Circuit.Cellar.192.Jul.2006_Martini.pdf

FEATURE ARTICLE

by Neal Martini

Compact Spectrum Analyzer

If RF testing is in your future, you’ll need a spectrum analyzer.But they don’t come cheap.

Your best bet is to follow Neal’s lead by building your own 4

× 4

system.

I t is funny how simple beginnings can

lead you down convoluted paths of learn-

ing and discovery. About a year and a half

ago, I was playing around with my auto-

matic garage door remote control unit. I

was experimenting with signal encoding

at the time, and I wanted to look at the

scheme used in my remote. One thing

led to another, and before I knew it, I was

trying to determine the RF frequency at

which the transmitter was operating.

I then built RF filters to look at sig-

nal levels and RF generators to aid in

prototyping and testing. Next came RF

mixing, frequency doubling, and so on.

You can see where I’m going here. I

became hooked on RF and all of the

art that goes along with getting RF cir-

cuits working properly. Currently, all

of my design-related pursuits are

focused on circuits and test equipment

at radio frequencies.

When it comes to RF test equipment,

the spectrum analyzer is the Holy Grail.

Unfortunately, both new and used RF

spectrum analyzers are extremely expen-

sive. There are a variety of home-brew

spectrum analyzers available on the ’Net.

Some are more sophisticated than others

because all of the circuit elements are

built from scratch. They feature complex

filter constructions, have precise PCB lay-

outs that turn PCB traces into inductors,

and include a lot of discrete components.

Other somewhat simpler designs fea-

ture integrated modules borrowed

from the TV/VCR tuner world.

These are easier to build, but they

lack sophistication and functionality.

As I was browsing the ’Net for an

alternative to a home-brew spec-

trum analyzer, I stumbled across the

Maxim MAX3550, which seemed

almost too good to be true. Not only

Photo 1a— My spectrum analyzer’s PCB is shown here with an optional keyboard and LCD, which I unplugged for

clarity.The 4 ″ × 4 ″ PCB includes the complete spectrum analyzer with a 5-V power supply. b— The spectrum ana-

lyzer can be controlled as a stand-alone device from the keypad or from a virtual keypad using a mouse.The output

spectrum can be a cursored trace on an oscilloscope or a more elaborate PC screen display with a few extra whis-

tles and bells.Alternatively, the power level at a single frequency can be displayed on the LCD.

does it perform much of the processing

required to construct a spectrum ana-

lyzer, it’s also highly integrated, simple

to lay out, and easy to control. When I

found this part, I knew I had to try to

design yet another spectrum analyzer.

The result of my effort is the single

board directly to an oscilloscope to dis-

play the output spectrum. All of these

combinations are possibilities. The circuit

is relatively easy to build, and construc-

tion is pretty forgiving when it comes to

the layout and part selection processes.

Read on if you want a low-cost spectrum

analyzer for your future RF endeavors.

PCB shown in Photo 1a.

You can use my spectrum analyzer

board as a stand-alone system with its

own keypad and display, or you can con-

nect it to a PC. If you do the latter, the

output will be displayed on the PC’s

screen and all of the keypad’s functional-

ity will be available with a mouse. You

can also connect the spectrum analyzer

× 4

ANALYZER FUNDAMENTALS

A spectrum analyzer is used to display

the power distribution of a signal as a

function of frequency. This type of display

is said to be in the frequency domain as

opposed to being in the time domain as

displayed on the screen of an oscilloscope.

To illustrate the value of a spectrum

analyzer, let’s look at some typical

signals being analyzed with spec-

tral analysis.

If the input to a spectrum ana-

lyzer is a pure sine wave, the out-

put spectrum might look like

what you see in Figure 1a. The

arrowed line at F 1 indicates that

the frequency of the sine wave is

P 1

P 2

F 1

Frequency

F 1

F 2

F 1

2F 1 3F 1 4F 1

Frequency

Figure 1— Compare the spectrum of a pure sine wave (a) to the

spectrum of two equal power pure sine waves (b) and the spec-

trum of a slightly distorted sine wave (c) .

www.circuitcellar.com

Issue 192 July 2006

CIRCUIT CELLAR ®

F 1 and the power level is P 1 . If

you have an oscilloscope that

operates at RF frequencies, you

can get the same information

from its screen. But oscillo-

scopes that operate at up to

1 GHz are extremely expensive.

Figure 1b illustrates the spec-

trum analyzer output that occurs

if your input signal is the sum of

two sine waves of different fre-

quencies and equal amplitude.

