Analog and Telecommunication Electronics

Transcription

1 Politecnico di Torino - ICT School Analog and Telecommunication Electronics B1 - Radio systems architecture» Basic radio systems» Image rejection» Digital and SW radio» Functional units AY /03/ ATLCE - B DDC This lesson describes architectures or radio receivers and transmitters, and points out some speciic problems, such as rejection o image requency and channel separation. The basic architecture here presented is heterodyne, based on requency translation. Reerences: Phillip E. Pace: Advanced Techniques or Digital Receivers Artech House, 2000, ISBN Chapters related with radio systems: 2- Receiver architectures; 3- Circuit components or signal conversion; Patrick D. van der Puije: Telecommunication Circuit Design John Wiley & Sons Canada, Ltd. (1992), ISBN-10: Chapters related with radio systems: 3- AM radio receiver 5- FM radio receiver 2016 DDC 1

2 Lesson B1: radio system architectures Basic radio systems architectures Heterodyne receivers The image problem & image rejection techniques Heterodyne transmitters Direct synthesis (PLL lesson group) Digital radio, sotware deined radio Identiication o unctional modules Description o unctional modules this lecture next lectures Reerences: Slides with notes 05/03/ ATLCE - B DDC As or any electronic systems, a radio receiver (or transmitter ) can use analog or digital signals and circuits. Digital circuits are more easy to design, reliable, and digital abrication technology can be less expensive, thereore current design emphasis is towards digital techniques. Also radio systems move rom analog to digital; the process is still on the way and the lesson presents beneits and critical issues. The lesson ollows a top down approach (like other parts ot this course): irst step is description o unction (high level block diagram), then the analysis moves down to more detailed descriptions: low level block diagrams, and inally circuit schematics, with selection or design o each device. There is always a separation between WHAT to do (the speciication, the user view o the systems), and HOW the task is accomplished (the circuit details). This architecture description identiies the various unctional units, which are presented and analyzed in detail in the ollowing lessons DDC 2

3 Which contents? INFO SOURCE TRANSMITTER ANTENNAS Focus o this lesson is on radio receivers Radio transmitters have structure and problems similar to receivers RECEIVER INFO USER 05/03/ ATLCE - B DDC The slide shows the architecture o a radio system, seen as a chain: Ino source transmitter antennas TX RX receiver Ino user This lesson describes the architectures or radio receivers and transmitters DDC 3

4 Expectations rom a radio receiver ANTENNA The input signal Va (rom the antenna) contains the wanted signal (usually quite weak), plus many types o noise and intererences. Va Vu Va RADIO RECEIVER Vu The receiver ouput signal Vu should be a copy o the original modulation, without noise and/or distortion. 05/03/ ATLCE - B DDC A radio receiver can be seen as a system which converts RF signals (Va in the diagram), collected by an antenna, and extract the inormation contained in one o the channels (e.g. a radio station), converted to a suitable representation (e.g. an electric signal Vu, able to drive earphones or other transducers). The useul signal/station is usually embedded in a variety o other signals, originated by other channels, intererences (EMI), and noise. The receiver must have some SELECTIVITY to isolate the required signal, and TUNING capability to select dierent channels. Beore the output, inormation is extracted rom RF signal by a DEMODULATOR. The electrical signals collected by the antenna may have very low levels (down to ew microv), or in the case o nearby transmitters medium levels (such as a ew hundred mv). In both cases they must be ampliied to the power level useul to drive transducers. The unctions required to the system are thereore - TUNING & SELECTIVITY (or channel separation), - AMPLIFICATION (to increase power towards output transducer) - DEMODULATION (inormation recovery) They can be applied also in a dierent sequence DDC 4

5 Elementary receiver ANT. Va DEMOD. Vu Input ilter (narrowband, variable F) Filter, ampliier, and demodulator must operate at variable requency. A B Tuning shits the resonant requency A o the ilter (e.g. rom A to B ) 05/03/ ATLCE - B DDC The block diagram shows the most simple architecture or a receiver: - the antenna, which collects RF signals (the desired signal is RFa in the ollowing) - a band-pass ilter, to isolate the good channel RFa, - a demodulator, to get the baseband signal rom AM, FM... The ampliier (middle block) is useul but at this point can be considered optional. To select dierent channel we must shit the response o the band-pass (e.g. using a resonant circuit, tuned by changing one o the L or C reactive parameters). To get good channel separation (selectivity, ability to listen only a speciic station), the input pass-band ilter needs steep requency response (high Q). This is not easy to achieve, especially i we consider also the tuning requirement DDC 5

6 The crystal receiver Antenna and input ilter Crystal demodulator Tuning changes the C o a resonant circuit Vu From 05/03/ ATLCE - B DDC This is an example o radio system based on the previous architecture: a Galena receiver. Galena is a natural mineral orm o lead sulide, which allows to build pointcontact diodes, and in turn AM demodulators. The antenna is a coil, which is also part o the RF band-pass resonant circuit. Tuning is achieved by changing C through a variable capacitor. This device can operate as AM receiver in LW, MW, and SW ranges. There is no ampliier, so the output power comes directly rom RF signal: the receiver can drive only earphones, and requires high power (or quite close) transmitters. Since no power supply is needed, the system remains always operational, without need or batteries or other power source. This radio architecture is now used in passive RF ID, which get operating power rom the RF signal DDC 6

7 -: Beats I we multiply two sine signals, the output is sum and dierence beats Werner s ormulae. a This is the RF signal we are interested in O O X IF Fo - Fa Fa Fo Fa + Fo Dierence beat Local Oscillator Sum beat 05/03/ ATLCE - B DDC The slide is a reresh on what happens when we multiply two sine signals, with requencies respectively a and o. In the time domain we can use Werner s ormula: sin(a t) sin(o t) =.5 [cos(a + o)t + cos(a - o)t] In the requency domain the output includes two terms: - a sum beat, with requency s = a + o, - a dierence beat with requency d = o a. Mixing can be seen as a translation o Fa to a+o and to o-a. I we multiply a signal with spectrum X() by a sinewave with requency o, the spectrum is translated to X(+o) and X(-o). This technique can be applied to a radio receiver, where a is the RFa signal rom the antenna, and o comes rom a local oscillator LO. Fa is translated to the dierence and sum positions. Fa can be any signal (instead o a simple sine-wave), its spectrum will be translated to positions corresponding to sum and dierence beats DDC 7

