Receiver Design – Tutorial

James B. Offner (Author)

Harris Corporation

Government Communications Systems Division

2400 Palm Bay Rd

Palm Bay, Florida 32905

Abstract––Numerous interrelated trade-offs are undertaken for

any receiver or receive chain design, which must be jointly

optimized for the intended operational environment. Some of

the requirements resulting from this environment are: noise

figure (NF), input 3rd order intercept point (IP3), input 1dB

compression point (P1dB), dynamic range, input desensitization

level, non-damage input power, out-of-band (OOB)

interference rejection, gain and output power. This paper

focuses on these requirements and the subtleties associated

with achieving them.

Keywords-receiver; receive chain; frequency plan; spur

analysis; spur; cascade analysis; Out-of-band; interference;

noise source; AGC; ALC; Monte-Carlo; intermodulation

I. INTRODUCTION

Receive (Rx) chain design is grouped into six key areas listed below, which are then expanded and treated more fully.

1. Frequency planning / spurious (spur) analysis 2. Cascade analysis of device Gains, NFs, IPs, P1dB’s

and damage levels 3. Non-standard noise sources other than from cascade

(wideband amplifier, image, LO, and reciprocal mixing)

4. Out-of-band interference rejection 5. Automatic Gain or level Control (AGC, ALC) 6. Statistical parameter variation, gain alignment and

compensation Many of these topics are interrelated, where optimization

of one area often negatively impacts one or more of the

others. An optimized receiver design globally optimizes all

six key design areas with equally weighted margins across

all parameters.

II. FREQUENCY PLANNING / SPUR ANALYSIS

One of the first things to do for any Rx chain design is to develop a frequency plan and perform a spur analysis (SA) on that plan. The frequency planning process determines the RF, IF, and LO frequency ranges. Usually the RF is given for a receiver, but sub-bands of RF may be more palatable to deal with by using a switched filter bank. Without the aid of a good frequency planning tool, this can be a time consuming, “bring me a rock” exercise, where the optimum plan may not be discovered or worse, the complexity, cost and performance are subpar. All Rx chains are susceptible to in-band and OOB spurious responses, which must be

managed to provide robust performance in a receiver intended for hostile signal and interference environments.

Filter quantity and complexity, as well as gain and phase linearity, are directly impacted by the frequency plan. The number of filters in a RF switched filter bank for a wideband receiver can be influenced by the IF selection, and an IF switched filter bank (selectable IFs after the first conversion stage) can reduce the RF filter quantity. Also, multiple conversion stages may reduce the total filter count and are often necessary to meet spur requirements for wide frequency ranges and large ratios of RF to final IF.

A frequency planning tool facilitates finding an IF that is “spur free” or exhibits the lowest spur levels possible, given the input signal and interferer levels, selected mixer spur responses, and required bandwidths (BW) or tuning range at each mixer port. More than one IF may be usable for a given conversion stage and the best one will optimize all of the above parameters as a group. Once a frequency plan is chosen, it can be further refined by modeling actual RF and IF filter responses and performing a SA given the RF, LO and IF frequency ranges developed using the frequency planning tool. The SA takes into account the filter’s rejection of OOB input levels, which can significantly improve the resulting output spur levels that are caused from these OOB input frequencies. The process of frequency planning and evaluating the plan via SA can be iterative, where the frequency plan may need updating based on SA results.

Down converters (DC) are either non-inverting (NIDC) using a low-side LO (1x-1), or inverting (IDC) using a high-side LO (-1x1), where ±1x±1 represents the MxN mixing product of M x Fin plus N x FLO at the IF output. An IDC often times yields better spur performance but at the price of a higher frequency LO and greater LO phase noise. Primary spurs to manage for either DC are those with M=-N, which includes the image response. The image response is removed by filtering or using an image reject mixer or both.

Spur management consists of four primary controls:

Signal power at the mixer input: determined by gain distribution and required NF

OOB power at the mixer input: determined by filtering and influenced by frequency plan

LO power: higher levels (within limits of chosen mixer) raises the mixer’s input IP, and hence, lowers spur levels, however, it does not increase P1dB significantly [1].

