, though slightly differently.
Since it is easier to build higher gain antennas at 1900 MHz than at
850 MHz, you can always compensate for any remaining path loss advantage
at 850 MHz by deploying higher gain 1900 MHz antennas. The problem
with this at a base station, however, is that with the highly efficient
antennas in use there (unlike the handset), higher gain implies greater
directionality, so the only way to use higher gain antennas is to
increase the sectorization, which seems certain to increase cost if
you don't need the added capacity as well. So, while I would agree
that building a 1900 MHz network in less densely populated areas
is likely to be more expensive than 850 MHz given equivalent coverage,
I don't agree at all that this extra expense necessarily needs to be
incurred by building more towers; it can also be incurred by spending
more money at the towers you've got.

And while the above necessarily includes some speculation about how
it works, what I can tell you for certain that it doesn't seem
to work your way, with Verizon having this incredible coverage advantage,
where I live because I did that experiment. I live in a coverage seam
which shows up at 1 bar on the T-Mobile map. I also know for
certain that my service comes from a cell site on top of a building
downtown where all 5 carriers have equipment (I know this because there
have been articles in the local paper about attempts to construct more
towers here which mentioned where the existing cell sites were). I
was a Sprint PCS customer several years ago when I moved into that house.
My phone worked okay on the second floor of my house, was marginal on the
ground floor, and wouldn't work at all in the basement except near the
windows. I eventually got fed up and, even though I was still under
contract with Sprint, I went to Verizon and bought a phone. I bought the
same model phone from Verizon (a Samsung, I think) that I had with Sprint
and took it home. It helped nothing. The maintenance display which
showed the received signal strength in db(m?) showed virtually no
difference; I think Sprint was in fact usually a db or so better, for
whatever that's worth. The phones were equally good upstairs, equally
marginal on the main floor and got service (or not) in pretty much the
same locations in the basement. There being no detectable improvement
with Verizon, I took that phone back (and no one in the shop seemed
surprised the phone didn't work well in my neighbourhood). I've since
had T-Mobile, and now Cingular service, and I've replaced Sprint with
Verizon for other reasons, but there's really little to choose between
them for coverage at my house. There is certainly no 850 MHz advantage
in the suburbs where I live.

In any case, if you want to persist in arguing that 850 MHz provides
a 2x distance advantage over 1900 MHz, you can't just quote the path loss
and stop. You've also got to explain why the antenna gains at
both frequencies would need to be identical, and my understanding of
the problem leads me to believe you'll have some trouble doing this.

>> Obviously in rural areas the 800MHz carriers have an advantage where
>> capacity isn't an issue, and distance is the limiting factor.

I believe there is a a cost advantage at 800 MHz, I just don't believe
the extra money for 1900 MHz necessarily needs to be spent on more towers.
It can also be spent on improving the coverage of the towers you've
got.

>> I remember in the late 80's a rural Nebraska cellular carrier (aptly
>> named "Nebraska Cellular") managed to provide excellent cellular service
>> along I-80 through almost the entire state with a minimal number of
>> towers thanks to 800MHz propagation and some VERY flat terrain!
>
> Yes, this is the big advantage of AMPS, at 800 MHz. The hope is that if
> AMPS ever gets turned off in those rural areas, that something will take
> its place, maybe something like Australia did with CDMA.

If there was an advantage for AMPS it was entirely due to the use of
3 Watt car phones (i.e. an 8-12 db advantage in power output over
a modern handset) with an antenna outside the car, as was common in
the late 80's. If you are power limited, as modern handsets are,
either CDMA or GSM will do better than AMPS under pretty much any
common set of circumstances.

Dennis Ferguson

Dennis Ferguson wrote:

> I think this is a bit bogus. I believe the "math" you are considering
> is the path loss alone which, as another post here pointed out, is about
> 7 db greater at 1900 MHz than at 850 MHz.

The "path loss" is constant with frequency if one considers a fixed gain
(electrical size) on one antenna and a fixed aperture (physical size) on
the other. The term 'path loss' was originally used (and misapplied)
modeling links with isotropic antennas at each end. Modeling that way
resulted in constant ERP but diminishing receive antenna aperture. The
result was an apparant loss of signal as frequency increased.
Interestingly, this was probably responsible for commercial and military
users of the radio spectrum considering everything of shorter wavelength
"200 meters and down" fairly useless. Actually though, energy is
conserved and no power is lost on a freespace path; the receive antenna
'bucket size' is just getting smaller with increasing frequency.

To empirically compare path attenuation one needs to know the ERP (which
takes care of both transmit power and sector antenna gain) and handset
antenna aperture.

Your comments about handset antenna differences are valid and without
knowing something about them, it's not possible to make a good
comparison of the effect of frequency on path length by comparing two
phones/systems.

Do note though, there is very solid measured evidence for the component
of attenuation due to frequency alone, as I posted earlier. In addition
to the 20*log(1900/850) 'pathloss' component there is also an absorption
component which truly is related to frequency.


> The same situation exists with cell phones, it is just that since
> I don't know the particular numbers

See above. Sector antennas are normally 10-12 dBi; they have roughly 120
degree beam width azimuthally but can have differing beam widths in
elevation.
Handset antennas are similarly hard to pin down, both due to
inefficiencies due to physical shortening you mention and also due to
ill-controlled polarization and siting errors when used in a typical way.

> And while the above necessarily includes some speculation about how
> it works, what I can tell you for certain that it doesn't seem
> to work your way, with Verizon having this incredible coverage advantage,
> where I live because I did that experiment.

