|
|
June 4, 1998
Video Switches route analog signals along paths of least resistance
Bill Schweber, Technical Editor
Analog video is very much alive and well despite the advance of digital video. By
properly using wideband, low-distortion switches and drivers, you can guide signals to
their intended destinations. Various switching configurations provide flexibility in
routing as well as path control.
There's little doubt that video is going digital. But this trend doesn't mean that
analog-video switching is an anachronism. Whenever you need to select, multiplex, or route
video signals near their source, such as a nondigital camera, or near their endpoint, such
as a display CRT, you likely are dealing with analog signals. Although you could use A/D
and D/A converters to force a transition from analog to digital mode or vice versa, such
converters are relatively expensive and power-hungry, especially if you don't otherwise
need that digital format.
In addition, there's the legacy factor. You cannot practically or economically switch
to digital more than 50 years of conventional analog TV, many thousands of installed
studios and control boards, and uncountable lengths of videotape supporting analog format.
Low-cost applications, such as security monitors, do not need the more expensive digital
video. Even in the high-end new world of digital-TV systems, analog signals still have
their place. The path between the set-top box output and the display is analog, as are one
or more of the inputs to the box in its role as multimedia-control center. For example, you need the box and its switching selector, along with a composite
NTSC, PAL, or SECAM signal, to handle the three links of the video-plus-stereo audio
output of a Nintendo 64. The system may even have a mix of composite video and
computer-friendly RGB signals.
Another technical reason for a system to use analog-video-signal
switching and routing is that such circuits are easier to build, troubleshoot, and
diagnose than digital ones. They operate transparently to the signal as long as it falls
within certain boundaries of bandwidth and magnitude without concern for data formats;
protocols; bit rates; or jitter, clocking, and synchronization problems. These signals
also can accept audio as well as video. Using virtually any convenient signal source and a
monitor, you can more easily verify their operation than you can for a digital switch.
Even better, you can use these switches in nonvideo applications, such as communications
or high-speed data-acquisition systems. These analog switches can also control and route
digital signals as well as analog ones, so one design provides tremendous flexibility.
Although terms such as "switch," "multiplexer,"
"router," and "crosspoint" refer to the general class of components
and subsystems that control signal path and flow, they each have specific implications. A
switch is the basic on/off signal path, whereas a multiplexer selects one of N inputs and
passes that input on to an output. A router is the complement of the multiplexer, taking
one input and sending it to a designated output or outputs. A crosspoint, array, or matrix
has multiple inputs and outputs and allows you to set up multiple input-to-output paths or
to direct one input to more than one output (Figure
1). These crosspoints usually provide an M×N switching function, where M is
the number of inputs and N is the number of outputs; these crosspoints need not be but
often are square.
People often use these terms somewhat imprecisely, and it's not unusual
to hear someone refer to a video multiplexer when they mean a crosspoint. Some of this
confusion comes from the architectural fact that the more complex functions build on the
simpler ones. Switches are the building blocks of multiplexers and routers, whereas
multiplexers and routers combine to develop the crosspoint function.
High-end projection systems, 3-D video headsets, and larger
videoconferencing systems use small switching matrices. They also have application in
RF/IF routing and switching, especially for base stations, and medical-imaging systems
employing dozens or hundreds of distinct transducers. Hotels and apartments use larger
crosspoint switches for sending video signals to different locations or in aircraft for
routing sound--and now even video--to seats.
Besides deciding on your basic architectural configuration, you need to
consider how to control the switching paths. If you have only a few switches, you can
control them directly from digital lines. However, as the switching configuration grows,
so does the number of control lines. Soon, you'll want to multiplex these control lines
and make them addressable. However, some ICs include built-in address-line decoders, or
serial interfaces that significantly reduce the number of control lines you need.
The route of some difficulty
Consumer products such as set-top boxes use highly integrated video-
switching components, providing a carefully defined set of functions and minimizing cost.
In contrast, studio and professional video installations opt for building-block
flexibility and somewhat better performance. Although the specifications for consumer
applications are somewhat less stringent than those for studios, the two applications are
beginning to converge in maintaining image quality.
