This page is an archive from our previous website. Please check out our new website where you can read new COMMENTARY eNewsletters, TELL IT LIKE IT IS blog posts or Press Releases.
Broadband is one more tool-not the only tool we will ever need.

Channelized vs. Broadband

Saturday, November 07, 2009

People brought up in a channelized world are often confused when faced with the new broadband technologies. Those of us who grew up with wired and then analog cellular phone systems know that a channel is a single voice path that is created when we dial a call. We have our own circuit or channel back to the hub (a wired switch) and outbound from that switch until our call is completed.


Wired telephone systems started out with one pair of wires that was assigned a number to each phone. When we dial a call, it is routed through switching centers until it is connected to the correct pair of wires on the other end. This system has worked well for years, though there are times such as Mother's Day when the switches run out of capacity to assign calls to the proper pairs. In a wired world we don't have to worry about bandwidth because everyone has their own.


When analog cellular was first turned on, it worked in a similar fashion. Once a call was placed, it consumed a single radio channel until it was completed. If the call was between two cell phones, it consumed one radio channel from the calling phone to the nearest cell site and one more radio channel from the cell site closest to the receiving phone. With cell to landline, it consumed one radio channel to the nearest cell site and one wired circuit from the nearest wired switch to the receiving landline phone. While this worked well, it also meant that cell site congestion could occur if too many people within the same cell site were trying to make or receive calls. The only real difference between a wired and wireless call was that as you moved around in the wireless world, the channel you were using might be switched to another channel at another cell site, but it would still occupy a dedicated channel until you were finished.


Once we got to digital, things began to happen differently. In the digital wired world it is possible to put two or more calls on the same phone line, or a voice call and DSL for Internet access on that single line with no degradation of voice quality or slowing of data rates. In wireless, once we got to GSM and CDMA, two types of digital, we could use the same amount of spectrum to process more calls than when we had to dedicate a single channel to each call. With GSM, the calls are broken up into time slices and two or more are combined, using what would normally have been a little more than one analog voice channel. In CDMA, many of the analog voice channels were combined into one broader channel on which multiple conversations could take place.


Going to digital increased the capacity of our wireless networks by ten times or more-the same amount of spectrum in a cell site could handle approximately ten times as many calls as when calls took up an entire channel. As digital systems advanced, the call capacity per amount of spectrum has increased.


But in second-generation systems, if you tried to put packets of data into the same amount of spectrum that carries several voice calls, you would find that data speeds were limited because bandwidth is limited. This is why data rates for CDMA2000 1X and GSM/EDGE are only about 100 Kbps. In the simplest of terms, if you are using an analog to digital voice converter (vocoder) that requires 8 Kbps of bandwidth and you want to send and receive data at 64 Kbps, you have to aggregate eight voice channels to have enough bandwidth.


Enter Broadband


When third-generation services became available, we began to see some decent wireless data rates. At first, UMTS (wideband CDMA) was capable of only about twice the speed of GSM/EDGE (256 Kbps) and this was in a bandwidth of 5 MHz. But this was quickly improved and today that same 5 MHz of spectrum can provide service for numerous voice calls along with a number of data sessions at speeds higher than 1 Mbps. And with HSPA plus, data speeds of up to 7 Mbps are claimed. On the CDMA side of the aisle where channels are 1.5 MHz wide, 1 Mbps+ data speeds were realized by using one channel (1.5 MHz) for voice services and another channel for data services (EV-DO). One reason CDMA can deliver the same data rate in 1.5 MHz as UMTS can deliver in 5 MHz is that CDMA does not mix voice with data. However, the UMTS network has more capacity because it is a 5-MHz channel where the CDMA system uses channels that are only 1.5 MHz wide.


As we move toward 4G or OFDMA technology, we will be able to move a lot more bits per second in the same amount of spectrum so our data capacity will be even greater. However, 4G networks were designed from the beginning as data networks. Any voice will be packetized as with VoIP, One Voice, or one of the other packetized voice systems. Packets will be packets regardless of what they contain. This is one reason 2G and 3G systems will be around for a long time. They can handle voice less expensively than 4G systems can today and they are already built out. If you add voice to a predominately data network, you have to worry about the exact timing of the packets that carry voice. They have to arrive on the other end quickly enough to be reassembled as audio that can be understood and has the characteristics of the speaker's voice. Data packets can be delayed or arrive in mixed order and when reassembled the data still reads or looks correct.


The world we are heading into is dramatically different from the world we are coming from. We grew up with voice systems and then data services were added. Now we are heading toward data services where voice might be added someday. But there will still be a bandwidth cost difference between delivering a voice call and delivering a data call. If we go back to a standard voice vocoder that takes 8 Kbps of bandwidth and measure it against streaming video that is being delivered at 2 Mbps, the theoretical cost difference is huge.


Let's say that the voice call costs $0.05 a minute and takes up 8 Kbps of capacity. Meanwhile, the video takes up 2 Mbps of capacity or the equivalent of 250 voice calls. What this means is that the data we download for that one video is worth $12.50 a minute! Yet we don't begin to pay that kind of money for our mobile data services. We can also see from this example that network operators need more spectrum to deliver data than to deliver voice services.