8 The heterodyne receiver a IF channel i = a - o Va Wideband input ilter O O X Filter and ampliier (narrowband, ixed requency i ) DEMOD. Vu i = a - o a O The input signal is shited to a ixed requency i = a - o. Tuning shits the requency O o the local oscillator 05/03/ ATLCE - B DDC In this diagram the RF signal rom the antenna is iltered (wideband input ilter) to remove outband noise, then multiplied by the Local Oscillator signal Fo; this causes a shit in the requency domain to the sum and dierence beat positions. The IF channel band-pass ilter isolates the dierence beat (at requency Fi), which is then ampliied and demodulated. The received signal is thereore the part o RF spectrum at Fa=Fo-Fi. Only this signal can go through the IF ilter. To receive another channel, the system moves the requency o the local oscillator to a new value Fo, and the mixer brings to IF the RF signal Fa = Fo +Fi. Channel separation is perormed by the IF ilter; which operates at ixed requency, and can be quite narrow. This is a HETERODYNE receiver. In the heterodyne receiver one o the beat (usually the dierence) is iltered, ampliied, and demodulated in the IF chain. All the units ater the mixer (the IF chain) operate at ixed requency, which makes possible to get better parameters, especially in the ilters DDC 8

9 Not so new Armstrong original drawing: 1 MHz RF to 100 khz IF (D. Marsh: Direct conversion receivers. edn europe, Oct.2000) Reginald Fessenden (1901): rom the Greek heteros (other) and dynamis (orce). (rom 05/03/ ATLCE - B DDC The principle o heterodyne (or use o beats) is well known; the slide shows the schematic diagram o an heterodyne receiver with electronic tubes (triodes). The irst triode receives at the input the sum o the antenna signal and local oscillator (through the transormer). Due to nonlinearity o the active device, the output includes product terms DDC 9

10 Beneits o heterodyne receivers Channel isolation achieved by ixed-requency IF ilter No need or tunable narrowband ilter Tuning achieved by shiting the LO requency Possible to cover wide requency range Ampliiers and demodulator operate at ixed IF Narrowband circuits more easy to design and test But Fi= Fo Fa or Fb Fo both Fa and Fb enter the IF chain image requency problem 05/03/ ATLCE - B DDC Beneits o heterodyne receivers: - Channel isolation is achieved by ixed-requency IF ilter; it can have high Q and narrowband response - Tuning is achieved by shiting the LO requency; that removes tunable high Q circuits, and makes possible to cover a wide requency range. - Ampliiers and demodulator operate at ixed requency (IF); such narrowband circuits are more easy to design and test. Main problem o heterodyne receivers - the same value o IF can be achieved as Fi = Fo Fa or Fb Fo: Two RF signals Fa and Fb are both moved to the IF requency and enter the IF chain. The unwanted one (Fb) is the IMAGE requency. Most part o variations in basic heterodyne architecture are introduced to handle the image problem, that is to remove the IF signal components coming rom image signals DDC 10

11 -: The image requency The mixer generates sum and dierence beats Sum beats can be easily iltered good (RFa) and image (RFb) both olded to IF Same requency: cannot be separated by IF ilters IF GOOD LO IMAGE 2LO LO-RFa RFb-LO RFa RFb RFa+LO RFb+LO SUM BEATS 05/03/ ATLCE - B DDC The mixer brings to the same IF requency two RF signals, RFa and RFb, with requency respectively: a = o i, and b = i o. The unwanted signal is called IMAGE (can be seen as the image o the good one, mirrored on the LO requency). The image could come rom other transmitters, and can be stronger than the useul signal, causing - Intererence (high noise) - Blocking (drive LNA/mixer into saturation) Since in the IF chain image and good signal are on the same requency range, they cannot be separated by iltering. Removing the image is the major problem in the heterodyne receivers, and can be accomplished in several ways: - Filters (on RF, beore the mixer) - Multiple-conversion - Image rejection mixers - Zero-IF architectures 2016 DDC 11

12 Image removal with RF ilter The RF bandpass ilter allows good channels to go through to the mixer, but blocks the images. IF GOOD RF + Other cannles IF LO IMAGE (o GOOD RF) The IF band-pass ilter keeps only one channel IF 05/03/ ATLCE - B DDC Image components can be removed when they have a dierent requency range, that is at RF level. The RF ilter allows all channels within desired bandwidth to reach the mixer, but blocks images. For multiple-channel receivers the RF ilter has wide bandwidth. The IF ilter isolates the desired channel DDC 12

13 Filters in the heterodyne receiver 1 IF channel I = o 1 1, 2 RF input ilter: Removes image requency 2 O O X DEMOD. Vu IF ilter: Removes adjacent channels (narrowband, ixed requency i ) i = o O Same eect as narrowband RF ilter 05/03/ ATLCE - B DDC The RF input ilter (band-pass, wide) removes noise and intererers outside the band used by received channels; all channels are still present at mixer input. The mixer translates the spectrum o the set o channels, to bring the desired channel to the IF requency. The IF signal is pass-band iltered and isolated rom other channels. The RF input ilter (wide band) isolates the set o channels, then the IF ilter (narrow band) isolates the single channel. Tuning, that is selection o desired channel, is achieved by shiting the LO requency Fo DDC 13

14 Noise and IF ilters Adjacent channels, rejected by IF ilter Good chalnnel; can go through RF and IF ilter Outband signals, wideband noise and image, removed by RF IR ilter 05/03/ ATLCE - B DDC Channel separation requires ilters with steep edges (high Q); this is diicult to achieve with variable requency at RF level, but rather easy on ixedrequency IF. The combined eect o requency translation and IF iltering is like placing a narrow band-pass ilter (with IF bandwidth) on the RF signal. When LO is moved, the eect is the same as a shiting a tunable band-pass with a high Q. The main beneit o the heterodyne technique is that IF ilters operate at ixed requency, and can comply more tight speciication (higher Q, better controlled shape) DDC 14

