The HSDPA Air Interface – A Peek In With Wireshark

This blog entry continues my reporting on HSDPA performance in
deployed networks. In part one I’ve been giving a general overview and
comparison of achievable speeds and delay times of UMTS and HSDPA. In part two I’ve presented an inter-packet space diagram for UMTS
to show how air interface bearer changes can be detected on the IP
layer. This part builds on the previous ones and discusses an inter-packet spacing diagram
for HSDPA.

HsdpainterspacegraphaverageconditioSo without further ado, let’s take a look at the inter-packet space diagram for a file download via HSDPA which is shown on the left. For details on how it was generated see the same analysis for UMTS. While the UMTS diagram shows the same spacing between each packet, the graph for HSDPA shows three distinctive horizontal lines. Inter-packet space for some packets is about 10 milliseconds, for others it is 20 milliseconds and for a few it is even 30 milliseconds. This is quite surprising at first as the throughput during the file transfer was stable at around 850 kbit/s.

At a speed of 850 kbit/s, the inter-spacing for IP packets with a size 1500 bytes (=12.000 bits as each byte has 8 bits) is 1 / (850.000 / 12.000) = 14 milliseconds. The diagram shows however, that some packets only have an inter-spacing of 10 milliseconds, while others have 20 or even 30. Taken the percental distribution into account that’s around 14 milliseconds on average.

So why three distinctive lines for HSDPA and not a single line as for the UMTS? My comments in the diagram already give a hint. On the air interface, HSDPA uses frames with a fixed length of 2 milliseconds. The amount of user data they carry (the transport block size) depends on the type of HSDPA terminal used and current transmission conditions. Based on the channel quality feedback of the terminal the base station (Node-B) selects an appropriate combination of modulation (QPSK or 16QAM) and coding. For this example I used a category 12 HSDPA card for which the highest transport block size is 3319 bits. For an IP packet with a length of around 12.000 bits at least 4 air interface transport blocks are required. At 2 milliseconds each the minimum time it takes to transfer a 1500 byte IP packet is thus 8 milliseconds. This is close to the first line in the diagram.

HSDPA has been designed to have a quick response time to packets which were not correctly decoded by the receiver. In fact the system even prefers a certain error rate over an error free transmission as it is more efficient to retransmit some packets than to reduce the transport block size to insert more error correction and detection bits. The time between reception, reporting the error and retransmission on the air interface is exactly 10 milliseconds. This is the explanation for the packets which were transmitted with an inter-packet space of 20 milliseconds and also for the few packets with 30 milliseconds. For the later ones at least one of the 4 required transport blocks had to be re-transmitted twice.

Note that it could also be possible that the scheduler in the base station could also have decided to change the transport block size during the transmission to produce these lines. However, I doubt this as the mobile station was not moving which means signal conditions were quite stable. When changing transport block sizes I would also expect a scheduler not to make such drastic changes which would result in some packets being transmitted in 4 or 5 blocks while others are transmitted in 10 (2 milliseconds each). Thus, I think my analysis above is more probable. No certainty, however, without a network analyzer on the HSDPA MAC layer.

In the next blog entry on this topic I will take a closer look at how small changes in antenna positioning can dramatically effect throughput and cell capacity. If you would like to find out more about UMTS and HSDPA in the meantime, my book is a good companion.

Using Wireshark to Peek Into the UMTS Air Interface

Since having arrived in Italy I’ve enjoyed using UMTS and HSDPA networks to connect to the Internet for my day to day work for reasonable prices. It has also allowed me to run a number of network traces to get some more insight into the performance of the UMTS and HSDPA air interface.  In the first part of this mini series, I’ve been looking at the data transfer rates and packet delay that I could get on some of the networks. While having been used to UMTS speeds for quite some time now I was even more positively surprised by HSDPA speeds of up to 1.5 MBit/s which I was able to reach in the network of Telecom Italia Mobile (TIM). For the details take a look here. Part two of the mini series focuses on how the network tracing tool Wireshark can be used to chase after some UMTS specific air interface phenomena.

Wireshark is a great tool to analyze any kind of network traffic over any kind of PC interface. The picture on the left (click to enlarge) shows a trace which was generated using Wireshark’s "TCP Round Trip Time Graph" statistics module and commented by me. In my opinion the name of this module is quite misleading as the graph does not show the TCP round tip time evolution during a transmission but rather the reception time delta between TCP packets over time. A better name for the analysis module would thus be TCP inter-packet spacing diagram. The y-axis of the diagram shows the time in seconds that has elapsed between two TCP packets while the x-axis represents the time. To generate the diagram I downloaded a large file from an FTP server.

