Home › Modern Spectrum Analyzer Signal Processing — BB60D, SP145 & R&S FSW

On this page

Modern Spectrum Analyzer Signal Processing — BB60D, SP145 & R&S FSW

30-second intuitionA modern spectrum analyzer is not simply an RF voltmeter sweeping a narrow analog filter. It is a frequency-selective RF receiver followed by high-speed data conversion and DSP. The architecture can vary substantially: the BB60D uses a superheterodyne chain with hardware preselection, the SP145 uses a hybrid direct-conversion/low-IF architecture with much faster FPGA processing, and a high-end analyzer such as the R&S FSW combines precision microwave front ends with very wide internal analysis bandwidth and dedicated vector-signal/standards applications.

RF input→preselection / gain / mixer→IF or low IF→ADC→FPGA / DSP→FFT / demodulation / real-time analysis→spectrum / IQ / EVM / persistence

The three instruments below are useful because they represent three levels of modern spectrum-analysis architecture:

Analyzer	Useful mental model
Signal Hound BB60D	compact superheterodyne + hardware-preselector analyzer with PC-hosted FFT processing
Signal Hound SP145	compact high-speed hybrid direct-conversion / low-IF real-time analyzer
R&S FSW	high-end benchtop signal-and-spectrum analyzer with very wide internal analysis bandwidth and specialized vector/standards applications

Official references:

At-a-glance comparison

Parameter	BB60D	SP145	R&S FSW
Frequency range	9 kHz–6 GHz	100 kHz–14.5 GHz	2 Hz–85 GHz; up to 90 GHz with FSW85 option
Real-time bandwidth	27 MHz	40 MHz	up to 800 MHz real-time
Max. analysis / I/Q bandwidth	27 MHz streaming	40 MHz streaming	up to 8.312 GHz internal analysis bandwidth
Sweep speed	24 GHz/s	200 GHz/s typical	architecture optimized for high-performance swept + wideband FFT analysis
Architecture	superheterodyne + preselector	hybrid direct conversion / low IF	multi-conversion precision signal/spectrum analyzer
Hardware preselection	Yes	No tunable preselector	Yes / model- and mode-dependent microwave preselection
Phase noise @ 1 GHz, 10 kHz offset	~−93 dBc/Hz typical comparison value	~−120 dBc/Hz typical	<−136 dBc/Hz specified on current models
Built-in GPS	No	Yes	external/reference options rather than field-GPS focus
Host architecture	USB 3.0 + PC	USB-C + PC	self-contained benchtop instrument
Best fit	weak signals, spurious/EMC, sub-6 GHz	fast surveys, telecom/X-band, VSA	demanding 5G/mmWave/radar/EVM/phase-noise analysis

The numbers are not perfectly equivalent across vendors because test conditions, options and definitions differ. The table is intended as an architectural comparison, not a replacement for the individual data sheets.

For the corresponding receiver architectures used in radios—superheterodyne, zero-IF, low-IF, image-reject and direct-RF sampling—see RF Transceiver Architectures — Interview Ready.

Part I — What is inside a modern spectrum analyzer?

1. Conventional swept analyzer

A classic superheterodyne spectrum analyzer can be thought of as

\[\boxed{ \text{RF input} \rightarrow \text{tunable LO + mixer} \rightarrow \text{fixed IF} \rightarrow \text{analog RBW filter} \rightarrow \text{detector} \rightarrow \text{display}. }\]

The LO sweeps across frequency.

At each LO frequency, a narrow portion of the RF spectrum passes through the IF resolution-bandwidth filter.

This architecture is conceptually similar to a radio receiver whose tuning frequency is continuously swept.

2. FFT-based analyzer

A digital analyzer instead digitizes a finite instantaneous bandwidth and computes

\[X[k] = \sum_{n=0}^{N-1} x[n]w[n]e^{-j2\pi kn/N}.\]

Here

$x[n]$ = sampled time waveform;
$w[n]$ = window function;
$N$ = FFT size;
$X[k]$ = complex spectral bins.

One FFT therefore measures many frequencies simultaneously.

3. Hybrid architecture

Modern analyzers often combine both concepts.

The RF front end tunes a finite block of spectrum into the ADC bandwidth, then FFT processing analyzes that block digitally.

To cover a wider span:

\[\boxed{ \text{tune RF block} \rightarrow \text{FFT} \rightarrow \text{retune} \rightarrow \text{FFT} \rightarrow \cdots }\]

The resulting blocks are stitched into one displayed sweep.

The BB60D is explicitly described by Signal Hound as a hybrid superheterodyne-FFT spectrum analyzer.

Part II — BB60D signal path

4. Main BB60D specifications relevant to DSP

Parameter	BB60D
RF frequency range	9 kHz–6 GHz
Real-time / usable instantaneous bandwidth	27 MHz
ADC rate	80 MS/s
ADC resolution	14 bit
Raw data transport	USB 3.0
Data stream	~140 MB/s
I/Q output rate at decimation 1	40 million I/Q pairs/s
Maximum sweep speed	24 GHz/s
Real-time FFT processing capability	up to ~1.2 million FFT spectra/s through the API

Signal Hound states that the BB60D continuously streams digitized RF/IF data to the host PC rather than performing all spectral processing inside the instrument.