15 Lesson B1: radio system architectures Heterodyne receivers The image problem Dual conversion receivers Complex mixer & I/Q processing Digital receivers Sotware Deined Radio Examples 05/03/ ATLCE - B DDC 2016 DDC 15

16 Remove the image by RF iltering i1 = o1 a a O1 b1 i1 A passband ilter beore the mixer can remove the image b1. i2 i2 i2 = o2 a a O2 b2 With high IF, the image b2 is more ar away, and the ilter can be less steep. 05/03/ ATLCE - B DDC A pass-band ilter on RF beore the mixer can remove the image. The RF ilter bandwidth must let all channels to go through, but provide high outband attenuation (high Q). Such ilter is more easy to design and build i the distance between desired channels and images is high, which brings towards high IF solutions. Some receivers use a set o RF input ilters, or dierent RF bands, to urther reduce noise and intererers. Numeric example, with RF Input channel: 2.5 GHz IF channel: 1 MHz LO requency: 2.5 GHz + 1 MHz = 2,501 MHz Image at: GHz Q o RF ilter: 2,500/2 = 1250!!! diicult to get IF channel: 10 MHz LO requency: 2.5 GHz + 10 MHz = 2,510 MHz Image at: 2.52 GHz Q o RF image removal ilter: 2,500/20 = 125 The second choice is more easible DDC 16

17 high or low IF requency? B i1 i1 B i1 = o1 a1 i2 i2 = o2 a1 a1 o2 b1 a1 o1 b1 Q = /B With high IF, a given bandwidth B needs high Q (or IF ilter) The IF channel is moved to a lower requency with a second beat: rom i1 to i2. Since i2 is lower, the same bandwidth B can be achieved with lower Q 05/03/ ATLCE - B DDC With higher IF image requencies are more ar rom the good signal (the actual distance is 2 Fi). On the other hand, a low IF can achieve channel separation with lower Q IF ilters. A structure which brings together the beneits o high IF (image rejection) and low IF (channel isolation) is the double conversion heterodyne DDC 17

18 Dual-conversion heterodyne receiver Va X X DEM. Wideband LNA + ilter O1 IF1 ilter +Ampli. O2 IF2 ilter +Ampli. i1 = a O1 a O1 High irst IF (IF1) easy image removal Low second IF (IF2) Simple channel ilter Tuning by shiting O1 or O2 i2 = i1 O2 i1 O2 i1b IF1 image risk on IF2 ( i1b )! 05/03/ ATLCE - B DDC Dual requency conversion joins the beneits o high and low IF: - Easy image removal, thanks to higher image / good signal separation. - Good channel isolation, thanks to narrow bandwidth o the IF2 ilter. Dual conversion releases the speciication o RF and IF1 ilters. RF ilter removes images & wideband noise; ilter BW and Q are related with 1st IF requency: High IF wide RF ilter (image is 2 x IF away) First IF (High): reduces noise and ar intererers (other channels) Second IF (low): - Further noise rejection - Remove adjacent channels (narrowband IF ilter) Tuning can be achieved by shiting either local oscillator (O1 or O2) DDC 18

19 IF ilters - a LC tuned circuits Medium complexity Requires tuning Multiple cells to control bandwidth Widely used in consumer equipment 05/03/ ATLCE - B DDC High-Q band-pass ilters are diicult to build with standard LC technology, and expensive. Achieving a speciied shape can be diicult. Available technologies include -LC resonators (Q is limited by losses) -Mechanical resonators (SAW, Ceramic, Quartz) Best perormance can be obtained with mechanical resonators. They are based on piezoelectric materials, and exploit mechanical resonance Electrical signals are converted to a mechanical wave which travels across a piezoelectric crystal or ceramic. The wave propagates across the device with minimum attenuation at the mechanical resonant requency, and at the other end is converted back to an electrical signal. SAWs use ceramic materials, less expensive then quartz lattice. (RF ilters or the cellular phone use this technology) DDC 19

20 IF ilters - b Surace Acoustic Waves (SAW) Exploit mech. Resonance High Q Low cost Quartz lattice ilters Combination o high-q mechanic resonators (quartz crystal) Used in proessional equipment Active ilters, Digital ilters Feasible or low IF channels 05/03/ ATLCE - B DDC High-Q band-pass ilters are diicult to build with standard LC technology, and expensive. Achieving a speciied shape can be diicult. Available technologies include -LC resonators (Q is limited by losses) -Mechanical resonators (SAW, Ceramic, Quartz) Best perormance can be obtained with mechanical resonators. They are based on piezoelectric materials, and exploit mechanical resonance Electrical signals are converted to a mechanical wave which travels across a piezoelectric crystal or ceramic. The wave propagates across the device with minimum attenuation at the mechanical resonant requency, and at the other end is converted back to an electrical signal. SAWs use ceramic materials, less expensive then quartz lattice. (RF ilters or the cellular phone use this technology) DDC 20

21 : - -: Complex mixers: SSB & image reject Werner s ormulae: sin a x sin b = ½ (cos(a-b) - cos(a+b)) A sin a x cos b = ½ (sin(a-b) + sin(a+b)) cos a x cos b = ½ (cos(a-b) + cos(a+b)) B Single Side Band beat: A + B = cos(a-b) No ilter Only dierence product term No sum term In TX chain: generation o SSB signals Image rejection mixer No need or image RF ilter (but needs phase shiters) 05/03/ ATLCE - B DDC Another approach to image requency removal is based on I/Q processing. The starting point is again Werner s ormulae: 1.sin 1 cos 2 = ½ [sin(1-2) + sin(1 + 2)] 2.cos 1 cos 2 = ½ [cos(1-2) + cos(1 + 2)] 3.sin 1 sin 2 = ½ [cos(1-2) - cos(1 + 2)] Adding 2 and 3 we obtain: cos 1 cos 2 + sin 1 sin 2 = cos(1-2) That is a single term o the 1 2 beat. Since it operates using In-Phase (I) and Quadrature (Q) components (the sine and the cosine), this is a irst example o I/Q signal processing DDC 21

