Receiver Architecture

Receiver basics
Channel selection why not at RF? BPF first or LNA first? Direct digitization of RF signal

Receiver architectures
Sub sampling Sub-sampling receiver noise problem Heterodyne receiver image problem Super-heterodyne receiver more image problem Image-reject receivers
Harley receiver Weaver architecture

Homodyne (direct conversion, zero-IF) DC offset zero IF) Digital IF Wide-IF double-conversion Sliding IF
Prof. C. Patrick Yue Slide 1

Fundamental Trade-off in Receiver

Using one or more IF stages to relax the filter requirements, but need to deal with images Using image reject mixers with I&Q LO signals to eliminate the need of band-pass filters (to enable higher level of integration) RF ICs typically employ a combination of simple mixing with some image filtering and image reject mixing

Prof. C. Patrick Yue

Slide 2

Channel Selection at RF?

GSM example: channel bandwidth is 200 kHz, RF carriers at 935960 MHz Filter Q = 10 * RF / BW = 10 * 950 M / 200k = 47 500! 47,500! (Impossible to achieve such high Q at RF, or too expensive!) Instead, we do band select, for GSM, the band is 25 MHz, so the filter Q required is 10 * RF / BW = 10 * 950 M / 25 M = 380, much more reasonable.

Prof. C. Patrick Yue

Slide 3

BPF First or LNA First?

Trade-off between suppressing inter-modulation products due to interferers vs. noise figure BPF first: better interferer rejection, but higher noise figure due to the insertion loss of the filter LNA first: better noise figure, receiver can be desensitized due to interferers Interferers are bigger problem so BPF first is adapted in all receivers problem,

Prof. C. Patrick Yue

Slide 4

Receiver Using High-Speed DSP

Directly sample the carrier at RF to facilitate the use of high-speed DSP. For input power ranging from 100 dBm (3.2 Vpeak) to 10 dBm (100 mVpeak), the ADC will need the following performance:

A 1-GHz, 15-bit ADC is impossible to implement with reasonable power in the near future

Prof. C. Patrick Yue

Slide 5

Sub-Sampling Receiver

Prof. C. Patrick Yue

Slide 6

Noise in Sub-Sampling Receiver

The noise floor is raised by a factor of 2m, where m is the sub-sampling factor Phase noise is increased by m2 at the output of the sub-sampler

Prof. C. Patrick Yue

Slide 7

Heterodyne Receiver

The KEY question in receiver design is at what stage to perform at pe o channel select? Easier to achieve high-Q BPF at lower frequency, which favors lower IF but IF, but Image rejection becomes difficult Another questions is should the LO (LO) be above or below carrier (RF)?

Mixer output

At mixer output At mixer output

Prof. C. Patrick Yue Slide 8

High-Side LO vs. Low-Side LO

Advantage of using high-side LO
the ease in tuning the LO over the desired band of frequencies choice of High Side LO is motivated by the ease in tuning the LO over the desired band of frequencies. Tuning of the LO is often done using a varactor. For a given voltage change (and varactor capacitance change), the LO frequency can be changed over a wider range of frequencies for a high-side LO compared to a low-side LO. Due to the limited linearity of the varactor, low side choice of the high-side LO results in improved linearity of the LO frequency with change in bias voltage. Due to this reason, the high-side LOs are more popular.

Advantage of using low-side LO

lower noise and power dissipation since operating at lower frequencies

Another important consideration in the choice of high-side LO versus p g low-side LO is in the image frequencies that will be picked up. The choice of high-side LO versus low-side LO might be made based on the relative quietness of the image band in each case.

Prof. C. Patrick Yue

Slide 9

Image Reject Filter in Heterodyne Receiver

Trade-off between sensitivity (image rejection) and selectivity (channel selection) dictates the choice of IF
High-IF image reject filter easier to implement and provides better sensitivity Low-IF channel select filter easier to implement and gives better selectivity

Both image reject BPF and channel select select BPF are difficult to implement on chip, which makes heterodyne receiver less attractive for chip hich heterod ne recei er attracti e monolithic RF transceiver
Prof. C. Patrick Yue Slide 10

Half-IF Problem

2LO

A spurious tone at 0.5*IF (half way between RF and LO) undergoes even harmonic 0 5*IF distortion and generates a spur at (RF + LO) at the input of the mixer, which mixes with the 2nd harmonic of the LO to produce noise at (RF LO). To suppress the half-IF phenomenon
minimize second-order distortion in the RF path (using differential circuits) maintain 50% duty cycle in the LO to reduce 2LO
Prof. C. Patrick Yue Slide 11

