When choosing a data acquisition board, many different selection criteria are
used. Speed, resolution, accuracy, and number of channels are all important
considerations. In addition, there are different architectures used within
the data acquisition device itself that can affect your decision. The particular analog
input architecture that you choose will affect how often the input channels are sampled
and the accuracy of your results.
The two most common architectures in analog input design are multiplexed and simultaneous.
Multiplexed architectures use one A/D converter for many channels. Conversely,
simultaneous architectures use an individual A/D converter for each channel.
For many applications, a simultaneous acquisition device is certainly the architecture of
choice due to its inherent speed and accuracy but, until recently, the cost of these devices
was somewhat prohibitive. Times have changed and simultaneous devices are now as cost
effective as a multiplexed device for most users.
Data Translation has designed a series of simultaneous data acquisition modules for USB
2.0. Useful in the following applications, the Simultaneous Series provide, high speed, highly
accurate measurements:
- Semiconductor device testing
- Nanotechnology testing
- Drug discovery
- Scientific analysis
- Automotive testing
The Simultaneous Series modules from Data
Translation provide a 16-bit A/D converter for
each analog channel. This allows the user to
correlate ultra high-speed measurements of up
to 2MHz at the exact same instant in time.
These modules are also designed with 500V
galvanic isolation to maximize signal integrity
and protect the PC. This series was developed
for customers seeking to correlate highly accurate,
high-speed measurements while eliminating
phase noise from channel-to-channel acquisition.
With this series, a common clock and trigger are
used for simultaneous and synchronous sampling
of all inputs. This means that all functions
of the data acquisition modules (A/D, D/A, DIO,
Counter/Timers, and Quadrature Decoders) can
be simultaneously triggered internally or externally.
The data can then be clocked either internally
or externally and streamed synchronously to host memory. The synchronous operation
allows all I/O data to be processed and correlated for all inputs and outputs. This is very valuable
in determining the response across a device-under-test (DUT) to stimuli at the same exact instant.
A more detailed discussion of the benefits of simultaneous acquisition versus the multiplexed
approach is provided here to help you make your buying decision.
SIGNAL BANDWIDTH INCREASED
Simultaneous acquisition dramatically increases signal bandwidth because each channel uses the
full throughput of an individual A/D converter. Compare the following tables for a 150 kHz module
in a multiplexed and simultaneous architecture. You can see that the signal bandwidth is consistent
across all channels with the simultaneous architecture, while it decreases linearly with each
additional channel in the multiplexed architecture.
Additionally, the Simultaneous Series from Data Translation is designed to accurately measure
higher bandwidth signal components. To accurately measure 16-bit accuracy, the front-end input
amplifier has a bandwidth of ten times the Nyquist limit. Below is an example of these design
characteristics for the DT9832A.
| The DT9832A has a sampling
rate for each channel of 2.0
MHz. This means that the
Nyquist limit allows signal frequencies
up to 1.0 MHz to be
adequately measured. The analog
input components have a signal
bandwidth that is ten times
the Nyquist limit or in this case,
greater than 10.0 MHz to minimize
roll-off and phase errors. |
 |
CHANNEL TO CHANNEL SKEW ELIMINATED
As you can see from the figures below, since multiplexed devices use one common amplifier
and A/D converter to sample all the input channels being used, a resulting time delay or skew
exists between samples. This phase noise is all but eliminated when using a single A/D converter
per channel as does the Simultaneous Series modules.
Simultaneous Sampling
eliminates time skew between channels and
simplifies both time and frequency based
analysis techniques.
Sequentially (Multiplexed) Sampling
may require software correction for detecting
certain patterns. |
 |
 |
The A/D design of simultaneous modules features built-in accuracy. A maximum aperture
delay of 35 ns (the time it takes the A/D to switch from track to hold mode) is well matched
at 5 ns across all track-and-hold circuits, virtually eliminating channel-to-channel skew that is
associated with multiplexed inputs. A maximum aperture uncertainty of 1 ns (the jitter or variance
in aperture delay) virtually eliminates phase noise in your data. |
BUILT IN ACCURACY
With a simultaneous A/D converter, all signal inputs are sampled at the exact same instant in
time. A common clock and trigger for all channels lets you accurately correlate signals using a
single pulse of the clock to acquire all channels. The hardware architecture has accuracy built
in. A maximum aperture delay
of 35ns (the times it takes the
A/D on the module to switch
from track to hold mode) is
well matched at 5ns across all
six track-and-hold circuits, virtually
eliminating the channelto-
channel skew that is associated
with multiplexed inputs. A
maximum aperture uncertainty
of 1ns (the "jitter", or variance
in aperture delay) virtually
eliminates time skew between
channels and simplifies both
time and frequency- based
analysis techniques.
HIGHER ACCURACY AT
HIGH SPEED
Simultaneous A/Ds are far more accurate than multiplexed A/Ds, especially at high speed,
because they eliminate several sources of error, including settling time, and channel-to-channel
crosstalk, explained below.