22 I/Q mixer in receivers Only dierence (or sum) beat here O X RF Filter and LNA /2 /2 X + DEMOD. The Q signal can be generated directly by LO Vu Need RF π/2 wideband phase shit 05/03/ ATLCE - B DDC This technique can be used in receivers to get rom the mixer only dierence (or sum) beat. In a receiver the I/Q mixer alone brings little beneit, since the dierence and sum beats have wide separation, and can be easily isolated by ilters in the IF chain (which must have narrowband band-pass or channel separation). Beneits can come rom the separate I/Q processing, as described in the next section DDC 22

23 Single Side Band (SSB) transmitter Vb LO /2 a X IFi + IFt B /2 X IFq Output Filter and Power Ampliier B V U a The spectrum o Vb is translated to Fa to build a SSB signal. O = a B 05/03/ ATLCE - B DDC A more interesting application is in SSB transmitters. A standard AM signal includes the carrier and two sidebands; since each sideband carries all inormation, higher eiciency is achieved by keeping only one o the sidebands. The beneits are - No power wasted or useless signals - Less spectrum occupation. The sidebands can be isolated by ilters, which should be quite complex and expensive. A more convenient solution is to use I/Q requency translation, with cancelation o one sideband. To achieve good cancelation, tight gain and phase rotation matching between the two mixing paths is required. The technique is known since a long time, but it was used only in proessional equipment, with individual trimming to match I/Q path parameters. With the improvements in RF IC technology, it is now used directly inside integrated systems without trimming. Another critical point are phase shiters. For the local oscillator (LO) it is possible to design circuits which directly generate sine and cosine at the same requency (DDS, Lesson B6). On the RF path we need a wideband phase shiter, which must comply high precision requirements or good cancellation DDC 23

24 : Image rejection mixer RF1 LO LO RF2 image IF IF good x Cos(LO) x Sin(LO) RF2 LO RF1 /2 Sin(RF1 LO) -Sin(LO RF2) Cos(RF1 LO) Cos(LO RF2) Sin(RF1 LO) Sin(LO RF2) + 2Sin(RF1 LO) 05/03/ ATLCE - B DDC Besides generating SSB signals, I/Q signal processing can be used to cancel the image signals in a receivers. The diagram in the slide reers only to dierence beats (sum beats can be easily removed by ilters). 1): two mixers multiply RF signal respectively by LO I (sine) and Q (cosine) components. The RF signal includes RF1 (good signal) and RF2 (unwanted image). Ater each mixer the IF signal includes both RF1 (the good one) and RF2 (the image), overlapped to the same requency. Due to the use o I/Q LO signal, the phases are dierent in the two branches. 2) A π/2 phase shit is applied to one channel; it generates image components with opposite polarity in the two channels. 3) the two channels are added; only RF1 beat (the good signal) survives. Image cancellation with this technique requires precise matching o mixer transer unction (Gain, Phase rotation) DDC 24

25 : - Hartley image rejection Sin(RF1, RF2) Cos(RF1-LO)+Cos(LO-RF2) O /2 X Sin(LO) Cos(LO) X /2 Sin( )+Sin( ) + Sin(RF1-LO) Sin(RF1-LO)-Sin(LO-RF2) IF π/2 phase shit (narrowband) No RF2 in IF 05/03/ ATLCE - B DDC The operations above described can be achieved with this structure (Hartley image rejection mixer). Critical issues are: - Phase shiters. A irst one can be embedded in the LO; the second operates on ixed requency (narrowband). - Channel balance to achieve cancelation (gain and phase) In summary, the main beneit o image rejection mixers is the removal o complex RF and IF band-pass ilters. The drawback is the need or tight balancing o gain and phase shit, to achieve good cancellation. This in turn requires good analog technology DDC 25

26 Receiver with image rejection mixer O X RF Filter and LNA /2 X + /2 DEMOD. Vu Image rejection mixer IF π/2 phase shit (narrowband) LO phase shit with digital techniques 05/03/ ATLCE - B DDC The image rejection I/Q mixer is here used in the receiver chain, ater RF ilter and LNA. Compared with single-branch mixer, this technique releases the requirements on RF input ilter. With dual conversion receivers, the I/Q image cancellation can occur at the irst or second IF conversion DDC 26

27 Receiver with image reject mixer 05/03/ ATLCE - B DDC Example o complete radio chain in a commercial IC (GPS receiver). In this diagram we can recognize several elements o an I/Q receiver. -The PLL loop or the LO -The VCO with direct sin/cos outputs -The image rejection unit -A Variable Gain Ampliier (controlled by the AGC unit) -Image rejection uses the I/Q Hartley structure DDC 27

28 Weaver image rejection Complex mixer: I/Q outputs 05/03/ ATLCE - B DDC The diagram represents another technique or image rejection. Here the circuit generated I and Q IF components. Indicating the π/2 rotation with the j operator, and considering input signal decomposed in I and Q components (a + jb), the circuit multiplies this signal with the LO, expressed as (c + jd). (a + jb) x (c + jd) = (ac-bd) + j (bc+ad) = I + jq, With: I = ac bd Q = bc + ad This structure is a COMPLEX MIXER DDC 28

29 : I-Q channels Va X O1 IF1 Ampliier V O /2 X X V Q V I Q channel I channel DEM. V V cos V Q V V Q IF2 Ampliier V cos t V I sin V I V sen t 05/03/ ATLCE - B DDC The receiver may use I/Q demodulation, to keep amplitude and phase inormation. I and Q components are obtained multiplying the received signal by sine and cosine reerences. In this way the signal is decomposed in I and Q components (at IF2 in this example). The I/Q demodulator can compute V and V. This technique is used or complex phase/amplitude modulations DDC 29

30 Zero- IF (or low-if) receiver a i Va Input ilter and LNA (variable F) O O X Low pass ilter DEMOD. Vu i = 0 O - a a - O O = a Received signal at a directly moved to baseband Lowpass ilter in IF channel Signal/Image overlapped separation not possible 05/03/ ATLCE - B DDC The heterodyne moves the signal spectrum to IF, usually lower than RF; Zero-IF receivers apply the same principle, but move the signal spectrum to DC (LO has the same requency as the desired channel). This is called Zero IF (ZIF). The main beneit o ZIF comes rom using low-pass IF ilters, with: - Better control o shape: changing IF bandwidth requires only to change the cuto requency o the LPF. - A LPF is more easy to design and build than band-pass inside an IC, making the architecture suitable or SOC. Critical issues in ZIF architectures are: - Oset (DC is a signal): high-pass ilters (to remove oset and DC unbalances) are not allowed. - LO to RF leakage in the mixer: causes a DC beat, which cannot be isolated rom actual RF (overlapped on the same requency band). - Image is the same signal spectrum lipped on the requency axis, and cannot be removed by ilters; image cancelation by I/Q processing becomes mandatory DDC 30