There are two components to

the signal, and the frequencies

and power levels of each compo-

nent are displayed. Even if you

were to use a high-frequency

oscilloscope, this information

would be difficult to discern from the

oscilloscope’s display.

Yet another example of spectrum

analyzer output is shown in Figure 1c.

This is the output that results when a

distorted sine wave is the input signal.

The components at the various fre-

quencies show the amount of harmon-

ic distortion contained in the signal.

A spectrum analyzer facilitates other

operations too. You can use one to deter-

mine filter responses, measure field

strength, tune antennas, locate noise

sources, and debug RF circuits. As

soon as you have a working spectrum

analyzer, you’ll wonder how you ever

accomplished anything without one.

mixing process produces both the

sum and difference frequencies,

which would result in unwanted

signals being shifted into the BPF’s

passband without the tracking

filter. Because this tracking filter

doesn’t have to be too narrow to

reject the image frequency compo-

nent, it lends itself to an easier

implementation than the tracking

filter in the swept filter approach.

Modern high-frequency receivers

use a double-conversion architec-

ture (see Figure 2c). The beauty of

this approach is that there are no

tracking filters to deal with. The

added complexity is a second mix-

ing stage. The front end fixed LPF

removes components outside the highest

frequency range of interest. The first

mixing stage moves the signal frequency

component of interest up in frequency

into the fixed passband of the first BPF.

The shift up in frequency causes the

image frequency to be above the fixed

passband of the input low-pass filter

(LPF), so there is no image frequency

content in the passband of the first BPF.

The second mixing stage brings the

frequency of interest back down. This

allows extremely narrowband final filters

to be used for maximum selectivity.

I used the double conversion architec-

ture for my spectrum analyzer design.

Why? Because it is relatively easy to

implement fixed filters and variable-

frequency VCOs.

Tr a c k i n g

filter

Input

Power

Tracking filter

Mixer

Input

BPF

Frequency

VCO

Mixer

Input

LPF

BPF

VCO

Figure 2a— The center frequency of a narrowband filter is swept across

the frequencies of interest. b— The VCO shifts the frequency of interest

into the passband of the BPF. c— The two mixers shift the frequency

into the passband of the BPF without needing a tracking filter.

tectures: swept filter, heterodyne with

tracking filter, and double conversion.

The swept filter analyzer varies the

passband of a band-pass filter (BPF)

over the frequencies to be covered (see

Figure 2a). It produces an output voltage

that’s proportional to the amplitude

levels of the various frequency compo-

nents. Although it appears simple, the

direct implementation of narrowband fil-

ters variable across 1 GHz is difficult.

The heterodyne with a tracking filter

approach is much easier to implement

(see Figure 2b). It has been the basic

approach used in radio receivers for years.

Basically, a voltage-controlled oscillator

(VCO) shifts the frequency of interest into

the passband of the fixed-frequency nar-

rowband BPF located at the output. The

purpose of the tracking filter is to elimi-

nate any signals located at the image fre-

quency before the signal enters the mix-

ing process. This is necessary because the

ARCHITECTURE CHOICES

Let’s look at some spectrum analyzer

architectures so you can better under-

stand my design. Figure 2 illustrates

three different spectrum analyzer archi-

SIGNAL PROCESSING

My spectrum analyzer is fairly sim-

ple (see Figure 3). Diagrams showing

Programming

connector

ICD

Input

1,274–2,111 MHz

BPF1

BPF2

BPF3

LCD

Mixer 3

Mixer 1

Mixer 2

F C = 1,226 MHz

BW = 65 MHz

F C = 45.75 MHz

BW = 5 MHz

F C = 110.7 MHz

BW = 330 kHz

IF 1

IF 2

0–1,000

MHz

PWR

Keypad

BPF4

1,180.25 MHz

33–37 MHz

VCO 1

MCU

VCO 2

VCO 3

F C = 10.7 MHz

BW = 7.5 kHz

PLL

PWR

Control

PLL

Control

RF Gain control

Four-channel

DAC

RS-232

Connector

to PC

MAX232A

To oscilloscope

vertical

To oscilloscope

external trigger

Figure 3— Three frequency conversion stages translate the frequency of interest into the passband of the final narrowband filters and log amplifiers.The PIC18F4520 micro-

controller constructs the spectrum and drives the various output devices.The PIC18F4520 microcontroller also controls all of the conversions and gains.