The graph shows that at the beginning of the transfer the inter-packet time was around 0.1 seconds or 100 ms. The TCP packet size of the download was 1500 bytes. These two values can now be used to calculate the transmission speed. One packet every 0.1 seconds means that on average 10 packets are received per second. As each packet has a size of 1500 bytes, 10*1500 bytes = 15.000 bytes are received per second. As each byte has 8 bit the resulting transmission rate is 15.000 bytes/s * 8 bits = about 120 kbit/s. This corresponds quite nicely to the maximum transmission rate that can be reached with a 128 kbit/s radio bearer with a CDMA spreading factor of 16.

At about the middle of the diagram the inter-packet spacing suddenly becomes much smaller, about 35 milliseconds. Doing the same calculation again results to 1/0.035 = 28.5 packets/s. 28.5 packets/s * 1500 bytes = 42.857 bytes/s. 42.857 bytes/s * 8 bits = 342.857 bits/s. This is close to the maximum transmission rate of a 384 kbit/s radio bearer with a CDMA speading factor of 8. In practice this means that at this point there were fewer people using the UMTS cell which in turn reduced the load of the cell. The network then decided to upgrade my radio bearer from a spreading factor of 16 to 8. As can be seen in the diagram the fun didn’t last long as the bearer was reduced again to 128 kbit/s for a short time. Then, I was upgraded again to 384 kbit/s for the remainder of the trace.

The same behavior could also have been achieved by physically moving through the network during the download and thus change reception conditions. As I was not moving with the mobile during the test the effect was definitely caused by changing load and changing interference conditions in the cell. Whether the load was caused by other people doing data transfers or by voice calls can’t be told from the diagram. What can be told however is that the network at this time was highly loaded as the network always assigns the best possible radio bearer.

So much for now. In the next entry of this mini series I am going present the results of the same test performed in a 3.5G HSDPA network. I can already promise spectacular and thought provoking results! Watch this space!

A Real Life Comparison of HSDPA and UMTS

These days I am totally unplugged but still always connected as I am staying in Italy for the moment, far away from my ADSL line back home. UMTS has kept me connected to the world in the past two weeks and I’ve been writing about my general experiences over at m-trends. So far, I’ve used a prepaid card from Wind to keep me connected. As Wind is not offering HSDPA (High Speed Data Packet Access) in their network yet the HSDPA card remained in the suitcase. Over the weekend, however, I’ve bought a prepaid card of Telecom Italia Mobile (TIM) where HSDPA is available. Their offer is not as good as Wind’s, giving me ‘only’ 500 MB for 20 euros over 30 days and an additional 1GB for an extra 20 euros should that not be enough. Still, for my purposes it should be more than enough. I am pretty much impressed by the speed increase HSDPA brings over plain UMTS. Also, responsiveness when clicking on a link in the web browser has noticeably increased as well. For the technical details read on.

The Hardware

HSDPA was standardized in a flexible way allowing data rates to grow as end user devices and networks become more capable. For my test, I used a Sierra Wireless Aircard 850, which supports HSDPA category 12 (inter-TTI of 1, QPSK only), i.e. a top speed of 1.8 MBit/s. Note that there are already category 6 mobiles and data cards available today promising speeds of up to 3.6 MBit/s by using 16QAM modulation in good coverage situations. However, my card is not capable of doing this yet. I am looking forward to compare the speeds of these two categories in a real network once I can get a hand on one. Enough about networks and terminal categories for the moment. For details you might want to take a closer look at my book 😉

Top Speed on Sunday Morning

There can be a big difference between theoretical maximum speeds and speeds that can be reached in a real environment. As I woke up early on Sunday morning I gave it a try when most other people were probably still sleeping, i.e. low overall radio network load from other people making phone calls and accessing the Internet. I was quite positively surprised in my first download test as the average speed for downloading a large file from the internet was about 1.5 MBit/s. Hey, that’s faster than my DSL line in Germany! It looks like TIM has not only upgraded their base stations to HSDPA but also ensured that the backhaul connection from the base station does not become the bottleneck.

I also downloaded the same file via the Wind UMTS network to be able to compare the behavior. As expected, the network load was also low and the download reached the highest possible UMTS download speed of 384 kbit/s. Also very nice but four times slower than via HSDPA.

The image above on the left shows a graph of the download as it happens. I started the download inside the apartment where radio coverage was far from ideal. Nevertheless, it can be seen in the graph that the download speed exceeded 1 MBit/s. Going to the balcony with the notebook after about half the download was finished improved the radio environment and the download speed even further.