5. Front-end gain and attenuation

Before digitization, the analyzer must choose an RF gain state.

Conceptually,

\[\text{RF} \rightarrow \text{attenuator} \rightarrow \text{gain stages} \rightarrow \text{mixer}.\]

The BB60D API exposes several gain and attenuation states.

The reference level is important because automatic gain/attenuation selection uses it to place the signal appropriately within the analyzer’s dynamic range.

Tradeoff

Too much gain:

ADC/front-end compression;
intermodulation;
false spurs.

Too little gain:

receiver noise dominates;
DANL worsens.

Thus

\[\boxed{ \text{reference level} \rightarrow \text{front-end gain state} \rightarrow \text{noise floor + linearity}. }\]

6. Preselection

Strong out-of-band signals can enter the mixer and create

intermodulation;
images;
second-order products.

The BB60D uses different front-end strategies with frequency.

Signal Hound documents:

below 30 MHz — a high-linearity low-noise low-frequency path;
30–130 MHz — push-pull amplification for improved second-order performance;
above 130 MHz — sub-octave hardware preselection.

The preselector is therefore physically ahead of the later conversion chain.

Its job is approximately

\[\boxed{ \text{reject unwanted RF before nonlinear stages}. }\]

This is one major reason preselection improves usable dynamic range in dense RF environments.

7. Superheterodyne conversion

The BB60D is a two-stage superheterodyne receiver.

Signal Hound documents two main intermediate frequencies:

\[f_{IF}=2380\ \mathrm{MHz}\]

and

\[f_{IF}=1220\ \mathrm{MHz},\]

selected according to RF input range.

The mixer relationship is

\[f_{IF}=|f_{RF}-f_{LO}|.\]

The LO therefore tunes so that the desired RF block lands at the appropriate IF.

8. BB60D IF mapping

The 2025 BB60D manual gives the following conversion regions:

RF input	IF
30–1810 MHz	2380 MHz
1810–4210 MHz	1220 MHz
4210–5590 MHz	2380 MHz
5590–6000 MHz	1220 MHz

This type of frequency planning is used to manage

image locations;
mixer products;
LO range;
filtering;
front-end linearity.

Part II-B — Signal Hound SP145 architecture

SP145 signal path

The SP145 represents a newer Signal Hound architecture than the BB60D.

Signal Hound describes it as a hybrid direct-conversion / low-IF spectrum analyzer:

$ \boxed{ \text{RF input} \rightarrow \text{gain / attenuation} \rightarrow \text{quadrature mixers} \rightarrow \text{low IF} \rightarrow \text{ADC} \rightarrow \text{FPGA filters / decimators} \rightarrow \text{corrected I/Q} \rightarrow \text{USB-C / PC}. } $

Unlike the BB60D, the SP145 does not use a tunable hardware preselector ahead of the receiver.

This is a deliberate trade:

$ \boxed{ \text{less front-end selectivity} \longleftrightarrow \text{much faster tuning + wider frequency coverage}. } $

SP145 sampling and FPGA processing

Signal Hound documents approximately

$ 245.76\ \mathrm{MS/s} $

of analog sampling inside the SP145.

The FPGA then performs digital processing and reduces this to approximately

$ 61.44\ \mathrm{million\ corrected\ I/Q\ samples/s}. $

The usable calibrated streaming bandwidth is

$ \boxed{ 40\ \mathrm{MHz}. } $

This is wider than the BB60D’s 27 MHz instantaneous/streaming bandwidth.

The important architectural change is that more filtering, correction and decimation are performed inside the SP145 FPGA before data reaches the host PC.

SP145 real-time operation

The SP145 can continuously stream without time gaps over its real-time bandwidth.

That matters for

high probability of intercept;
burst signals;
frequency hopping;
spectrum monitoring;
RF drive testing;
airborne measurements.

Its published sweep speed reaches approximately

$ 200\ \mathrm{GHz/s} $

under the stated wide-RBW condition.

Compared with the BB60D:

$ \frac{200}{24}\approx8.3. $

So the SP145 can scan a wide frequency range dramatically faster.

SP145 phase noise

A major improvement is analyzer LO phase noise.

Signal Hound’s current data sheet gives a typical value near

$ L(10\ \mathrm{kHz}) \approx -120\ \mathrm{dBc/Hz} $

for a 1 GHz center frequency.

The older BB60D comparison value is about

$ -93\ \mathrm{dBc/Hz}. $

This difference is very important when measuring

close-in spurs;
PLLs/VCOs;
high-order QAM;
EVM;
weak signals near strong carriers.

A low DANL does not compensate for poor phase noise close to a carrier.

SP145 DANL and linearity

Typical SP145 DANL values are frequency dependent and reach roughly the high

$ -160\ \mathrm{dBm/Hz} $

range through much of the microwave region.

Signal Hound also quotes strong typical third-order intercept performance, with IIP3 values around +30 dBm in lower-frequency ranges and somewhat lower at higher microwave frequencies.

This gives the SP145 good intrinsic linearity.

However, because it lacks BB60D-style hardware preselection, a very strong out-of-band transmitter can still enter the receiver and contribute to compression or intermodulation.