Issue 192 July 2006 61

www.circuitcellar.com

CIRCUIT CELLAR ®

what the frequency spectrum looks

like at various stages of signal process-

ing (assuming the analyzer is locked

on an 800-MHz input sine wave) are

posted on the Circuit Cellar FTP site.

The front end of the process is basically

the double conversion architecture.

The first mixer/VCO moves the fre-

quency of interest up to 1,226 MHz,

the passband of the BPF1. For the

800-MHz hypothetical input signal,

VCO1 is set to 2,026 MHz.

Notice that there is no LPF at the

front end of the spectrum analyzer

like that in the double conversion

approach. The LPF needs to be only

2 GHz to reject the image frequency

components. I didn’t include an LPF

because my applications typically

don’t have content higher than 2 GHz.

If your application is different, you

can add an external BNC-connected

LPF. Also note that the VCOs are

phase-locked-loop (PLL) controlled for

high stability and ease of control. More

on PLLs later.

The second mixing stage shifts

the frequency of interest down to

45.75 MHz, the center of the BPF2’s

passband. The BPF2’s output is a 6-MHz

wide region of the spectrum centered

at 45.75 MHz. As you’ll learn when I

describe the hardware, many of the

center frequency and filter bandwidth

choices were dictated by the parts I

selected.

I chose two resolutions for the spec-

trum analyzer: one for broad frequency

looks and one for close frequency

examination. The coarse spectrum res-

olution bandwidth is 330 kHz. The

finer resolution bandwidth is 7.5 kHz.

These resolutions are readily available

in band-pass filters with center fre-

quencies of 10.7 MHz. The problem is

that the signal of interest is 45.75 MHz,

so another conversion stage is required

to bring the frequency of interest

down to 10.7 MHz. Mixer 3 provides

the required down conversion.

The 10.7 MHz centered content is

now delivered to BPF3 and BPF4, each

with their respective bandwidths. The

final task is to convert the output

voltages of the two filters to power

levels. Log amplifiers provide this

functionality by producing an output

voltage proportional to the log of the

input voltage. The log amplifier out-

puts are delivered to the microcon-

troller for display. I used a Microchip

Technology PIC18F4520 microcon-

troller for this project.

The PIC18F4520 microcontroller

controls the frequency scanning of the

RF portion of the spectrum analyzer. A

coarse frequency stepping of 330 kHz

is accomplished by varying the VCO1’s

frequency. The finer-resolution 7.5-kHz

stepping operation is accomplished by

varying VCO3. There are high- and

low-sensitivity operating modes.

Figure 4— The MAX3550 integrates many of the spectrum analyzer’s front-end functions.The Philips Semiconductors SA612A mixer shifts the signal of interest to 10.7 MHz for

filtering.The Analog Devices AD8307 amplifier performs the conversion to decibels relative to 1 milliwatt (dBm).

www.circuitcellar.com

Issue 192 July 2006

CIRCUIT CELLAR ®

These two modes of operation are

controlled by varying the RF gain in

the first mixing stage (also under the

control of the PIC18F4520 microcon-

troller).

My spectrum analyzer can display

its output spectrum on an LCD, an

attached PC, or directly on an oscillo-

scope. The PIC18F4520 microcon-

troller uses a DAC to generate the sig-

nals to drive the oscilloscope. An RS-

232 interface is included to communi-

cate with a PC. The PIC18F4520

microcontroller also responds to com-

mands from the optional keypad and

drives the LCD to display the power

level of a user-selected frequency. The

system includes a connector so you

can use a firmware development sys-

tem to in-circuit program the micro-

controller.

is the IC that got me interested in tak-

ing a crack at a home-brew spectrum

analyzer.

There are several ICs on the market

that are said to have single-chip TV

functionality, but I haven’t found any

with the level of integration and ease

of use of the MAX3550. The chip con-

tains all of the components necessary

to perform double conversion tuning.

It has the necessary variable gain RF

amplifiers, double-balanced mixers, a

band-pass filter, and PLL frequency

synthesizers.