Switches for analog signals are traditionally defined by such parameters
as 3-dB bandwidth, on-resistance, off-resistance, and leakage current. Unfortunately,
it's tricky to relate these factors to video-signal performance. The increased widespread
use of analog-video switches in mass-market applications has driven vendors to add
specifications that relate directly to image quality. These specifications include
differential gain and phase error, bandwidth flatness (usually defined as 0.1-dB
roll-off frequency), dynamic range, channel-to-channel crosstalk at specified frequencies,
and SNR.
As a result of IC-process improvements and clever IC designs that reduce
the effects of internal parasitics, video-switching components are now achieving
performance that was impossible to obtain with ICs just a few years ago. It's now
practical to get devices with 100 to 200 MHz, 3-dB bandwidth and flatness to 50 MHz.
Designers used to consider 0.1% and 0.1º, respectively, to be good differential gain and
phase error, but these figures are now five to 10 times better. Meanwhile, crosstalk at 5
MHz is now less than 60 dB. If you need even better crosstalk performance, you can use
"T"-configuration switches but at a cost (see sidebar "Configuration suits isolation to a T").
Depending on your application, switching speed and switching glitches
that occur when you change the video path may be concerns. If you are doing pixel-by-pixel
switching, such as in a studio "blue-screen" setup, such slow switching and
large glitches will cause visible smearing and artifacts. These parameters, however, are
less a concern if your application involves switching under manual control between sources
or switching during the retrace time between full frames. Too much switching glitch,
though, can cause false sync triggering, so be careful here.
Even if you don't at first glance think you need these superior
specifications in components, you may find that your overall system needs them. Video
signals typically travel through a multistage signal-processing chain, and errors
accumulate at each stage along the way. Your error budget will
likely show that you need individual component performance that is 10 times better than
final system-performance goals. This better performance is a measure of insurance for your
design and also permits easier worst-case analysis. It also gives you some psychological
margin because video signals--somewhat like their audio counterparts--often have
hard-to-correlate relationships between their hard technical specifications and their
perceived image quality and deficiencies.
Don't forget the multichannel aspects of switching, either. Unless you
are handling just one composite-video channel, you have to look at channel-to-channel
matching and performance. A high-quality video monitor could have four signals, for R, G,
B, plus sync, for example, and some vendors offer four-channel switches for this
requirement; one example is the $4.40 (1000) Maxim MAX498 (Figure
2). On-resistance values that are closely matched (to within 1 or 2ohm) and flat RON versus frequency determine how well the various
channels or RGB color components track each other.
Although there's a strong trend in analog components toward lower
quiescent and operational power, as well as toward single-supply operation, video-related
designs are not on the cutting edge of this trend. There are good reasons for this
scenario. First, relatively few video-switching designs and their associated systems are
battery-powered. Thus, although power consumption is always a designer's concern because
of dissipation, it's not a problem because of battery life. Thus, whereas nearly all
circuits benefit from lower power operation, saving a few milliwatts is not a top priority
in these designs.
Further, the video world is not a high-impedance, low-current
environment. The signal lines in video systems usually have 75ohm impedance. Some, derived from RF applications, may have 50ohm lines. Thus, signal current levels are approximately 1 mA. This figure is
comparable with, and often much higher than, the component's quiescent-current-consumption
values. A switch's dissipation of a few microwatts doesn't adversely affect your thermal
or supply budget.
The video world
The single-supply trend is also less
pronounced in the video world. Because most systems are line-powered, it's easier to
provide bipolar supplies than it is with a battery-powered system. More important,
single-supply line drivers and related analog components with low distortion at high speed
are less available and more costly than bipolar-supply devices (see sidebar "Do you know how to drive?"). However, this
aversion to single-supply designs is changing as better single-supply op amps with
adequate specifications and good rail-to-rail performance appear on the market.