Mixer type: class I, II and III (+7 dBm, +17 dBm and +27 dBm nominal LO drive levels) for low, medium and high mixer input IP

The measured power in a given spur will vary as

(∆P[MxN])dB = (∆PRF)dB x |M|, (1)

where ∆PRF is the change in input RF power of the signal

producing the spur [2] [3]. At higher input powers (positive

978-1-61284-080-2/11/$26.00 ©2011 IEEE

∆PRF) the spur level increases. The ratio of spur power to

input RF power varies as

(∆R[MxN])dB = (∆PRF)dB x (|M| – 1). (2)

Additionally, [2] states that for doubly balanced diode

mixers, the LO drive level can also be taken into account to

predict changes in spur levels from measured values as

(∆R[MxN])dB = (|M| – 1) [∆PRF - ∆PLO], (3)

where ∆PLO is the change in LO power. Spur levels decrease

for lower input powers (negative ∆PRF) and higher LO

power (positive ∆PLO).

Single vs. dual or triple conversion depends on the BW

required at each mixer port: a tracking filter or switched

filter bank may allow a dual conversion, where triple would

otherwise be necessary. The choice of an IF frequency may

be dictated by component capability (i.e., performance and

cost). A trade-off must be made between continuous tuning

with increased intermod (IM) levels for tracking filters vs.

fixed-tuned filter quantity in a filter bank when dealing with

high power inputs.

Some final notes on frequency conversion:

Harmonics of the LO should be filtered, otherwise a SA for each LO harmonic as the actual LO frequency should be performed with the desired input and output frequency ranges (this type of SA should also be done when using sub-harmonic mixers).

Assess the MxN mixer spurs resulting in a dual or higher conversion from LO#1 leakage through the first conversion stage and mixing with LO#2 in the second mixer (a ±1x±1 CW product could exist, which requires significant IF filtering if not discovered during the frequency planning stage).

For multi-stage conversions, assess spurs that are OOB at the 1

st stage IF but higher than the in-band

requirement. These spurs may become in-band at the final output IF. For example: two relatively poor spurs, a -1x2 produced in the 1

st conversion stage

and a -1x3 in the 2nd

stage, can combine to become in-band at the final IF output.

When de-hopping a spread waveform, better spur performance is usually obtained by de-hopping with the highest frequency LO (smallest percentage BW)

III. CASCADE ANALYSIS.

When beginning a new receiver/Rx chain design, a

rough cascade analysis is usually done first, which is

followed by the other key design areas, the exact order

dictated by overall requirements. Once a top level gain/loss

(G/L) budget has been done, with the resulting gain

distribution satisfying the basic in-band NF, IP,

compression (P1dB), and non-damage requirements, then

approximate levels will be known at the mixers. With these

levels in hand, a frequency plan/spur analysis can be

performed or updated with more accurately predicted spur

levels. As the design progresses, more detail is added to the

cascade analysis, such as actual part values, gain variation

due to device tolerance and gain versus frequency and

temperature (see the final section below for an expanded

discussion on parameter variation, alignment and

compensation). Space limitations prohibit deriving and

detailing the equations used to obtain overall performance

values for a string of cascaded devices which make up a

subsystem. However, most of these equations are readily

available in the literature and on web sites [4].

Some commercially available analysis programs

estimate the overall P1dB of a subsystem by approximating

the Pout vs Pin compression curve using individual device

PSAT and P1dB values. P1dB of a subsystem is not a fixed or

typical number of dBs below its IP3 as is often used in

approximations for an individual device, the cascading

mechanism and equations being different for the two

parameters. Input IP3 degrades two to three times faster

than does P1dB (in dBm) with additional devices of equal

contribution. The relative softness or hardness of the

compression curve depends on ∆ = PSAT – P1dB of the

device. ∆’s greater than 3 dB yield soft curves and are

indicative of low power solid state devices and high power

traveling wave tubes (TWT). High power solid state PAs

(SSPA), linearized TWTs, and mixers have ∆’s on the order

of 1 dB, representing a hard curve. A piece-wise linear

curve is approached as ∆ tends to 0 dB, where device gain is

constant below an input power of PSAT – Gain (small signal)

and output power is constant above. This is not realistic for

any device and care must be exercised when using a

program which models compression using the ∆ method. If

the value is set arbitrarily too low or a default value near 0

dB is used, the resulting prediction for P1dB will be too high

and possibly not discovered until test.

