The Physical Layer - Systems (Part 2)

The modems did modulation and demodulation. The voice would go to form through other modulators and demodulators to the coders, which encodes in decodes. So everything depends on packets through the packet search network. 

There's more in codec like the ability to do compression and extra bits for error detection, etc. That was the first step forward. Moving to a point infrastructure, adapting it, and adding modems and codecs on an edge, an enabler for us to be able to access the internet through our own infrastructure. 

Shannon limit is about 35 kbps based on the average length of the local loops. Modems also differed in the number of bits per symbol and the bits used for error correction. The DSL (digital subscriber line, or broadband) idea is still an improvement on previous idea but relying on same infrastructure, but we try to increase the bit rate to around 30 kbps. It's the same exact cable at 3 kHz, but how does DSL increase bit rate? 

DSL idea was to remove filters or apply them over a large range to get a higher band (tuning the filters / applying frequency division multiplexing over this range) this is all we needed to do. This led us to the ability for voice conversation and internet access by manipulating these filters. 

Removing these filters and applying them over 1mHz range then we can apply and receive data better. It is split into bits of 4312.5 Hz each, 256 frequency bands. This is OFDM (Orthogonal Frequency Division Multiplexing). Channel 0 is used for voice, channel 1-4 are not used and then the remaining 250 channels (excluding 2 channels) are used for data. 

Some of these channels are used for upstream (data) and then receiving data is used for downstream channels. Companies can manipulate what they give you based on how much you pay. Typically replies carry more than request (so then you get more downstream/download speed)/. Providers offer 1/4/8 Mbps downstream and 0.256/0.512/1 Mbps upstream. Splitter decides what to do (frequency division multiplexing/taking signals and sending them to service provider/wait for the service provider). Remember we are talking about systems. 

There is an issue with DSL, performance of system would depend on my end, and the signal quality will degrade over distance. If you are close to the local office, you may enjoy good quality. NID is network interface device, mark of the end of  phone company's property and customer premises. Splitter is the analog filter does the frequency division multiplexing (channel 0 for voice, channel 1 for data, etc.) We've seen the most popular cases in the modulation lectures. 

The next system is definitely an improvement, Fiber To The Home. We get potentially few Mbps potentially. How is this used? These days, we need lots of bandwidth. Fiber Optic Cables replace phone lines as the communication channel to the end office. We get one wavelength that we all use for upstream (λup) and downstream (λdown) communication. 

Typically we have a few homes served by the same fiber through time division multiplexing. Fibers uses the splitter so that only one fiber reaches the office for every 100 houses. 

We do time division multiplexing, so each house gets a small amount of time to use the fiber. Same thing for downstream fibers. It's good enough, given the high capacity of fiber. Each customer will get a specific timeslot, requiring synchronization between the clocks of the different houses, assignment of time periods (different ones to different ones to different houses).  OFDM is putting the bands to each other so they don't interfere with each other. 

First generation carried analog signals related to the network. Dealing with voice in analog form means that we are not digitizing before modulating. We cannot encrypt if I don't first convert signals into binary which is a disadvantage. If I don't digitize things, since we talked encryption, I also miss on possible compression. It is possible in some cases to compress the content to send less signals to reveal the content. You can correct the signal/interpret the signal correctly with high probability if it was digitized. 

In cellular technology, since everything was analog, we are not able to send and receive messages. This drew its origins from push to talk systems. In these systems we own the users and using the same frequency. Sometimes we use the same frequency to send or receive, you either send or receive, and nothing else. Each exchange goes through a separate frequency band vs. the others. 

You have cellular towers (base stations or base station controllers) or you can have a controller and this Base Station Controller is Relayed to the Mobile Switching center, carries infrastructure and it has the brain. The switching center carries the databases for every user and knows which user is connected to which base station. You can have multiple MSCs connected to each other or connected to PSTN (Public Switched Telephone Network). 

We have cells, or hexagons and you can represent then the way you want. We have base stations connected to MSCs and the BSC is connected to the base station and the MSC. You get some channel to control then negotiate and get data channels for voice conversations. We want to establish a connection end to end. How do we come to know? We can get it by theory and experimentation. The higher the frequency the lower the coverage area, and vice versa. Now we can take it as a few km of coverage. Everything else is through cables, making multiple devices speak in the same time. 

