ONSEMI AN1406

AN1406/D
Designing with PECL
(ECL at +5.0 V)
The High Speed Solution for the
CMOS/TTL Designer
http://onsemi.com
Prepared by
Cleon Petty
Todd Pearson
ECL Applications Engineering
APPLICATION NOTE
This application note provides detailed information on designing with Positive Emitter Coupled Logic (PECL) devices.
Introduction
PECL, or Positive Emitter Coupled Logic, is nothing
more than standard ECL devices run off of a positive power
supply. Because ECL, and therefore PECL, has long been
the “black magic” of the logic world many misconceptions
and falsehoods have arisen concerning its use. However,
many system problems which are difficult to address with
TTL or CMOS technologies are ideally suited to the
strengths of ECL. By breaking through the wall of
misinformation concerning the use of ECL, the TTL and
CMOS designers can arm themselves with a powerful
weapon to attack the most difficult of high speed problems.
It has long been accepted that ECL devices provide the
ultimate in logic speed; it is equally well known that the
price for this speed is a greater need for attention to detail in
the design and layout of the system PC boards. Because this
requirement stems only from the speed performance aspect
of ECL devices, as the speed performance of any logic
technology increases these same requirements will hold. As
can be seen in Table 1 the current state–of–the–art TTL and
CMOS logic families have attained performance levels
which require controlled impedance interconnect for even
relatively short distances between source and load. As a
result system designers who are using state–of–the–art TTL
or CMOS logic are already forced to deal with the special
requirements of high speed logic; thus it is a relatively small
step to extend their thinking from a TTL and CMOS bias to
include ECL devices where their special characteristics will
simplify the design task.
System Advantages of ECL
The most obvious area to incorporate ECL into an
otherwise CMOS/TTL design would be for a subsystem
which requires very fast data or signal processing. Although
this is the most obvious it may also be the least common.
Because of the need for translation between ECL and
CMOS/TTL technologies the performance gain must be
greater than the overhead required to translate back and forth
between technologies. With typical delays of six to seven
nanoseconds for translating between technologies, a
significant portion of the logic would need to be realized
using ECL for the overall system performance to improve.
However, for very high speed subsystem requirements ECL
may very well provide the best system solution.
Transmission Line Driving
Many of the inherent features of an ECL device make it
ideal for driving long, controlled impedance lines. The low
impedance of the open emitter outputs and high input
impedance of any standard ECL device make it ideally
suited for driving controlled impedance lines. Although
designed to drive 50Ω lines an ECL device is equally adept
at driving lines of impedances of up to 130Ω without
significant changes in the AC characteristics of the device.
Although some of the newer CMOS/TTL families have the
ability to drive 50Ω lines many require special driver circuits
to supply the necessary currents to drive low impedance
transmission interconnect. In addition the large output
swings and relatively fast output slew rates of today’s high
performance CMOS/TTL devices exacerbate the problems
of crosstalk and EMI radiation. The problems of crosstalk
and EMI radiation, along with common mode noise and
signal amplitude losses, can be alleviated to a great degree
with the use of differential interconnect. Because of their
architectures, neither CMOS nor TTL devices are capable of
differential communication. The differential amplifier input
structure and complimentary outputs of ECL devices make
them perfectly suited for differential applications. As a
result, for systems requiring signal transmission between
Table 1. Relative Logic Speeds
Logic
Family
Typical Output
Rise/Fall
Maximum Open Line
Length (Lmax)*
10KH
1.0ns
3″
ECLinPS
400ps
1″
FAST
2.0ns
6″
FACT
1.5ns
4″
* Approximate for stripline interconnect (Lmax = Tr/2Tpd)
 Semiconductor Components Industries, LLC, 1999
September, 1999 – Rev. 2
1
Publication Order Number:
AN1406/D
AN1406/D
to increase the performance of their designs without having
to resort to more complicated architectures or costly, faster
logic. ECL logic has the capability of significantly reducing
the clock skew of a system over an equivalent design
utilizing CMOS or TTL technologies.