Note: The numbers used above are based on published per-minute costs and may not represent actual costs. For example, Barney, one of our consulting partners, provides a more real-world scenario for his iPhone "Andy, consider this. Last month I used 126,881 KB for $30 on my iPhone, which comes to $.000236 per KB. If I used my 450 non nights and weekends minutes in the plan for $39.99, each minute would cost $.0888. However, I only used 83 minutes so each minute cost me $0.4819, nearly 50 cents a minute. So 83 minutes at 8 KB is 664 KB for $40 or $.06 per KB. So voice costs me about 250 times per KB more than data." However, for a network operator, the cost of delivering data is higher than the cost of delivering voice because it requires more bandwidth, and many data plans are capped at a specific amount of data per month as opposed to the iPhone pricing model.


Typically, the next question is why we can get to 2 Mbps or even 5 Mbps of data in a wired system using a single wire and still fit in at least one voice channel. The answer is that a copper wire provides more bandwidth than we need for voice and, over the years, engineers have learned how to use that extra bandwidth to deliver DSL and other services. Another way to look at this is that ten years ago cable operators offered 30 or 40 channels. Today, on the same cable in your house, they offer 150 channels, 5-10 Mbps of data, plus one or two voice calls. They have changed their back-end plant, but the cable in your house is exactly the same as it was ten years ago. It now has twenty times the capacity because it already had that bandwidth capability, we just didn't have the technology to use it.


So you might say that today 1 MHz of spectrum has more capacity than ten years ago and you would be correct. However, there is a theoretical limit to the amount of data that can be crammed into 1 MHz of spectrum and it is a lot less than can be pushed through a coax cable. In order to duplicate the capacity of a single coaxial cable using 4G technology, we would need more than 1.5 GHz of spectrum[1].


Wire and cable also have finite bandwidth limitations, but we can think of this as every cable and wire having its own spectrum assigned to it, and two cables or wires next to each other can reuse the same spectrum. Spectrum that is confined in a container such as a cable or a wire can be used over and over again, but wireless signal reuse is limited. If you put six TV cables in the same room, you could connect them to six different TV sets and watch six different TV shows. If you tried to put six access points in that same room, all operating in the same portion of spectrum, chances are pretty good that none of them would work.


Just as there is a time and place for different types of wireless services-Bluetooth for the last 30 feet, Wi-Fi for the last 300 feet, wide-area networks for the last mile-it should be recognized that broadband services don't have to carry all of the voice and data services we want and need. It is more practical and less expensive today to use 2G and 3G services for voice and save 4G for high-speed data. We don't need 4G to hold a voice conversation or to send a test message or tweet, but we do want 4G when we watch the latest video on YouTube, share a video with our friends, or move a file between our office and our handheld in only a few seconds.


We will have all of these technologies and more available to us for a long time. I would like to think that we are smart enough to understand that we don't have to use the newest technology when it is overkill for what we are doing. The object of smart devices and smart networks, as I see it, is that the networks will have a variety of capabilities and the devices will select the best one at the time for what we want to accomplish. Broadband is one more tool-not the only tool we will ever need.


Andrew M. Seybold


[1]RG6 coax cable typically used for in-home cable wiring has a rating of up to 1.5 GHz, see

COMMENTS: This is an archived post. Commenting is no longer available.

Frank Bulk - 11/07/2009 21:25:11

You must have read the same article that I did, that talked about how the wireless carriers are valuing the voice, but that its data that they are marketing at relatively a fraction of the cost (per bit), but is driving costs.

Maxim Chartrand - 11/09/2009 16:58:16

Hi Andrew. In the Barney's calculation, can you explain what's exactly the 8 KB used to calculate the equivalent bandwidth of a voice call ?

Barney Dewey - 11/09/2009 18:16:40


There are different vocoder rates for converting voice to data. All that have been used by cellular operators have been more efficient than VoIP. That is a big reason voice has been separate from data on the vast majority of cellular systems. As technology moves forward, trade-offs and new techniques are allowing voice to be used on cellular data networks. Over the years there have been many different vocodar rates, some of which have provided good voice quality and some that have not, we use an vocoder rate of 8 Kbps because it is an good average across all of the different technologies. Some technologies are more efficient and others suffer poorer voice quality.

I believe the 8 Kbps average shows how little data (bandwidth) voice uses compared to other applications on cellular networks. Voice requires a very low latency while most data applications can tolerate much higher latency. Most of the earlier cellular data technologies had high latency. They provided good data (email, etc.) connections but they could not support voice. The new data technologies (e.g., LTE) have addressed the latency issue. However, the standards groups have not yet agreed on voice-over-data standards for LTE. They are getting close (I hope).

I hope this helps. Let me know if I can provide additional comment.


Maxim Chartrand - 11/09/2009 19:32:36


Thanks for your explanation. I agree with the principle of your calculation, but I think there is something wrong about the 8 kbps...

An 8 kbps (kilobit per second) voice channel would generates 4,980 KB (kilobytes) of data for 83 minutes of calls. This means that the cost per KB would be $0.008 and not $0.06 as you mentioned. So, in your example, the cost per KB for voice would be like 33 times more than data.


Barney Dewey - 11/10/2009 10:13:43


My math was wrong. It looks like it should be (8,000 [bps] * 60 [sec] *83 [min] / 8 [bits-to-bytes] = 4,980 KB (kilobytes). So, 4,980 [KB] / $40 = $0.008/KB. I think this agrees with your math.

So if I used all 450 minutes the voice would have cost $0.0015/KB or about 6.4 times the cost of data. The network operators may have a higher cost to connect a voice call (equipment and connection fees) than deliver data but it still seems like they make more on voice than data especially if they require data users to include bulk voice messages.