Channel Selection by Adjusting LO Frequency

Tunable high-Q bandpass filters are difficult and expensive to implement, so rather th t i th BPF center f th than tuning the t frequency, th LO is changed the i h d depending on the desired channel
For example, FM radios use 10.7 MHz as the fixed IF, to tune to the station at 101.3 MHz, 101 3 MHz the LO is adjusted to 112 0 MHz. 112.0 MHz
Prof. C. Patrick Yue Slide 12

Super-heterodyne Receiver

To relax the trade-off between sensitivity (image reject) and selectivity (channel select), we can introduce a second IF to the heterodyne receiver architecture, which results in a super-heterodyne receiver p y A super-heterodyne receiver is a heterodyne receiver with dual IFs A super-heterodyne receiver relaxes the bandpass filter Q at each stage by having more filter stages
Qoverall = 10 950 M 100 M 10 M 10 10 = 95 100 500 = 4.75e 6 100 M 10 M 200k
Prof. C. Patrick Yue Slide 13

Secondary Image Problem in Superheterodyne Receiver

The input spurious tone at IM2 (at input of 1st mixer) = RF + 2LO2 will p p ( p ) cause a secondary image tone at LO2 + IF2 at the input of the 2nd mixer. Example: given that RF = 950 MHz, LO1 = 1050 MHz, and LO2 = 110 MHz
IF1 = 100 MHz and IF2 = 10 MHz IM2 = IF2 + LO2 + LO1 = 10 + 110 + 1050 = 1170 MHz (= RF + 2LO2)
Prof. C. Patrick Yue Slide 14

Another Source of Secondary Image

The input spurious tone at 2LO1 2LO2 RF will also cause a secondary image tone at LO2 + IF2 at the input of the second mixer Notice that the difference between 2LO1 2LO2 RF and LO1 is the same as p p the spurious tone in the previous slide

Prof. C. Patrick Yue

Slide 15

Channel Selection in Super-heterodyne Receiver

There are three approaches
Variable LO1 with fixed LO2
Requires very high precision in frequency synthesizer (temperature variation can be a big problem)

Fixed LO1 with fixed LO2

Requires very wide tuning range in in frequency synthesizer

Variable LO1 with variable LO2

Requires both LOs to track each other

Prof. C. Patrick Yue

Slide 16

Channel Selection in Super-heterodyne Receiver

Variable V i bl LO1 and fixed LO2 d fi d

Using a variable LO1and fixed LO2 makes the task of channel selection extremely challenging Example: In GSM we need to zero down on to a frequency such as 935.20 MHz with increments of 0.2 MHz Fractional h F ti l change in LO1frequency is 0.2 MHz / 950 MHz ~ 0.02% i LO1f i 0 2 MH MH 0 02% Total change in LO1 is 25 Hz / 950MHz ~ 2.5%
Prof. C. Patrick Yue Slide 17

Channel Selection in Super-heterodyne Receiver

Fixed LO1 and variable LO2

Using a fixed LO1 makes it easier to design LO1. Fixed frequency oscillators have the advantage that a lower phase noise can be obtained due to the lower PLL bandwidth that can be used Using a variable LO2 makes the task of channel selection much easier. Fractional change required in LO2 is 0.2 MHz / 100 MHz ~ 0 2% 02 0.2%. Total change required in LO2 is 25 MHz / 100 MHz ~ 25%
Requires wide tuning range VCO in the frequency synthesizer, need to the use of varactor in conjunction with switching capacitor arrays
Prof. C. Patrick Yue Slide 18

More Filters => More Images => More Filters

High-Q filters are hard to build and expensive Use more filters each with lower Q But that requires more mixing in the receive chain which leads to image q g g problems and needs more filters So, So are there any other way to suppress image without using filters?

Prof. C. Patrick Yue

Slide 19

Image Reject Receiver

If the desired input is cosine p

<= Use cosine as the in-phase, I carrier

<= Use sine as the quadrature-phase, Q carrier

(RF < LO < IM because we are using high-side LO)

Prof. C. Patrick Yue Slide 20

Principle of Image Reject Receiver

I and Q path use LO th t are 90 out of phase, cosine vs. sine d th LOs that t f h i i
Prof. C. Patrick Yue Slide 21

Principle of Image Reject Receiver

90 degree shift is added to the Q path, such that the output due to the image path signal is 180 out of phase with respect to the image in the I path
Prof. C. Patrick Yue Slide 22

Image Reject Receiver

Question: Remember quadrature modulation using I&Q as two separate channels? How would it work in a image reject receiver architecture?
Prof. C. Patrick Yue Slide 23

