Components using internal digital signal processing must not be
overdriven since any loading in excess of the digital level range would
result in strong distortion of the signal (clipping level). The full-scale
amplitude therefore
plays a far more important role in digital than in analog applications.
The clipping level must be determined for all digital components with analog
input stage. If digital outputs are accessible, this is accomplished by
increasing the level of a 997-Hz sinusoidal input signal until the peak value
of the digital output signal equals the largest data word (full scale).
The level thus obtained defines the full-scale amplitude of the digital system
and is used as a reference value in a variety of measurements.
The setup supplies an analog output signal of 997 Hz. The level is set
to 1 V.
As the clipping level serves as a reference in a variety of other measurements,
it is expedient to use the Ref Volt function of UPD/UPL. The data can
then be entered in dBr in the VOLTAGE line of the GENERATOR panel, which
does away with the need for constantly converting the levels to the clipping level.
To determine the full-scale amplitude as described above, the level of the
generator signal is increased in the Ref Volt line until the analyzer
indicates the peak value of 0 dBFS. In doing this, it must be ensured that the
full-scale value is not exceeded in none of the channels.
The clipping level thus obtained can then be transferred to all setups used for
measurements on that particular DUT; any other level entries are made in dBr in
the VOLTAGE line.
7.2. Linearity of A/D Converters
o
LINS_AD.SAC
A 997-Hz sinusoidal signal is
applied to the input of the DUT. The level of this signal is decreased in steps
of 5 dB starting from the full-scale amplitude. The output signal is measured
and represented graphically versus the input signal. As the signal disappears
in the noise with decreasing level, narrowband measurement using a third-octave
bandpass filter is performed.
In the setup described here, converter linearity is measured by means of a
level sweep from 0 dBr to -120 dBr. With a linear response of the converter, a
diagonal is obtained as shown in the graphic display of Fig. 11.
Since deviations from the nominal characteristic are difficult to
recognize in the above type of representation, level nonlinearity is measured in
most cases, which is described in the next setup.
To drive the DUT at full-scale level, the clipping level determined in
the previous measurement is to be entered into the Ref Volt line of the
GENERATOR panel. This level serves as a reference for all level values defined
in dBr in the sweep lines.
The reference level used for x-axis scaling in the graphic display shown above
is in this setup automatically transferred from the GENERATOR panel into the Reference
line under x Axis.
7.3. Nonlinearity of A/D Converters
o
LINDS_AD.SAC
Same as previous setup, but showing deviation from ideal characteristic.
This type of measurement
is defined by AES 17, the test parameter being referred to as level-dependent logarithmic
gain. A linearity measurement is performed, the first result is however
recorded only at 5 dBFS. For each test step, the logarithmic gain, ie ratio of output
amplitude to input amplitude, is to be determined and represented graphically
versus the input level. The resulting diagram shows the deviation of the
converter transmission characteristic from the nominal linearity
characteristic.
Measurements are to be performed selectively using a third-octave bandpass
filter.
For this measurement, too, the clipping level determined in accordance
with 7.1 "Clipping Level" is to be entered into the Ref Volt line
of the GENERATOR panel. This level serves as a reference for all level values
defined in dBr in the sweep lines.
In the ideal case, a straight line is obtained in the graphic display, any
deviation from the ideal characteristic of the converter can be read in dB.
This type of measurement
however involves a physical problem, ie referring the digital output voltage of
the converter to the analog input voltage at every test point. Audio Analyzers
UPD and UPL have an internal "conversion factor" of 1 FS ^ 1 V.
With this factor, a straight line would be obtained but it would not coincide
with the zero line. The gain factor of the DUT must, therefore, be taken into
account in addition. This is done by means of the NORMALIZE function,
which is included in the DISPLAY Panel (see Fig. 12). Here the gain can be
entered directly, it is however easier in most cases to transfer this value
from the graphic display. To this end, a cursor is placed on the linear section
of the curve and the cursor value is transferred to the NORMALIZE line
by selecting the item o Cursor.
o
LINS_DA.SAC
For linearity measurements of D/A converters, the information given
under 7.2 "Linearity of A/D Converters" applies analogously.
In addition it should be noted that for this measurement the digital input
signal should contain a dither with a triangular probability density function
and a level of 1 LSB. This dither is set in the setup.
The reference value used for graphic display of the level values in dBr
is obtained from the gain ratio of the converter, ie the ratio of digital input
amplitude to analog output amplitude. With this setup, the reference value is
easiest taken from the Reference line by selecting the MAX item.
The maximum value measured will thus be taken as a reference. In this setup,
this value corresponds to maximum level of the DUT since the generator sweep is
started at 0 dBFS.
7.5. Nonlinearity of D/A
Converters
o
LINDS_DA.SAC
This measurement too is defined by
AES 17. The information given under 7.3
"Nonlinearity of A/D Converters" applies analogously.
For this measurement too the digital input signal should contain a
dither with a triangular probability density function and a level of 1 LSB.
This dither is set in the setup.
The measurement procedure is the same as described under 7.3 for
"Nonlinearity of A/D Converters".
In this case too the NORMALIZE function is needed; the gain of
the converter is easiest transferred from the graphic diagram as described for
the A/D converter above.
7.6. Signal Delay in
Analog and Digital Systems
! This
measurement function is available only in Audio Analyzer UPL !
o
DEL_AA.SAC
o
DEL_AD.SAC
o
DEL_DA.SAC
o
DEL_DD.SAC
This measurement is used to
determine the signal delay between the input and the output of a digital system.
In accordance with AES 17, a pulse-shaped signal is applied to the DUT. The
input and the output signal are displayed on an oscilloscope from which the
delay can be read. The measurement is used whenever digital signal processing
takes place, also on DUTs with analog or analog/digital interfaces.
The measurement is performed with Audio Analyzer UPL in compliance with
AES 17. Compared with conventional dual-channel oscilloscopes, UPL offers the
advantage that the two stereo channels can be measured simultaneously, thus
allowing any delay between the two channels to be detected immediately.
This is possible because of the fact that UPL, using the Waveform function,
can be triggered not only to the measurement channels but also to a burst
signal supplied by the generator. This measurement function ensures that the
measurement is started for the two channels exactly time-synchronously with the
issue of the test signal. Since the test signal is applied to the input of the
DUT, exact triggering to the input signal of the DUT is performed. Internal
delays of UPL are taken into account and do not affect results.
For this test, a sine burst with a level of -20 dBFS is generated as prescribed
by AES 17. The burst consists of a 1-kHz signal which is output 10 times,
followed by an interval of 90 ms.
The signal delay is measured with the aid of the cursor. To this end,
the cursor is placed at the point at which the signal departs from the zero
line. The graphic window shows the level measured for each cursor position, so
this procedure is very easy to perform. The delay is then indicated directly in
the second cursor display window (the zero point in the graphic display
corresponds to the start of the test signal).
If the measurement is to be performed for both channels, TRACE B is to
be set to the FUNC CH 2 measurement function. The second cursor can then
be used for channel 2. It is further possible to display the time difference
between the two cursors directly by appropriate setting using the softkeys at
the bottom of the screen.
With equipment performing filtering of the signal, it may occur that the test
signal is attenuated while settling to steady state. To be able to observe this
effect in greater detail, a burst with several signal periods is used. Fig. 13
shows an example of such a measurement.
In addition to signal delay, AES 17 describes determination of the polarity
between the input and the output signal. This is likewise performed with the
setup described here. Polarity reversal is indicated if the displayed output
signal does not start with the positive half-wave as is the case with the test
signal from the generator.