31 ZIF with I-Q channels Q channel Va Input ilter and LNA (variable F) O /2 X X I channel V Q V I DEM. & IMAGE REJECT V V The signal is decomposed in I and Q components directly at RF. I/Q processing mandatory or image rejection baseband Ampliier & demod. DC is in-band! critical 05/03/ ATLCE - B DDC ZIF radio architectures are widely used inside ICs, mainly because lowpass ilters are more easy to design and build than band-pass. ZIF radio systems can be recognized by the low-pass ilters in the IF chain (instead o the usual band-pass). Image is the good RF with inverted spectrum; cannon be iltered at RF level. I/Q mixers with image rejection are mandatory or ZIF structures 2016 DDC 31

32 ZIF beneits an problems Not new (homodyne), but not easy to build Beneits Low-pass IF ilters: more easy to build inside Ics Problems Requires I/Q processing or image rejection» Second mixer DC errors overlapped with signals»use dierential structures to get low DC errors» Careull matching o various branches» Diicult and expensive with discrete components previously used in proessional equipment» More easy with Ics currently widely used in consumer systems 05/03/ ATLCE - B DDC Oset and balance must be tightly controlled, thereore these circuits use dierential signaling and structures. ZIF radio systems can be recognized by the low-pass ilters in the IF chain (instead o the usual band-pass). Image is the good RF with inverted spectrum; cannon be iltered at RF level. I/Q mixers with image rejection are mandatory or ZIF structures 2016 DDC 32

33 Example o ZIF with I-Q channels Dierential chain or DC balance Commercial device (receiver channel) LPF: ZIF architecture 05/03/ ATLCE - B DDC In this diagram (commercial device MAX2820) we can recognize: - the ully dierential structure, rom RF to outputs - a Variable Gain Ampliier (VGA) at RF ront-end, with external control - a PLL-based LO (requency synthesizer) - the I/Q mixer - a phase shiter or the LO - LPF in the IF chain (meaning it is a ZIF system) - Variable gain IF ampliiers 2016 DDC 33

34 Lesson B1: radio system architectures Heterodyne receivers The image problem Dual conversion receivers Complex mixer & I/Q processing Digital receivers Sotware Deined Radio Examples 05/03/ ATLCE - B DDC All the system previously analyzed are based on analog building blocks (ilters, ampliiers, mixer, local oscillator). The same unction can be obtained also with digital circuits, which give beneits in terms o design ease, abrication cost, and lexibility. Moving rom analog to Digital requires an Analog-to-Digital converter (ADC) somewhere along the receiver chain. The ADC position deines the boundary between analog and digital processing: all units beore the ADC operate on analog signals, the ones ater it process digital samples. The various positions correspond to dierent tradeo between complexity, perormance, power consumption. This section describes the various choices or moving rom the analog structures to digital ones, discussing the respective beneit and drawbacks DDC 34

35 First step towards digital radio IF channel A/D conversion ater demodulation. Va Wideband ilter O X IF ilter and Ampliier (ixed F) DEMOD. Analog Demodulator A/D out Many applications use directly digital data 05/03/ ATLCE - B DDC The basic concept o digital (radio) systems is to exploit the beneits o digital technology by moving complexity rom the analog domain to the digital domain. This trend applies to all types o electronic systems, and to any application. In the case o radio systems the limits and drawbacks come or the high carrier and bandwidth o RF signals (which bring towards high sampling rate), and the need to preserve the SNR (which means high processing resolution, that is many bits). Processing rate and resolution impact complexity and power (that makes analog circuits still the best choice, in some cases). The ollowing slides analyze the various solutions or digital radio systems with increasing perormance (and complexity), up to ull SDR. A irst choice is to place the ADC at the end ot the receiver chain, ater the demodulator. On demodulated signal represented by numeric samples, it is possible to apply error correction, encryption or other unctions that can be easily implemented on digital signals. Carrying out similar operations with analog circuitry i not impossible; just more complex and expensive, due to signal degradation caused by noise intrinsically added at each analog processing step DDC 35

36 More digital - 2 IF channel Va O X A/D DEMOD. Vu Wideband ilter IF ilter and Ampliier (ixed F) Digital demodulator The digital demodulator can use complex algorithms The same HW support dierent types o modulation The A/D converter must operate at high requency 05/03/ ATLCE - B DDC A second choice is to place the A/D converter at the end o the IF chain, between the IF output ampliier and the demodulator. The ADC will operate on IF band, requiring higher sampling rate than in the previous case. The digital demodulator provides high versatility; it can handle complex modulations, and use proprietary demodulation schemes in sotware, achieving improved noise immunity and robustness. The same hardware can operate on dierent modulation schemes (provider they require the same processing till IF). Dierent inormation (e.g., voice, data, video) can be easily muxed and demuxed. Since digital processing can be carried out by a processor (microp or DSP) with SW deined algorithm, this can be seen as a irst step towards sotware radio DDC 36

37 More digital - 3 IF (digital) channel Va O X A/D DEMOD. Vu Wideban LNA and ilter (variable F) IF ilter and demodulator (all digital) The digital IF ilter increases the computational load, but allows to modiy the IF parameters in the SW 05/03/ ATLCE - B DDC The next step is to place the A/D converter at the input o the IF chain, just ater the mixer. The ADC operating requirements are the same, but now the digital processor must carry out also digital iltering unctions. Again, the main beneit is increased lexibility in the IF ilter: a digital ilter is more easy to build (and modiy) than an analog one. Frequency and shape changes require just reprogramming (in the processor SW or o a FPGA). This structure requires more computational power than the previous ones, which in turns means more energy rom the power supply DDC 37