But do you know that the ERP of the two systems was identical and that
the effective aperture of the two handset antennas, as well as the
receiver system temperature were identical?

> In any case, if you want to persist in arguing that 850 MHz provides
> a 2x distance advantage over 1900 MHz, you can't just quote the path loss
> and stop. You've also got to explain why the antenna gains at
> both frequencies would need to be identical, and my understanding of
> the problem leads me to believe you'll have some trouble doing this.

This is certainly true. Since there really isn't any 'path loss' one has
to know a lot about the system in order to measure it. However, this has
been measured extensively over real paths and is the source of COST231,
Lee and other models which do in fact show considerable frequency
dependent attenuation in non-LOS environments.

g

On Mon, 13 Nov 2006 18:13:55 -0800, g <[email protected]> wrote in
<[email protected]>:

>Dennis Ferguson wrote:

>> In any case, if you want to persist in arguing that 850 MHz provides
>> a 2x distance advantage over 1900 MHz, you can't just quote the path loss
>> and stop. You've also got to explain why the antenna gains at
>> both frequencies would need to be identical, and my understanding of
>> the problem leads me to believe you'll have some trouble doing this.
>
>This is certainly true. Since there really isn't any 'path loss' one has
>to know a lot about the system in order to measure it. However, this has
>been measured extensively over real paths and is the source of COST231,
>Lee and other models which do in fact show considerable frequency
>dependent attenuation in non-LOS environments.

The key words there are "non-LOS". In such cases, differences vary
considerably from location to location, and no frequency generalization
is truly valid; i.e., there are cases where 1900 will work better than
850 and vice versa. As in some many other things, it all depends.

--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/Cingular_Wireless_FAQ>

John Navas wrote:
>
> The key words there are "non-LOS". In such cases, differences vary
> considerably from location to location, and no frequency generalization
> is truly valid; i.e., there are cases where 1900 will work better than
> 850 and vice versa. As in some many other things, it all depends.

Yes, different environmental models are used to approximate the median
characteristics. And on average, when a large number of measurements are
taken, higher frequencies will incur greater attenuation.

In typical environments here really is a significant, statistical and
practical difference between the two bands.

g

On Mon, 13 Nov 2006 20:37:03 -0800, g <[email protected]> wrote in
<[email protected]>:

>John Navas wrote:
>>
>> The key words there are "non-LOS". In such cases, differences vary
>> considerably from location to location, and no frequency generalization
>> is truly valid; i.e., there are cases where 1900 will work better than
>> 850 and vice versa. As in some many other things, it all depends.
>
>Yes, different environmental models are used to approximate the median
>characteristics. And on average, when a large number of measurements are
>taken, higher frequencies will incur greater attenuation.
>
>In typical environments here really is a significant, statistical and
>practical difference between the two bands.

The empirical studies I've seen are anything but conclusive.

--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/Cingular_Wireless_FAQ>

John Navas wrote:

> The empirical studies I've seen are anything but conclusive.

I don't know what studies you are referring to but
http://www.radyn.com/Papers/gmc/Foliage_Attenuation.pdf might interest you.

This is particularly a millimeter wave study but pages 8-9 and
particularly paragraph 4 on page 9 are relevant and compare results from
several studies which were within 1 dB of each other.

From these, I would say it is conclusive that there is a significant
frequency dependency to attenuation through foliage.

g

On Mon, 13 Nov 2006 21:37:25 -0800, g <[email protected]> wrote in
<[email protected]>:

>John Navas wrote:
>
>> The empirical studies I've seen are anything but conclusive.
>
>I don't know what studies you are referring to but
>http://www.radyn.com/Papers/gmc/Foliage_Attenuation.pdf might interest you.

I'm talking real world surveys of signal propagation in various
environments, particularly urban settings.

>This is particularly a millimeter wave study but pages 8-9 and
>particularly paragraph 4 on page 9 are relevant and compare results from
>several studies which were within 1 dB of each other.
>
> From these, I would say it is conclusive that there is a significant
>frequency dependency to attenuation through foliage.

Sure, but that's only a small factor in very complex circumstances. On
the other hand, higher frequencies are generally better at penetrating
smaller openings. Etc. That's why it's not valid to generalize from
limited data

--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/Cingular_Wireless_FAQ>

Here Sprint PCS and verizon are both PCS. The area covered by Sprint
PCS is larger than verizon. The quality of that coverage is probably
about the same. The two cellular carriers uscellular and cingular cover
a lot more area, but their coverage is not better in the areas also
covered by Sprint PCS .

SMS wrote:

> Yes, I could add the profit and the margins, quarter after quarter.
>
> However it really doesn't tell you about sacrificing profits for market
> share. AT&T and Cingular had low margins at the height of the TDMA to
> GSM conversion because of the high expenses of the network conversion,
> compared to Verizon which has enjoyed very high margins. Now Cingular's
> margins are increasing because the network conversion is complete.

Following up on my own post, I saw a report on the non-voice revenue for
the various carriers, which drives home the point about why the Cingular
ARPU is lagging.

Verizon has an ARPU of $50.59, with non-voice revenue representing $7.16
of the total.

Cingular has an ARPU of $49.76, with non-voice revenue representing
$6.32 per user.

If Cingular raises it's non-voice revenue to the same level as Verizon,
they would have an ARPU of $50.60, one cent more than Verizon. Cingular
has been slow to roll out their high speed data network, which is why
they lag Verizon in non-voice revenue, but that will change in the next
six months to one year.