Mixed-signal DSP enables satellite systems

Electronic News, August 18, 1997 by Michael McDonald

In the past two years, the success of digital satellite technology has pushed mixed-signal digital signal processing (DSP) to a new level. In return, the technological advances of mixed-signal DSP have enabled satellite system manufacturers to expand their market through lower cost and higher performance products. An additional benefit of this advance in technology is lower costs for systems besides broadband technologies, such as digital cable and high-definition television (HDTV). This drives the success of these markets, which in turn drives advances in the technology. The net result to the consumer is an upward spiral of superior technology at lower cost.

The complexity of these devices is significant. The SNR of a signal coming from a satellite is often only 4dB. There are echoes found in many of today's cable networks as well as in terrestrial broadcasts (signals bouncing off buildings, or improperly terminated cables); therefore, receivers for cable signals and terrestrial signals require equalizers and echo cancellers. Both burst noise ("impulse" noise, often caused by electrical equipment, switches, and other devices) and Gaussian noise ("white" noise caused by the random movement of electrons interacting with the signal) impact the quality of the signal, and must be corrected. Market requirements determine the amount of information that must be sent over a single channel, which also increases the real-time performance requirements of these devices.

The requirements from the STB manufacturers also drive the technology. They are looking for highly integrated solutions to shorten design cycle times and fully integrated solutions that can be easily installed on a board and require very little design effort. The successful vendors will be those that can take today's integration standards to a new level.

A Mixed-Signal Example

A mixed-signal device from Siemens is an example of the level of integration needed for broadband technology. This device is a fully integrated QPSK processor for the satellite market. As can be seen in the block diagram below, the device incorporates two advanced analog to digital converters (ADCs), and performs all additional signal processing in the digital domain.

The ADCs digitize the baseband "I" and "Q" signals supplied by the tuners for digital satellite TV. They contain information which is sent at data rates ranging from less than 10 to more than 40 MegaSymbols per second (MS/s); as a result, the ADCs require a sampling frequency between 5MHz and 90MHz. This flexibility is necessary, because the amount of information being sent will vary depending upon the content and the desired strength of the signal. The ADCs must adjust accordingly to handle the high data rates, while not using excessive energy with lower data rates.

The automatic gain control (AGC) is the first digital block. It ensures the ADCs are working at maximum efficiency, and enables the ADCs to work with very poor SNR signals or adjust for the different gain in channels.

The timing block automatically identifies the symbol rate. This is helpful for decoding channels when no information is known, such as the first time a set-top box is used. Once this information is established, it can be stored in an external EEPROM, reducing acquisition time the next time the channel is chosen.

The carrier block adjusts for phase rotation. This occurs when the tuner is not at the same frequency as the satellite. In previous systems, this was solved with a feedback loop to the tuner. The phase rotation is handled with internal math equations, which eliminates the feedback loop seen in other systems. The matched filter is designed to match the filter characteristics found at the satellite up-link station where the signal is sent to the satellite. By using a matched filter, the processor passes only the information that is likely to be the signal. The matched filter of the transmitter also "softens the corners" of the digital data signals. This is necessary to limit the signal bandwidth, so the signal is not allowed to interfere with signals in neighboring channels.

One of the problems mentioned earlier is noise. When the device receives the signal, it de-interleaves the bits, necessary because the up-link facility has spread out the signal. If a burst error occurs, the de-interleaver spreads out the error over several blocks. Although more blocks now contain an error, the errors in each of the blocks are now small enough to be corrected through error-correction algorithms.

The processor also continuously removes redundant bits from the data stream. These redundant bits are inserted by the up-link facility into the bit stream to eliminate problems associated with Gaussian noise, and several methods exist to generate such a "convolutional code." By concatenating two error correction methods, one algorithm for white noise and one for burst noise, the error correction ability is better than either one individually. In a sense, it is the proverbial "belt and suspenders."