In addition, although 5V designs have largely supplanted 12V systems,
designers are reluctant to go to 3V for video-analog-signal processing. This lowest
voltage value aggravates your SNR challenge and makes it harder for components to provide
the head room and current sourcing and sinking that fast-slewing video circuits thrive on.
Depending on your application, you may need only a relatively small
switch or multiplexer, or you may choose to build a matrix of 4×4 or smaller. For
example, the $1.50 (1000) Pericom PI5V330 is a quadruple two-channel
multiplexer/demultiplexer with 200-MHz, 3-dB bandwidth. One digital line simultaneously
controls the device's four switch pairs. You can use the digital line to select one of two
sources or route a signal to one of two paths. Because this IC is a switch without
integral line drivers, it is not constrained to one direction of signal flow; thus, you
can use this same IC in various multiplexer/demultiplexer parts of your video-switching
system.
If you need multiplexing only or building blocks for larger
configurations, the $4.80 (1000) LT1204 four-input video multiplexer from Linear
Technology Corp can directly drive 75ohm cables using its
internal 75-MHz current-feedback amplifier. With 0.1-dB gain, flatness to 75 MHz, and
channel-to-channel switching time of 120 nsec plus 40 mV/50-nsec switching transients,
this device is suitable for studio monitors. You can effectively combine multiple LT1204s
into larger routing matrices, because it contains a circuit that bootstraps the feedback
resistors when it is in disable mode, thus raising the impedance of the circuit and
minimizing cable mistermination.
National's $14.40 (100) CLC533 is a 4-to-1 multiplexer with active input
and output stages. The company designed the device for infrared and CCD imaging systems
and communications receiver I/Q processors, but its specifications are commensurate with
video-switching needs. It features 3-dB bandwidth of 110 MHz (at a gain of 2), 80-dB
isolation at 10 MHz, and distortion lower than 80 dB at 5 MHz; settling time to 0.01% is
24 nsec.
As the size of the multiplexer or array you need increases, one option
you have is to combine smaller multiplexers into larger configurations to produce a custom
version. This approach allows you to tailor the final product to your needs. It also lets
you maximize isolation between channels, because the ICs you employ have higher isolation
between them than the isolation available between channels on a single IC.
However, the build-your-own ap-proach has its weaknesses, beyond the
simple fact that building a custom version is more work
for you to design, evaluate, and manufacture. You'll probably need to add a multiplexer or
serial input register to simplify control of the various array switches and reduce the
number of control-interface lines you need. Without careful buffering between stages, the
many switch-input capacitances can sum to produce a relatively large capacitive load on a
channel. This load affects switch distortion and drive requirements and may even cause
oscillation.
Your alternative is to look at standard product offerings that meet a
larger part--or even all--of your configuration requirements or that are designed as
building blocks. Maxim's $1.70 (1000) MAX4111 family, for example, comprises a series of
1×1 through 4×1 multiplexers. The MAX4141 constituent has 330-MHz, 3-dB bandwidth,
with 0.1-dB gain flatness to 150 MHz (Figure 3).
Despite its 4×1 size, it targets use in larger arrays, because it includes a constant
high input impe-dance and a disable function that puts the output into a high-impedance
state; its open-loop design prevents oscillation that may occur from capacitive loads. As
an extra benefit, its switching glitch, already 13 mV, is a positive voltage. This fact
makes it less likely that the rest of the system will confuse that voltage with standard
negative-going video sync pulses.
You're not limited to N×1 devices, either. In addition to basic
multiplexers, Harris Semiconductor offers the $18.99 (1000) HA456, an 8×8 video
crosspoint switch with fully buffered inputs and outputs that allows you to connect video
signals to any or all of its outputs. The 3-dB bandwidth of this device is 120 MHz, and
crosstalk at 10 MHz is 55 dB. The same device incorporates both serial and parallel
control via 3 bits. The unity-gain output buffers of each channel slew at 200V/µsec and
can drive 400ohm, 5-pF loads to ±2V.