The skew introduced by a logic device can be broken up
into three areas; the part–to–part skew, the within–part skew
and the rise–to–fall skew. The part–to–part skew is defined
as the differences in propagation delays between any two
devices while the within–device skew is the difference
between the propagation delays of similar paths for a single
device. The final portion of the device skew is the
rise–to–fall skew or simply the differences in propagation
delay between a rising input and a failing input on the same
gate. The within–device skew and the rise–to–fall skew
combine with delay variations due to environmental
conditions and processing to comprise the part–to–part
skew. The part–to–part skew is defined by the propagation
delay window described in the device data sheets.
Careful attention to die layout and package choice will
minimize within–device skew. Although this minimization
is independent of technology, there are other characteristics
of ECL which will further reduce the skew of a device.
Unlike their CMOS/TTL counterparts, ECL devices are
relatively insensitive to variations in supply voltage and
temperature. Propagation delay variations with
environmental conditions must be accounted for in the
specification windows of a device. As a result because of
ECLs AC stability the delay windows for a device will
inherently be smaller than similar CMOS or TTL functions.
The virtues of differential interconnect in line driving
have already been addressed, however the benefits of
differential interconnect are even more pronounced in clock
distribution. The propagation delay of a signal through a
device is intimately tied to the switching threshold of that
device. Any deviations of the threshold from the center of
the input voltage swing will increase or decrease the delay
of the signal through the device. This difference will
manifest itself as rise–to–fall skew in the device. The
threshold levels for both CMOS and TTL devices are a
function of processing, layout, temperature and other factors
which are beyond the control of the system level designer.
Because of the variability of these switching references,
specification limits must be relaxed to guarantee acceptable
manufacturing yields. The level of relaxation of these
specifications increases with increasing logic depth. As the
depth of the logic within a device increases the input signal
will switch against an increasing number of reference levels;
each encounter will add skew when the reference level is not
perfectly centered. These relaxed timing windows add
directly to the overall system skew. Differential ECL, both
internal and external to the die, alleviates this threshold
sensitivity as a DC switching reference is no longer required.
Without the need for a switching reference the delay
windows, and thus system skew, can be significantly
several boards, across relatively large distances, ECL
devices provide the CMOS/TTL designer a means of
ensuring reliable transmission while minimizing EMI
radiation and crosstalk.
Figure 1 shows a typical application in which the long line
driving, high bandwidth capabilities of ECL can be utilized.
The majority of the data processing is done on wide bit width
words with a clock cycle commensurate with the bandwidth
capabilities of CMOS and TTL logic. The parallel data is
then serialized into a high bandwidth data stream, a
bandwidth which requires ECL technologies, for
transmission across a long line to another box or machine.
The signal is received differentially and converted back to
relatively low speed parallel data where it can be processed
further in CMOS/TTL logic. By taking advantage of the
bandwidth and line driving capabilities of ECL the system
minimizes the number of lines required for interconnecting
the subsystems without sacrificing the overall performance.
Furthermore by taking advantage of PECL this application
can be realized with a single five volt power supply. The
configuration of Figure 1 illustrates a situation where the
mixing of logic technologies can produce a design which
maximizes the overall performance while managing power
dissipation and minimizing cost.
ECL Serial
Data >200MHz
Serial/Parallel
Conversion
CMOS/TTL Parallel
Data <50MHz
Low Frequency
Information Processing
CMOS/TTL Parallel
Data <50MHz
Parallel/Serial
Conversion
ECL Serial
Data >200MHz
Figure 1. Typical Use of ECL’s High Bandwidth,
Line Driving Capabilities
Clock Distribution
Perhaps the most attractive area for ECL in CMOS/ TTL
designs is in clock distribution. The ever increasing
performance capabilities of today’s designs has placed an
even greater emphasis on the design of low skew clock
generation and distribution networks. Clock skew, the
difference in time between “simultaneous” clock transitions
throughout an entire system, is a major component of the
constraints which form the upper bound for the system clock
frequency. Reductions in system clock skew allow designers
http://onsemi.com
2
AN1406/D
design. After all the inclusion of ECL requires two
additional negative voltage supplies; VEE and the
terminating voltage VTT. Fortunately this is where the
advantages of PECL come into play. By using ECL devices
on a positive five volt CMOS/TTL power supply and using
specialized termination techniques ECL logic can be
incorporated into CMOS/TTL designs without the need for
additional power supplies. What about power dissipation
you ask, although it is true that in a DC state ECL will
typically dissipate more power than a CMOS/TTL
counterpart, in applications which operate continually at
frequency, i.e.. clock distribution, the disparity between
ECL and CMOS/TTL power dissipation is reduced. The
power dissipation of an ECL device remains constant with
frequency while the power of a CMOS/TTL device will
increase with frequency. As frequencies approach 50MHz
the difference between the power dissipation of a CMOS or
TTL gate and an ECL gate will be minimal. 50MHz clock
speeds are becoming fairly common in CMOS/TTL based
designs as today’s high performance MPUs are fast
approaching these speeds. In addition, because ECL output
swings are significantly less than those of CMOS and TTL
the power dissipated in the load will be significantly less
under continuous AC conditions.