Speeds at Other Times

To see how the network load impacts download speeds I ran the same test again at around noon on Sunday. This time my download speed was about 750 kbit/s or about 90 kbyte/s. The corresponding graph for the download is shown in the third image on the left. Note that I did not download the whole file which is why the download graph is not as long as in the previous image. Not quite as high in the morning but still quite respectable.

Web Browsing

The next test on my list was web browsing. I connected one notebook to the Internet via TIM’s HSDPA network and another one via Wind’s UMTS network. Then I surfed to a number of graphics intensive pages such as those from Nokia, CNN and a couple of German news magazines to compare first page display times and overall download times. While UMTS is by all means capable of delivering a good web browsing experience, HSDPA is by far quicker. All pages I tried always started to be shown a couple of seconds earlier on the notebook with the HSDPA connection than on the notebook with the UMTS connection. Needless to say that the time until the complete page is downloaded is also faster. I have to try again when at home with an ADSL connection in reach but I am pretty sure I would not be able to tell the difference between a DSL line and an HSDPA connection for web surfing except for the channel establishment delay described below.

Uplink Speed

TIM has also upgraded its radio network to support uplink speeds of 384 kbit/s. Note that this is not HSUPA (High Speed Uplink Packet Access) yet but plain 3G standards pushed to the limit. Even under average reception conditions, sending an eMail with a 2MB attachment was very quick with an average uplink data rate of about 350 kbit/s. Compare that to most 3G only networks today which usually support 64 kbit/s or 128 kbit/s at the most. 1 MBit/s ADSL connections usually have a 128 or 180 kbit/s uplink. So in this respect, current HSDPA even have a speed advantage in the uplink over a typcial 1 MBit/s DSL line.

Round Trip and Channel Establishment Delay

Round trip delays have also decreased a bit. While 3G connections usually have around 120-130 ms round trip delay times, I measured 90-100 ms to the first hop in the network over the HSDPA connection.

During the test it was also interesting to see that there is still a noticeable delay of 2.1 seconds in ping times or web page access time when no packets were transferred for some time. This is due to the fact that the network releases the HSDPA radio connection after some time of inactivity to reduce the power drain on the mobile’s battery and also the channel usage on the air interface. I experimented a bit and it seems TIM has set the transition timer to 15 seconds. Unless TIM has a stupid network implementation which drops the user to PMM IDLE state after this time, the 2.1 seconds are the time required for the transition from the FACH to HSDPA (DCH).


I am very impressed by the performance of HSDPA. Even my first generation category 12 data card exceeds a download speed of 1.5 MBit/s in a lightly loaded network and still over 700 MBit/s under normal network load conditions during the day. Uplink speeds beyond 350 kbit/s are very impressive as well. With further enhancements like category 6,7 and 8 handsets in the near future, multiple antennas in end user devices, enhanced receivers, improved signal processing, etc., etc., both end user speeds and overall wireless network capacity will continue to grow over the next couple of years. And beyond that, 3GPP Long Term Evolution is already in the pipe which ensures speeds will continue to rise. After all, the You-Tube generation needs as much bandwidth and speed as they can get!

Note: Click on the "HSDPA" category link below next to the date to see all articles on further tests which have followed afterwards.

VoIP capacity over HSDPA

One of the main issues with VoIP over 3G networks is that the number of possible simultaneous calls per cell is much lower today than the number of calls that can be transported over 3G networks in circuit switched mode. This is due to the fact that the radio interface has been optimized on every layer to squeeze through as many circuit switched voice calls as possible. VoIP calls on the other hand are transported over IP which makes it impossible to specifically adapt each layer of the air interface for the application as each protocol layer is independent from the one above and below.

Another disadvantage to transport voice over IP is it’s requirement for real time data transmission. As voice data can be compressed quite well, the required bandwidth is quite small. In order to keep the delay acceptable a single IP packet only carries around 20 milliseconds of speech data. At this rate, the additional information generated by the air Interface, IP, UDP and RTP headers is almost the same as the actual voice data. This doubles the bandwidth required to transport a voice call over IP compared to transporting it over optimized circuit switched channels over the air interface.

As if this was not enough there is yet another problem that plagues VoIP over wireless: While most other IP applications benefit from retransmission of lost or damaged air interface frames, this is most unwelcome for VoIP as it’s better to loose a couple of frames rather than to wait for the retransmission. As the lower layers are not application aware, however, it’s not possible to carry voice and data of other applications over the same connection and treating them differently on the air interface.

