How Does HARQ Compare to Wi-Fi ACK?

Here's a thought that I recently had when I looked at how the Hybrid Automatic Retransmission Request (HARQ) functionality works in HSPA and LTE: From a conceptual point of view HARQ is quite similar to the Acknowledgement mechanism of Wi-Fi. Here, the reception of each packet has to be confirmed by the receiver by returning a MAC Ack(nowledgement) frame back to the receiver. This is done in a way that the ACK package has precedence over any other packets that are waiting in the queues of other users of the system. If the ACK is not received, the sender automatically retransmits the packet with the same or a different modulation and coding scheme.

The HARQ mechanism of HSPA and LTE is pretty similar: Each transmission has to be immediately acknowledged on the MAC layer as well. If a NAK or nothing is received the transmission is repeated. When one goes into the details, of-course, there are fewer similarities. With HARQ, the system can use incremental redundancy to send a different version of the packet with different error detection and correction bits. In addition, several HARQ processes run concurrently so a transmission failure of a single packet does not stop the overall transmission. And then, HARQ uses an 'out of band' channel for the feedback, while the Wi-Fi Ack is a normal packet on the air interface.

German Spectrum Auction At €1.5 Billion At The End of Week 2

For about one and a half weeks or so the German spectrum auction was not very exciting for external observers as little activity could be spotted. On Wednesday morning, though, things started to heat up.

Until that time, the total amount that was bid at this point was still below 400 million euros. Then suddenly, activity spiked and within a day the proceeds were almost up to one billion euros. The main activity for the moment is focused on the 800 MHz digital dividend band. By the end of Friday, the few MHz available there accounted for around €1.2 billion of the total €1.46 billion bid for everything so far.

While in other bands there is an ample amount of spectrum for everyone, only 30 MHz is available there, and consequently not enough space for the four contenders each wanting at least a 10 MHz chunk. So while a single 2x5MHz chunk is priced at over €200 million there at the moment, the same amount of spectrum can be had for 'as little' as €7 million in the 2.6 GHz band.

Voice Quality of Conference Calls

Every time I attend conference calls where people dial in from all over the world, my ears are usually suffering if there's more than one speaker. It's for various reasons:

  • The volume different people have on the call is widely different. So while you have to listen very closely to understand some participants, others can be heard so loudly that your ear-drums almost pop out when the call goes from one extreme to the other.
  • Most of these issues are caused not only because different telephone networks seem to interconnect on different volume levels. It's also because every participant has a different phone, some use the hands-free mode and, for international calls, some countries use a different voice codec that at some point is converted into the codec used in the country of the phone bridge.
  • Add to that some echo when people are not muted, background noise such as babies crying, dogs barking and cars passing by and the perfect storm approaches.
  • There are always people on a conference calls who are on the move and their mobile phones often try much too hard to filter out the background noise, resulting in shriek peaks and hard to understand participants.
  • Automatic announcements that people are leaving and and re-joining the conference call due to patchy network coverage every couple of minutes doesn't make things much easier, either.

And on top of all of that put non-native speakers with sometimes heavy accents and after an hour your (or at least my) head starts spinning. So what's the solution to this?

I think it's wideband audio conference calls with heavy pre-processing of the individual call legs. As you can't get that over the standard telephone network, it must be Internet based, maybe, and that's already a big compromise, with telephone dial-in for those who for one reason or another can't access the Internet (on the move, stupid company firewall, etc.). I guess anyone who've once enjoyed the difference between wideband and narrowband speech knows what I am talking about. And on top of that the conference server or the clients should do some intelligent pre- or post-processing of the signal coming from the different participants. Is it really that hard to have everybody's voice arriving at the same level?

Anyone aware of such a system?

Femto Search and Rejects

3G femtocells are an interesting topic but I haven't had much time yet to take a look at the details of how mobility management works in practice. There's lots of activity in 3GPP to standardize mobility management around femtocells (or Home NodeBs how they are called there) in Release 8 and beyond. However, there are already already femtos on the market today and they have to work together with pre-Release 8 mobiles. So I've had two fundamental questions: How can mobiles find the femtos when they are on the 3G macro layer and how does the femto get rid of users which do not belong to the subscriber group, i.e. everyone except the owner and his/her family and friends?

