Communications IC Architectures
Unit 10: Novel Architectures - The Software Radio
Unit Contents
- 10.1 Radio Communications - Where is it going?
- 10.2 What are the Technological Improvements Possible?
- 10.3 Novel Radio Architectures
- 10.3.1 DSSS
- 10.3.2 The ASH Receiver
- 10.3.3 The Technology Level
- 10.3.4 The Component Level
- 10.3.5 The circuit Level
- 10.4 Summary
- 10.5 Self Assessment
10.1 Radio Communications - Where is it going?
There are many directions in which the development of radio and related communications system are developing. This chapter will take a top-down look at the systems aspects, then a bottom-up look at the technological requirements.
10.1.1 Modulation Methods
Commercial broadcasting has been dominated from the start by amplitude modulation (AM), and this remains the case today, both for radio and television. Frequency modulated radio (FM) has been in existence for over 40 years, and FM television, through satellites, for about 10 years. These have become the de facto standards, although it is interesting to note that the AM systems persist, and will only disappear from the UK scene if government action de-licences them.
Similarly with radio telephony, most private mobile radio (PMR) and shipping and aircraft communications are carried out using SSB, for long range HF radio in ships and aircraft, FM for shorter range communications such as taxis, police etc. and, rather anachronistically, AM for VHF aircraft communications.
Mobile cellular phones have had a shorter lifetime and faster development; from the large units of a decade ago to the very small hand portable units today. The early systems were analogue; virtually all phones are digital today.
We can expect all analogue systems to be phased out within 10 years. This has already been determined for television through proposed legislation, and may well be legislated for other forms.
Digital Audio Broadcasting (DAB) is already available, although take up has been slow. The bands are around 220MHz, with a digital modulation which gives excellent quality. The receiver consists of a superhet with conventional front end, followed by a baseband conversion. The data is transmitted in digital form, so analogue to digital conversion is not required, except in the sense of conversion of the Quadrature Amplitude Modulation (QAM) to digits. In principle, DAB is useable in a car, although commercial receivers have not yet been shown.
Digital television is already available, through satellite and terrestrial
links, and appears to work well.
The real issues with these new systems are that the digital modulation offers
advantages to the hardware manufacturer in reductions of unit cost. The software
manufacturer, i.e. the programme makers, can at least see a different and larger
market for their products. For the consumer, the main effect has been the many
new channels available, and the apparent low cost. TV 'set top boxes' are often
'free', i.e. the cost is bundled with the rental of the facilities. Mobile phones
are essentially marketed in the same way.
This has interesting implications on the design of hardware. It is possible using analogue systems to deliver sound or TV with a quality which is beyond the ability of the viewer/listener to distinguish from perfect. Even by objective measures, receiver noise performance and bandwidth may approach technological limits.
If we consider for example a TV satellite receiver, it has a noise performance of typically <0.7dB, at 11GHz. It has a synthesised tuning capability of >1300MHz, with uniform noise performance. It has a broadband FM discriminator of good performance and linearity. It has a reasonably secure encryption/decryption system. All this is delivered in an acceptable package, with remote control, guarantee etc. Most importantly, it is 'free', although it has a real cost, hidden in rentals, of about £120. This is truly remarkable. Such a performance, only available in military system just over a decade ago, would have cost many thousands of pounds.
We cannot expect similar advances through the move to digital. This is partly because they are not technologically possible, and partly because the end customer cannot see such great benefits again. The move from three terrestrial channels only to perhaps 30 including satellite today (UK figures) has perhaps been useful; but the number of really good programmes has not increased proportionally. A move to 300 or 3000 channels will not give sufficient perceived benefits to attract customers, who cannot watch that much anyway.
Similarly the technical improvement, while real, is small in customer perceptions. The slow take up of DAB is an example where the technological improvement is real, but the quality benefit over FM sound is not perceived in a market where the full cost penalty must be borne by the consumer.
Thus the technical direction for the next decade or so must be in system cost
reductions. It must become cheaper for the consumer to receive or use the service,
while maintaining or improving standards.