HSDPA And Intelligent Scheduling Come To The Rescue

While I knew all this for some time now and was thus a bit pessimistic about mid-term success of VoIP over 3G and WiMAX networks, Harri Holma and Antti Toskala describe in their book about HSDPA, or 3.5G as sometimes called in the press, that VoIP capacity is not necessarily lower than 3G circuit switched capacity per cell. Compared to an average of around 64 simultaneous circuit switched calls per cell as referenced in their book, they present a study which results in at least equal or even higher VoIP capacity in an HSDPA enabled cell. So how’s this possible with all the difficulties mentioned before? Here are the main principles they used for their calculations:

HSDPA Capacity

Due to the use higher order modulation for mobile stations with good reception conditions, better error coding and fast re-transmission, total capacity of an HSDPA cell is twice as high compared to a 3G UMTS only cell.

Use of AMR

Many VoIP implementations today use the G.711 codec for digital voice transmission which requires a bandwidth of 64 kbit/s. For HSDPA cell capacity, the authors used the AMR codec instead, which is also used for circuit switched wireless calls today, which only requires around 12 kbit/s to achieve the same voice quality.

Header Compression

Compressing IP headers of VoIP frames is absolutely essential for capacity. Thus the authors have assumed the use of Robust Header Compression (ROHC) for their simulation. This is quite realistic for the future as ROHC between the mobile station and the RNC is already in the 3GPP standards.

Intelligent Scheduling

HSDPA packets have a transmission duration of 2 milliseconds. A 2ms packet, however, can hold several VoIP packets. To achieve the highest cell capacity the traffic scheduler has to hold enough packets destined for a user to fill up a full air interface frame before they are sent. While this increases the total VoIP capacity of the cell it also has the disadvantage to introduce unwanted speech delay. For their simulation the authors did not queue more than three VoIP packets for a single user. This introduces a maximum additional delay of 60 milliseconds.

Fast HSDPA retransmission

The retransmission problem for VoIP described above is reduced by HSDPA by it’s fast retransmission scheme. A faulty packet can be retransmitted within 10 milliseconds. If air interface parameters are used to ensure that at most two retransmissions are required before the packet can be deciphered correctly on the other end, a maximum additional delay of 20 milliseconds can appear.


Based on the assumption that an additional latency of 80 milliseconds is acceptable to the user, the authors show that HSDPA network can have the same or even better voice capacity than 3G networks have today for circuit switched calls. It’s still some way to go until we are at this point as enhancements have to be made on all parts of the network. But this study impressively guides the way forward!

Ericsson And Telstra Experiement With 200 km UMTS Cell Range

In this press release Ericsson and Telstra (Australia) report that they have successfully tested a range update of Telstra’s W-CDMA UMTS/HSDPA network operating in the 850 MHz band to support cell ranges of up to 200 km. The press release says that downlink speeds of 2.3 MBit/s were achieved over this distance.

It would have been nice if the press release would have gone a bit more into the details of how this was achieved as that sort of range and speed can not be achieved with the typical cell site on a rooftop transmitting at 10 watts and a standard mobile phone in the hands of a user. It is more likely that a base station with high transmit power on an elevated position like a hill was used in combination with a stationary handset, power amplifier and directional antenna.

It would also have been interesting to hear some details from Telstra on where they plan to deploy this. Australia is a big country so I guess there is quite an opportunity this way to bring high speed internet to people living far away from cities where broadband Internet is available either by conventional UMTS coverage, DSL or cable. Also, this offers interesting opportunities to cover ship routes along costs.

The technical background: Looks like this is the result of Ericsson’s recent Release 7 work item in 3GPP on "Extended WCDMA Cell Range up to 200km" which was reported to completed in December 2006 in TSG#34. According to the work item, a Node-B (base station) so far was only able to report propagation delays on the random access channel in the order of 768 chips, or a range of about 60 km. The work item description further says that changing this parameter in the radio network has no impact on currently deployed terminals, hence, the measure is backwards compatible.

Note that for conventional network deployment scenarios, being able to report propagation delays for the random access channel of up to 60 km is more than enough given the fact that due to capacity reasons and propagation in urban environments, UMTS cells are usually spaced just 2 km or even less apart.

3G Licensens Of T-Mobile U.S. Are Incompatible With The Rest Of The World Today

UMTS is operated on the 2.1 GHz band (or UMTS operating band I) pretty much everywhere around the globe. The U.S., however, is a special case. There, the band is already occupied for other uses. Thus, operators are using the 1900 MHz band both for 2G and 3G wireless (UMTS operating band II) and in addition the 850 MHz band (operating band V), again both for 2G and 3G. It looks like T-Mobile ran a bit out of luck when it came to 3G as they had to resort to a frequency band which is not used by anyone else so far.