Then this book, "Femtocells – Technologies and Deployment" by Jie Zhan and Guillaume de la Roche came my way. I haven't had time yet to go through it in detail but it looks highly interesting and informative and I could answer my questions with it within minutes:

Cell-Reselection to a Femto: To make a cell reselection to another 3G cell, it needs to be part of the neighbor cell list of the cell. As there could be many femtos inside one macro network and provisioning them automatically might not be a straight forward approach, one option is to select a couple of Primary Scrambling Codes and declare them as neighbors in every macrocell or at least on those macrocells in which femtos are located. This works even if there are many femtos inside the coverage area of a macrocell as not all of the femots are overlapping and hence the PSCs can be reused. If the femtos scan their surroundings when they start up they can help to avoid the PSC overlapping issue.

How to get rid of non-femto subscribers: The femto deployments I have heard of so far are closed-subscriber-group femtos, i.e. only registered people have access. But since todays mobiles know nothing of femtos how can you ensure only those remain in the cell that are supposed to be there? The book gives this as one of the potential solutions: For the femtos a certain range of location area codes (LACs) are reserved. If a non-femto subscriber mobile finds the cell and tries to perform a location update it gets a location update reject with cause code #15 (no suitable cells in location area). The mobile then goes back to the macro layer and puts the LAC in the forbidden LAC list on the SIM. The 3GPP UMTS RRC spec says It's only removed when the mobile is switched-off or after a significant amount of time has passed (12-24h). A bit of a disadvantage here: If the user of femto-A passes femto-B during the course of a day, the mobile will try to register with femto-B and will be rejected. In case the LAC was the same as that of femto-A the mobile will not try to reselect to femto-A until the forbidden LAC list is cleared. In other words, the user comes home and the mobile will not use the femto.

Agreed, there's much much more to the topic, those where just my two most burning questions concerning femtos.

What’s Your Experience With OperaMini on the iPhone?

Looks like Apple has decided to let OperaMini into their AppStore and within just a couple of days it has become hugely successful according to Engadget here. Being a long time user of OperaMini and knowing about its strength and advantages in bandwidth constrained and high outage environments when moving in trains, cars, etc. I can imagine why everybody seems to rush to it. But not having an iPhone myself I'd be interested from you what your experiences are if you tried it!

Dealing With Limited Uplink Power

Here's a little comparison of how UMTS and LTE deal with limited uplink power of mobile devices which I think it is quite interesting:

When uplink power for a UMTS E-DCH (HSUPA) transmission reaches a maximum, the number of simultaneously used codes can be reduced, a more conservative coding can be employed for additional redundancy and the modulation order can also be changed.

In LTE, modulation and coding can also changed as needed. And in addition, there's a third parameter: LTE uses an OFDM air interface, or to be more precise, SC-FDMA in the uplink direction. In other words, many subcarriers are used for the data transmission which are grouped into consecutive Resource Blocks (RBs) in case of uplink transmissions. When the mobile device reaches its maximum power level and the network detects this, it can reduce the number of RBs assigned in the uplink direction. This way the mobile can concentrate it's power on fewer RBs and hence it has more power available on the narrower channel it now uses. From a network point of view this is much better than leaving the number of RBs as they are and reduce modulation and coding as the RBs that are removed can be assigned to other devices also requesting resources in the uplink direction. For details see the power control section in this excellent book.

More Wi-Fi VPN Options

For those of you using public Wi-Fi hotspots now and then and who are a bit worried after my previous post on cookie theft there are several VPN options that protect you from eavesdroppers. In this post I talked about installing a PPTP server on your windows machine at home to redirect all your traffic while you are away via your home network. That's not everybody's cup of tea, however, as you need a DSL or cable connection with a fast uplink and a PC running all the time. So for those of you looking for an alternative on the net here are two:

Both offer a number of different options ranging from PPTP, which Windows already has a client for, up to a full OpenVPN SSL with certificates and all bells and whistles you can imagine. Both VPN offers also work with Linux and the OpenVPN configuration especially with Ubuntu (I tried with Jaunty) it is quite straight forward.