An external bandpass filter can therefore be valuable in severe RF environments.

SP145 residual responses

A useful subtle distinction is:

BB60D: stronger hardware preselection and very low residual responses;
SP145: much better phase noise, bandwidth, speed and frequency range.

The SP145 manual specifies residual responses around

$ <-103\ \mathrm{dBm} $

under its stated condition.

Thus the SP145 is not simply “better at everything.”

For extremely weak CW/spur searching below 6 GHz in a crowded environment, the BB60D preselector can still be valuable.

Internal GPS

The SP145 includes internal GPS.

When locked, Signal Hound specifies approximately

$ \pm1\ \mathrm{ppb} $

typical frequency-reference accuracy and GPS-based timestamping.

This is useful for

mobile RF surveys;
drive testing;
geolocated interference hunting;
distributed monitoring;
field measurements.

This is one of the clearest ways the SP145 differs from a conventional desktop-only analyzer.

BB60D versus SP145 — parallel architecture

Stage	BB60D	SP145
Input selection	sub-octave hardware preselection above ~130 MHz	wideband input, no tunable preselector
RF conversion	two-stage superheterodyne	hybrid direct conversion / low IF
Sampling	80 MS/s, 14-bit ADC	~245.76 MS/s internal sampling
FPGA role	packetization/high-rate transport	substantial filtering, correction and decimation
Host I/Q	up to 40 M complex pairs/s; 27 MHz useful BW	61.44 M corrected I/Q/s; 40 MHz useful BW
Real-time BW	27 MHz	40 MHz
Frequency ceiling	6 GHz	14.5 GHz
Sweep speed	24 GHz/s	200 GHz/s
Phase noise	modest	much better
Field timing	external reference	integrated GPS
Strong OOB environment	advantage BB60D	external filtering may help
Fast/wide RF survey	adequate	advantage SP145
X-band analysis	no	yes

Part II-C — High-end example: Rohde & Schwarz FSW

Why include the FSW?

The BB60D and SP145 are compact analyzers that rely heavily on an external PC.

The R&S FSW shows what happens when the same RF/DSP concepts are pushed into a high-performance benchtop instrument designed for

5G NR;
Wi-Fi / wideband OFDM;
radar;
satellite;
phase-noise characterization;
amplifier measurements;
mmWave;
very-wideband vector-signal analysis.

The signal-processing ideas remain recognizable, but almost every subsystem becomes more sophisticated.

FSW frequency and bandwidth scale

Current FSW models span from

$ 2\ \mathrm{Hz} $

up to

$ 85\ \mathrm{GHz}, $

with the FSW85 extendable to approximately

$ 90\ \mathrm{GHz} $

through the corresponding frequency-extension option.

Depending on model and bandwidth options, the internal analysis bandwidth can reach

$ \boxed{ 8.312\ \mathrm{GHz}. } $

Real-time analysis bandwidth can reach approximately

$ 800\ \mathrm{MHz}. $

This is a fundamentally different class from the 27–40 MHz real-time bandwidth of the compact Signal Hound units.

FSW architecture — conceptual view

A useful simplified architecture is

$ \boxed{ \begin{aligned} RF &\rightarrow \text{precision attenuator / preselector / preamp}
&\rightarrow \text{multi-stage frequency conversion}
&\rightarrow \text{wideband IF}
&\rightarrow \text{high-dynamic-range ADC system}
&\rightarrow \text{FPGA / DSP fabric}
&\rightarrow \begin{cases} \text{swept spectrum}
\text{FFT spectrum}
\text{real-time spectrum}
\text{I/Q capture}
\text{standards demodulation}
\text{phase-noise analysis} \end{cases} \end{aligned} } $

The exact internal implementation varies by FSW model and bandwidth option, but the conceptual difference is clear: the FSW is designed to preserve high dynamic range over far wider instantaneous bandwidths while supporting many specialized measurement applications internally.

Phase noise and close-in dynamic range

Current FSW models specify phase noise better than approximately

$ -136\ \mathrm{dBc/Hz} $

at a 1 GHz carrier and 10 kHz offset under the vendor’s specified condition.

That puts the analyzer in a different class for

oscillator characterization;
blocking measurements;
radar;
EVM on high-order modulation;
weak-sideband measurements.

This also illustrates why a more expensive analyzer can outperform a cheaper analyzer even when both show a similar DANL far from the carrier.

Wideband vector signal analysis

The FSW can capture very wide complex I/Q bandwidths and then perform vector analysis.

For 5G NR, dedicated measurement applications can analyze quantities such as

EVM;
channel power;
frequency error;
per-channel/per-signal power;
modulation quality;
allocation-specific measurements.

For very wide signals, the analyzer’s internal analysis bandwidth determines whether the entire waveform can be captured in one coherent record.

This matters for

wide NR carriers;
carrier aggregation;
Wi-Fi 7 / very-wide OFDM;
radar chirps;
satellite payload signals.

Dedicated measurement applications

A high-end signal analyzer is more than a general-purpose FFT display.

The FSW platform supports dedicated applications for areas such as

3GPP LTE;
5G NR uplink/downlink;
WLAN including 802.11ad/ay;
OFDM vector-signal analysis;
phase-noise measurement;
amplifier characterization;
multi-standard radio analysis;
real-time spectrum analysis.

These applications understand the waveform structure and compute measurements that cannot be obtained from a simple power-versus-frequency trace alone.

Multi-standard radio analysis

A modern RF product may transmit several signals or standards simultaneously.

The FSW multi-standard architecture can analyze multiple radio standards in one measurement context.

This is useful when studying

coexistence;
intermodulation;
carrier aggregation;
adjacent-channel interaction;
transmitter coexistence.

A compact USB spectrum analyzer can often capture the raw I/Q, but the high-end analyzer integrates much more of the demodulation and standards analysis directly in the instrument software.

Real-time spectrum versus wide analysis bandwidth

These two numbers should not be confused.

For the FSW:

$ B_{analysis} \le 8.312\ \mathrm{GHz} $

in supported wideband configurations,

while real-time spectrum-analysis bandwidth is smaller:

$ B_{RT} \le 800\ \mathrm{MHz}. $

Analysis bandwidth asks:

How wide a complex waveform can I capture/analyze coherently?

Real-time bandwidth asks:

Over what bandwidth can I continuously process spectrum without acquisition gaps at the specified real-time performance?

This same distinction also applies to lower-cost analyzers.

Preselection tradeoff at extreme bandwidth

At very wide analysis bandwidths, a traditional narrow YIG/preselector can itself become the bandwidth bottleneck.

High-end analyzers therefore change or bypass parts of the preselection path in certain ultra-wideband modes.

The engineering trade remains the same:

$ \boxed{ \text{more instantaneous bandwidth} \longleftrightarrow \text{less aggressive analog preselection}. } $

This is the same trade visible between BB60D and SP145, only implemented at a much higher performance level.

BB60D vs SP145 vs FSW — what changes architecturally?

Question	BB60D	SP145	R&S FSW
Where is most DSP?	host PC	FPGA + host PC	internal high-performance DSP/CPU/FPGA
Instantaneous analysis	tens of MHz	tens of MHz	hundreds of MHz to multi-GHz
Front-end strategy	strong preselection	wideband low-IF	precision multi-path/multi-conversion front end
VSA capability	raw/calibrated I/Q + software	stronger VSA capability + I/Q	extensive integrated VSA/standards apps
Phase-noise performance	basic	strong	metrology/high-end class
Frequency coverage	sub-6 GHz	through X band	microwave/mmWave through 90 GHz
Field portability	excellent	excellent	bench/lab instrument
GPS field work	external	built in	not the primary design goal
Best value	weak-signal sub-6 measurements	fast/wide portable RF analysis	advanced communications/radar/mmWave R&D

Which instrument for which measurement?

Choose BB60D when

measurements are below 6 GHz;
weak-signal hunting matters;
hardware preselection is valuable;
EMC/spurious work is important;
27 MHz real-time bandwidth is sufficient.

Choose SP145 when

you need 6–14.5 GHz;
sweep speed matters;
phase noise/EVM performance matters;
you need X-band;
field/GPS measurements matter;
40 MHz real-time/IQ bandwidth is sufficient.

Choose an FSW-class analyzer when

you need 100 MHz to multi-GHz-wide vector analysis;
you need high-order-QAM EVM with large margin;
phase-noise characterization is critical;
you need integrated 5G/WLAN/radar applications;
you need mmWave coverage;
the analyzer must be substantially better than a state-of-the-art DUT.

A useful hierarchy

The three analyzers can be viewed as

$ \boxed{ \begin{array}{c} \text{BB60D}
\text{compact RF spectrum / real-time analyzer} \end{array} } $

$ \Downarrow $

$ \boxed{ \begin{array}{c} \text{SP145}
\text{fast portable microwave real-time / vector analyzer} \end{array} } $

$ \Downarrow $

$ \boxed{ \begin{array}{c} \text{R&S FSW}
\text{wideband high-dynamic-range signal + spectrum analysis platform} \end{array} } $

The underlying mathematics—mixing, sampling, filtering, FFTs, detectors and I/Q analysis—is the same. What changes is how much RF selectivity, bandwidth, dynamic range, phase-noise performance, processing power and measurement intelligence the instrument can provide.

Part III — ADC and digital transport

9. ADC

After analog frequency conversion/filtering, the IF reaches a 14-bit ADC operating at 80 MS/s.

Signal Hound notes that the ADC uses built-in dithering to improve linearity and reduce coherent ADC spurs.

The fundamental ADC operation is

\[x[n]=x(nT_s), \qquad T_s=\frac{1}{f_s}.\]

\[f_s=80\ \mathrm{MS/s},\]

the raw sample interval is

\[T_s=12.5\ \mathrm{ns}.\]

10. Why a 14-bit ADC does not mean 14-bit spectral accuracy

ADC resolution is only one contributor.

Actual analyzer dynamic range also depends on

RF gain;
mixer IP3/IP2;
phase noise;
ADC noise;
ADC spurs;
clock jitter;
preselector rejection;
calibration;
FFT processing.

A strong signal can create internally generated spectral products even though the ADC has many nominal bits.

See ADC & DAC — Data Converters for ENOB, SFDR, quantization and jitter.

11. FPGA packetization

Signal Hound documents that the digitized IF samples are passed to an FPGA.

The FPGA primarily performs high-rate data handling/packetization before sending the samples through the Cypress FX3 USB interface.

Conceptually:

\[\boxed{ ADC \rightarrow FPGA \rightarrow USB\ 3.0 \rightarrow PC\ memory. }\]

This architecture moves much of the computational burden from dedicated analyzer hardware into software running on the host.

12. Why USB throughput matters

At 80 million samples/s with high sample resolution, the data rate is substantial.

Signal Hound quotes approximately

\[140\ \mathrm{MB/s}\]

over USB 3.0.

If the host cannot process or transport data continuously, the analyzer cannot maintain its intended real-time behavior.

Thus a PC-based real-time spectrum analyzer is a hardware + USB + memory + CPU/GPU system.

Part IV — From samples to spectrum

13. Time-domain block

Suppose the analyzer collects

\[N\]

samples.

The observation time is

\[T_{obs}=\frac{N}{f_s}.\]

The raw FFT bin spacing is

\[\boxed{ \Delta f=\frac{f_s}{N}. }\]

Increasing $N$ gives finer frequency spacing but increases observation time and computational work.

14. Windowing

A finite-length sample record is equivalent to multiplying an infinite signal by a window.

A rectangular truncation produces spectral leakage.

Therefore the analyzer uses a window:

\[x_w[n]=x[n]w[n].\]

The FFT then becomes

\[X[k]=FFT\{x[n]w[n]\}.\]

The choice of window changes

main-lobe width;
equivalent noise bandwidth;
sidelobe level;
amplitude accuracy;
scalloping loss.

15. Flat-top window

Signal Hound swept-analysis modes commonly use a flat-top window when amplitude accuracy is the priority.

Advantages:

excellent amplitude accuracy;
very low scalloping loss.

Disadvantages:

wider equivalent noise bandwidth;
lower frequency resolution for a given FFT length;
more processing for a requested narrow RBW.

Signal Hound recommends the flat-top RBW shape when accurate CW marker amplitude is important.

16. Nuttall window

Signal Hound BB/SP real-time processing uses Nuttall-style windowing in relevant real-time modes.

Compared with flat-top:

narrower effective bandwidth;
faster processing;
favorable sidelobe behavior;
larger scalloping error.

Signal Hound states that a tone halfway between FFT bins may read up to roughly 0.8 dB low in real-time Nuttall processing.

For accurate integrated channel power this is much less problematic because energy across neighboring bins is summed.

Part V — Resolution bandwidth

17. What RBW means in an FFT analyzer

In an old analog analyzer, RBW is approximately the bandwidth of an IF filter.

In an FFT analyzer, RBW is produced digitally through

\[\boxed{ \text{sample rate} + \text{FFT length} + \text{window ENBW}. }\]

Thus RBW is not simply FFT-bin spacing.

18. BB60 FFT-size relationship

Signal Hound’s API documentation gives an internal FFT-size estimate of the form

\[N_{best} = \frac{80\times10^6\,B_w}{RBW},\]

where $B_w$ depends on the selected window.

Documented factors include approximately

Window	Window bandwidth factor
Nuttall	2.02
Flat-top	3.7702
CISPR	2.65

The requested FFT is then adjusted to a suitable power-of-two length and may be zero-padded.

19. Smaller RBW

A narrower RBW:

separates closely spaced signals;
lowers integrated noise in each resolution filter;
improves apparent DANL.

But it also requires

longer time records;
larger FFTs;
longer sweep time.

For thermal-like noise,

\[P_n=kTB.\]

Reducing RBW by a factor of 10 ideally lowers integrated noise by approximately

\[10\ \mathrm{dB}.\]

Part VI — FFT bins and scalloping loss

20. FFT bins

The FFT evaluates the spectrum at discrete frequencies

\[f_k=k\frac{f_s}{N}.\]

If a CW signal lands exactly on a bin, its amplitude can be represented cleanly.

If it lies between bins, its energy distributes into multiple bins.

21. Scalloping loss

Scalloping loss is the variation in measured peak amplitude as a tone moves between FFT-bin centers.

The amount depends strongly on the window.

Therefore

“marker peak amplitude” is partly a DSP/windowing problem.

For precision amplitude measurement, use an amplitude-accurate window such as flat-top where supported, or use properly integrated channel-power measurement.

Part VII — Video bandwidth

22. RBW versus VBW

A useful distinction is

\[\boxed{ RBW \rightarrow \text{frequency-domain resolution} }\]

while

\[\boxed{ VBW \rightarrow \text{smoothing of detected amplitude}. }\]

Spike documentation states that after RBW processing, the signal is converted to amplitude and filtered by the VBW filter.

When

\[VBW=RBW,\]

no additional VBW filtering is performed in that processing path.

VBW must be less than or equal to RBW.

23. Why lower VBW makes noise look smoother

Noise changes rapidly from acquisition to acquisition.

A lower VBW suppresses fast amplitude fluctuations.

It does not provide the same effect as narrowing RBW.

A smooth trace can still represent the same underlying integrated noise power.

Part VIII — Detectors

24. Why a detector is needed

Many FFTs may contribute to one displayed sweep bin.

The analyzer must decide how to combine them.

Signal Hound’s BB API fundamentally provides

average detector;
min/max detector.

25. Peak / min-max behavior

With min/max detection, the analyzer maintains the minimum and maximum spectral values observed for each output bin.

The maximum trace approximates traditional positive-peak detection.

This is useful for

intermittent signals;
pulsed RF;
frequency hopping;
EMC precompliance;
finding rare peaks.

26. Average detector

Average processing combines repeated spectral measurements.

For RMS/channel-power work, average detection should be performed in linear power rather than simply averaging dB values.

If powers are $P_i$,

\[P_{avg} = \frac{1}{M}\sum_iP_i,\]

then convert to dB:

\[P_{avg,dBm} = 10\log_{10} \left( \frac{P_{avg}}{1\ \mathrm{mW}} \right).\]

Spike explicitly recommends linear-power video units with average detection for RMS power measurements.

Part IX — Trace processing

27. Normal trace

Every new sweep replaces the previous one.

Useful for stable CW or continuously changing signals.

28. Max hold

For each frequency bin,

\[X_{max}[k] = \max \left( X_1[k],X_2[k],\ldots \right).\]

Max hold is powerful for

hopping transmitters;
intermittent interference;
transient emissions.

But it is not an instantaneous spectrum.

It is the historical maximum at each frequency.

29. Averaging

Repeated spectra can be averaged to reduce variance.

This is useful for

noise-like signals;
stable carriers;
amplitude measurement.

But averaging can hide transient events.

Part X — Real-time spectrum analysis

30. Swept mode versus real-time mode

Swept

The analyzer tunes through a span.

A transient at a frequency not currently being observed can be missed.

Real time

The analyzer continuously digitizes the full instantaneous bandwidth.

Within the BB60D’s 27 MHz real-time bandwidth, there are no acquisition gaps under supported real-time conditions.

31. Overlapping FFTs

The BB60 real-time implementation uses 50% overlapping FFTs.

If an FFT record contains $N$ samples, the next FFT begins roughly halfway through the previous record.

Conceptually:

\[\begin{array}{cccccccc} |&FFT_1&&|\\ &&|&FFT_2&&|\\ &&&&|&FFT_3&&| \end{array}\]

Overlap prevents events occurring near one FFT boundary from being missed or heavily attenuated by the window.

32. Probability of intercept

Signal Hound specifies 100% probability of intercept for a full-amplitude event as short as approximately

\[4.8\ \mu s\]

at an RBW of approximately

\[631\ \mathrm{kHz}.\]

This number is configuration-specific.

POI is not simply “how fast the screen refreshes.”

It is determined by the continuous acquisition and FFT/window processing.

33. Why screen refresh rate is not FFT rate

The analyzer may process enormous numbers of FFTs internally but only update the display tens of times per second.

Therefore

\[\boxed{ \text{FFT processing rate} \neq \text{display frame rate}. }\]

Many FFT results are compressed into

a spectrum trace;
persistence/density information;
a waterfall frame.

Part XI — Persistence and waterfall displays

34. Persistence display

A real-time persistence display records how often signal energy appears at a frequency and amplitude.

Conceptually define a 2-D histogram

\[H(f,P).\]

Every FFT contributes samples into this histogram.

Color then represents

occurrence count;
probability/density;
recency depending on display implementation.

This makes intermittent signals visible even when a normal trace would miss their temporal behavior.

35. Waterfall

A waterfall adds time as another dimension:

\[\boxed{ x=f, \qquad y=t, \qquad color=P. }\]

It is particularly useful for

frequency hopping;
drifting oscillators;
pulsed interference;
burst communications;
intermittent spurs.

Part XII — I/Q streaming

36. Why stream I/Q?

A spectrum is only one representation of the digitized signal.

Streaming complex I/Q preserves

amplitude;
phase;
modulation;
timing.

This allows external processing such as

demodulation;
vector signal analysis;
custom FFTs;
matched filtering;
direction finding;
pulse analysis;
coherent measurements.

37. Complex I/Q representation

A complex sample is

\[z[n]=I[n]+jQ[n].\]

Instantaneous power is proportional to

\[|z[n]|^2 = I^2[n]+Q^2[n].\]

Signal Hound’s API documents calibrated I/Q power calculation in this form.

38. BB60D I/Q sample-rate table

Signal Hound documents the following approximate I/Q configurations:

Decimation	I/Q sample rate	Maximum software-filter bandwidth
1	40 MS/s	27 MHz
2	20 MS/s	17.8 MHz
4	10 MS/s	8 MHz
8	5 MS/s	3.75 MHz
16	2.5 MS/s	2 MHz
32	1.25 MS/s	1 MHz
64	625 kS/s	500 kHz
128	312.5 kS/s	250 kHz
256	156.25 kS/s	140 kHz
512	78.125 kS/s	65 kHz
1024	39.0625 kS/s	30 kHz
2048	19.53125 kS/s	15 kHz
4096	9.765625 kS/s	8 kHz
8192	4.8828125 kS/s	4 kHz

This is a concrete example of digital downsampling + filtering.

Part XIII — Digital downconversion and decimation

39. Conceptual DDC

A digital downconverter can be represented as

\[x[n] \times e^{-j2\pi f_0n/f_s} \rightarrow \text{low-pass filter} \rightarrow \text{decimate}.\]

The result is lower-rate complex baseband I/Q.

This reduces data rate when only a narrow portion of the 27 MHz instantaneous bandwidth is needed.

40. Decimation

If the original sample rate is $f_s$ and the decimation factor is $M$,

\[f_s' = \frac{f_s}{M}.\]

Before throwing away samples, a low-pass/anti-alias filter must restrict bandwidth.

Thus

\[\boxed{ \text{filter first} \rightarrow \text{decimate second}. }\]

Part XIV — Wide swept spectrum

41. How a 6 GHz sweep is built from a 27 MHz receiver

The ADC cannot see 6 GHz simultaneously.

The analyzer instead performs many local measurements.

Conceptually:

\[\begin{aligned} &f_{LO,1}\rightarrow FFT_1\\ &f_{LO,2}\rightarrow FFT_2\\ &f_{LO,3}\rightarrow FFT_3\\ &\cdots \end{aligned}\]

The useful portions are combined to make a wide spectrum.

This explains why

instantaneous bandwidth;
span;
sweep speed

are three different specifications.

42. Instantaneous bandwidth

For the BB60D,

\[B_{inst}\approx27\ \mathrm{MHz}.\]

Within that window, all frequencies can be observed concurrently.

Outside it, the LO must retune.

Part XV — Channel power and occupied bandwidth

43. Channel power

A digital analyzer calculates channel power by integrating spectral power across the defined channel.

If each FFT bin represents linear power $P_k$,

\[\boxed{ P_{channel} \approx \sum_{k\in channel} P_k }\]

with corrections for the window/RBW representation.

Signal Hound explicitly recommends average/RMS-style detection in linear power for this calculation.

44. Why you should not sum dBm directly

dBm is logarithmic.

This is wrong:

\[-50\ \mathrm{dBm} + -50\ \mathrm{dBm} \neq -100\ \mathrm{dBm}.\]

Instead:

convert each bin to watts/mW;
add powers;
convert back to dBm.

Two equal $-50$ dBm signals sum to approximately

\[-47\ \mathrm{dBm}.\]

45. Occupied bandwidth

Occupied bandwidth finds the frequency interval containing a selected percentage of the integrated signal power.

It is therefore based on integrated spectrum, not simply the visible -3 dB bandwidth.

46. ACP / ACPR

Adjacent-channel power is calculated by integrating power in specified neighboring channels.

A common ratio is

\[ACPR = 10\log_{10} \left( \frac{P_{adjacent}}{P_{main}} \right).\]

This measurement reveals effects from

PA nonlinearity;
phase noise;
modulation;
analyzer dynamic range.

Part XVI — Dynamic range

47. DANL

Displayed average noise level is the analyzer’s displayed noise floor under specified conditions.

Thermal-noise-like contributions scale with RBW:

\[P_n\propto B.\]

Therefore DANL must always be interpreted together with RBW and front-end settings.

48. Third-order intermodulation

With two strong tones,

\[f_{IM3,1}=2f_1-f_2\]

and

\[f_{IM3,2}=2f_2-f_1.\]

These spurs may be produced by

the DUT;
the analyzer itself.

The analyzer must have enough internal linearity that its distortion is below the DUT’s distortion.

49. Intermodulation-free dynamic range

Signal Hound gives the common approximation

\[\boxed{ IMFDR = \frac{2}{3}(TOI-DANL) }\]

when normalized consistently.

The BB60D manual gives an illustrative typical value around

\[104\ \mathrm{dB}\]

at 1 GHz, -10 dBm reference level, normalized to 1 Hz RBW.

This is not the same as display range.

Part XVII — Images, spurs and spur rejection

50. Mixer images

A superheterodyne receiver can respond to frequencies other than the desired RF because multiple RF/LO combinations can produce the same IF.

This is why preselection and image rejection are essential.

51. Residual spurs

An analyzer can generate internal spurs through

mixer products;
ADC nonlinearity;
LO leakage;
clock coupling;
digital clocks;
internal harmonics.

A displayed line is not automatically an external RF signal.

A useful test is to change

center/span;
reference level;
attenuation;
preselector configuration;
input termination.

A true external signal and an internally generated spur often behave differently.

52. Spur rejection processing

The BB API offers a spur-reject option in standard swept analysis.

Signal Hound recommends it for steady CW/slowly varying measurements where minimizing image/spurious responses is more important than capturing extremely short transients.

This illustrates a broader principle:

Some analyzer “clean-up” functions trade temporal fidelity for improved spectral certainty.

Part XVIII — Phase noise and clocking

53. Analyzer LO phase noise

A perfectly pure CW signal should appear as a delta function.

Real oscillators have phase fluctuations.

The analyzer LO itself therefore broadens or masks nearby weak signals.

Measured phase noise contains contributions from both

DUT oscillator;
analyzer local oscillator.

Near a strong carrier, analyzer phase noise can dominate the apparent noise skirt.

54. External 10 MHz reference

The BB60D provides a dedicated external 10 MHz reference input.

This allows synchronization to

GPS-disciplined oscillator;
rubidium standard;
laboratory reference;
common multi-instrument reference.

A common reference is useful for

repeatable frequency accuracy;
synchronized RF experiments;
multi-instrument measurements.

Part XIX — Calibration and amplitude correction

55. Why calibration is required

The measured FFT amplitude is affected by frequency-dependent

gain;
attenuator loss;
mixer conversion loss;
filter response;
ADC scale;
temperature.

Calibration coefficients compensate these effects.

Thus the displayed spectrum is not simply

\[20\log_{10}|FFT|.\]

It is closer to

\[\boxed{ \text{raw spectrum} + \text{window correction} + \text{gain/attenuation correction} + \text{frequency-response calibration}. }\]

56. Temperature

Signal Hound notes that analyzer calibration depends on operating temperature.

For the BB60C/D API architecture, reinitiating the device after significant temperature change allows recalibration at the new operating temperature.

This is a reminder that digital processing does not eliminate analog drift.

Part XX — What the display trace actually means

57. A single displayed point may contain much processing

One pixel/bin on the screen may represent

\[\boxed{ \text{many ADC samples} \rightarrow \text{window} \rightarrow \text{FFT} \rightarrow \text{calibration} \rightarrow \text{multiple overlapping FFTs} \rightarrow \text{detector} \rightarrow \text{VBW / averaging} \rightarrow \text{trace processing}. }\]

Therefore never interpret the plotted line without knowing analyzer settings.

58. Settings table

Setting	What it changes physically / numerically
Center frequency	RF/LO tuning
Span	number/range of tuned FFT blocks
Reference level	RF gain/attenuation state
RBW	FFT length/window equivalent noise bandwidth
VBW	post-detection smoothing
Detector	how repeated FFT results are combined
Sweep time	acquisition/dwell opportunity
Window	leakage, scalloping, ENBW
Trace mode	normal/max-hold/average history
Real-time mode	continuous FFT acquisition within instantaneous BW
IQ bandwidth	DDC/filter/decimation configuration

Part XXI — Measurement examples

59. Measuring a CW tone accurately

Use

reference level just above the signal;
flat-top window;
suitable narrow RBW;
VBW = RBW initially;
marker peak.

Why flat-top?

Because its low scalloping loss provides better amplitude accuracy for a tone not exactly centered on an FFT bin.

60. Measuring weak tone beside strong tone

Important settings:

lower reference level if possible without compression;
use preselection;
narrow RBW;
evaluate analyzer phase noise;
verify the weak line is not an internal spur.

The limiting mechanism may be analyzer phase noise rather than thermal noise.

61. Capturing a short burst

Use

real-time mode;
suitable RBW;
persistence/waterfall;
max detector.

Do not assume a very narrow RBW is always beneficial: a narrower filter has a longer time response and may reduce visibility of very short events.

62. Measuring broadband noise

Use

average detector;
linear-power units;
controlled RBW;
channel-power integration.

Avoid peak detector if the desired quantity is RMS/average noise power.

63. Measuring a modulated communication channel

Use

appropriate span;
average/power detector;
channel-power integration;
occupied bandwidth;
ACP/ACPR;
I/Q streaming or vector-signal analysis if modulation quality is required.

Spectrum amplitude alone cannot provide EVM because EVM requires complex amplitude/phase information.

Part XXII — Spectrum analyzer versus VSA

64. Spectrum analyzer

Primarily answers:

How much energy exists versus frequency?

Output:

\[P(f).\]

65. Vector signal analyzer

Uses complex I/Q:

\[z(t)=I(t)+jQ(t)\]

to recover

constellation;
EVM;
phase;
frequency error;
modulation symbols.

Modern analyzers such as the BB60D and SP145 can provide spectrum plus calibrated I/Q for software-based analysis, while higher-end instruments such as the FSW also perform extensive vector/standards analysis internally.

Part XXIII — What you should know

66. Ten important concepts

If you work with a modern spectrum analyzer, understand:

superheterodyne frequency conversion;
instantaneous bandwidth versus total tuning range;
ADC sampling and dynamic range;
FFT-bin spacing versus true RBW;
window functions and scalloping loss;
RBW versus VBW;
peak versus average/RMS detection;
swept versus real-time acquisition;
I/Q streaming and decimation;
analyzer-generated noise, phase noise and spurs.

67. Three-analyzer mental model

The BB60D baseline architecture can be summarized as

\[\boxed{ \begin{aligned} RF &\rightarrow \text{attenuator / preselector}\\ &\rightarrow \text{superheterodyne conversion}\\ &\rightarrow 27\ \mathrm{MHz\ IF}\\ &\rightarrow 80\ \mathrm{MS/s},\ 14\text{-bit ADC}\\ &\rightarrow FPGA\\ &\rightarrow USB\ 3.0\\ &\rightarrow \text{PC DSP}\\ &\rightarrow \text{FFT / detector / VBW / trace / persistence}. \end{aligned} }\]

That mental model transfers directly to many modern USB-, PCIe-, and Ethernet-connected real-time spectrum analyzers.

Part XXIV — References

Signal Hound

Rohde & Schwarz

Measurements & Instruments · ADC & DAC · RF & Microwave · Signal Integrity & Power Integrity · 3GPP Standards · RF Bands