38 Mixing and sampling A mixer shits the input signal requency Sampling can be seen as product o the signal with a sequence o δ. O X The spectrum o a δ is a sequence o δ The product o input signal with the undamental requency o the δ stream corresponds to the mixing operation Va S/H A/D DEMOD. O Local oscillator 05/03/ ATLCE - B DDC Sampling corresponds to multiplying a signal by a sequence o pulses, spaced by the sampling interval Ts. The spectrum o a sequence o deltas is a sequence o delta (in the requency domain), spaced by the sampling rate Fs = 1/Ts. The sampling thereore can be seen as multiplying a signal by a set o sine waves, corresponding to the various lines o the spectrum o deltas. The product o the signal by the undamental sampling rate Fs corresponds to mixing the signal with Fs. Only one o the X(t) x Fs beat can go trough the IF ilter; all other products with higher order harmonics are iltered out DDC 38

39 Test B1-a: beats and spectrum Draw the spectrum o a voice signal (300 Hz 3 khz) sampled at 16 ks/s Mixing 1 and 2 create 1 2 and Sampling corresponds to multiplying a signal by a sequence o pulses, spaced by the sampling interval Ts. The spectrum o a sequence o deltas is a sequence o delta (in the requency domain), spaced by the sampling rate Fs = 1/Ts. Sampling creates multiple copies o the spectrum Discuss what happens in baseband Aliasing o out-band components 05/03/ ATLCE - B DDC The sampling correspond to multiply by a pulse, which has several (actually ) harmonics. It creates multiple copies o the spectrum DDC 39

40 Heterodyne requency translation B1 Dierence beat: IF signal Local Oscillator LO1 RF signal B1 Sum beat LO1 IF bandpass ilter 05/03/ ATLCE - B DDC Here the useul RF signal has a spectrum close to LO1 with bandwidth B1. In a standard heterodyne structure this RF signal beats with a local oscillator signal LO1, and the signal spectrum is moved to the IF requency range. IF channel must have a bandwidth at least B1. IF signal is sampled at Fs > 2 B1, and converted to digital. The digital signal contains all inormation available in the original RF signal, with no aliasing. The digital processing can shape the IF transer unction and isolate single channels within the IF requency range 2016 DDC 40

41 Nyquist sampling and reconstruction Signal Spectrum S Sampling rate F S = 1/T S F S 2F S 3F S 4F S Anti-aliasing input ilter F S -S F S F S + S 4F S S Anti-aliasing reconstruction (output) ilter Aliases caused by sampling 05/03/ ATLCE - B DDC The slide shows the eect o sampling on signal spectrum with the classic Nyquist approach. The bandwidth o input anti-alias ilter is related with useul signal bandwidth B1. Compliance with the Nyquist-Shannon rule (Fs > 2 B1) guarantees that aliases are not overlapped, and the original signal can be recovered by the reconstruction low-pass ilter. In a radio receiver the useul signal is RF, usually with rather high carrier requency. Applying the Nyquist rule to the carrier leads to unreasonably high sampling rates. The actual minimum sampling rate is related with signal bandwidth, not with carrier requency. This can be seen with heterodyne structures, and can be applied using sub-sampling DDC 41

42 Frequency translation by sampling Sampling rate RF signal S > 2 B1 F S 2F S 3F S B1 4F S Reconstruction ilter Sampled signal S - 3F S 5F S -S S - F S S 05/03/ ATLCE - B DDC Here we use the same RF signal: spectrum just above 3Fs, with bandwidth B1. The signal now is directly sampled at Fs > 2 B1. The sampling creates the set o aliases. Since B1 < Fs/2, these aliases are not overlapped (that is there is no inormation loss). One o the aliases (S 3 Fs in this example), correspond to baseband. The sampling process can thereore be seen as a requency conversion. Sampling at a rate lower than the carrier causes a requency translation o signal spectrum. I the Nyquist rule (towards signal bandwidth, not carrier) is ulilled, the various aliases do not overlap (no inormation loss), and one o them (usually baseband) can be isolated by the output reconstruction ilter. The ADC sampling rate is thereore related to signal bandwidth, but sampling is aected by a jitter (noise in the time domain, which moves the actual sampling instant). The time precision o the sampling circuit (sampling jitter) is related to carrier requency, as discussed later DDC 42

43 Subsampling and spectrum olding F S = 1/T S Sampling rate F S 2F S 3F S 4F S RF signal S S - 3F S 5F S -S S - F S 05/03/ ATLCE - B DDC The generation o aliases can be represented with a olding process. The spectrum is drawn on a strip o paper, which is olded at Fs/2, Fs, and multiples o Fs/2. The signals present in each old are copied (by the sampling) to all other segments. This is SPECTRUM FOLDING. Folding shows how a spectrum S is moved to baseband by sampling. To avoid aliasing, the sampling rate must be higher than twice the signal bandwidth (or any position o the carrier) DDC 43

44 Single signal subsampling and olding 0 F S F S /2 3F S /2 2F S F S 2F S 0 F S /2 3F S /2 05/03/ ATLCE - B DDC Spectrum olding overlaps several parts o the spectrum; in this eample the signal (blue line) has signiicant power only in the Fs-3Fs/2 range. All ranges KFs-(K+0.5) Fs are olded and overlapped. To preserve the signal: - Folding width must be larger that the signal BW. - other olds must have 0 (or very low) signal power DDC 44

45 Subsampling and olding Unwanted Out-o-Band Signals RF ilter From: Critical techniques or High Speed A/D converters - Pentek 05/03/ ATLCE - B DDC Spectrum olding overlaps several parts o the spectrum; to keep (in each old) only the useul signal, this must be iltered (pass-band old BW = Fs/2) to eliminate/reduce outband noise and intererers. The sampling rate is related with the bandwidth (set o channels). Ater the ADC, iltering and channel separation is carried out by digital processing. Example: A set o 20 channels, 15 MHz each, has a total bandwidth 300 MHz. To avoid aliasing, must be sampled at least at 600 MHz (or better at higher rate, e.g 1 Gs/s). The individual channels are isolated ater ADC by digital iltering DDC 45