Power levels at which damage occurs throughout the Rx

chain should be determined using the compression

characteristics of each device and not their linear gains. For

specified high level, non-damage inputs an input limiter

may be necessary to protect the front end. However, even

though the front end is protected, its saturated output level

may not protect downstream components, and a lower

power downstream limiter may also be required to protect

components from high saturation levels of preceding stages.

When assessing damage levels throughout the Rx chain,

keep in mind that the PSAT and P1dB values used in the

analysis for the basic G/L distribution are worst-case (lower

bound) values and a higher bound set of values is needed for

the non-damage assessment. When a device is guaranteed to

provide minimum values, it will by definition exceed those

values most of the time. As a result, when using the min

values, maintain at least 2 or 3 dB of damage margin.

Receivers typically work over a large range of input

powers and often receive simultaneous in-band signals,

some of which are at the very bottom of the power range

while at the same time others exist at the upper end. This

scenario stresses the receiver’s sensitivity and power

handling capability at the same time; high power can cause

suppression of small signals, reducing S/N and removing

AM, and it also creates IMs, which may fall on a weak

desired signal. This is the “near-far” problem of receiving

weak distant signals in the face of strong, local in-band

interferers. The interferer may only be in-band at RF or

through the first IF, but the receiver must remain linear

wherever the interferer exists. SFDR3 (3rd

order spur free

dynamic range) is a figure of merit that gives the difference

(in dB) between a threshold level PT and the CW carrier

power that causes a single 3rd

order IM to equal the noise

power in a given BW. This parameter is defined by the

receiver NF, IP3 and noise power BW as

SFDR3 = ⅔(IIP3 – NF – BW + 174) – POffset, (4)

where IIP3 is the input IP3 in dBm, both NF and BW are in

dB, and POffset is the difference in dB between PT and the

noise power in BW [5]. This equation is usually shown with

POffset = 0 dB. However, it could be set to 10 dB (e.g.) to

account for a minimum signal-to-noise (S/N) necessary in

BW, raising the minimum useful signal level and reducing

the dynamic range. Note that the IM power in BW will

reduce the minimum S/N to 7 dB in this example.

There are several dynamic range definitions, and it is not

always clear which one is invoked when a specification

simply states that the receiver dynamic range must exceed a

certain value. Some of these definitions are:

SFDR3 (as defined above)

SFDR2: 2nd

order SFDR, the difference between a threshold level PT and the CW carrier level that causes a sum or difference frequency to equal the noise power in a given BW. Many times PT is set equal to the noise power in BW.

P1dB – PNOISE: distance between the input P1dB point and the noise power in a given BW referred to the receiver’s input

PDESENSITIZATION – PNOISE: difference between the input power which causes a specified amount of degradation (desensitization) and the noise power in a given BW referred to the receiver’s input

PCAR_MAX – PNOISE: difference between the maximum input carrier power for some specified degradation and the noise power in a given BW referred to the receiver’s input

PCAR_MAX – PCAR_MIN: difference between the maximum and minimum input carrier powers for specified degradation(s)

Instantaneous SFDR3 – same as SFDR3 but with all variable gain amps and attenuators fixed to respond to strong and weak signals with the same gains.

IV. NOISE POWER.

Several additional noise sources should be considered

beyond those typically assessed in a generic cascade

analysis. While noise floor and noise power within the noise

bandwidth (NBW) are important parameters to determine,

maintaining their values within acceptable limits does not

necessarily guarantee adequate system performance without

also evaluating and controlling these additional sources

discussed below.