Using the same frequency by 2 devices that are physically close to each other may mess up the signal. We can pick other frequencies to avoid interference. Within the cell, I have some number of frequencies that I can use. If I have a cell that covers half the city, I still have the same band cells in the electromagnetic spectrum to use. Within this cell, let's say we have 200 channels (only use 200 channels covering a large channel) but I will only be able to serve 200 customers simultaneously.

 We might have to rely on multiple cells, smaller cells, since they might have 200 channels to use concurrently, serving 200 customers, but we have to make sure that these said frequencies should not be used in neighboring cells. Each cell serves some of these channels, and we reuse these frequencies for some cells. 

If I increase the cell size, I potentially can serve less customers, and I spend less on my infrastructures (less cellular towers). However if there is demand, we need to reduce the cell size in order to increase the capacity of the system, and make sure that the frequency bands/channels are not reused in frequency channels. This is a tradeoff. The neighbor of the neighbors may use the same frequency band. There are design issues, 1. serving the number of customers, to use you need to reduce cell size. 

BSC is the base station controller connects the base station to the MSC and they relay the request to the MSC (which is the "brain") Each cell uses some set of frequencies not used by its neighbors. MSCs are handoffs with infrastructure, used not only for switching, but also used for pricing. MSCs communicate with BSCs, with each other and the PSTN using Packet Switching.

As you move from one cell to another, you have to associate yourself with a property base station. Base stations will advertise themselves and your device may hear multiple base stations at the same time. You associate yourself with it, leading to a notification to the MSCs. If you move from a cell to another, you get to another base station, and it has a few steps such as authorization, authentication, etc. 

The smaller the cell size, the more number of customers you can serve (since you can reuse the frequency channels). In the layout, I am using ABCD as the set of frequencies using the cell. You can reuse the same pattern. The smaller the cell size the larger the capacity of the system due to frequency reuse. 

Let's talk about the AMPS channel, first generation cellular technology. There are 832 full duplex channels each consisting of a pair of simplex channels. You can categorize them as control channels, paging channels, access channels, and data channels.

 Control channels are unidirectional. I follow an AT&T device and provides it ID. The control channel signal strength is going to be used by the mobile device to figure out what to associate with. This is used by BSC to get the highest signal power. 

Paging channels, are unidirectional and from BSC to mobile, and used by BSC to inform mobile of a new call. 

Access channels are bidirectional and they are used to setup and inform the voice code that you are placing. They are used for the call setup and the channel assignment. The system will hand them the organization. When you get a response from the destination, you are going to hear it from channel and the system is going to handle the organization. 

Data channels are also bidirectional and you can talk or hear over what the other party is saying. 

Every mobile device has a serial number, a phone number, and it scans a preprogrammed list of 21 control channels to find the most powerful signal and associate them with the corresponding BSC. This control information is sent in a digital form. You are sensing these paging channels and you are also sensing these control channels.

To place a call, you transmit the number and its identity to access channel. Then BSC receives the signal and informs the MSC. Then, the MSC looks for an idle data channel for the call and informs the mobile on the control channel about the idle channel. Finally, the mobile phone switches automatically to selected data channel and waits until the called party picks up the phone. 

To intercept a call, all idle phones listen to paging channel to detect messages directed at them. When a call is placed, a packet is sent to the callee's home MSC and find out where it is and message is sent to BSC in callee's current cell, then BSC broadcasts on paging channel "Unit 14, are you there?". Mobile responds with "Yes" and BSC responds with "you have a call on channel 3" and mobile switches to data channel 3 and rings for the user to pick up the call. Data channel at the callee doesn't have to be the data channel at the caller.

1G is analog Voice doing frequency division multiplexing. The clear thing that is missing is turning the voice into digital and this is the purpose of 2G. This way we can optimize, compression, encryption, better error detection.