If your needs extend beyond 8×8 arrays, you don't necessarily have to
combine smaller arrays to achieve this size. Analog Devices offers the $90 (100) AD8116
16×16 array, an IC with an internal 256-point matrix that replaces eight 8×4 or 64 4×1
devices (Figure 4). The 128-lead IC has a 200-MHz,
3-dB bandwidth and can drive 150ohm loads with 0.01%/0.018
differential gain/phase error. You control this array via an 80-bit serial word, and the
control port lets you daisy-chain the array to others and create arrays as large as
256×256 input/output.
Specialized markets
Beyond general-purpose multiplexers and arrays of various dimensions,
the mass-market needs of consumer applications such as set-top boxes require cost- and
architecture-optimized video-switch configurations, and some vendors recognize this fact.
Vishay-Siliconix (formerly, Temic Semiconductors), for example, offers the $7.82 (1000)
DG884 8×4 crosspoint array but also provides the DG894 component video selector. You can
use the serial I2 bus or direct control lines to control this 200-MHz IC, which
lets you switch a cluster of signals from one of several sources to a single output group.
In a typical configuration, these signals can be RGB video, plus video sync, plus left and
right audio (Figure 5). You can also switch Y
(luminance) and C (chrominance) video signals or S-VHS, in addition to the basic RGB
signals.
Similarly, SGS-Thomson offers both matrix and application-specific
devices. The TEA6415C 8×6 matrix is unusual in that it offers 6-dB gain and I2
bus control. However, the company's $5 (1000) STV6410 audio/video switch matrix is more
complex and focused; Figure 6 shows a typical
application. This 15-MHz, 64-pin IC, also I2-bus-controlled, has a 5×4
composite video array, a 5×3 Y and C array, a 2×1 RGB array, video muting, and
video-sync bottom clamping. For the audio portion of a user system, this IC switches left
and right stereo with a 5×4 array with mutable outputs (and two of these paths have level
adjustment) and combines left and right into a mono sound output.
Don't worry if you are facing the prospect of routing analog's nemesis,
namely digital video and its high bit rates. Vendors such as Gennum Corp (www.gennum.com) offer ICs and chip sets that integrate
the unique needs of standards-compatible serial digital video at 540 Mbps. For example,
the company's $18 (100) GX9533 8×9 crosspoint switch forms the core of a larger routing
array. Supporting this array are the $12.49 (100) GS9024 equalizer, which compensates for
the distortion effects of as much as 350m of cable; the $21.51 (100) GS9035 video
reclocker, which tolerates high input jitter and yields a low-jitter output clock; and the
$5.15 (100) GS9028 digital-video cable driver with two isolated 540-Mbps outputs.
Don't neglect basics
You can use the best components available, but always remember you're
dealing with fast analog signals in a physically tight environment and with relatively
large switching loads. Power-supply bypassing right at the video switch or multiplexer is
essential to getting the performance potential of these components and to maintaining
signal integrity. Terminate all lines with the appropriate impedance, and don't leave
unused inputs or outputs unterminated. Use vendor evaluation and reference designs to the
extent you can, and ask if CAD files are available.
Crosstalk is among the nastiest parameters to control. Even if the
component itself has low crosstalk, nearby digital signals, shared voltage references,
ground noise, and other wideband signals can affect video lines. Don't even think of using
wire-wrapped boards or breadboards in your prototype; their capacitance gives severe
crosstalk. Similarly, sockets add parasitic capacitance and inductance, which causes
crosstalk, ringing, and even oscillation.
You may need to use microstrip or stripline layout techniques to
maintain line impedance on the circuit board. Surface-mount components are preferable to
through-hole components and boards. You'll need a multilayer pc board with a large,
low-impedance ground plane. Keep all pc-board traces short and straight between the signal
source and destination; when you have to make a turn, use a rounded corner instead of a
sharp 90º turn.
Be prepared to invest in suitable video-test equipment for
"big-picture" signal analysis if your application demands high performance. In
addition to an oscilloscope, you need the multichannel equipment for measuring crosstalk,
as well as specialized units, such as video-measurement sets that analyze image quality in
the time, frequency, and color-vector domains. Otherwise, you have to fall back on the
subjective opinions of anyone who sees the image on a screen, and you know there's no
accounting for taste when it comes to image quality. The dispute is analogous to those
over audio quality with the unresolvable perspectives of "golden ears" versus
instrumentation results and discrepancies associated with the numbers versus the perceived
sonic colorations.
