Receiver Noise Figure

Pass Loss

Tx Output

Pass Loss Rx Noise Figure

Noise Floor

Prof. C. Patrick Yue, ECE, UCSB

Power Spectrum of Global System for Mobile (GSM)

In-band

Close-in interferers i t f
Prof. C. Patrick Yue, ECE, UCSB

Sensitivity vs. Selectivity

Desired Channel to select Input

MIXER
Output

IF Filter
if Received channels after frequency translation

rf Received Channels at RF

if

LO

Received channels after frequency translation

Sensitivity
The minimum (available) signal power needed at the receiver input to provide adequate SNR at the receiver output to data demodulation Noise Insertion Loss Inter-modulation products

Selectivity
Blockers (in-band and out-of-band) Phase Noise Image-Rejection (will be discussed with radio architecture)
Prof. C. Patrick Yue, ECE, UCSB

Required Receiver Sensitivity A Qualitative View

What is the required receiver NF to achieve a certain level of sensitivity?
Transmit Power Path Loss Receiver S iti it R i Sensitivity Noise Figure Required SNR Noise Floor

To find Receiver NF
Transmit Power FCC regulated Path loss Receiver sensitivity govern by standards and applications Required SNR depends on BER requirement and modulation scheme Noise floor thermal noise or circuit noise limited depending on the modulation schemes

Prof. C. Patrick Yue, ECE, UCSB

Receiver NF Requirement Calculations

IEEE 802 11a WLAN 802.11a FCC limits the PSD in 5GHz to 2.5 mW/MHz Channel bandwidth is 16 MHz Transmit Power i 40 mW or 16 dB T it P is W dBm Thermal noise floor 174 dBm/Hz X 16 MHz = 102 dBm Total SNR budget is 16 dB ( 102 dBm) = 118 dBm dBm (102 dB ) dB To cover ~300 ft. at 5 GHz results in a path loss of 86 dB
i.e. Receiver sensitivity is 70 dBm (802.11a (802 11a specification is 65 dBm ) 65

GSM (DCS-1800 ) cellular FCC limits the PSD in 1.8 GHz to 5 mW/kHz Channel bandwidth is 200 kHz Ch l b d idth i kH Thermal noise floor 174 dBm/Hz X 200 kHz = 121 dBm Required SNR for GSM is 9 dB
to keep BER < 103

Required SNR for 64QAM (54Mbps) is 27 dB

802.11a packet length is 8 kb Worst packet loss < 10%, (1 BER)8000 = 1 10% BER = 105

GSM receiver sensitivity specification is 102 dBm 102 Receiver noise figure requirement = Receive sensitivity Noise floor Required SNR = 102 (121) 9 = 10 dB

Receiver noise figure requirement = Tx Power Path Loss Required SNR Noise floor = 16 + 102 86 27 = 5 dB
Prof. C. Patrick Yue, ECE, UCSB

Receiver Sensitivity for GSM

Prof. C. Patrick Yue, ECE, UCSB

One More Receiver Sensitivity Calculation Example

Prof. C. Patrick Yue, ECE, UCSB

Fundamental Concepts in RF Systems

Receiver sensitivity
Noise Figure Signal to noise ratio (SNR) Thermal noise floor

Receiver selectivity
Nonlinearity y
gain compression inter-modulation desensitization cross modulation

Phase noise and blockers

Receiver spurious-free dynamic range (SFDR)

Lower limit set by sensitivity Upper limit set by selectivity

Prof. C. Patrick Yue, ECE, UCSB

Key Receiver Metrics

At any input signal level, the receiver must achieve a minimum Signal-to-Noise Ratio (SNR)
Detection schemes need a minimum signal-to-noise ratio for adequate performance f
Some analog detectors (AM detectors) improve gradually with increasing SNR Digital detectors improve rapidly past a threshold SNR

Dynamic range
The range of input power (signal and interferer) over which the receiver performs adequately
Measured by performance of the base-band transducer (speaker/video display etc) For system analysis, Bit Error Rates or final SNR are used

Smallest signal level is the receiver sensitivity Largest signal determines the upper limit of dynamic range (What does largest signal mean? We will come back to this point later)

Prof. C. Patrick Yue, ECE, UCSB

Receiver Architecture Considerations

Heterodyne is a well-proven architecture
Monolithic implementation (low-cost integration) is a challenge owing to the large number of BPFs required

Alternative architecture suitable for integration will be studied later The architecture as shown is a consequence of available technologies
For example, if low loss, tunable front-end BPFs could be manufactured for channel select, the receiver could be replaced by one mixer

The components shown are usually common to all architectures with possibly different requirements Front-end circuits (e. g. LNA & Mixer) are critical design challenges and technology drivers in wireless applications