During FCC frequency auctions last year, T-Mobile received frequencies in the what seems to be the new 1700/2100 MHz band (UMTS operating band IV). Here’s a report from Unstrung that describes this detail. The 1700 MHz part is used for the uplink while the 2100 MHz part of the spectrum is used for the downlink (network to mobile). I guess this is a bit confusing because speculations have been going on if T-Mobile will be compatible with UMTS devices sold in the rest of the world in the areas where they deploy 2100 MHz. Well no, they are not because the 2100 MHz part is just the downlink part of their spectrum. The uplink is on 1700 MHz and not on 1920-1980 MHz as for UMTS operating band I devices.

Here’s the table of UMTS operating bands from the standards (3GPP TS 25.101). Take a look on line 4. The frequency ranges match with those in the Unstrung report about the auction I linked to above.

Therefore be careful! Some people are saying that T-Mobile U.S. uses 2100 MHz but it is slightly off the European band. Well, that’s not accurate. The 2100 MHz portion is inside the frequency range used in the rest of the world. The uplink however, is totally off mark.

I am not sure if T-Mobile U.S. will be happy with these frequencies both long and short term. Not even the latest and greatest data cards supporting multiple UMTS bands like the Globtrotter from Option supports band IV today. Also, I wonder if the band will be used in other regions of the world in the future. If not, T-Mobile might have a big problem with 3G handset vendors as the market for band IV devices will be quite small. Also, the use of yet another frequency range for 3G in the U.S. will fracture the market even more.

The Wireless Industry’s Best Kept Secret: The Price of a Base Station

While being at the 3GSM congress last week, I used the opportunity to visit my publisher and to pick up a couple of good books which will keep me entertained in the weeks to come. One of these books is "HSDPA/HSUPA for UMTS" by Harri Holma and Antti Toskala.

If you ‘only‘ want to get a good overview of HSDPA (and UMTS), you might want to take a look at my book first. If, however, you want to get the nitty gritty details, read the standards (like I did before writing my book, but not really recommendable as you’ll die of boredom or confusion unless you are a die hard like me) or Harri’s and Antti’s book. In chapter 7 on HSDPA bit rates, capacity and coverage, they feature an interesting mathematical formula on how to calculate the CAPEX (capital expenditure) cost per giga byte transmitted over HSDPA depending on the price per base station (or price per TRX to be exact).

They concluded that at a price per base station (Node-B) with six transceivers of around €100.000 (which includes the partial price of the RNC and core network serving this base station) the CAPEX cost would be around two to four Euros per GB of data traffic. An interesting number! Be careful, however, as the OPEX (operational expenditure) part of the cost is still missing. Also, the formula does not take partly loaded networks into account. They also give the price per GB of data traffic for other network / base station prices as well.

So what’s the cost of a UMTS base station these days? I did some research on this on the Internet but came up almost empty handed. Seems to be quite a well kept secret. The only reference I could find on UMTS and GSM base station prices is in an article on Unstrung from 2004. Here, Brett Simpson of Arete Research LLC is quoted giving a price for a UMTS base station in 2004 of $24.000. It seems rather low to me. Anyone got other sources?

For more on network capacity and cost take a look at my "The 1 kb/s 3G surfer" blog entry.

3GSM: Public 3G Networks Get Their Fair Share of Load

3G has been around for a couple of years but I guess even at the congress, networks have only been lightly loaded in the past as most people were still using 2G phones.

This year, things are a lot different. Except for ‘proud’ Berry owners, most other people these days carry a 3G phone and are using it heavily to make calls. I’ve also seen people skipping the Wifi coverage, which is a bit slow at times from what I have heard, and instead use their 3G PC card to access the Internet.

One of the toughest 3G environments in the world must be hall  8 this week with all major mobile phone manufacturers being present and showing their latest and greatest phones using the four public 3G networks. There must be hundreds of phones in this hall using the networks simultaneously and still they manage to show their online demos with good performance.

Every now and then I go online as well to get my eMails and to post my blog entries with my mobile phone. The network feels a bit slower in the halls than elsewhere but still I get my work done.

I think this speaks for a number of things. Firstly, all operators on site must have made sure their networks have enough capacity. It also shows UMTS is able to perform well in such demanding environments. And lastly, I think that 3G network use for both voice and data this year at the conference is most likely is higher than the use of the GSM networks.

The public does not seem too far behind. Yesterday, T-Mobile announced a revenue of 1 billion (dollars?) from data services excluding SMS in their group last year. Agreed, this is only a tiny fraction of their overall revenue, but data use has increased 8 times over the previous year and the amount of data transfered is doubling every quarter. Looks like competitive and attractive prices for mobile data finally get the mobile Internet train rolling.