2G, or digital voice is multiplexed using both FDM and TDM. The most popular standard is GSM (Global System for Mobile Communications). Voice is digitized and sampled for analog to digital and digital to analog conversions. 

GSM architecture is similar to AMPS architecture. You now have SIM cards that are placed on your phone equipped with some key used to authorize your ability to use this network. If you change your phone's sim card to other network, then you use the other network.

GSM has 124 full-duplex channel, and each full duplex channel consists of a pair of simplex channels, each 200 KHz are going to be split between different users. The first generation is on Frequency Division Multiplexing, and the Second Generation is Frequency Division and Time Division Multiplexing. Transmitting and Receiving does NOT happen in the same time slot. So we have 992 full duplex connection since each simplex connection supports 8 separate connections using TDM. You cannot talk or listen at the same time you either send or receive. 

For example, you are given as a device an up time slot on some of the channels and a down timeslot on some of the channels. 

Problem with TDM is that since TDM is done in a static way (give each user a timeslot up and down) then these timeslots are wasted when not in use since these slots are already assigned to you. At least you can serve a few hundred customers at the same time. Now we are able to digitize voice and get the benefits of that.

Now let's move on to 3G and see how it improves over 2G. Problem with 2G is that we lose some system complexity using TDM. 3G changes how multiplexing, by relying on CDMA (Code Division Multiple Access). It accommodates the internet and the most popular standard is the Universal Mobile Telecommunications System. In CDMA, each device would get an orthogonal code, and these codes are used to modulate your signal, so rather than getting a specific frequency, we get a code that we use to modulate the signal. I can interpret want you're saying given my knowledge on the code. There are standards and terminologies, and UMTS is the most popular standard. 

There are advantages to CDMA, since it uses the wasted timeslots, simplifies the frequency planning (no having to make reservations, as long as I know your code I can accommodate you). By varying the number of chips, I can give more chips to someone who cares more about reliability vs time efficiency/speed/doesn't want to spend a lot of time sending/receiving content. 3G is digital voice and data. 

Now let's talk about 4G. 4G turned all the way back to frequency division multiplexing. This does not really increase the capacity of the system, just makes the views of the system more flexible, since it creates better utilization. 4G and 5G are not relying on CDMA anymore, but OFDN (Orthogonal Frequency Division Multiplexing) and TDM. 

The advantage of LTE is that it simplified the architecture to IP-Base and increased the capacity and speed relying on different ranges for communication and this was the main contribution, simplify the interfaces, turn to IP the exchanges ran more smoothly. The way this is done is using OFDN.

With 5G, things are very different. It's not just in there to increase the capacity, but accommodating a massive amount of devices that are trying to report data (IoT, lifeline communication in times of disaster). Another target is high capacity, relying on a larger portion of the electromagnetic spectrum, relying on higher frequencies. There is an empty space in the EM spectrum that accommodate many cases that require high capacity. 

5G has a popular triangle where at one corner you have 5 high capacity service, then massive connectivity, then low delays (needed for applications like self-driving cars, etc. in order to provide with fast enough feedback on something happening). 5G is pretty flexible, includes major capabilities with regard to throughput, offering them on a pick and choose basis. It kind of suits and adapts itself to the application at hand. The portion of higher frequency is in the mm range rather than cm, the ones that we use for Wi-Fi. This has implications, since you need line of sight for these. If you make use of this mm wave range, then your base stations cannot cover km. In the mm range you get very high capacity compared to what we have right now. There are, however, other directions for researching this area. 

5G will rely on more antennas, and we are able to tune in and be with different devices. If someone else speaks up the left, I would not care, so whatever he's saying I can capture and create multiple things depends on the size of the antenna.  We use spatial multiplexing, splitting the wireless channel into multiple subchannels. Spatial multiplexing is a MIMO wireless protocol that sends separate data signals or streams between antennae to enhance wireless signal performance or functionality. With 5G your device may be communicating with multiple other devices because of the advances in multiplying and dividing. Here is the triangle: 

There is a beam which is an access point creating multiple beams, splitting the space, in any direction I can use frequencies I need. So it not only splits frequencies, it splits the space. Just tune to whatever comes from an area.