Prof. C. Patrick Yue, ECE, UCSB

Functions of Receiver Components (1)

T/R Switch RF Band Select Filter On O PCB On-Chip Balun LNA Image Reject Filter Channel Ch l Select IF RF Mixer Filter VGA 90 LO1 (Fixed)
o

ADC

LO2 (Tuned) ADC Q

IF Mixer Anti-alias Anti alias LPF

RF band select filter

typically a ceramic filter Used to filter and reduce incident power levels of distant interferers at the LNA allows the entire RF Band (all possible useful channels) into the receiver rejects out-of-band signals and attenuates image signals out of band

Transmit / Receive Switch

connects the antenna to the receiver or transmitter in a time-division duplexed systems

Balun
Balanced to Un-balanced differential to single-ended converter
Prof. C. Patrick Yue, ECE, UCSB

Functions of Receiver Components (2)

T/R Switch RF Band Select Filter On O PCB On-Chip Balun LNA Image Reject Filter Channel Ch l Select IF RF Mixer Filter VGA 90 LO1 (Fixed)
o

ADC

LO2 (Tuned) ADC Q

IF Mixer Anti-alias Anti alias LPF

Low Noise Amplifier (LNA)

Front-end amplifier used to amplify the signal with minimum degradation in the SNR amplifies the signal to reduce impact of noise from latter stages

Image Reject Filter

Ceramic (or SAW) band-pass filter used to provide filtering of distant interferers Used primarily to reject the image frequency of local oscillator (LO1) Allows the entire RF Band (all possible useful channels) into the receiverattenuates image-signals before mixing

RF Mixer
converts the incoming RF signal to intermediate frequency (IF) is the difference between the RF and LO1 Usually have stringent linearity and noise requirement
Prof. C. Patrick Yue, ECE, UCSB

Functions of Receiver Components (3)

T/R Switch RF Band Select Filter On O PCB On-Chip Balun LNA Image Reject Filter Channel Ch l Select IF RF Mixer Filter VGA 90 LO1 (Fixed)
o

ADC

LO2 (Tuned) ADC Q

IF Mixer Anti-alias Anti alias LPF

Channel select filter

Select the desired the channel and rejects adjacent channels Typically requires a SAW filter with high attenuation to suppress out of band tones

Intermediate frequency variable gain amplifier (IF VGA)

adjusts the received signal level so that it maps to the dynamic range of the based-band circuits such as the ADC

IF mixers
Down-converts the I & Q signals to base-band for signal processing In the b I th above example, we assume that the i l th t th image f frequency i greatly attenuated b the is tl tt t d by th channel select filter and therefore image-reject mixers are not used.
Prof. C. Patrick Yue, ECE, UCSB

Receiver Requirements
Gain and stability requirements
Power gain, voltage gain, stability measures

Low-noise requirements
Noise figure or temperature Desensitization (impact of non-linearity on noise performance)

Linearity requirements y q
Intercept points, gain compression

Prof. C. Patrick Yue, ECE, UCSB

Noise in Receiver
Receiver noise level directly limits sensitivity Receiver sensitivity = minimum input power that the receiver can detect Noise figure of cascaded stages g g
Noise figure of RF receivers from antenna to ADC output Noise figure of passive networks Noise figure of ADC g

Prof. C. Patrick Yue, ECE, UCSB

Input-Referred SNR in Terms of Noise Factor

Prof. C. Patrick Yue, ECE, UCSB

Receiver Sensitivity Min. Pin to Achieve Required SNR (1)

Prof. C. Patrick Yue, ECE, UCSB

Receiver Sensitivity Min. Pin to Achieve Required SNR (2)

Prof. C. Patrick Yue, ECE, UCSB

Receiver Sensitivity Min. Pin to Achieve Required SNR (3)

Prof. C. Patrick Yue, ECE, UCSB

Noise Figure for Cascaded Stages

Prof. C. Patrick Yue, ECE, UCSB

Noise Factor of Passive (Lossy) Networks

Prof. C. Patrick Yue, ECE, UCSB

Noise Figure Calculation of BPF Followed by LNA

Prof. C. Patrick Yue, ECE, UCSB

SNR of Analog-to-Digital Converter

Typically, ADC is characterized using SNR at the output rather than NF To determine the NF of an ADC we need to compute the degradation in ADC, SNR due to quantization noise after the signal passes through the ADC
Prof. C. Patrick Yue, ECE, UCSB

Noise Figure of Analog-to-Digital Converter

NF ADC = SNR in P / N i in Noise Noise N i out = in = Pout / Noise out SNR out Noise in