A. Total Noise Power

Total noise power from amplifiers over their individual

NBWs, can be much greater than the single fixed NBW

used for system evaluation. Crystal filters commonly

employed at lower frequency IFs can have 3 dB BWs which

are a tiny fraction of the IF amplifier(s) NBW. For example,

consider a 10 KHz crystal filter used in a 21.4 MHz IF strip

with amplifiers having significant gain out to 300 MHz. The

noise power over 300 MHz is 45 dB greater than that over

10 KHz and could easily compress or saturate the output of

an IF strip with relatively high gain and low P1dB. Even if

the total (average) noise power is below P1dB, the 3 noise

peaks for additive white Gaussian noise (AWGN) are 9.5

dB higher given N0 = 2 [6]:

Pn (3) = [3(N0)½]

2 BW

(5)

= 9 N0 BW, (6)

where N0 is noise density (W/Hz) and BW is in Hz. These

peaks can get clipped, making it no longer AWGN. It is

important to realize that noise saturation can occur even

though the signal is far below P1dB.

B. Image Noise

Image noise, without proper filtering prior to each mixer,

can increase the standard cascaded NF by up to 3 dB for

each conversion stage. To sufficiently lessen the impact of

image noise, an image reject filter must be placed close to

the mixer’s input. It is not sufficient to provide image

rejection to signals only (i.e., input filter followed by gain

prior to the mixer). The wideband response of the amplifiers

will fold over their noise generated at the image frequency

into the IF and can impact the system, even though the

image signals and noise prior to the filter have been

adequately suppressed.

C. Wideband, Unfiltered LO Noise

LO noise leaks through the mixer to the IF (LO-IF

isolation) and adds to the cascaded noise floor. The LO

chain’s output noise density at the input to the mixer LO

port must be assessed. High gains in the LO chain followed

only by a low pass filter (LPF) to remove harmonics

guarantee additional noise will be added to the IF, degrading

the NF predicted by the cascade analysis. A bandpass filter

(BPF) or highpass filter (HPF)

D. Reciprocal mixing

Reciprocal mixing occurs as a result of the transfer of

the LO’s phase noise to each of the receive signals (in dBc)

via the convolution process of the mixer. Degradation to a

weak signal can occur when high and low power in-band

signals are present that are closely spaced in frequency. The

added phase noise from the LO on the high power signal

may “cover up” the nearby lower power signal, significantly

reducing its S/N. For extremely large dynamic ranges

(PCAR_MAX – PCAR_MIN), the AWGN of the LO may also

reduce the S/N of weak signals, independent of frequency

separation. The equivalent C/kT of the weak receive signal

in the presence of a large tone from reciprocal mixing is [7]:

(

) {(

)

*(

) +

}

(7)

.

The (C/kT)LO term is either the LO carrier to noise density,

when assessing LO AWGN impacts, or it is the single side

band phase noise at a given frequency offset in dBc. Each

term above is in dB and must be converted to watts to do the

calculation and then back to dB again (10 log X) to give the

equivalent C/kT result in dB-Hz.

V. OUT-OF-BAND INTERFERENCE.

There are several mechanisms by which strong OOB

interferers at the input can degrade the performance of the

Rx chain and are similar in nature to how large in-band

signals degrade performance. The mitigation approach for

most of these involves filtering and higher intercept and

compression point devices. Since most microwave filters

have a re-entrant passband (PB) at 2Fo or 3Fo, an additional

low pass filter (LPF) should also be used to ensure signals

in one of these unprotected PBs don’t degrade performance.

The filter’s ultimate rejection, which is limited by leakage

around it and isolation of its individual elements, is another

reason to use additional filters, especially with very high

power OOB interferers. The magnitude of an OOB

interferer can be 200 V/m in accordance with MIL-STD-

461/464, incident power levels from some radar systems can

be orders of magnitude greater, and local TV stations can

have EIRPs up to 5 MW in the UHF frequency range.

A. Spurious Responses

Spurs in-band to the IF output can also result from high

power OOB interferers (includes image spurious response).

The power at some OOB frequencies may be many 10’s of

dBs above the strongest desired in-band signals at the mixer,

which can produce very high spur levels relative to the

weaker desired signals. For example, consider a -3x2 spur

where the OOB interferer power is filtered to not exceed the

mixer’s P1dB point of 0 dBm at its input (60 dB above a

desired -60 dBm signal). The mixer data sheet shows the

-3x2 spur level to be -50 dBc for an input of -10 dBm. For

the interferer level of 0 dBm, the spur relative to the

interferer’s power increases by 20 dB to -30 dBc in

accordance with (IAW) (2) (i.e., ∆P x (|M|-1) = [0 dBm –

(-10 dBm)] x (3-1)). The desired signal is 60 dB below the

OOB interferer, so the spur level with respect to the desired

carrier rises 60 dB to +30 dBc. Obviously, keeping the OOB

power just below the mixer’s P1dB point is not sufficient.

