|
Using USB as a
Data Acquisition Interface
by Nick Wright and Bob Judd,
Measurement Computing
Learn about the similarities and differences of USB 1.1 and USB 2.0
and how they impact system performance.
 The Universal Serial Bus (USB) has become the de facto low-cost PC
interface standard for consumer applications. But only very recently has
it become a popular interface for data acquisition applications. As the
new kid on the block, many data acquisition buyers are unfamiliar with
the capabilities, advantages, disadvantages, and terminology required to
fairly evaluate USB as a data acquisition interface.
USB 1.1 vs. USB 2.0
USB 2.0 is the latest version of the USB specification. To maintain
backward compatibility with existing devices, the USB 2.0 operating
specification is a superset of the USB 1.1 specification. USB 2.0 does
everything that USB 1.1 does and is completely compatible with 1.1. In
addition it supports much faster transfers. The maximum data transfer
rate in 2.0 is 480 Mb/s, 40 times faster than the highest speed
supported in 1.1.
Data Transfer Modes
The USB can be used with a large number of applications. To do so,
developers of the USB have created four different data transfer modes:
control, interrupt, bulk, and isochronous. Each of these has its
advantages and disadvantages, and the application typically will
determine which of the modes a USB device uses.
Control Mode
Control transfers are used for configuration and must be supported by
all devices. Data acquisition devices requiring very limited data
transfer capability may be configured to use only the control mode
although most will take advantage of other data transfer modes.
Interrupt Mode
Interrupt transfers typically are used when data must be transferred
within a specific amount of time, such as in many data acquisition
applications or in a keyboard or mouse interface. Each USB port provides
multiple interrupt transfer pipes. The guaranteed timing makes the
interrupt mode ideal for use in most data acquisition applications.
A single data acquisition device can use multiple interrupt transfers to
provide higher data transfer rates. Since the number of interrupt pipes
and overall bus bandwidth is limited, using multiple interrupts reduces
the number of independent devices that may be attached to a particular
USB port.
Bulk Mode
Bulk transfers typically are used when the transfer rate is not
critical, such as writing to a printer. Bulk transfers will use
bandwidth that currently is not allocated to other transfer types, so
they will not slow down critical operations.
However, if the bus is busy, bulk transfers get the lowest priority. As
most data acquisition applications require data transfers to be made on
a periodic basis and cannot allow a loss of data, the bulk mode is not
recommended for most data acquisition applications.
Isochronous Mode
Isochronous transfers are streaming transfers used for constant rate
data, typically audio or video. There is no facility for retransmitting
data that has been received with errors, so it is not appropriate for
data that must be accurate. This limits the utility of the isochronous
mode in most data acquisition applications.
Overall, the interrupt mode provides the best fit for most data
acquisition applications. The isochronous mode should be ruled out for
most applications due to the lack of data integrity it offers.
The bulk and, to a lesser extent, the control modes could be used to
implement many data acquisition devices. But since they do not offer
guaranteed data transfer rates, there is a substantial risk of data loss
if the data acquisition device does not provide very large data buffers.
As you investigate data acquisition devices from various vendors, ask
which transfer mode is used. If the device uses a mode other than
interrupt, you should inquire about possible system limitations if you
plan to connect more than one device to a given USB port.
Data Transfer Rates
The current USB specification accommodates systems running at three
different data transfer rates. Both USB 2.0 and 1.1 support data
transfers referred to as low speed and full speed.
In addition to these rates, 2.0 supports a bus speed referred to as high
speed. Table 1 depicts the theoretical maximum transfer rate as well as
other specifications related to low-, full- and high-speed transfers.
Table 1. USB 2.0 Theoretical Maximum Transfer Specs
These are theoretical maximum limits that
will not necessarily be achieved due to various processing and control
overhead requirements.
Based on overhead and processing delay issues, these transfer rates are
not achievable in practice. Also remember that most data acquisition
applications are based on 12-b or 16-b words rather than bytes, so these
numbers must be scaled prior to conversion from bytes per second to
samples per second.
Low Speed
Low-speed transfers use a bus transfer rate of 1.5 Mb/s. This is the
actual rate at which bits are transferred over the bus and does not
include the overhead required for control and error checking. The actual
data transfer rates possible in a low-speed system will be considerably
less than 1.5 Mb/s.
Low speed also limits the modes of transfers allowed to the interrupt
and control modes. Data packets can only be 8 B maximum. The maximum
latency for interrupt transfers can be as low as 10 ms. A maximum of 24
B/ms can be transferred with a control endpoint and 8 B/10 ms with an
interrupt endpoint.