Hartleys Image Reject Receiver

Prof. C. Patrick Yue

Slide 24

Quadrature Generators
Frequency dividers (covered in previous lecture) Polyphase filter

Prof. C. Patrick Yue

Slide 25

Polyphase Filter (1)

Prof. C. Patrick Yue

Slide 26

Polyphase Filter (2)

Prof. C. Patrick Yue

Slide 27

Practical Consideration in Polyphase Filter Design

(lowers the output amplitudes)

Prof. C. Patrick Yue

Slide 28

Practical Implementation of Multi-Stage Polyphase Filter

Using a ring topology to allows easy cascading of multiple stages

Prof. C. Patrick Yue

Slide 29

Practical Implementation of Multi-Stage Polyphase Filter

With Qi and QBi absent, the polyphase filter reduces to the simple RC-CR The Th current configuration allows easy cascading of multiple stages t fi ti ll di f lti l t
Prof. C. Patrick Yue Slide 30

Practical Implementation of Multi-Stage Polyphase Filter

With Qi and QBi absent, the polyphase filter reduces to the simple RC-CR The Th current configuration allows easy cascading of multiple stages t fi ti ll di f lti l t
Prof. C. Patrick Yue Slide 31

Practical Implementation of Multi-Stage Polyphase Filter

Multi-stage broaden the bandwidth over which the amplitude of I and Q paths matches.
Prof. C. Patrick Yue Slide 32

Image Rejection Ratio in Hartley Receiver

For perfect image rejection, this term is equal to zero

Why 4A2?

5% of amplitude imbalance (~0 5 dB) and 1 of phase mismatch result in (~0.5 approximately 30 dB of IRR
Prof. C. Patrick Yue Slide 33

Image Rejection Ratio

Amplitude imbalance and phase mismatch in the I and Q paths limits the IRR
Gain mismatch can usually be limited to < 0.5~1.0 dB and phase error is around 12 Mismatch in mixers and LPF also d Mi t hi i d LPFs l degrade IRR d

Typical image rejection ratio achievable on-chip is about 2530 dB using Hartley architecture

Prof. C. Patrick Yue

Slide 34

Polar Plot of IRR

Prof. C. Patrick Yue

Slide 35

Applying Polyphase Filters in Hartley Receiver

Quadrature LO can be generated by passing the output of the frequency synthesizer through a polyphase filter
A 90 phase shift between I & Q signals is achieved by shifting the I-signal with 45 and the Q-signal with +45

Prof. C. Patrick Yue

Slide 36

Adding the I and Q Paths

Prof. C. Patrick Yue

Slide 37

Complete Implementations of Hartley Receiver

Prof. C. Patrick Yue

Slide 38

Practical Considerations
Image Rejection Requirement
An overall image suppression of 5570 dB is needed in most receiver (what determines the required image rejection?) With an appropriate choice of IF frequency (high enough), an image suppression of 30-40 dB can be achieved by the RF band select filter or the image reject filter (if necessary) Hartley s Hartleys architecture provides an additional 25 30 dB image rejection 2530 bringing the overall image rejection to 5570 dB Other mismatches in the I&Q paths limits the practical IRR to 25 to 30 dB
Mismatches between I&Q mixers, LPFs I&Q LO mismatches in amplitude and phase

Polyphase filter yp
Multi-stage improves I & Q matching over a wider bandwidth, but increases noise (thermal noise in the resistors) and power consumption (amplitude reduction) Practical implementation rarely uses more than 2 stages of RCCR
Prof. C. Patrick Yue Slide 39

Weaver Architecture

We have seen that a 90 phase shift is introduced between I & Q signals by mixing the incoming signal using cosine and sine LOs We need to achieve another 90 phase shift, such that the down-converted desired signal will remain in phase while the one due to the image spur will be become 180 out of phase. In Weaver architecture, the second down-conversion mixing is also performed with I&Q LOs to achieve another 90 phase shift between the I&Q paths
Prof. C. Patrick Yue Slide 40

Second Quadrature Mixing in Weaver Architecture

Note that a negative sine is used as the LO for the Q path to provide the +90 phase shift because the incoming desired signal is a positive sine wave and the image tone is negative sine wave
Prof. C. Patrick Yue Slide 41

Weaver Receiver

What happens if the desired input signal is a sine wave? pp p g Weaver architecture is sensitive to relative phase of the RF input and the LOs To support quadrature modulation (incoming signals in cosine and sine), the second mixing requires quadrature mixing to preserve the desired signal

Prof. C. Patrick Yue

Slide 42

Secondary Image Problem in Weaver Receiver

Recall that there are two secondary image tones which can cause interference at IF2 the one above LO1 will appear as a negative sine wave at the input of the second mixer, so it will under go a total of 180 phase shift after the second down-conversion 180 down conversion the one below LO1 will appear as a positive sine wave (just like the desired signal) at the input of the second mixer, and it will become in-phase with its I-path counterpart BPF is used to remove this tone before the second mixer Prof. C. Patrick Yue

Slide 43

Homodyne Receiver

LO frequency is the same as the incoming RF carrier frequency Also known as direction conversion or zero-IF receiver No more image problem to worry about But it will work only if the desired signal has symmetrical sidebands (known as double sideband modulation)
Prof. C. Patrick Yue Slide 44

Homodyne Receiver

Done in baseband using DSP g

Use quadrature mixing to separate the upper sideband signal and the lower sideband signal g Quadrature mixing also removes the problem due to phase mismatch between the carriers and LO Use DSP to reconstruct the desired signal in the baseband Note that is the input signal has a 2 MHz bandwidth, the I&Q paths each need to have a bandwidth of 1 MHz
Prof. C. Patrick Yue Slide 45

Quadrature Down-Conversion for Homodyne Receiver

If phase modulation is used, quadrature mixing converts the information in to the relative phase between I and Q signals. In the abo e e a p e, t e ca e is modulated us g Q S , hence t e above example, the carrier s odu ated using QPSK, e ce (t) is /4, 3/4, 5/4, or 7/4
Prof. C. Patrick Yue Slide 46

Homodyne Receiver Pros and Cons

Advantages
Remove the need for image reject BPF between LNA and mixer Channel select can be performed using LPF in stead of BPF

Disadvantages
DC offset (A BIG PROBLEM!!)
LO leakage Strong interferer Time-varying offset LO with non-50% duty cycle y y Even order distortion in LNA

1/f noise I&Q gain and phase mismatch More stringent dynamic range and reverse isolation

Prof. C. Patrick Yue

Slide 47

DC Offset

DC offset can be as large as 10 mV due to various sources The desired signal can be much small, e.g. 0.5 mV In order for the ADC to be able to resolve the desired signal, the IF amplifier needs to provide sufficient gain so the desired signal reaches the full scale of the ADC, e.g. 500mV With a gain of 1000, the DC offset will clearly saturate the ADC

Prof. C. Patrick Yue

Slide 48

DC Offset Due to LO Leakage

LO signal can leak through the LO port to the RF port due to parasitic couplings and cause self mixing

Prof. C. Patrick Yue

Slide 49

DC Offset Due to Strong Interferer

A strong interferer can leak through the RF port to the LO port due to parasitic couplings and cause self mixing

Prof. C. Patrick Yue

Slide 50

DC Offset Due to Time-Varying Offsets

LO signal can leak through the antenna, radiate into the air and reflect from the surrounding and reach the RF port of the mixer Good LNA reverse isolation can suppress this effect

Prof. C. Patrick Yue

Slide 51

DC Offset Due to Non-50% Duty Cycle LOs

If the duty cycle of the LO is not 50%, the output of the mixer will have a DC offset ff t

Prof. C. Patrick Yue

Slide 52

DC Offset Due to Even Order Distortion in LNA

In homodyne receivers we also need to consider even order distortion characterized by IIP2 Consider two strong interferers closely spaced in frequency being received atthe antenna. Second order distortion results in a difference frequency to appear at t S d d di t ti lt i diff f t t the output of the LNA. The mixer exhibits a finite amount of direct feedthrough; hence the difference frequency signal would end up at the output of the mixer corrupting the desired signal. For homodyne applications, the LNA should be designed t h d i d to have hi h IIP2 i addition t hi h IIP3 high in dditi to high IIP3. Use differential circuit to reduce even order distortion
Prof. C. Patrick Yue Slide 53

Reducing DC Offset with AC Coupling

Requires huge AC coupling capacitor (too big to fit on a chip) Very slow settling time, for example, a 200-Hz cut-off implies a settling time of 2 ms, which can too long for most system Only O l works with zero DC modulation which has no data below the cut-off k ith d l ti hi h h d t b l th t ff frequency of the HPF

Prof. C. Patrick Yue

Slide 54

Reducing DC Offset with Offset Cancellation

Some systems such as TDMA, inherently contain time intervals during which the receiver is idle, which could be used to perform offset cancellation The output DC voltage accumulated on the capacitor during the idle time could be measured and subtracted from the output voltage of the mixer resulting in cancellation of the DC offset If the offset cancellation is performed at a sufficient rate, the time-varying DC time varying offset can also be handled kT/C noise due to the switch must be considered An alternative is to have two sets of mixers so that, at any given moment, one is used while the other is having its offset cancelled sed hile ha ing
Prof. C. Patrick Yue Slide 55