Additional input filtering is needed to drop the spur power

to at least 30 dB below the minimum desired signal level.

While this requires a 60 dB spur level reduction, it only

requires 20 dB of additional filter rejection. Recall the

earlier spur level discussion: absolute spur power at IF

(dBm) resulting from an OOB interferer varies IAW (1) as

∆P x |M|. Here we see that for a 3xN spur, an OOB power

change of only 20 dB yields the 60 dB spur reduction.

B. Compression

Compression of the front end and down-stream IF

components can occur until sufficient rejection is provided.

Compression from a large signal produces small signal

suppression of weak signals resulting in receiver

desensitization. The effective gain of a compressed stage is

reduced, degrading its ability to keep weak signals above

the noise floor of succeeding stages. For QAM signals, the

distance between the inner and outer points of the

constellation become compressed, causing degradation.

A weak signal experiences reduced gain from two

factors when passing through a stage that is driven into

compression by a large signal: device gain compression and

small signal suppression. Gain compression is determined

from the typical Pout vs. Pin curve as measured with the large

signal causing the compression. Small signal suppression is

an additional amount of up to 6 dB that only happens to

weak signals [8]. Gain seen by the weak (suppressed) signal

is given below where compression is < 4 dB:

GSUPPRESSED_SIGNAL ≈ GSS (small signal) – GC

(compression) – (small signal suppression). (8)

For a device at or driven beyond saturation (i.e., > 4 dB

compression), the suppressed signal gain is:

GSUPPRESSED_SIGNAL ≈ GSS – 10 dB – P, (9)

where P is the amount the interferer power is above PSAT at

the input, and the 10 dB is comprised of 4 dB (compression

at saturation) and 6 dB (max small signal suppression). The

total gain reduction, GR, of a stage is:

GR ≈ 0 to 4 dB (compression)

+ 0 to 6 dB (suppression) + P, (10)

where P only applies for a device at or beyond saturation.

A rule of thumb to ensure the gain seen by a weak signal is

not degraded more than 1 dB is to keep large signals 2 to 3

dB below input P1dB.

C. 3rd

Order Intermods

IMs (in-band) that result from OOB carriers can

dominate over those created solely from in-band carriers.

The IM level depends on the IP3 of each stage and the

carrier levels throughout the chain, which depends not only

on the in-band G/L of each stage but also filter rejection. As

a result it is a little more involved to calculate the IM ratio

(IMR), the IM level compared to the desired carrier, than for

a standard in-band cascade. The individual IM generating

carrier levels can have different values at the Rx chain input

and typically experience different rejection amounts. Two-

tone and three-tone IM levels generated in any stage are

calculated as described below, with carrier levels defined at

the same location as IP3 (input or output).

Two-tone IMs are those generated from two carriers with

power in Watts defined as follows [9, 10]:

2

3

2

2

1

)2( 21 IP

PPP

FFIM

and 2

3

2

21

)2( 12 IP

PPP

FFIM

(11)

where P1 and P2 are the power in Watts at frequencies F1

and F2, respectively, and IP3 is the 3rd

order intercept point

in watts of the device creating the IMs.

For powers in dBm, the equations become:

321)2( 2221

IPPPP FFIM (12)

321)2( 2212

IPPPP FFIM (13)

The 2-tone IMs are 6 dB lower in power than an IM

resulting from three carriers (3-tone IM), when the same

individual carrier power is used in both cases. The power in

a 3-tone IM is defined below [9, 10]:

2

3

321)(

4321 IP

PPPP FFFIM W (14)

3321)( 26321

IPPPPP FFFIM dBm (15)

Management of these IMs is accomplished by filter

rejection, device IP3 and gain, and judicious gain

distribution throughout the Rx chain, especially through the

first mixer stage. For very large input levels additional

filters may be necessary to improve the ultimate rejection of

the front-end filter(s).