As a result, devices using low-speed transfers typically are not used
for data acquisition applications requiring more than a few kilosamples
per second. Many of the currently available USB-based data acquisition
products including Measurement Computing’s PMD-1208LS and the QUANCOM®
USBOPTOREL16 use low-speed transfers.
Full Speed
A full-speed transfer offers a maximum data transfer rate of 12 MB/s.
Once again, this is the actual speed that bits are transferred over the
bus, and actual usable data transfer rates will be less. All four
transfer modes are allowed although only control and interrupt are
supported by the stand-ard Windows human interface device (HID) driver.
Data packets can be 64 B maximum for the control, interrupt, and bulk
transfers and 1,023 B maximum for isochronous transfers. Maximum latency
for interrupt transfers can be as low as 1 ms. A maximum of 832 B/ms can
be transferred with a control endpoint, 1,216 B/ms with a bulk endpoint,
64 B/ms with an interrupt endpoint (up to 19 interrupt endpoint pipes
are available), and 1,023 B/ms with an isochronous endpoint.
The full-speed transfer mode implements data acquisition devices with
sample rates in the 100-kHz range based on interrupt transfers.
Measurement Computing’s PMD-1608FS and the softDSP SDS 200 use
full-speed data transfers. Faster devices could be developed based on
bulk or isochronous transfers, but the interrupt transfer method is the
most reliable way to ensure accurate data without risking data loss or
first in first out (FIFO) overruns.
High Speed
A high-speed transfer offers a maximum data transfer rate of 480 Mb/s
and is only supported by USB 2.0 ports. Like the low- and full-speed
specifications, this is the actual rate at which bits are transferred
over the bus, and with overhead and error checking, the attainable data
transfer rates will be less.
High speed supports all four transfer modes. Data packets can be 64 B
maximum for control transfers, 512 B maximum for bulk transfers, 1,024 B
maximum for interrupt transfers, and 1,024 B maximum for isochronous
transfers. Maximum latency for interrupt transfers can be as low as 125
µs.
A maximum of 15,872 B/ms can be transferred with a control endpoint,
53,248 B/ms with a bulk endpoint, 24,576 B/ms with an interrupt
endpoint, and 24,576 B/ms with an isochronous endpoint. USB-based data
acquisition products based on high-speed USB transfers could be expected
to support sample rates in the megahertz range.
The USB bus cannot keep up with the 80-MB/s transfer rate possible over
the PCI or CompactPCI buses. However, a high-speed USB should
accommodate the vast majority of today’s data acquisition applications.
The preferred method would be the interrupt mode because it can
guarantee timing of the acquisition. Currently, we are not aware of any
acquisition device that takes advantage of the capabilities of
high-speed data transfers although new products are released quite
frequently, and by the time you read this article, there may be.
Mixing Different-Speed Devices on a Single Port
If you are using one USB device or a single USB device per computer
port, there are no issues concerning mixing devices of different speeds.
If you have a USB 2.0 port, you simply plug in any USB device. The
system will recognize the device’s speed and communicate accordingly.
If you plug a high-speed 2.0 device into a 1.1 port, the device should
be recognized, installed, and actually work. However, you will not
necessarily achieve the sample rates that you would obtain from a 2.0
port.
To connect more than one USB device to a single port on your computer,
you must install a USB hub. Hubs are available in both 1.1 and 2.0
configurations. You may plug any combination of 1.1 and 2.0 devices into
either hub, and they should work.
You also may mix low-speed, full-speed, and high-speed devices on a
single hub without typically reducing the performance of the higher
speed devices. To take advantage of high-speed transfers, the device,
the hub, and the computer port all must be 2.0. The USB has been
designed to minimize, if not eliminate, the penalty for mixing older and
newer technology.
About the Authors
Nick Wright is a hardware design engineer at Measurement Computing. He
has been developing microcontroller-based hardware and firmware for more
than nine years, concentrating on communications buses and measurement
systems. Mr. Wright also has held engineering positions at Microchip
Technologies and Advanced Instruments. e-mail:
nick.wright@mccdaq.com
Bob Judd is vice president of marketing at Measurement Computing. He has
been involved in the test and measurement industries for more than 20
years, serving in engineering and marketing capacities at companies
including Keithley MetraByte, Datel, and Northrop. e-mail:
bob.judd@mccdaq.com
Measurement Computing, 16 Commerce Blvd., Middleboro, MA 02346,
508-946-5100.
FOR MORE INFORMATION
on Universal Serial Buses
www.rsleads.com/406ee-182
|