SETTLING TIME
Not only is the channel-to-channel skew virtually eliminated with the simultaneous A/D architecture,
but so is the time needed to discharge the built-up capacitance on each input of a multiplexed
A/D.
Figure 5 and 6 show an equivalent RC circuit for a multiplexed A/D versus a simultaneous A/D.
As you can see, in a multiplexed architecture, each channel is tied to the same A/D. A minimum
settling time is required for the switched voltage to reach the actual input signal level. If you do
not factor the setting time into your sampling rate, some portion of the signal from the previous
channel can "cross over" to the next channel (especially when using high source impedance),
and generate erroneous results. This effect is most serious when the signal on the previous
channel is much larger than the signal on the present channel, and when you try to sample at
high speeds.
Because simultaneous architectures have a separate A/D for each input, settling time between
channels is not an issue. That means you can acquire data at very high speeds virtually error free!
SOURCE IMPEDANCE AND SETTLING TIME
Let's look at the multiplexed A/D a little closer. Source impedance (Rs) and capacitance (C),
when multiplied together, determine one time constant (t).
Assume that you have a 10 V input on channel 0 and a 0 V input on channel 1. Also assume
that the input impedance on channel 1 is 1 kOhm (Rs = 1 kOhm). Since it typically takes 9 time
constants for the multiplexer to settle to within 0.01% accuracy of the voltage you want, the
time that it takes to settle when switching from 10 V on channel 0 to 0 V on channel 1 is 9 * 100
ns (1kOhm * 100pf), or 900 ns.
Now, assume that you have a source impedance of 10 kOhm (Rs = 10 kOhm) on channel 1. If
the current is 100 pf, the settling time for one channel is 9 * 1 us (10kOhm * 100pf) or 9 µs - a
large source of error if you're sampling too fast!
Simultaneous A/Ds eliminate switching between inputs and, therefore, eliminate the settling
time problems associated with high source impedance, allowing you to measure highly accurate
data at high speeds.
CHANNEL-TO-CHANNEL CROSSTALK
Another problem with multiplexed systems which is eliminated with simultaneous A/Ds is channel-
to-channel crosstalk. Crosstalk occurs when the signals on one or more multiplexed channels
couple, or interfere, with the signal on the channel that is being measured. Crosstalk is
inherent to the multiplexing process, and gets worse as you increase the channel count and/or
the signal frequency.
Crosstalk occurs because parasitic capacitance across each open switch couples a portion of
each channel signal going to the output, distorting the desired signal. Figure 8 shows an example
of channel crosstalk, where the 5 pf capacitor can cause crosstalk between channels. In this
example, if the switch for one channel is on and the switches for the other 15 channels are off,
the crosstalk is 75 pf (5 pf x 15 channels).
ADDING OP AMPS TO YOUR SIGNAL CONDITIONING
If you are using a multiplexed A/D, you also need to be aware of problems that can occur when
adding operational amplifiers (op amps) to your signal conditioning circuitry. Slow-speed op
amps have long settling times, which as described previously, can introduce errors into your
measurement. High-speed op amps, on the other hand, will ring when they are hit with the
switch transients (100 pf capacitor) from the multiplexer because it takes time for the voltage to
settle, as shown in Figure 9.
LESS APERTURE JITTER
Aperture jitter (uncertainty) can be measured by simultaneously inputting a +/- 10 volt sinusoidal
signal at a specified frequency across all A/D channels. The sampled data is then analyzed
to characterize the A/D converter's ability to sample all channels at precisely the same
moment.
Since a sinusoidal signal's rate-of-change is greatest at the zero-crossings (locations on the signal where negative-to-positive voltage transitions occur), aperture jitter is most effectively measured there.
A sinusoid is determined by the following equation:
V(t) = pSin(2πft)
where,
p = peak voltage of sine wave
f = frequency of sine wave
V = voltage
t = time (in seconds)
dV
dt = 2πfp
In order to find the voltage change near the zero-crossing, take the derivative of the above
equation evaluated at t = 0.
Using this equation with an aperture uncertainty of 1 ns, a 10 kHz signal should yield a voltage
change near the origin of 628 uV or ~2 bits of error for a 16-bit A/D converter.
Data Translation offers a full line of simultaneous acquisition boards for USB 2.0.
These products include:
DT9836 & DT9832 Series - Simultaneous,
high performance, isolated analog inputs
provide throughput rates up to of 2.0MHz
per channel. Analog outputs, digital I/O,
quadrature decoders, and counter/timers
included. |
 |
DT9816 & DT9816-A - Low Cost
Simultaneous A/D Module. Part of the
ECONseries of low-cost, multi-function,
non-isolatedUSB modules. |
DT9840 Series - Real-time simultaneous
A/D and D/A DSP multi-function USB module
enclosed in a Sleek Box. Uses the Texas
Instruments TMS320C6713 DSP chip. |
Products & Services
