Spectrum: span, axis, window, trace, markers, distortion, gate and level

The Spectrum keys appear while the spectrum is on screen (Split, Spectrum or Spectrogram under Display). They are kept with your other settings. The keyboard shortcuts can zoom and move the span, change the axis, window and trace mode, and toggle markers or the peak table.

Which frequencies are shown

• Span is the width of the frequencies shown. Narrower zooms in about the centre and Wider zooms out, in steps of 1, 2 and 5. The narrowest span is a few frequency steps of the spectrum, since zooming further would show nothing new.
• Centre frequency moves the span a division lower or higher. It works once you have zoomed in.
• Full span goes back to the whole spectrum, from 0 Hz to half the sample rate.
• With the pointer over the spectrum, the mouse wheel zooms in and out, keeping the frequency under the pointer where it is.

The frequency of each labelled line is written along the bottom of the spectrum.

Frequency axis

• Linear spaces frequencies evenly: each division is the same number of hertz. Best for harmonics, which sit at equal steps.
• Log gives each factor of ten in frequency the same width, as audio work prefers, so low frequencies get as much room as high ones. A log axis cannot reach 0 Hz, so it starts at the spectrum’s first step above it.

Changing the axis shows the full span again.

Window

The spectrum is worked out from one capture. The capture starts and ends part way through a cycle, and that cut spreads some of each frequency into its neighbours, a smear called leakage. A window fades the capture in and out to reduce it, at the cost of wider peaks. Each window strikes a different balance:

Window	Good for
Hann	General use. The default.
Flat top	Reading a level accurately: within 0.1 dB wherever the tone falls. Widest peaks.
Blackman-Harris	Seeing a small tone next to a large one: the least leakage.
Hamming	Telling apart two close tones of similar level.
Rectangular	One-off events, or a signal that fits the capture a whole number of times.

With Hann, a tone that falls halfway between two frequency steps reads up to 1.4 dB low. Choose Flat top when the level matters more than telling close tones apart.

Under the window, the keys show two figures:

• Steps: the spacing between the frequencies worked out, the sample rate divided by the number of samples used. A lower sample rate gives finer steps over a smaller range.
• Resolution bandwidth: how wide a slice of frequencies each step really takes in, which depends on the window. It is about 1.5 steps for Hann and nearly 4 for flat top. Two tones closer than this merge into one peak.

Trace

Trace chooses whether the spectrum shows each capture on its own, or builds up over many:

• Normal: each capture on its own.
• Average: the average power of each frequency over the last few captures. Noise looks the same height but far smoother, so a small steady tone stands out of it. Averages sets how many captures, from 2 to 256. Until that many have arrived, the average is of those so far; after that each new capture counts for its share, so the trace follows slow changes.
• Max hold: the highest level each frequency has reached. Catches a tone that comes and goes, or a sweep.
• Min hold: the lowest level each frequency has reached. Shows what is always there.

Restart starts the average or hold again. It also starts again by itself when you change the window, the sample rate or the trace settings.

The line under the keys says how many captures are in the trace. When you stop, the trace is worked out again from the captures kept, so stepping back through them steps the trace back too. The spectrum measurements and the hover readout read the trace shown.

Markers

Markers puts two lines, A and B, on the spectrum, as the time cursors do on the trace. Each reads the frequency and level where it stands, and the measurements show B minus A: how far apart two tones are, and how much weaker one is. For example, put A on a square wave’s fundamental and B on its 3rd harmonic to read about 9.5 dB between them.

• Drag a marker line on the screen, or use Marker A and Marker B: each press moves a marker a tenth of a division.
• A to peak puts A on the strongest peak.
• A to next peak moves A to the next weaker peak; press again to walk down the peaks, and round to the strongest again.

A marker reads the nearest frequency step of the spectrum shown, so with Average or a hold chosen it reads that trace.

Peak table

Peak table lists the five strongest peaks beside the measurements, strongest first, with each one’s frequency, level, and level relative to the strongest (dBc, decibels relative to the carrier). Choose a frequency in the table to put marker A on it.

A peak is listed only when it stands well clear of the noise (20 dB above the typical level of the spectrum) and rises at least 6 dB above the dips either side of it, so noise and the shoulders of a large peak are not counted.

Distortion

Distortion shows how pure a signal is, beside the measurements. It takes the strongest tone as the fundamental, finds its harmonics (2, 3 and up to 10 times its frequency), and gives:

Figure	What it means
THD	Total harmonic distortion: the harmonics together, against the fundamental. Lower is purer.
THD+N	The same with the noise added in.
SNR	Signal to noise ratio: the fundamental against the noise, harmonics left out. Higher is better.
SINAD	Signal to noise and distortion: the fundamental against everything else.
SFDR	Spurious-free dynamic range: the fundamental against the largest other tone.
ENOB	Effective number of bits: how many bits a perfect converter with the same SINAD would have.

Under them, each harmonic’s level is given below the fundamental in dBc. A harmonic above half the sample rate folds back to a lower frequency (see Sampling); it is still measured where it lands, and marked folded.

The figures are worked out from the spectrum shown, so Average steadies them. Choose Blackman-Harris or Flat top for SNR, SINAD and ENOB: the other windows leak a little of each strong tone across the whole spectrum, which counts as noise and makes those three read low (the panel says so). THD and the harmonics read well with any window but Rectangular. A harmonic too small to see is given as below noise. While Distortion is on, the figures are also logged in the Trend window and included in Copy report.

Measuring a board’s own test signal? Its sample rate is often an exact multiple of the test signal’s frequency, so every capture samples the same points of each cycle. The harmonics then read a little high (the square’s 7th can read -15 dBc instead of -17 dBc) and can change as the signal drifts. For textbook figures choose a rate that is not a whole multiple of the signal’s frequency.

Some figures to compare with: a pure sine has a THD near 0 %, a square wave about 43 % (counting harmonics up to the 10th), and a perfect 10-bit converter an ENOB of 10 bits.

Gate to cursors

Normally the spectrum is worked out from the whole capture. Gate to cursors works it out from only the stretch between the two time cursors, to look at one part of a signal on its own: a burst, the ringing after an edge, or the moment something changed.

1. Turn on Cursors under the screen and place them around the part you want.
2. Press Gate to cursors. The stretch used is shaded on the trace, and a line under the keys says how many samples it holds.

It uses every sample between the cursors, at least 64; with fewer, the spectrum goes back to the whole capture and says so.

A shorter stretch can tell apart only tones further apart: the resolution bandwidth under the keys grows as the gate shrinks (64 samples at 20 000 samples a second give about 470 Hz with Hann). Widen the cursors, or lower the sample rate, to separate closer tones. A burst also reads at its true level when gated: across the whole capture its energy is spread over the quiet parts too, so it reads low.

Level scale

• Reference level is the level at the top of the screen. Lower and Raise move it by one division.
• Level per division is the level of each of the eight divisions up the screen: 1, 2, 5, 10 or 20 dB, or volts in steps of 1, 2 and 5.

Level unit

Every level is worked out from the RMS value of each frequency, as bench spectrum analysers do. A sine of 1 V peak is 0.707 V RMS.

Unit	Means	A 1 V peak sine reads
dBV	Decibels relative to 1 V RMS	-3.0 dBV
dBu	Decibels relative to 0.775 V RMS	-0.8 dBu
dBm	Decibels relative to 1 mW, into 50 ohms	+10.0 dBm
V	Volts RMS, on a linear scale	0.707 V RMS

dBm assumes the signal drives a 50 ohm load, as radio equipment does. The scope itself does not load the signal that way; the unit is there so readings match other equipment.

Choosing a unit also sets that unit’s usual scale. The hover readout and the Spectrum peak level measurement use the unit chosen here.