The Signal Generator can produce signals of the following types:
Setting the audio output is described in Getting Started.
The RMS signal level can be set for any of the signal types with 0.1 dB
resolution relative to digital full scale. The arrow buttons on the RMS Level
spinner change the value in steps of 1dB, or any required value can be
typed directly into the level box.
Sine waves can be generated with frequencies between 10.0 and half the
soundcard sample rate, e.g. 24 kHz
for a soundcard operating at 48 kHz.
The frequency is controlled by entering a value in the Frequency box, or
using the arrow buttons to increment or decrement the value in steps of 0.5 Hz
for frequencies below 200 Hz and steps of 1 Hz thereafter.
The exact frequency that has been generated is shown at the bottom right
corner of the frequency display when the generator is running.
The RMS level can range from -90 to -3.0 dB FS (-3.0 dB FS is the maximum RMS level for a sine wave before clipping, at this level the peaks are at 0 dB FS).
The frequency can also be controlled via the graph cursor by checking
the "Frequency tracks cursor" box. When this box is checked the signal generator
frequency is linked to the position of the graph cursor and will change to
follow the cursor frequency as it is moved - the changes are smooth with no
phase discontinuities.
When the "Add dither to output" box is selected the generator adds 2 lsb pk-pk
triangular dither to the output to remove quantisation noise spikes. The level at which
the dither is added is controlled by the sample width selector to the right of the check box.
N.B. When using the JavaSound drivers audio data is usually limited to 16 bit precision.
Dither is beneficial if making very precise distortion measurements
of an electronic device such as a receiver, processor or equaliser.
It is usually not required when making acoustic measurements as the quantisation
artefacts it removes are far below the acoustic noise floor. The
Graphs below show the effect of the dither option during a loopback test of a soundcard
playing a 1 kHz tone at -6 dB FS. The first plot is without dither, the second plot is
with dither. Addition of dither cleans up much of the noise that was apparent below
-120 dB FS, especially at high frequencies, making the true harmonic distortion levels
more visible.
When the "Lock frequency to FFT" box is checked the generator output frequency
is locked to the nearest FFT bin centre for the current RTA FFT length. This allows
a rectangular FFT window to be used for maximum spectral resolution of the RTA plot.
The exact frequency that has been generated is shown at the bottom right corner of
the frequency display when the generator is running.
The square wave generator allows duty cycles between 1% and 99% in 1% steps.
The frequency that is generated is adjusted to ensure there is an even number
of samples in the period, so that the spectrum of a 50% duty cycle square wave
will only have odd harmonics. The actual frequency that has been generated is
shown at the bottom right corner of the frequency display when the generator
is running, at higher frequencies this can be significantly different to the
frequency that was entered.
The dual tone generator is to facilitate intermodulation distortion measurements.
It has presets for SMPTE, DIN and CCIF signals and allows custom signals to be
generated with ratios of 1:1 or 4:1. Note that for valid IMD results with custom signals
f2 must be higher than f1.
The CEA-2010 Burst generator produces a 6.5 cycle, Hann-windowed tone burst at the selected frequency. If the Repeat checkbox is selected the burst will be repeated. This signal is used for testing the maximum power output of subwoofers by using an RTA to observe the levels of the distortion components produced when the signal is playing, usually testing at 63, 50, 40, 31.5, 25 and 20 Hz. The limits for the distortion components are shown in the table below, where f0 is the test signal frequency.
Start Freq (Hz) | End Freq (Hz) | Limit (dB) | Comment |
---|---|---|---|
16 | 1.59*f0 | 0 | Fundamental |
1.59*f0 | 2.52*f0 | -10 | 2nd harmonic |
2.52*f0 | 3.78*f5 | -15 | 3rd harmonic |
3.78*f0 | 5.61*f0 | -20 | 4th and 5th harmonic |
5.61*f0 | 8.50*f0 | -30 | 6th - 8th harmonic |
8.50*f0 | 10 k | -40 | Higher order harmonics |
The highest level of the fundamental for which none of the limits are exceeded
is the maximum output level at that test frequency. When the CEA burst signal is playing
the RTA shows the limits as an overlay, provided the frequency of the burst is not more than
1,176 Hz. The recommended RTA settings for 44.1 kHz or 48 kHz sample rate are:
For 88.2 kHz or 96 kHz use FFT length of 131,072. Refer to the CEA-2010 standard for details of the measurement procedure, or search for guides on the Internet.
The Pink Noise generator uses white noise filtered through a -10 dB/decade
filter generated from a weighted sum of a series of first order filters, as
devised by Paul Kellet circa 1999. Stated accuracy is within 0.05 dB above
9.2 Hz at 44.1 kHz sample rate.
The Full Range option outputs the filtered noise directly, giving the widest bandwidth and the greatest low frequency content. The Speaker Calibration option applies 2nd order (40 dB/decade) filters at 500 Hz and 2 kHz, producing a signal with its energy centred on 1 kHz. Subwoofer Calibration applies filters at 30 Hz and 80 Hz. Both are broadly in line with the THX test signal recommendations. Custom Filtered allows low and/or high cut filter frequencies to be set arbitrarily, subject to a minimum bandwidth of 1 octave.
REW automatically adjusts the signal levels for the various options and filter settings so that the RMS values reflect the setting in RMS Level. Note that as Pink Noise has random variations some clipping of peaks will occur at RMS levels above approximately -10 dB.
Periodic Noise (PN) sequences are ideally suited for use with spectrum and
real time analysers (RTA's). They contain every frequency the analyser can
resolve in a sequence length that matches the length of the analyser's FFT.
Their great benefit is that they produce the desired spectrum shape without
requiring any averaging or windowing, so the analyser display reacts much
more rapidly to changes in the system than it would if testing with Pink
or White random noise, making them ideal for live adjustment of EQ filters.
The PN sequences REW generates are optimised to have a crest factor (ratio
of peak level to rms level) that does not exceed 6 dB.
Use Pink PN when measuring with an RTA or White PN with a Spectrum analyser.
The Length control must be set the same as the length of the FFT used
by the analyser. If it is set shorter than the analyser FFT there will be
notches in the analyser display, as the periodic noise will not contain
some of the frequencies the analyser is looking for. If it is set longer
the extra frequencies will give a noisy display requiring more averaging.
The images below show the effect of correct and incorrect settings of the PN
length for a loopback measurement with 1/48 octave RTA that is using an FFT
length of 65536 (64k).
Length 32768, shorter than FFT
Length 131072, longer than FFT (no averaging)
Length 65536, matching FFT
When using the REW RTA the PN length is automatically set the same as the FFT length.
The Save PN to WAV file button generates a 16-bit stereo wave file containing the PN sequence in both channels. The file duration is approximately 1 minute, the level is per the RMS Level setting of the signal generator. This file can be used to generate a test disc to be played on a system whose response is to be measured. Make sure that the current soundcard sample rate corresponds with the format of the disc to be made - for example, 44.1kHz should be used if generating a CD, or 48kHz for a DVD. When measuring the system the sample rate and FFT length must be the same as used for the test disc.
The Signal Generator can produce sweeps with configurable start frequency/
level, end frequency/level, duration and linear or logarithmic progression.
Sweep duration can be up to 60 seconds. If the "Loop" box is checked the sweep
will repeat continuously.
The Measurement Sweep signal is used by REW when measuring system response.
It consists of a logarithmic sweep from the start frequency to the end frequency. The
sweep duration is set using the Length control. If the start frequency is below 20Hz
the signal begins with a linear sweep from DC to 10Hz, followed by a logarithmic sweep
from there to the end frequency. This signal is selected automatically to make sweep
measurements.