It is clear that PECL can be a powerful design tool for
CMOS/TTL designers, but where can one get these PECL
devices. Perhaps the most confusing aspect of PECL is the
misconception that a PECL device is a special adaptation of
an ECL device. In reality every ECL device is also a PECL
device; there is nothing magical about the negative voltage
supply used for ECL devices. The only real requirement of
the power supplies is that the potential difference described
in the device data sheets appears across the upper and lower
power supply rails (VCC and VEE respectively). A potential
stumbling block arises in the specified VEE levels for the
various ECL families. The 10H and 100K families specify
parametric values for potential differences between VCC and
VEE of 4.94V to 5.46V and 4.2V to 4.8V respectively; this
poses a problem for the CMOS/TTL designer who works
with a typical VCC of 5.0V ±5%. However, because both of
these ECL standards are voltage compensated both families
will operate perfectly fine and meet all of the performance
specifications when operated on standard CMOS/TTL
power supplies. In fact, Motorola is extending the VEE
specification ranges of many of their ECL families to be
compatible with standard CMOS/TTL power supplies.
Unfortunately earlier ECL families such as MECL 10K
are not voltage compensated and therefore any reduction in
the potential difference between the two supplies will result
in an increase in the VOL level, and thus a decreased noise
margin. For the typical CMOS/TTL power supplies a 10K
device will experience an ≈50mV increase in the VOL level.
Designers should analyze whether this loss of noise margin
could jeopardize their designs before implementing PECL
formatted 10K using 5.0V ± 5% power supplies.
reduced while maintaining acceptable manufacturing
yields.
What does this mean to the CMOS/TTL designer? It
means that CMOS/TTL designers can build their clock
generation card and backplane clock distribution using
ECL. Designers will not only realize the benefits of driving
long lines with ECL but will also be able to realize clock
distribution networks with skew specs unheard of in the
CMOS/TTL world. Many specialized functions for clock
distribution are available from Motorola (MC10/100E111,
MC10/100E211, MC10/100EL11). Care must be taken that
all of the skew gained using ECL for clock distribution is not
lost in the process of translating into CMOS/TTL levels. To
alleviate this problem the MC10/100H646 can be used to
translate and fanout a differential ECL input signal into TTL
levels. In this way all of the fanout on the backplane can be
done in ECL while the fanout on each card can be done in the
CMOS/TTL levels necessary to drive the logic.
Figure 2 illustrates the use of specialized fanout buffers to
design a CMOS/TTL clock distribution network with
minimal skew. With 50ps output–to–output skew of the
MC10/100E111 and 1ns part–to–part skew available on the
MC10/100H646 or MC10/100H641, a total of 72 or 81 TTL
clocks, respectively, can be generated with a worst case
skew between all outputs of only 1.05ns. A similar
distribution tree using octal CMOS or TTL buffers would
result in worst case skews of more than 6ns. This 5ns
improvement in skew equates to about 50% of the up/down
time of a 50MHz clock cycle. It is not difficult to imagine
situations where an extra 50% of time to perform necessary
operations would be either beneficial or even a life saver. For
more information about using ECL for clock distribution,
refer to application note AN1405/D – ECL Clock
Distribution Techniques.