Acknowledgments
Thanks to Will Drachler and Scott Pavlik of Analog Devices Inc; Dave
Fullagar and Ron Clark of Maxim Integrated Products; and Garth Powell, Amir
Sheikholeslami, and Christophe Prugne of SGS-Thomson Microelectronics for their insight
and perspectives.
- Travis, Bill, "Take account of errors in designs
using analog switches and multiplexers," EDN, Jan 4, 1996, pg 61.
- Schweber, Bill, "Line drivers and receivers push
signals through cable's reality," EDN, Aug 1, 1996, pg 44.
Do you know how to
drive?
A designer's life would be easier if video signals required just
switching and routing. Unfortunately, they need much more than that. Video drivers have
the task of placing a single-ended or differential video signal on 50 or 75ohm lines for transmission of distances ranging from a few centimeters to many
meters. You must match the driver's impedance to the line to minimize reflections of
high-speed video signals. Many switches include the driver, simplifying your design and
reducing overall component count. Note, though, that these buffered switches can handle
only signals that flow in one direction, whereas an unbuffered switch can usually accept
bidirectional signals and has no designated input or output port.
In a representative configuration, the line driver performs its task
operating at a gain of 2, to compensate for voltage division of the video signal between
the line's matching impedance and its termination impedance (Figure
Aa). Because video drivers often require this gain value, some vendors offer such
drivers with a fixed gain or with internal resistors that you can select to provide this
gain value, thus eliminating the need to add external, discrete gain-setting components.
(Incidentally, de-pending on who you talk to, the name for this standard matching-resistor
technique can be "back-matching," "reverse-termination," or even
"double-terminated.")
Depending on the noise situation, the distance between the source and
the receiver, and the quality of their common ground, you may also want to consider
implementing a fully differential signal path. This approach minimizes the effects of
common-mode noise but at the cost of twice as many amplifiers at the drive end (Figure Ab).
Although nearly every vendor of basic video switches and their larger
variations also offers line drivers, other linear-IC vendors, such as Burr-Brown (www.burr-brown.com), Elantec (www.elantec.com), Micrel (www.micrel.com), and Texas Instruments (www.ti.com), offer only the driver function, and you may find
the best video line driver for your situation with one of these nonswitch vendors. Some
vendors also offer specialized video driver configurations. For example, the $2 (1000)
SPT9400 from Signal Processing Technologies (www.spt.com)
is a triple video driver that accepts a Y (luminance) and a C (chrominance) signal and
provides corresponding buffered Y and C outs, plus a composite-video output derived from
the Y and C inputs, all in a 12-pin SSOP.
Configuration
suits isolation to a T
If the approximately 60-dB isolation of a conventional switch is
insufficient for your application, such as ultrasound and automated test equipment, don't
despair. Systems have long used the T-configuration switch, which even predates
solid-state switches, to provide the highest isolation in high-power transmitter/receiver
systems, such as radar. In this configuration, three internal switches are in the signal
path, although the input-output appearance of the switch is the same as for the non-T
configuration (Figure A). A well-designed T-switch is as
close to a mechanical reed switch as you can get but obviously much faster.
Maxim's MAX4545 series, for example, has switches with 300-MHz
bandwidth, 80-dB crosstalk, and 10-MHz off-isolation, all in a 50ohm load; insertion loss is 1 dB at 100 MHz. You can get this family with four spst
switches, dual spst switches (one NO, one NC), or two spdt configurations in a single
package. Prices start at $1.80 (1000). Because the T-configuration requires three times as
many switches internally per function, it consumes more die area and is thus more costly;
fortunately, T-switches such as the Maxim device are pin-compatible with conventional,
non-T switches, so you can choose what you need with minimum layout dislocation.
|