Since the ADC only performs digitization of the input voltage and thus does not provide any gain, i Pin = Pout, the output signal can b t t d as th input signal id i i.e. th t t i l be treated the i t i l plus quantization noise. Expressing NF in log form, we obtain:
NF = Noise
out ( dBm

) Noise

in ( dBm

Assume that the ADC noise is completely due to quantization error, then
Noise
out

( dBm ) = Pin SNR ADC

Assume that the noise at the input is due t th A th t th i t th i t i d to thermal noise, then l i th
NF ( dB ) = Pin SNR ADC Noise = Pin SNR ADC kTB
in

One can also express NF as the power ratio of quantization noise (at the output) and thermal noise (at the input) which results in:
NF = NP 02( rms ) A p , ADC 4 kTR s B 1 ) 4 = 2 2 N 12 R s
2 V FS

Quantizati on Noise a t Output 1 = kTB Thermal No ise at Inp ut

( A p , ADC =

Prof. C. Patrick Yue, ECE, UCSB

Effect of Over-Sampling on ADC NF

Increasing the sampling frequency reduce noise which has the same effect as noise, increase the ADC resolution
Over-sampling by a factor of 4 results in 6 dB reduction in noise, or effectively 1 more bit
Prof. C. Patrick Yue, ECE, UCSB

Receiver Requirements
Gain and stability requirements
Power gain, voltage gain, stability measures

Low-noise requirements
Noise figure or temperature Desensitization (impact of non-linearity on noise performance)

Linearity requirements y q
Intercept points, gain compression

Prof. C. Patrick Yue, ECE, UCSB

Interferers in Global System for Mobile (GSM)

In-band

Close-in interferers i t f
Prof. C. Patrick Yue, ECE, UCSB

The Large Signal

Prof. C. Patrick Yue, ECE, UCSB

Interferers

Prof. C. Patrick Yue, ECE, UCSB

Linear Systems
The Th system S i linear if and only if t is li d l if:
x1 x2 y1 y2 Then: x1+ x2 x1 y1+y2 y1 x S y

In other words in a linear system, the output can be expressed as a linear combination of the responses to the individual inputs. In a linear system with no initial stored energy, the output time function is related to the input time function by the convolution integral:
y (t ) =

x( ) h (t ) d

where h(t) is the system response to a unit impulse. In frequency domain: Y j = X j H j

( ) ( ) ( )

System transfer function

H( j ) =

h( t) e

j t

dt
Prof. C. Patrick Yue, ECE, UCSB

Source of Non-linearity
A system is time-invariant if a time shift in the input results in the same time shift in the output:
x(t) y(t) Then: x(t-) x(t ) y(t-) y(t )

In linear time invariant systems, a single frequency input can only generate a single frequency output output. A linear time variant system, e.g. a mixer, can generate frequency components that do not exist in the input, which cause non-linearity. Device non linearity non-linearity
ID = K(V GS - VT)2

Ic = Ics. E (qVBE /KT) Exp ( V

Prof. C. Patrick Yue, ECE, UCSB

Gain Expansion or Compression due to Nonlinearity

For time-variant, memoryless systems, we can assume F ti i t l t
x(t) NL y(t)

y(t) = a1 x(t) + a2x2(t) + a3x3(t) + ...

x(t) = Acos(ct) y(t) = a1Acos(ct) + a2A2cos2 (ct) + a3A3cos3(ct) + ...

a3 3 a2 2 a3 3 a2 2 = ---- A + a1 A + 3---- A cos ( ct ) + ---- A cos ( 2 c t ) + ---- A cos (3 c t ) 2 4 2 4

From this equation we see that the output signal consists of a component at the applied fundamental frequency c and spurious signals at dc, the second harmonic 2c, and the third harmonic 3c. The amplitude of the fundamental component can be greater than a1A (the gain if the two-port is linear) if a3>0 and smaller than a1A if a3 < 0. This property is called gain expansion or gain compression.

Prof. C. Patrick Yue, ECE, UCSB

1-dB Compression Point

Aout
a 3<0 Ao u t = output amplitude @ c 1 dB

Aout A

A1 dB A

Gain at fundamental frequency = 20log |a1 + 0.75a3A2| Linear Gain = 20log |a1| At the 1-dB compression point, the actual gain is 1dB below the linear gain 20log |a1 + 0.75a3A2| = 20log |a1| 1 Therefore for , A-1dB2 = 0.145a1/a3 (a3 < 0) or
A1dB = 0.145 a1 a3

P-1dB = A-1dB2 / 2R
Prof. C. Patrick Yue, ECE, UCSB

Blocker and Desensitization

Blocker: If input signal to the receiver consists of a weak desired signal at c1 accompanied by a strong interferer at c2 (the blocker). The blocker tends to reduce the average gain experienced by the desired signal:

Meaning that the effective signal gain at c1 (desired signal) is reduced by

For large enough A2, the receiver is desensitized as the output at c1 is g g p overwhelmed by the blocker.
Prof. C. Patrick Yue, ECE, UCSB

Inter-modulation
Intermodulation products due to two input tones:

4 a1 3 a3

= AIP3

IIP3 = AIP32 / 2R

When A = IIP3, the 3rd order term = fundamental at the output (gain ( i compression is neglected, i 9/4*a3*A3 << a1A) i i l t d i.e. 9/4*
Prof. C. Patrick Yue, ECE, UCSB

Signal Corruption due to IIP3 of Interferers

Aint, in Asig, in AIM3,out Aint, out Asig, out

Given Asig, in, Aint, in and IIP3, we want to find the ratio of the signal to IM3 i e Asig, out / AIM3,out at the output IM3, i.e. To find the dynamic range, we refer Asig, out / AIM3,out to the input

Prof. C. Patrick Yue, ECE, UCSB

IP3 Calculation and Graphical Interpretation

To express IIP3 in terms of the input and output signal amplitudes, take the ratio of the first and third terms from the previous expression and express in terms of IIP3
Aint,out AIM 3,out =
2 Aint, in 2 AIP 3

(Slope = 1) Aint,out

(Slope = 3) Aint,in AIM3,out

AIM3, in

Aint,in

= AIP3
Prof. C. Patrick Yue, ECE, UCSB

Signal Corruption due to IIP3 of Interferers

Aint, in Asig, in AIM3,out

Aint, out

Asig, out

Asig ,out Asig , in

Aint,out Aint, in

Asig ,out =
Asig ,out AIM 3,out

Aint,out Aint, in
=
2 AIP 3

Asig , in

AIM 3,out

2 Aint, in

2 AIP 3

Aint,,out

3 Aint, in

Asig , in

Given Asig, in = 1Vrms, Aint, in = 1mVrms, and IIP3 = 10dBm (AIP3 = 70mVrms on 50 )

Asig ,out AIM 3,out

(1m )

(70m )3 = 4.9 = 13.8dB 3

Prof. C. Patrick Yue, ECE, UCSB

Relation between 1-dB Compression Point and IIP3

A1dB = 0.145 a1 a3

(with single tone input)

IIP 3 =

4 a1 3 a3

(with dual tone input)

0.145 A1dB = = 0.33 IIP 3 4/ 3

IIP 3 (dB) = A1dB (dB) + 9.6

Prof. C. Patrick Yue, ECE, UCSB

IIP3 of Cascaded Stages (I)

Prof. C. Patrick Yue, ECE, UCSB

IIP3 of Cascaded Stages (II)

The term consisted of 2 (due to 2nd order distortion) can be neglected if gain of stage 1 has a band-pass characteristic

Prof. C. Patrick Yue, ECE, UCSB

Dynamic Range

DR = P1dB kTB(dBm) F SNRmin DR = IIP3 9.6 kTB(dB ) F SNRmin 9 6 kTB(dBm)

Prof. C. Patrick Yue, ECE, UCSB

Spurious Free Dynamic Range

Prof. C. Patrick Yue, ECE, UCSB

Receiver Selectivity
A measure of the receiver s ability to reject signal outside of the desired receivers band. We have seen that IM3 due to two adjacent channels creating spurious tone in the designed signal band Blockers and phase noise of the local oscillator (LO) signal also degrades receiver selectivity.
Phase noise i a measure of spectral purity of th LO signal Ph i is f t l it f the i l Blocker

Prof. C. Patrick Yue, ECE, UCSB

Phase Noise Requirement

Input Spectrum BW Desired f c Phase Noise
S x (f) () f

BL

Receiver
Signal LO Vout

Desired Signal Blocker Mixed Inband

Receiver Output

Local Osc. Output f L0 fc

C/Im in

-PN ( fc ) f

Assume that the receiver is noiseless, therefore required SNR is determined by C/Imin (Carrier / Interferer ratio)
C / I min ( dB ) = S signal Sblock + PN fc (dBc )

= S signal Sblock PN fc (dBc / Hz ) 10 log( BW ) g( PN fc (dBc / Hz ) = S signal Sblock C / I min ( dB ) 10 log( BW )

Prof. C. Patrick Yue, ECE, UCSB

Phase Noise Requirement Calculation

PCS 1900 (North America version of GSM)

Desired signal at fo can be as small as 99dBm with 43-dBm blocker at 600kHz GSM required SNR is 9 dB Channel bandwidth is 200 kHz PN (at 600kHz offset) = 99 ( 43) 9 10log(2e5) = 118 dBc/Hz
Prof. C. Patrick Yue, ECE, UCSB