Part–Part
Skew = 1ns
Output–Output
Skew = 50ps
E111
Differential ECL
Differential
ECL Input
H641
TTL
Outputs
1 of 9
H641
TTL
Outputs
9 of 9
Figure 2. Low Skew Clock Fanout Tree
PECL versus ECL
Nobody will argue that the benefits presented thus far are
not attractive, however the argument will be made that the
benefits are not enough to justify the requirements of
including ECL devices in a predominantly CMOS/TTL
http://onsemi.com
3
AN1406/D
for both the 10H and 100K ECL standards. As mentioned
earlier any changes in VCC will show up 1:1 on the output
DC levels. Therefore any tolerance values for VCC can be
transferred to the device I/0 levels by simply adding or
subtracting the VCC tolerance values from those values
provided in Table 2.
The traditional choice of a negative power supply for ECL
is the result of the upper supply rail being used as the
reference for the I/0 and internal switching bias levels of the
technology. Since these critical parameters are referenced to
the upper rail any noise on this rail will couple 1:1 onto them;
the result will be reduced noise margins in the design.
Because, in general, it is a simpler task to keep a ground rail
relatively noise free, it is beneficial to use the ground rail as
this reference. However when careful attention is paid to the
power supply design, PECL can be used to optimize system
performance. Once again the use of differential PECL will
simplify the designer’s task as the noise margins of the
system will be doubled and any noise riding on the upper
VCC rail will appear as common mode noise; common mode
noise will be rejected by the differential receiver.
PECL Termination Schemes
PECL outputs can be terminated in all of the same ways
standard ECL, this would be expected since an ECL and a
PECL device are one in the same. Figure 3 illustrates the
various output termination schemes utilized in typical ECL
systems. For best performance the open line technique in
Figure 3 would not be used except for very short
interconnect between devices; the definition of short can be
found in the various design guides for the different ECL
families. In general for the fastest performance and the
ability to drive distributive loads the parallel termination
techniques are the best choice. However occasions may arise
where a long uncontrolled or variable impedance line may
need to be driven; in this case the series termination
technique would be appropriate. For a more thorough
discourse on when and where to use the various termination
techniques the reader is referred to the MECL System
Design Handbook (HB205/D) and the design guide in the
ECLinPS Databook (DL140/D). The parallel termination
scheme of Figure 3 requires an extra VTT power supply for
the impedance matching load resistor. In a system which is
built mainly in CMOS/TTL this extra power supply
requirement may prohibit the use of this technique. The
other schemes of Figure 3 use only the existing positive
supply and ground and thus may be more attractive for the
CMOS/ TTL based machine.
MECL to PECL DC Level Conversion
Although using ECL on positive power supplies is
feasible, as with any high speed design there are areas in
which special attention should be placed. When using ECL
devices with positive supplies the input output voltage levels
need to be translated. This translation is a relatively simple
task. Since these levels are referenced off of the most
positive rail, VCC, the following equation can be used to
calculate the various specified DC levels for a PECL device:
PECL Level = VCCNEW – |Specification Level|
As an example, the VOHMAX level for a 10H device
operating with a VCC of 5.0V at 25°C would be as follows:
PECL Level = 5.0V – |–0.81V|
PECL Level = (5.0 – 0.81)V = 4.19V
The same procedure can be followed to calculate all of the
DC levels, including VBB for any ECL device. Table 2, on
page 4, outlines the various PECL levels for a VCC of 5.0V
Table 2. ECL/PECL DC Level Conversion for VCC = 5.0V
10E Characteristics
0°C
Symbol
100E Characteristics
25°C
85°C
0 to 85°C
Min
Max
Min
Max
Min
Max
Min
Max
Unit
VOH
–1.02/3.98
–0.84/4.16
–0.98/4.02
–0.81/4.19
–0.92/4.08
–0.735/4.265
–1.025/3.975
–0.880/4.120
V
VOL
–1.95/3.05
–1.63/3.37
–1.95/3.05
–1.63/3.37
–1.95/3.05
–1.600/3.400
–1.810/3.190
–1.620/3.380
V
VOHA
—
—
—
—
—
—
—
–1.610/3.390
V
VOLA
—
—
—
—
—
—
–1.035/3.965
—
V
VIH
–1.17/3.83
–0.84/4.16
–1.13/3.87
–0.81/4.19
–1.07/3.93
–0.735/4.265
–1.165/3.835
–0.880/4.120
V
VIL
–1.95/3.05
–1.48/3.52
–1.95/3.05
–1.48/3.52
–1.95/3.05
–1.450/3.550
–1.810/3.190
–1.475/3.525
V
VBB
–1.38/3.62
–1.27/3.73
–1.35/3.65
–1.25/3.75
–1.31/3.69
–1.190/3.810
–1.380/3.620
–1.260/3.740
V
http://onsemi.com
4
AN1406/D
consumption. In addition, this extra power is consumed
entirely in the external resistors and thus will not affect the
reliability of the IC. As is the case with standard parallel
termination, the tolerances of the VTT and VCC supplies
should be addressed in the design phase. The following
equations provide a means of determining the two resistor
values and the resulting equivalent VTT terminating voltage.
Parallel Termination Schemes
Because the techniques using an extra VTT power supply
consume significantly less power, as the number of PECL
devices incorporated in the design increases the more
attractive the VTT supply termination scheme becomes.
Typically ECL is specified driving 50Ω into a –2.0V,
therefore for PECL with a VCC supply different than ground
the VTT terminating voltage will be VCC – 2.0V. Ideally the
VTT supply would track 1:1 with VCC, however in theory
this scenario is highly unlikely. To ensure proper operation
of a PECL device within the system the tolerances of the VTT
and the VCC supplies should be considered. Assume for
instance that the nominal case is for a 50Ω load (Rt) into a
+3.0V supply; for a 10H compatible device with a VOHmax
of –0.81V and a realistic VOLmin of –1 .85V the following
can be derived:
R1 = R2 ({VCC – VTT}/{VTT – VEE})
R2 = ZO ({VCC – VEE}/{VCC – VTT})
VTT = VCC (R2/{R1 + R2})
For the typical setup:
VCC = 5.0V; VEE = GND; VTT = 3.0V; and ZO = 50Ω
R2 = 50 ({5 – 0}/{5–3}) = 125Ω
R1 = 125 ({5–3}/{3–0}) = 83.3Ω
checking for VTT
IOHmax = (VOHmax – VTT)/Rt
IOHmax = ({5.0 – 0.81} – 3.0)/50 = 23.8mA
IOLmin = (VOLmin – VTT)/Rt
IOLmin = ({5.0 – 1.85} – 3.0)/50 = 3.0mA
VTT = 5 (125/{125 – 83.3}) = 3.0V
ZO
Rpd
If +5% supplies are assumed a VCC of VCCnom –5% and a
VTT of VTTnom +5% will represent the worst case. Under
these conditions, the following output currents will result:
VEE
Open Line Termination
IOHmax = ({4.75 – 0.81} – 3.15)/50 = 15.8mA
IOLmin = ({4.75 – 1.85} – 3.15)/50 = 0mA
ZO
Rs
Using the other extremes for the supply voltages yields:
Rpd
IOHmax = 31.8mA
IOLmin = 11mA
VEE
RS = ZO
Series Termination
The changes in the IOH currents will affect the DC VOH
levels by ≈±40mV at the two extremes. However in the vast
majority of cases the DC levels for ECL devices are well
centered in their specification windows, thus this variation
will simply move the level within the valid specification
window and no loss of worst case noise margin will be seen.
The IOL situation on the other hand does pose a potential AC
problem. In the worst case situation the output emitter
follower could move into the cutoff state. The output emitter
followers of ECL devices are designed to be in the
conducting “on” state at all times. If cutoff, the delay of the
device will be increased due to the extra time required to pull
the output emitter follower out of the cutoff state. Again this
situation will arise only under a number of simultaneous
worst case situations and therefore is highly unlikely to
occur, but because of the potential it should not be
overlooked.
ZO
Rt
Rt = ZO
VTT
Parallel Termination
VCC
ZO
R1
R2
VEE
Thevenin Parallel Termination
Figure 3. Termination Techniques for
ECL/PECL Devices
Thevenin Equivalent Termination Schemes
The Thevenin equivalent parallel termination technique
of Figure 3 is likely the most attractive scheme for the
CMOS/TTL designer who is using a small amount of ECL.
