High Speed Physical Downlink Shared Channel
Modulation and Data Rate
Reduction of Response Time
Improvement of Channel Coding (Turbo Codes)

High Speed Packed Data

A problem with UMTS was that it consisted of a compromise between circuit switched connection and packet switched connection. For Packed Switched Data, dedicated physical channels have been assigned as we have described. This already proved to be problematic with GPRS because packet transmission tends to involve large amounts of data being transmitted in batches rather than continuously as with voice. It was therefore required to keep releasing channels or they won’t really be used.

For this reason, a different, new approach was considered when developing UMTS.

High Speed Physical Downlink Shared Channel

A new type of channel was introduced, the shared channel. This was not made available to a user “permanently” but could be made available to any user at very short notice for a specific period of time if they needed it.

The so-called High Speed Physical Downlink Shared Channel (HS-PDSCH) had a fixed spreading factor of 16. Assuming that no dedicated channels can be used by a UMTS channel, up to 15 such HS-SDSCH channels can be operated in parallel. Another feature of these channels is the duration of the frames. They are only 2 ms long. The Node-B can now provide a user with not just one, but up to 15 (i.e. all) channels. This alone makes the data rate 15 times higher.

Let’s imagine 15 channels. With a dedicated channel, this corresponds to a road on which trucks drive, as in the following picture, which the user can “load” with their data. However, he always only has one truck available. The user keeps the road and the trucks even if he has no data to transfer. This means that empty trucks are always driving. With HSDPA, individual trucks are not assigned to roads. It is possible to use up to 15 trucks at the same time as shown in the figure below. If no more data is transferred, the trucks can be made available to one or more other users. This type of allocation not only results in faster data transfer, but also better utilization of overall capacity.

So that shared channels can be assigned, 4 so-called High Speed Shared Control Channels (HS-SCCH) are introduced. They have an SF of 128. A terminal device must be able to read all 4 channels in parallel.

Modulation Rate and Data Rate

To further increase the data rate, the possible modulation was also increased. Instead of the previous QPSK (2 bits per symbol), it was now possible to use 16 QAM (4 bits per symbol). Combined with different coding rates for error correction, there are many possible data rates.

Modulation	Coding Rate	5 Channels (Mbit/s)	10 Channels (Mbit/s)	15 Channels (Mbit/s)
QPSK	1/4	0,6	1,2	1,8
	1/2	1,2	2,4	3,6
	3/4	1,8	3,6	5,4
16-QAM	1/2	2,4	4,8	7,2
	3/4	3,6	7,2	10,7
	1	4,8	9,6	14,4

This means that a speed of 14.4 Mbit/s was possible with very good conditions (error-free transmission, 15 channels and 16 QAM).

However, the quality of the channels was not always good and changed very quickly. Therefore, HSDPA uses an uplink channel, the High Speed Dedicated Physical Control Channel (HS-DPCCH), to inform the E-Node B whether a data frame was received correctly and to provide information about the channel quality. Accordingly, the E-Node B can adjust the modulation and/or the coding rate to improve the transmission.

Reduction of Response Time

Another improvement of HSDPA was the response time. As described above, this had already been reduced by GPRS from 500 ms to UMTS R99 100-200 ms. The first measure was the already mentioned reduction of the frame length to 2 ms. Furthermore, a frame was divided into 3 so-called slots. The channel assignment via the HS-SCCH is designed in such a way that after just two slots the information about the channel to be used is known and the E-Node B begins transmitting the data while the HS-SCCH is still running. The UE now has 5 ms to decode the data and send an acknowledgment. This comes in good time so that after just 10 ms, a frame that may have been transmitted incorrectly is sent again. However, the identical frame is not sent. For example, if you are working with error coding and puncturing, you send the same frame but puncturing different bits. When receiving, the new frame can then be combined with the old frame and the probability of error correction becomes significantly higher. Such a procedure is called Hybrid Automatic Repeat Request HARQ.

Due to the extremely short time in which errors are corrected and forwarded in the E-Node B, the response time is reduced to 10-20 ms.

Improvement of Channel Coding (Turbo Codes)

Another improvement that was introduced with HSDPA and was also used in CDMA2000 were so-called turbo codes. In 1992, a French team achieved a new type of error coding. In addition to the original bit sequence, a bit sequence with convolutional code was sent. Furthermore, the sequence was delayed, shuffled/Interleaved and also coded convolutionally. There are therefore three versions that will be sent. The result shown by the French team was so good that it was hard to believe at first. The results were close to the theoretical optimum, which is known as the Shannon limit.