Other one-chip solutions for TV

tuning require external VCO tuning

inductors and external varactors. The

MAX3550 has eight digitally selec-

table tank circuits and the necessary

varactors on the chip. Maxim has

somehow found a way to integrate the

required inductors and capacitors on

the chip, which isn’t an easy task over

the wide frequency operating range.

The operating frequency range extends

from 50 to 878 MHz, the standard TV

frequency space, while providing 60-dB

RF gain control. The response across

this frequency range is flat, typically

0.3 dB. The high-quality local oscilla-

tors have superior phase noise per-

formance of –86 dBc/Hz at 10 kHz.

The integrated filter achieves 68 dBc

of image rejection. Device program-

ming and configuration are easily

accomplished with a standard three-

wire interface to the microcontroller.

I am actually using the chip from

10 to 956 MHz, sacrificing some accu-

racy at the very high and very low fre-

quencies. All that’s needed to support

this part are a handful of SMD resis-

tors and capacitors to form the PLL

loop filters and to do normal bypass-

ing. A 4-MHz crystal is also needed.

The MAX3550 requires a single 5-V

supply.

Another thing that attracted me to

the MAX3550 is that Maxim offers an

evaluation kit for testing it. The beauty

of this was that I could copy the cir-

cuit design in the kit, which ensured

that the part would perform as speci-

fied. The evaluation kit also came

with a piece of software to exercise

the part using a PC. This enabled me

HARDWARE FRONT END

Figure 4 (p. 62) shows the RF por-

tion of the spectrum analyzer. The

front end centers on the MAX3550

broadband TV tuner IC, which is a

48-pin QFN, 7 mm × 7 mm IC. This

www.circuitcellar.com

Issue 192 July 2006

CIRCUIT CELLAR ®

to copy the values that were used for

setting up the various PLL divider reg-

isters and to ensure specified perform-

ance once again.

The BPF in the chip actually has a

movable center frequency that’s used

to avoid a problem that can occur in

the double-conversion process. It turns

out that in the process, unwanted beat

frequencies are generated from har-

monics of the two local oscillators.

The MAX3550 allows the center fre-

quency of its internal BPF to be shift-

ed slightly when these beat frequen-

cies come into play. This moves the

beat frequencies outside the output

BPF’s passband. I used the software

that came with the evaluation kit to

decide how to control the inter-

nal BPF’s center frequency.

The MAX3550’s output is fed

into a 6-MHz wide EPCOS

band-pass SAW filter centered

at 45 MHz. This filter, which is

contained in a five-pin SIP

package, has an out-of-band

rejection of higher than 50 dB,

which is excellent. I used it

because it’s the same filter used in the

evaluation kit. I wanted to ensure

there weren’t any impedance-matching

issues. The filter has a steep response

and is designed for use with standard

TV channel spacing.

components are required. The SA612A,

which can receive inputs at –119 dBm,

has a reasonable third-order intercept

that’s typically –13 dBm. It also contains

a high-frequency common collector

transistor that can be configured as a

local oscillator for the mixing process.

Before I describe the actual circuit,

let’s look at what makes up a PLL.

Figure 5 shows a PLL-controlled VCO.

As you can see, the voltage V TUNE con-

trols an oscillator’s frequency of opera-

tion. The VCO output is divided by

factor N . A fixed frequency reference

is divided down by factor R . These

two divided-down signals are fed into

the phase comparison box whose out-

put is an error signal proportional to

the difference in frequencies of

the two divided-down signals.

The error signal is passed

through a loop filter. V TUNE is

changed until the divided-down

VCO frequency matches the

divided-down reference frequen-

cy. At this point, the loop is

said to be in lock, and the VCO

output remains stable locked to

CONVERSION DOWN

The conversion of the 45.75-MHz cen-

tered signal down to 10.7 MHz is accom-

plished with a Philips Semiconductor

SA612A double-balanced mixer/oscil-

lator and a National Semiconductor

LMX2306 PLL frequency synthesizer.

The SA612A contains a differential

input mixer with an input impedance

that closely matches that of the EPCOS

filter output, so no impedance-matching

Refence

input

Error

V TUNE

÷ R

Loop

filter

Phase

compare

VCO

Output

÷ N

Figure 5— An error signal is generated in the comparison of divided-down

versions of the VCO output and a reference input.The loop filter conditions

the error signal.It corrects the VCO tuning voltage until the output gets to

the desired frequency.

www.circuitcellar.com

Issue 192 July 2006

CIRCUIT CELLAR ®

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: