|
|
|
Abstract |
Full Duplex DAQ comparison
E. Olson, Sep 28 2005
Test, measure, and characterize the
full duplex performance
of various data acquisition hardware and software drivers. The device's
ADC and
DAC will be tested as a complete
system in a loopback fashion. The sample rate accuracy, frequency
response, distortion, and quantization measurements will be analyzed and
compared.
|
|
|
|
Procedure |
The DAQ device becomes the black
box under test that is illustrated in the diagram to the right. The input
and output channels are connected together with an external loopback cable (or
optionally with an internal "volume" mixer setting). The
AC coupling
option is enabled to remove the effects of the DC offset from the distortion
measurements.
The measurements are performed at all of the standard sample rates for a
complete picture of the DAQ's capabilities and filter response. The
baudline test and measurement application notes below describe the procedural
details:
|
|
|
|
Devices |
The following device reports contain technical information, spectral response
plots, detailed measurements, and analysis:
|
|
|
|
Sample Rate |
The following table of measurements were made using the
sample rate stability
technique at the highest native sample rate supported by the DAQ device.
The in/out rate and error columns are absolute measurements of
the ADC/DAC clock. The loop error column uses a tone generator
loopback method for a high accuracy measurement of the relative difference
between the ADC and the DAC clocks.
| device |
in rate |
out rate |
in error |
out error |
loop error |
| CS4236B |
47972.865 |
47973.211 |
-565.313 PPM |
-558.104 PPM |
+0.0000 PPM |
| SB Vibra16X |
44086.273 |
44086.296 |
-311.270 PPM |
-310.748 PPM |
|
| SB16 PCI |
48000.103 |
48000.103 |
+2.1396 PPM |
+2.1396 PPM |
+0.0000 PPM |
| SB128 |
48003.136 |
48003.136 |
+65.3333 PPM |
+65.3333 PPM |
+0.0000 PPM |
| SB Live! |
48001.674 |
48005.204 |
+34.8750 PPM |
+108.417 PPM |
+0.0000 PPM |
| VIA 8235 |
47999.130 |
47999.245 |
-18.1208 PPM |
-15.7271 PPM |
+0.0000 PPM |
| SiS 7012 |
48006.40 |
48006.42 |
+133.33 PPM |
+132.75 PPM |
+0.0000 PPM |
| Maestro 2E |
47185.71 |
48000.94 |
-16964.4 PPM |
+19.5833 PPM |
-16991.0 PPM |
| iMic v0.06 |
48007.95 |
48007.94 |
+165.625 PPM |
+165.417 PPM |
±2.4842 PPM |
| iMic v3.00 |
48002.45 |
48008.97 |
+51.0417 PPM |
+186.875 PPM |
-130.653 PPM |
| Edirol UA-25 |
44108.671 |
44108.671 |
+196.62 PPM |
+196.62 PPM |
+0.0000 PPM |
The devices without a loop error entry are half duplex. The
± symbol represents a PPM range because the frequency was
fluctuating.
Most full duplex devices have a zero PPM loop error which signifies that
their ADC and DAC clocks are in lockstep. Large errors such as the ESS
Maestro 2E's -17000 PPM are a sign of a major problem. The SB Live's
in / out error mismatch of 73 PPM represent a driver fragment problem.
A lot of internal details about the clock structure, sample rate conversion,
and driver robustness can be learned from rate analysis.
|
|
|
|
Distortion |
The following table of measurements were made using the
sine distortion
technique at the highest native sample rate supported by the DAQ device.
This is a full duplex test that uses a loopback of the tone generator to
measure the various SNR, THD, SINAD, ENOB, and SFDR metrics. See the
distortion measurement
window for more information. The stereo crosstalk column is a
measure of channel bleed through that uses a sine wave channel and a silent
channel as signal sources.
| device |
rate |
SNR |
THD |
SINAD |
ENOB |
SFDR |
crosstalk |
| CS4236B |
48000 |
+72.47 dB |
-73.19 dB |
+69.80 dB |
+11.302 bits |
+79.84 dB |
-67.35 dB |
| SB Vibra16X |
44100i |
+72.74 dB |
-74.58 dB |
+70.55 dB |
+11.426 bits |
+77.69 dB |
-78.08 dB |
| SB Vibra16X |
44100o |
+75.24 dB |
-73.22 dB |
+71.10 dB |
+11.518 bits |
+76.95 dB |
-75.26 dB |
| SB16 PCI |
48000 |
+78.60 dB |
-77.05 dB |
+74.74 dB |
+12.122 bits |
+79.87 dB |
-81.33 dB |
| SB128 |
48000 |
+78.88 dB |
-80.66 dB |
+76.67 dB |
+12.442 bits |
+87.61 dB |
-77.51 dB |
| SB Live! |
48000 |
+78.63 dB |
-77.99 dB |
+75.29 dB |
+12.213 bits |
+80.55 dB |
-81.17 dB |
| VIA 8235 |
48000 |
+64.31 dB |
-60.89 dB |
+59.26 dB |
+9.551 bits |
+65.30 dB |
-81.37 dB |
| SiS 7012 |
48000 |
+69.84 dB |
-70.95 dB |
+67.35 dB |
+10.894 bits |
+72.69 dB |
-86.33 dB |
| Maestro 2E |
48000 |
+17.70 dB |
-30.52 dB |
+17.48 dB |
+2.611 bits |
+25.62 dB |
-66.47 dB |
| Maestro 2E |
48000o |
+71.92 dB |
-65.77 dB |
+64.83 dB |
+10.475 bits |
+68.58 dB |
-73.64 dB |
| iMic v0.06 |
48000 |
+84.06 dB |
-85.99 dB |
+81.91 dB |
+13.312 bits |
+93.81 dB |
-90.55 dB |
| iMic v3.00 |
48000 |
+79.18 dB |
-77.60 dB |
+75.31 dB |
+12.216 bits |
+78.74 dB |
-64.76 dB |
| Edirol UA-25 |
44100a |
+88.77 dB |
-86.24 dB |
+84.31 dB |
+13.711 bits |
+90.58 dB |
-82.09 dB |
| digital |
no dither |
+99.88 dB |
-106.45 dB |
+99.02 dB |
+16.154 bits |
+115.27 dB |
-inf.00 dB |
| digital |
dither |
+96.91 dB |
-103.54 dB |
+96.06 dB |
+15.662 bits |
+112.73 dB |
-inf.00 dB |
The "i" and "o" suffixes in the sample rate column represent isolated testing
of the input ADC (i) and the output DAC (o) components.
The digital device is the baseline level of performance for a lossless 16-bit
channel.
|
|
|
|
Analysis |
Visually comparing the spectrogram sine sweeps and the average window filter
response for different audio devices reveals many interesting details.
See the galleries below for an insightful view into the similarities and
differences of the DAQ devices under test.
gallery
|
|
|
|
Conclusion |
All of the tested devices fall far short from the theoretical ENOB of
16-bits.
The quality and performance of different audio devices varies wildly.
Consumer sound cards have an incredible number of flaws and artifacts, some
subtle and some not. Performance as a function of sample rate is a major
variable. Quantization errors and inter channel delay also play a
critical role.
The most important lesson we learned in the creation of this application note
is the value of conducting a thorough test and measurement procedure for all
of a device's control parameters. Many sound cards, that for years were
thought to be top performers, had numerous flaws lurking under the
surface. The only way to uncover them was with a systematic and
exhaustive search.
If you are an OEM please keep in mind that
SigBlips can be contracted to help in
your hardware and system development. You can send us hardware to run the
series of DAQ comparison tests on. We can build an automated GoNoGo test
system customized for your equipment manufacturing. We can write or
repair audio drivers. Maximize performance and minimize design flaws by
involving us in your development process.
|
|
|
|
|