As mentioned earlier this technique will consume more
power, however the absence of an additional power supply
will more than compensate for the extra power
Because of the resistor divider network used to generate
VTT the variation in V will be intimately tied to the variation
in VCC. Differentiating the equation for VTT with respect to
VCC yields:
dVTT/dVCC = R2/(R1 + R2) dVCC
http://onsemi.com
5
AN1406/D
Again for the nominal case this equation reduces to:
impedance backplane is not available the differential
outputs can be distributed via twisted pair of ribbon cable
(use of ribbon cable assumes every other wire is a ground so
that a characteristics impedance will arise). Figure 4
illustrates common termination techniques for twisted
pair/ribbon cable applications. Notice that Thevenin
equivalent termination techniques can be extended to
twisted pair and ribbon cable applications as pictured in
Figure 4. However for twisted pair/ribbon cable
applications the standard termination technique picture in
Figure 4 is somewhat simpler and also does not require a
separate termination voltage supply. If however the
Thevenin techniques are necessary for a particular
application the following equations can be used:
∆VTT = 0.6 ∆VCC
So that for ∆VCC = ±5% = ±0.25V, ∆VTT = ±0.15V.
As mentioned previously the real potential for problems
will be if the VOL level can potentially put the output emitter
follower into cutoff. Because of the relationship between the
VCC and VTT levels the only situation which could present
a problem will be for the lowest value of VCC. Applying the
equation for IOLmin under this condition yields:
IOLmin = ({VOLmin – VTT}/Rt
IOLmin = ({4.75 – 1.85} –2.85)/50 = 1.0mA
From this analysis it appears that there is no potential for the
output emitter follower to be cutoff. This would suggest that
the Thevenin equivalent termination scheme is actually a
better design to compensate for changes in VCC due to the
fact that these changes will affect VTT, although not 1:1 as
would be ideal, in the same way. To make the design even
more immune to potential output emitter follower cutoff the
designer can design for nominal operation for the worst case
situation. Since the designer has the flexibility of choosing
the VTT level via the selection of the R1 and R2 resistors the
following procedure can be followed.
Let VCC = 4.75V and VTT = VCC – 2.0V = 2.75V
Therefore:
R1 = R2 = ZO/2
R3 = R1 (VTT – VEE)/(VOH + VOL – 2VTT)
VTT = (R3{VOH + VOL} + R1{VEE})/(R1 + 2R3)
where VOH, VOL, VEE and VTT are PECL voltage levels.
Plugging in the various values for VCC will show that the
VTT tracks with VCC at a rate of approximately 0.7:1.
Although this rate is approaching ideal it would still
behoove the system designer to ensure there are no potential
situations where the output emitter follower could become
cutoff. The calculations are similar to those performed
previously and will not be repeated. The same equations
with the change R1 = R2 = ZO can be used to calculate a “Y”
termination for differential outputs into separate microstrip,
strip or coaxial cables.
R2 = 119Ω and R1 = 86Ω thus:
IOHmax = 23mA and IOLmin = 3.0mA
Plugging in these values for the equations at the other
extreme for VCC = 5.25V yields:
VEE
Rpd
VTT = 3.05V, IOHmax = 28mA and IOLmin = 5.2mA
Although the output currents are slightly higher than
nominal, the potential for performance degradation is much
less and the results of any degradation present will be
significantly less dramatic than would be the case when the
output emitter follower is cutoff. Again in most cases the
component manufactures will provide devices with typical
output levels; typical levels significantly reduces any chance
of problems. However it is important that the system
designer is aware of where any potential problems may
come from so they can be dealt with during the initial design.