46 Folding with spurious and noise Sampling rate F S = 1/T S F S 2F S 3F S 4F S Useul signal Outband spurious and noise Folding overlaps spurious and noise in baseband UNUSABLE! 05/03/ ATLCE - B DDC A RF signal contains also out-o-band components and wideband noise. Subsampling without RF iltering signals brings to overlapping o various spectrum segments. Outband signals become inband noise, which cannot be isolated by any reconstruction ilter DDC 46

47 Subsampling and iltering F S = 1/T S F S 2F S 3F S 4F S Useul signal Bandpass ilter Translated signal Lowpass ilter OK 05/03/ ATLCE - B DDC For correct subsampling operation, the RF signal must be iltered beore the sampling. The bandpass ilter keeps only the useul signal components, removing wideband noise and other outband signals. To keep aliases separated, the sampling rate must be higher than twice the bandwidth. The original signal translated to baseband can be recovered by the reconstruction ilter. Further narrowband iltering (e.g. or channel separation) can be carried out by digital processing o the translated signal DDC 47

48 Which sampling rate? Sampling rate Feasible RF ilter Too narrow ilter Multichannel RF > 2 B1 F S 2F S 3F S Baseband signal (still multichannel) Channel ilter (Digital) 05/03/ ATLCE - B DDC The useul RF signal here includes several channels (FDM). Single RF channels are usually quite narrow (e.g 100 khz), and it is very hard to isolate channels at RF level (requires high Q ilters). Subsampling based on channel bandwidth (e.g. at 200 khz or slightly higher) is not easible. The solution is similar to double conversion: RF ilter isolates a block o channels (RX band, bandwidth B3), rather wide, without tight requirements on ilters. The channel block is subsampled at Fs > 2 B3 and converted to digital; these operation move the signal spectrum o the whole channel block to baseband. Channel isolation is achieved by digital processing o subsampled signal. The digital processing allows to select the ilter shape and BW, even with high Q. The digital processing circuit operate at the sample rate Fs DDC 48

49 Subsampling and sampling jitter Sampling rate depends on signal bandwidth Subsampling receivers use less power Sampling jitter depends on carrier Amplitude error = (time jitter) x (slew rate) Slew rate = V x ω(carrier) Phase error in I/Q chains Sampling jitter is a critical parameter Need or low jitter sampling clock high-speed Sample/Hold circuits (low aperture jitter) 05/03/ ATLCE - B DDC A digital radio moves signals to digital domain as soon as possible. The radiorequency ilter must be large enough to keep all useul channels (proessional receivers have many ilters, to divide the RF in several bands with better SNR). Sampling rate depends o RF ilter bandwidth; lower rates reduce processing requirements (and power), but put more severe specs on the ilter (the aliases generated by sampling must not overlap). Sampling occurs with a time error, called aperture jitter (Tja). The errors in sampling time become amplitude errors through the signal slew rate. The ADC sampling rate is related to signal bandwidth, but time precision o the sampling circuit (aperture jitter) is related to carrier requency. Numeric example: RF band: MHz BW = 60 MHz Suitable sampling rate: 200 Ms/s Sampling jitter error Ej = Tja x SR = Tja x ω V at 850 MHz carrier, percentage error, or ull scale signal: Ejp = (Tja x ω V)/2V)/100 = 2016 DDC 49

50 Oversampling, iltering and decimation A: Complete input signal signal o interest other spectral components (useless) B: Useul input signal Nyquist rate Filtering, resampling at lower rate s 2 s C: Signal ater decimation s and decimate (get rid o redundant samples) 05/03/ ATLCE - B DDC Sampling the channel block requires a sample rate ar higher than the minimum required by a single channel. I we are interested only in the inormation o a single channel, that bunch o samples is heavily redundant. The system operates with OVERSAMPLING. Oversampling allows to use wide BW anti-alias ilters, but increases the digital processing requirements. The useul inormation or a single channel has a ar smaller BW, and can be represented with lower sample rate. The number o samples can be reduced with a DECIMATION process (actually a digital ilter). With oversampling and decimation, complexity o anti-alias iltering is moved rom analog to digital domain. The analog pre-adc antialias ilter is replaced by digital post-adc processing and decimation. The main beneits is reduced complexity o analog part (the RF band-pass ilter); the cost is an increased complexity and higher power consumption or the digital processing. To preserve phase inormation, these operation must be carried out on separate I/Q channels DDC 50

51 Digital downconversion (GSM receiver) A B C 05/03/ ATLCE - B DDC The diagram represents an integrated circuit (AD6655, rom Analog Devices) used to build the receiver section o GSM cell phones with various standards (TD-SCDMA, WiMax, WCDMA, CDMA2000, GSM, EDGE, LTE). The device includes two receiver chains (diversity), each with digital I/Q downconversion, decimation (to reduce the number o processed samples), reconstruction (adding I/Q components). A: complete input signal B: useul input signal, sampled at high rate C: decimated signal rebuilt rom I/Q components; useul bandwidth preserved, lower sample rate. The AD6655 is a mixed-signal intermediate requency (IF) receiver consisting o dual 14-bit, 80 MSPS/105 MSPS/125 MSPS/150 MSPS ADCs and a wideband digital downconverter (DDC). The AD6655 is designed to support communications applications where low cost, small size, and versatility are desired DDC 51

52 How many bits do we need? RF signals have wide dynamic (µv V) Nonlinearity distortion, intermodulation, Two approaches: Many bits /12, 14, high power & cost AGC with analog or digital control loop Variable Gain Ampliier (VGA) Vu O X A/D PROC. Gain control 05/03/ ATLCE - B DDC A key parameter in complex digital processing (such as ilters) is the resolution (or precision) o the computation (the number o bits). To get a higher resolution the ADC must provide a high number o usable bits, over all possible signal values. Instead o using wide dynamic ADCs, some systems use variable gain ampliiers (VGA) beore the ADC. Gain is controlled by eedback rom the ADC output; as it approaches ull scale, the gain is reduced DDC 52