D. Harmonics

2nd

harmonic generation within the IF strip is a spurious

mechanism that is often overlooked. RF input frequencies

that are offset from the tuned RF frequency by ½ the IF

become in-band at the IF in two ways:

Mixer 2x-2 (or -2x2) spur from RF input and

2nd

harmonic of IF/2 Both produce CW tone spurs for BPSK modulation (i.e.,

0/180 becomes 0/360).

The 2nd

harmonic is 6 dB below 2nd

order IMs (i.e.,

F1+F2, F2-F1) and experiences a 2:1 reduction (dBm) with

lower interferer levels [11].

Pspur = Pint – (IP2 – Pint) – 6 dB

= 2Pint – IP2 – 6 dB (16)

A HPF or BPF is used in the IF strip to mitigate for less than

octave instantaneous IF BWs, but high IP2H (2nd

harmonic

IP) is the only mitigation approach for larger BWs.

A similar effect occurs for 3rd

harmonic generation

within the IF strip, where RF input frequencies that are

offset from the tuned RF frequency by 2/3 the IF become in-

band at the IF in two ways:

Mixer 3x-3 (or -3x3) spur from RF input and

3rd

harmonic of IF/3 (i.e., IF – 2/3 IF) In this case the 3

rd harmonic is 9.5 dB below 3

rd order two-

tone IMs and experience a 3:1 reduction (dBm) with lower

interferer levels [10, 11].

Pspur = Pint – 2(IP3 – Pint) – 9.5 dB

= 3Pint – 2IP3 – 9.5 dB (17)

These spur generating frequencies are more easily filtered at

RF and in the IF strip than the ½ IF offsets for 2nd

harmonic

generation. Higher IP3H devices also help to mitigate.

VI. AGC/ALC.

Automatic Gain or Level Control is employed in many

receivers as a means to increase dynamic range and hold

input power to the demodulator constant. There are two

main types of AGC by location:

Post-demod: reacts to the demodulated signal level (coherent AGC) + noise and any interference, primarily within the data filter BW. This type responds to the desired signal and is relatively insensitive to undesired signals and interference (due to the narrow data filter BW). The demod must be locked to the signal for a meaningful output to exist, and the AGC gain is typically at maximum prior to lock (no signal present).

Pre-demod: reacts to Signals + Noise + Interferers + Distortion products (spurs/IMs) in the wider IF BW prior to the AGC’s detector. This type is used to mitigate front end compression, desensitization, or small signal suppression at the expense of NF degradation. The benefits of its use are a trade-off based on the EMI/RFI environment.

Either type can be an analog (continuously variable

attenuation/gain) or digital (step attenuator) implementation.

Any AGC can be “captured” by or AGC on undesired

power. Strong OOB signals can easily exceed the weakest

desired signal power at the AGC detector due to insufficient

ultimate rejection of filters. An additional narrowband filter

may be necessary to prevent capture by undesired signals.

Total noise power can also capture a pre-demod AGC.

The IF BW can be many times larger than the occupied BW

of the desired signal, which affects the accuracy of the

AGC. As the desired signal level approaches the total power

in the noise at the AGC detector, the gain control of the

AGC stops responding to the signal and maintains the S+N

constant, which is dominated by the total noise power. As a

result the signal continues to drop and eventually falls below

the operational range of the demod, even though adequate

S/N in the NBW may be present. This effect limits the

useful range of pre-demod AGCs from the maximum signal

level down to a level above the total noise power, which

typically is adequate to prevent compression.

Another example of AGC noise capture comes from the

earlier noise power discussion. Here a high gain IF strip has

multiple gain stages with a narrowband XTAL filter after

the first stage to limit IM generation. An AGC detector at

the end of this IF strip can easily be captured by noise with

filter-to-amp NBW ratios of 40 dB to 50 dB.

One final note concerns AGC control (or variable gain)

elements. Some voltage variable attenuators have the

undesirable characteristic that their input IP3 and P1dB do

not remain constant as their attenuation is varied. Worse,

some can exhibit both increasing and decreasing values (i.e.,

their behavior is not monotonic). This makes it difficult to

predict overall receiver performance with one gain setting.