Both VPNs are not free but if you compare the power cost over a year if you leave an extra machine running at home, the extra cost for the external VPN might just be negligible.

Observe the German Spectrum Auction in Near-Real Time

A quick one today: For all of you interested in the German spectrum auction going on these days for 800, 1800, 2100 and 2600 MHZ spectrum, the Bundesnetzagentur has a web page which is updated after each bidding round. For each 90 minutes round it shows the the highest bidder for each block. Today was day one and the current total price is €116 million.

The LTE Band Challenge

With LTE, the complexity of including a reasonable number of different frequency bands in a mobile device not only for LTE but also for GSM and UMTS is once again getting trickier. Here's a how I see things from a historical point of view and where I think we are heading:

Once upon a time the wireless frequency landscape was quite simple. When GSM started in Europe, there was only a single frequency band in the 900 MHz band which all network operators used. Sure there was the legacy analog network in the 450 MHz band but nobody seriously thought about working on dual mode devices. GSM or bust! Things got a bit more complicated when the second band in the 1800 MHz range was opened for GSM at the end of the 90's and but it didn't take take device manufacturers long to come up with dual mode devices. In the US things were pretty similar but the remainder of this post continues with a Europe point of view.

Since then, things have gotten much more complicated. With UMTS, things started well for some time with 2100 MHz being 'the' 3G band around the world, except for the US. In the US, UMTS and GSM are used in the 850 and 1900 MHz ranges and these days also on the 1700/2100 MHz band combination. In Europe in the meantime, UMTS in the 900 MHz band has also taken off in some countries. I guess this was the point where the number of bands used around the world and the number of bands supported in a single mobile device really started to diverge. Today, the state of the art from a European point of view is the following combination:

  • Quad band GSM support (850, 900, 1800 and 1900 MHz)
  • UMTS tri-band (2100, 900 and one of the US bands)

And now with LTE just around the corner things are about to get even more complicated. Here's the bands where I think LTE will see the day of light in the next two to three years:

  • In Europe LTE will likely start on 2600 MHz and potentially also on 1800 MHz and 2100 MHz
  • And then there's the digital dividend band in the 800 MHz range which is likely to be used in some countries to bring broadband connectivity with LTE to rural areas.
  • In Japan, LTE will be used on 2100 MHz with an additional band likely to follow.
  • In the US, the situation is even more divergent. Each network operator seems to have its own band. Verizon uses a 10 MHz block in the 700 MHz range and another operator has another block in the same range but with exchanged uplink/downlink assignments. Some operators might launch LTE in the 1700/2100 MHz band combination and there are speculations of a satellite backhaul based LTE network with its own frequency range. Finally, there are rumors of Clearwire jumping from WiMAX to LTE in the 2600 MHz band but with TD-LTE.

From my point of view, this frequency diversity is far from ideal for everyone involved. For users it's an issue as global roaming capabilities of devices will get worse and worse. Also, especially in the US, it will be difficult for users to switch between networks by changing SIM cards and subscriptions while keeping the device. For network operators and device manufacturers it's also far from ideal as some will have trouble getting good devices as volumes are just too low to reach good prices. There might be multi-frequency LTE devices tailored for the US market but since almost every operator uses different legacy network technologies and frequency ranges the potential band and technology combinations for GSM, CDMA, UMTS and LTE are huge.

So what's the way out of this? To me it looks like it's in the hand of device manufacturers as the number of frequency bands will not shrink anytime soon. The question is if the ever growing number of bands and backwards compatibility combinations change the device design?

  • Is it physically possible today to support so many bands? Software defined radios have been discussed for many years but as far as I know antennas and filters are not so easily to be adapted to different frequency ranges with software only.
  • Or could the radio part of the device in the future be built in a way that it can easily be interchanged?
  • How about exchangeable radio modules? With this approach I would in the future select a SIM, a mobile device and an RF module and maybe one or two extra for international roaming? Or will we just have to live with the situation as it grows worse?
  • And then, there's still Wi-Fi which, at least so far, can be used universally around the world. Most smartphones today have Wi-Fi built there's no ubiquitous coverage and logging into foreign Wi-Fi networks automatically is still a dream.

As things are I don't see a good solution yet. As always, comments are welcome!