10.2 What are the Technological Improvements Possible?
10.2.1 The System Level
The section or Wireless LANs described the Direct Sequence Spread Spectrum (DSSS) approach. This has been the most significant new receiver architecture devised in recent years. Although it has some similarities to the superhet, the spreading effect means that channel selection by digital means is in some respects simpler than in the superhet, especially in low cost systems. Future mobile phones may well use variants of this, in Code Division Multiple Access, (CDMA) although the system is more complex and difficult to handle than some of its early proponents believed. However, there is little doubt that CDMA will become very important in the next few years. LANs and WANs may be the major users of DSSS, for low cost and reconfigurability, although compatibility issues remain.Where the signal sent is digital in form, the need for accurate ADCs in receivers
is reduced to some extent, although the dynamic range issues remain, and are
frequently overlooked by designers.
10.2.2 The Software Radio
As the first chapter of this module indicated, there is a move towards general-purpose radios. Several companies have built radio receivers and transmitters based on a common set of building blocks, with a strong software element for configuration. There are many limitations to this approach, but these are steadily being overcome as the technologies involved advance.
Receivers have been designed by Watkins-Johnson, STC Microwave and Siemens which are essentially digital software reconfigurable. These are relatively large units, intended for military applications.
In the military field, there is a great attraction to the 'Antenna-ADC-DSP' approach. The DSP can be expanded in an almost limitless way, so that a receiver may simultaneously watch hundreds or thousands of channels. Such receivers have presumably been built, at immense cost, and for applications where current ADC performance is acceptable, i.e. usually radar or surveillance applications. 'Tuning' is entirely by digital frequency selection in the DSP. While there may ultimately be a spin off from this technology to domestic applications, the cost and size implications ensure that this will not be soon.
In more cost conscious but still high performance applications, there has been a move towards very small units which are add-ons, sometimes plug-ins, for PCs. The add-on contains all the R.F. circuitry, through to the ADC and interface to the PC. The PC then handles some, but clearly not all, of the filtering and signal processing. Of these, Raytheon offer a PCMCIA card version, which is probably the smallest feasible structure at present, while ICOM offer the PCR-1000, which is an external add-on, working through the parallel port of a PC. A similar, but not identical product is the 'WinRadio', from Enformatica. This may be external, or may be supplied on an ISA card. This gives a multiple simultaneous channel capability; up to 8 channels in a single PC. In each of these units, the PC screen offers the 'front panel' of the radio, with tuning and bandwidth functions, memories etc. Scanning is possible and signal recognition.
These radios use a relatively straightforward radio architecture, with the PC giving the final filtering DSP functions. More radical architectures are also possible.
The need for multi-standard software radios has come from the many new radio
standards such as GSM, UMTS, PCS, and others. These standards differ in band
coverage, in frequency band and in modulation technique. What they have in common
is that all need the usual front-end components of R.F. amplifier, mixer, synthesiser,
I.F. filters and amplifiers and ADC or ADCs. On the transmit side the architectures
again have many commonalities. It would be cheaper for manufacturers to sweep
up all of the DSP sections of the radio in a single chip design, which would
ideally be capable of taking an ADC output in any modulation protocol, and converting
it to analogue speech and digital code output. The same chip, either through
a separate channel or by means of multiplexing, could provide the same facilities
for transmit.
10.3 Novel Radio Architectures
10.3.1 DSSS
The DSSS receiver is the most novel structure to appear in recent years. The format is shown in the chapter on WLANs, and should be compared with the conventional superhet.10.3.2 The ASH Receiver
One other new architecture which has been introduced, especially to the low cost market for alarms remote controls etc., is the ASH - Amplifier Sequenced Hybrid. This has some aspects in common with the super-regenerative receiver, and some with the superhet. However, it would be more correct to distinguish it from the superhet by noting that the ASH uses time diversity rather than the frequency diversity effectively used in the superhet.
Fig 1 : The ASH Receiver
The ASH architecture is shown above. SAW filters are used in this structure, but alternatives could be used. The R.F. amplification is all at the incoming frequency, but instability is avoided by operating the amplifiers in separated time slots as shown. The similarity to the super-regenerative receiver is really superficial; the new architecture stands alone. Selectivity is from the front-end filter, and dynamic range is limited in the simpler versions, but this is nevertheless a significant new low cost receiver structure.
The diode on the diagram above should be interpreted as a detector of some
form, either amplitude, FM or an ADC/DSP combination.