Input power levels must be swept (especially for devices

which are not monotonic) to exercise the AGC control

element while checking for compression and IMs. PIN

diode attenuators typically do not share this behavior.

VII. DEVICE PARAMETER VARIATIONS, ALIGNMENT, AND

GAIN COMPENSATION.

Parameters should be evaluated over statistical

tolerances, and frequency and environmental variations via

Monte-Carlo (MC) simulations with gain alignment and

compensation applied for each trial of random gain settings.

Gain tolerance build up should be removed, and gain

compensation applied, often throughout the Rx chain for

best performance.

MC analysis is necessary to even find the worst-case (or

3) performance condition for anything other than a simple

Rx chain which does not use gain alignment, compensation,

or AGC. When an analysis is done using nominal values,

actual performance can be several dB (e.g., 5 dB) worse

than predicted, yielding production problems. On the other

hand, designing with absolute worst-case values

(simultaneously) will produce units that exceed

requirements by as much as 8 dB (depending on the

parameter and gain distribution), which leads to an over-

designed unit and higher production costs. MC analysis can

make the difference between being able to produce a unit at

a reasonable cost versus a no bid for a very difficult set of

requirements.

Once a MC analysis has been completed, the Rx chain

should be evaluated at the condition which yielded the 3

performance for any parameter(s) that are non-compliant.

Often the offending part revealed is not the same as that

shown under nominal or even absolute simultaneous worst-

case condition (i.e., all high or low gains). Then optimize

the gain distribution and/or other parameters, as appropriate

(e.g., increase IP of the actual dominate component), and re-

run the analysis.

VIII. CONCLUSION.

Six key Rx chain design areas (spurious, cascaded

elements, non-standard noise, OOB interference, AGC, and

MC analysis) have been presented with several subtleties of

each discussed. The approach has been somewhat a design

check-list of topics to address for receiver design. The

importance of MC analysis to achieve a realizable design

that is cost effective cannot be understated.

Author. James B. Offner is an RF

Systems Engineer with Harris

Corporation in Melbourne, FL with

over 30 years of experience, working

on large multidiscipline programs

from mission requirements

determination and system analysis

through HW/SW implementation. He has contributed to the

development of mobile tactical and large fixed strategic

satcom terminals, and recently was the RF System Architect

/Analyst for the Army’s strategic MET terminal

development. He has performed analysis for development of

Navy shipboard terminals, operating with heavy EMI/cosite

interference and has developed analysis programs in use at

Harris for RF subsystem design. He was the Chief Systems

Engineer for the development of a vehicular, on-the-move

communication system, via aircraft relay, using multiple

phased array antennas. Jim earned his B.S.E.E. from

Michigan State University in 1977.

REFERENCES

[1] Daniel Cheadle, “Selecting Mixers for Best Interemod Performnace, 1993 Watkins-Johnson Co. Catalog Article

[2] William F. Egan, “Practical RF System Design”, Wiley-Interscience, 2003, pp. 171-180.

[3] Stephen A. Mass, “Microwave Mixers”, Artec House, 1996, pp. 151-154.

[4] William F. Egan, “Practical RF System Design”, Wiley-Interscience, 2003, pp. 49-53 and 91-122.

[5] William F. Egan, “Practical RF System Design”, Wiley-Interscience, 2003, pp. 137-139.

[6] John G. Proakis, “Digital Communications”, McGraw-Hill, 1983, pp. 93.

[7] Jim Offner, Internal Harris document, 2011.

[8] Robert M. Gagliardi, “Satellite Communications”, Van Nostrand Reinhold, 1984, pp. 201-203.

[9] Stephen A. Mass, “Microwave Mixers”, Artec House, 1996, pp. 154-158.

[10] Hinrich Heynisch, “Useful Design Criteria Predict TWT Intermod, MICROWAVES, March 1980

[11] Keneth A. Simons, “The Decibel Relationships Between Amplifier Distortion Products”, Proceedings of the IEEE, VOL. 58, NO. 7, July 1970