53 SDR: Sotware Deined Radio IF sampling X A/D O RF (under)sampling DEMOD. μp, DSP A/D DEMOD. Analog: LNA, ilter O μp, DSP Digital: DSP or EPLD 05/03/ ATLCE - B DDC The block diagram in the slide shows a radio system where all unctions (besides RF ilter and LN Ampliier) are carried out by digital units, ater the ADC. When the digital processing is carried out by microprocessors or DSPs, the unctions are casted in the SW. This creates the term SW RADIO. The actual meaning is almost all digital, including the LO (usually a Numeric Controlled Oscillator, NCO). Digital processing can be carried out by DSP and SW, or by a FPGA. Systems which operates up to 5 GHz with ZIF architecture are commercially available as ICs; higher perormance can be achieved with multi-chip boards. The slide shows a heterodyne architecture (top, the ADC operates on IF) and a direct sampling structure (bottom, the ADC operates on RF signal) DDC 53

54 Direct sampling I-Q SDR RF A/D /2 O DEMOD. Analog components - ilter -LNA A/D DSP: BANDPASS O = RF + IF IF processing Bandpass ilter Universal radio HW: Can change requency, modulation, application (GSM, GPS, UMTS, ) 05/03/ ATLCE - B DDC The slide shows the block diagram o a direct sampling digital I/Q receiver (no mixer or downconverter). Image cancellation can be achieved by I/Q processing. As discussed or previous slides, sampling rate is related with RF signal bandwidth (RF ilter bandwidth, to avoid RF aliasing). Sampling jitter is related with carrier requency. Sampling moves the RF signal to IF; IF ilter and demodulator operate on digital samples. This actually represent the block diagram o a universal receiver, where the digital part can be programmed or dierent IF bandwidth or modulation types. It can operate on any signal compatible with the rontend (Antenna, RF ilter, LNA) DDC 54

55 ZIF I/Q Va A/D /2 O DEMOD. Analog components - ilter -LNA A/D DSP: LOWPASS O = RF Baseband proc. Lowpass ilter Universal radio HW: Can change requency, modulation, application (GSM, GPS, UMTS, ) 05/03/ ATLCE - B DDC The only dierence rom previous slide is the LO requency: here the RF is moved directly to baseband (DC); we get a ZIF system This receiver used LPF in the IF chain, and is more easy to build as monolithic IC. The added problems are related to DC oset & drit; now DC in within useul signal bandwidth DDC 55

56 Lesson B1: radio system architectures Heterodyne receivers The image problem Dual conversion receivers Complex mixer & I/Q processing Digital receivers Sotware Deined Radio Examples 05/03/ ATLCE - B DDC 2016 DDC 56

57 Positioning RX dual conversion Single requency conversion Double requency conversion IEEE JSSC, VOL. 33, NO. 12, Dec Thomas H. Lee: A 115-mW, 0.5- m CMOS GPS Receiver with Wide Dynamic-Range Active Filters 05/03/ ATLCE - B DDC This diagram shows heterodyne receivers with single and double requency conversion. Local oscillators LO1 and LO2 are obtained rom the same reerence with a PLL synthesizer The output uses low resolution ADC (2 bits), and VGA to increase the dynamic range DDC 57

58 Positioning Receivers I/Q ZIF (same paper) 05/03/ ATLCE - B DDC This circuit uses I/Q structure with direct conversion (ZIF). IF ilters become low-pass DDC 58

59 Positioning RX image reject mixer 05/03/ ATLCE - B DDC In this commercial IC, we can recognize -I/Q Local Oscillator, with PLL -I/Q architecture (not a ZIF, since the IF ilters are bandpass) -VGA controlled by an AGC circuit Image rejection is achieved by combining the I/Q components DDC 59

60 Complex architectures (lesson A0) From 05/03/ ATLCE - B DDC This is an example o modern radio architecture (receiver and transmitter). Most o the processing is carried out with digital circuits, and the analog RF ront-end has the task to handle RF signals limiting noise and intererers (receiver) or avoiding spurious emission (transmitter). These radio architecture will be analyzed in detail in the irst lesson o part B (B1); the ollowing ones are detailed descriptions o each unctional unit and o the inside circuits DDC 60

61 IF ilter Complete RX Image reject mixer Mixer II RF input ampliier PLL synthetizer LPF and A/D converter 05/03/ ATLCE - B DDC Another commercial circuit, with I/Q irst mixer or image rejection, and dierential circuits in the IF chain, rom LNA output to VGA input DDC 61

62 The transmitter chain Conventional Generate carrier Modulate Filter and Ampliy (PA) Reverse heterodyne receiver chain Generate baseband modulated signal Filter and translate to RF Filter and ampliy (PA) Next Direct generation o RF (high speed DAC) Digital correction o PA nonlinearity 05/03/ ATLCE - B DDC The previous slides describe only receivers; transmitters have similar architectures, with inverted signal path: - First operation: direct synthesis o I/Q baseband components, using DACs; - Next: combination o I/Q and requency shit as required (mixers and ilters), - Final: output Power Ampliier (PA) towards the antenna. Digital processing occurs or baseband components generation, and can be used to compensate the nonlinearity o the PA DDC 62

63 Complete RTX structure 05/03/ ATLCE - B DDC The slide shows the complete structure o a dual-band cell phone - Separate LNA and PA or each band, - Dierential signals in RX ater the irst mixer, - I/Q baseband 2016 DDC 63

64 ATLCE - B1 05/03/2016 Details o receiver I/Q chain Section o the receiver chain: - ilters, ampliiers, PLL, mixer - A/D converters - Numeric demodulator Signal low 05/03/ ATLCE - B DDC This is a detail o the receiver II IF chain. - dierential signals are extensively used; - the second mixer generates I and Q IF components; - anti-alias LPF are placed beore the ADC. This same block diagram can represent also a ZIF system DDC 64

65 I/Q transmitter structure Section o the transmission chain: - D/A converters, - ampliiers, ilters - PLL, mixer, Signal low 05/03/ ATLCE - B DDC Details o the transmitter chain - the digital circuits generate I and Q components, ed to a pair o DACs. - DAC output is smoothed by low-pass ilters (reconstruction ilters) - the I and Q signal are requency- translated by a pair o mixed, and combined to get the TX-IF signal. - A second mixed brings the modulated signal to the proper carrier position. - a VGA provides power control beore the PA 2016 DDC 65

66 Example o ZIF transceiver LPF : ZIF architecture 05/03/ ATLCE - B DDC The diagram represents a commercial ZIF transceiver The IF ilters (both RX and TX) are low-pass DDC 66