10.3.3 The ASH Receiver
Technology advances are likely also to be in cost. The moves from current 150mm wafers to 200mm and 300mm have been relatively slow when compared with the rapid strides made in earlier times. The reason for this is that the 'customers', in this case the semiconductor manufacturers, see little advantage, and many problems, in the larger wafers. Although 200mm systems have been available for almost a decade, they are by no means universal. 150mm systems are probably still the most numerically superior, while 100mm lines still exist in numbers. 300mm lines are almost non-existent, several years after the technology became available.
GaAs remains the technology of choice for the microwave amplifiers, mixers
and switches, although for little else. Silicon bipolar is used where possible,
and SiGe is extending this to several GHz, both for analogue and digital applications.
Silicon CMOS remains dominant for digital systems, which are now reaching 1
GHz clock rates. The two main technological themes of the future will be the
impact of SiGe bipolars, and the synergy of bipolar and CMOS in R.F. circuits.
10.3.4 The Component Level
Memory devices are already at 0.18 micron geometries. There are apparent fundamental problems in using large-scale techniques below this figure. The reason for this is the wavelength of 'light' needed. Monochromatic deep UV has long been the workhorse in the industry, and even with raster techniques has reached this limit due to its wavelength. Electron beams are capable of better resolution, and the photoresists are also available, but the technique is slow. Another possibility is focussed ion beam milling, although here again speed is an issue. It may be that this limit really is 'fundamental', at least in the large scale production sense. There is already a general acceptance that analogue circuits benefit very little from going below 0.8 microns geometries.
The most likely technological improvement is the use of E-beam defined gates
for extreme frequency operation. This has been used for GaAs for some years.
Many GaAs processes have 0.25 micron gates, defined by e-beam techniques, while
the major parts of the chip are defined using conventional techniques, to 2
micron rules. Extending this to CMOS, for R.F. performance geometries may be
fairly relaxed at say 0.5 micron, with 0.1 micron or better gates. This would
give performance at least to 10GHz.
10.3.5 The Circuit Level
The combination of silicon (or SiGe) and CMOS is probably the most interesting
area of new work in circuits. The two device type (bipolar and MOS) are
'complementary' in a synergistic way. Although there have been a small number
of publications in this area, one paper has been singled out as an example;
the student is strongly recommended to read the original.
Fig 2 : The Circuit Level
The circuit above shows an interesting combination of bipolar devices, for
lowest R.F. noise, and MOS for control of the function. This indicates the likely
way ahead for R.F. design, i.e. a combination of the best technologies for the
task in hand. It should be noted that in this diagram only a single device,
Q43, is in the RF path, while all the others provide bias, switching or a secondary
follower function.The circuit above shows an interesting combination of bipolar
devices, for lowest R.F. noise, and MOS for control of the function. This indicates
the likely way ahead for R.F. design, i.e. a combination of the best technologies
for the task in hand. It should be noted that in this diagram only a single
device, Q43, is in the RF path, while all the others provide bias, switching
or a secondary follower function.
Fig 3 : The Circuit Level
A similar transmit amplifier from the same paper also shows a simplicity of design. The critical issue in designs such as this is in careful consideration of transistor geometries and current paths. Serious loss mechanisms (see below) can occur in the metallisation which result in higher noise in receive and loss of signal level on transmit.
Fig 4 : The Circuit Level
The mixer from the paper quoted used a 'Gilbert Cell' type of multiplying structure,
with a differentially driven local oscillator input and a fully differential
output. This is ideal in R.F. terms, since objectives such as good power supply
rejection are easily achieved.
Fig 5 : The Circuit Level
A similar transmit mixer, albeit operated at higher signal levels, included all inductors on the chip. Since the product was a low-cost wireless LAN device, this was clearly important in achieving target costs.
Fig 6 : The Circuit Level
The chip is shown in block diagram form above. Only minimal external components would be needed for this design.
10.4 Summary
The rate of real introductions of genuinely new architectures for radio remains low. The rate of innovation in circuits however has probably never been higher. The combination of consumer demand, technology supply and high profit business should drive this market sector for several more years before demand levels off.
10.5 Self Assessment
Read the paper suggested above, and review earlier design work against a competitive SiGe BiCMOS process.