1/f Noise in Homodyne Receiver

1/f noise from the mixer presents a severe problem in the design of homodyne receivers since the 1/f noise spectrum falls in the same band as the down-converted output signal

Prof. C. Patrick Yue

Slide 56

I&Q Mismatch in Homodyne Receiver

Ideal I&Q constellation

I&Q constellation gain and phase mismatch

Phase mismatch is a more severe problem than gain mismatch

Deviation from the quadrature phase difference means that some of the I g pp , signals will appear in the Q channel and vice versa, which reduce SNR in both channels Can be compensated with DSP in the baseband using known data as training (or calibration) sequence

Gain mismatch can be compensated by the variable gain IF amplifiers in the I&Q paths
Prof. C. Patrick Yue Slide 57

Channel Selection in Homodyne Receiver

Case 1: relaxes LPF noise requirements, demands higher linearity from IF amp Case 2: LPF needs to low noise figure IF amp linearity requirement is relaxed figure, Case 3: high linearity required in both IF amp and ADC, LPF performed in digital domain Which one will you choose?
Prof. C. Patrick Yue Slide 58

Requirements on Homodyne Receiver

Linear LNA (low IIP2 and IIP3) Linear mixers (to suppress DC offset) LOs LO operating at quadrature with precisely 50% d cycle i d ih i l duty l DC offsets in the range of uV Low 1/f Noise High degree of isolation and stability

Prof. C. Patrick Yue

Slide 59

Digital IF Receiver

Commonly used for multi-band, multi-mode cellular phone applications The second stage of mixing and filtering in a super heterodyne (dual IF) architecture is performed in the digital domain After the first i Aft th fi t mixer, th signal i digitized b th A/D the i l is di iti d by the The quantization and thermal noise of the A/D cannot exceed a few uV for a good receiver The li Th linearity of the A/D must be sufficiently high to suppress the it f th tb ffi i tl hi h t th intermodulation outputs from corrupting the desired signal Choice of the first IF is dictated by the speed of the A/D Typically, Typically first IF is around 75 MHz with the ADC running at 150 to 200 MS/s and 9 to 11 bit resolution
Prof. C. Patrick Yue Slide 60

Full Implementation of a Digital IF Hartley Receiver

But we still need image reject BPF filter before the first mixer, which means that we need to go off chip can we do better? off-chip,

Prof. C. Patrick Yue

Slide 61

Wide-IF Double Conversion Receiver

The wide IF double conversion architecture is Weaver architecture with the IF2 = 0 As in a Weaver architecture, there are two stages of down-conversion The secondary image that plagues Weaver architecture is suppressed by using IF2=0 As in a homodyne receiver, additional mixers (hence the name double conversion) are required to correctly detect the signal The first LO is fixed and the second LO is tuned to the desired LO Because the first LO is fixed, easier trade-offs may be obtained with regard to a LO fixed trade offs phase noise As in the case of homodyne architecture, channel select BPF is eliminated LO leakage problem is greatly suppressed since LO1 is not equal to the carrier g p g y pp q frequency of the desired signal The only filter therefore required is the front-end band select filter Very good for monolithic implementation

Prof. C. Patrick Yue

Slide 62

Wide-IF Double Conversion Receiver

Prof. C. Patrick Yue

Slide 63

Sliding-IF Receiver Architecture

Both the first and second LO are generated by the same frequency synthesizer
Prof. C. Patrick Yue Slide 64

Sliding-IF Receiver Architecture

Vary LO1 and LO2 together to perform essentially the same function as direct conversion No external IF filtering Channel selection at baseband with LPF Very high IF of 1GHz 3GHz image i 2GH away from 5GHz signal 3GH i is 2GHz f 5GH i l Inherent bandpass filtering of 3GHz: 23dBc RF mixer: 5-4 = 1GHz (IF) and 5+4 = 9GHz No image-reject mixers required

Prof. C. Patrick Yue

Slide 65

Frequency Synthesizer for Sliding-IF Architecture

LORF is the first LO LOIF is the second LO

Divid-by-4 in the divider chain produces I&Q LOIF with excellent q quadrature p p properties

Prof. C. Patrick Yue

Slide 66

Receiver Architecture Trade-offs

Prof. C. Patrick Yue

Slide 67

References
1.

2.

Prof. M. Perrott, MIT http://ocw.mit.edu/OcwWeb/Electrical-Engineering-and-ComputerScience/6-776Spring-2005/CourseHome/index.htm Prof. L. Larson, UC San Diego ECE 265A and 265B lecture notes

Prof. C. Patrick Yue

Slide 68