ZO
Rt
Rpd
VEE
Rt = ZO
Standard Twisted Pair Termination
VTT
ZO
Rt
Rt = ZO/2
Rt
Parallel Twisted Pair Termination
Differential ECL Termination
Differential ECL outputs can be terminated using two
different strategies. The first strategy is to simply treat the
complimentary outputs as independent lines and terminate
them as previously discussed. For simple interconnect
between devices on a single board or short distances across
the backplane this is the most common method used. For
interconnect across larger distances or where a controlled
VTT
ZO
R1
R2
Rt = ZO/2
Thevenin Twisted Pair Termination
R3
VEE
Figure 4. Twisted Pair Termination Techniques
http://onsemi.com
6
AN1406/D
Noise and Power Supply Distribution
Since ECL devices are top rail referenced it is imperative
that the VCC rail be kept as noise free and variation free as
possible. To minimize the VCC noise of a system liberal
bypassing techniques should be employed. Placing a bypass
capacitor of 0.01µF to 0.1µF on the VCC pin of every device
will help to ensure a noise free VCC supply. In addition when
using PECL in a system populated heavily with CMOS and
TTL logic the two power supply planes should be isolated
as much as possible. This technique will help to keep the
large current spike noise typically seen in CMOS and TTL
drivers from coupling into the ECL devices. The ideal
situation would be multiple power planes; two dedicated to
the PECL VCC and ground and the other two to the
CMOS/TTL VCC and ground. However if these extra planes
are not feasible due to board cost or board thickness
constraints common planes with divided subplanes can be
used (Figure 5 on page 8). In either case the planes or sub
planes should be connected to the system power via separate
paths. Use of separate pins of the board connectors is one
example of connecting to the system supplies.
For single supply translators or dual supply translators
which share common power pins the package pins should be
connected to the ECL VCC and ground planes to ensure the
noise introduced to the part through the power plane is
minimal. For translating devices with separate TTL and
ECL power supply pins, the pins should be tied to the
appropriate power planes.
Another concern is the interconnect between two cards
with separate connections to the VCC supply. If the two
boards are at the opposite extremes of the VCC tolerance,
with the driver being at the higher limit and the receiver at
the lower limit, there is potential for soft saturation of the
receiver input. Soft saturation will manifest itself as
degradation in AC performance. Although this scenario is
unlikely, again the potential should be examined. For
situations where this potential exists there are devices
available which are less susceptible to the saturation
problem. This variation in VCC between boards will also
lead to variations in the input switching references. This
variation will lead to switching references which are not
ideally centered in the input swing and cause rise/fall skew
within the receiving device. Obviously the later skew
problem can be eliminated by employing differential
interconnect between boards.
When using PECL to drive signals across a backplane,
situations may arise where the driver and the receiver are on
different power supplies. A potential problem exists if the
receiver is powered down independent of the driver.
Figure 6, on page 8, represents a generic driver/receiver
pair. A current path exists through the receiver’s VCC plane
when the receiver is powered–down and the driver is
powered–on, as shown in Figure 6. If the receiver has ESD
protection, the current will flow though the ESD diode to
VCC. If the receiver has NO ESD protection, the current will
flow through the input transistor and emitter–follower
base–collector junctions to VCC. The amount of current
flow, in either case, will be enough to damage both the driver
and receiver devices. Either of these situations could lead to
degradation of the reliability of the devices. Because
different devices have different ESD protection schemes,
and input architectures, the extent of the potential problem
will vary from device to device.
Another issue that arises in driving backplanes is
situations where the input signals to the receiver are lost and
present an open input condition. Many differential input
devices will become unstable in this situation, however,
most of the newer designs, and some of the older designs,
incorporate internal clamp circuitry to guarantee stable
outputs under open input conditions. All of the ECLinPS
(except for the E111), ECLinPS Lite, and H600 devices,
along with the MC10125, 10H125 and 10114 will maintain
stable outputs under open input conditions.
Conclusion
The use of ECL logic has always been surrounded by
clouds of misinformation; none of those clouds have been
thicker than the one concerning PECL. By breaking through
this cloud of misinformation the traditional CMOS/TTL
designers can approach system problems armed with a
complete set of tools. For areas within their designs which
require very high speed, the driving of long, low impedance
lines or the distribution of very low skew clocks, designers
can take advantage of the built in features of ECL. By
incorporating this ECL logic using PECL methodologies
this inclusion need not require the addition of more power
supplies to unnecessarily drive up the cost of their systems.
By following the simple guidelines presented here
CMOS/TTL designers can truly optimize their designs by
utilizing ECL logic in areas in which they are ideally suited.
Thus bringing to market products which offer the ultimate
in performance at the lowest possible cost.
http://onsemi.com
7
AN1406/D
CMOS/TTL
+5.0V PLANE
*
**
**
CMOS
Sub System
TTL
Sub System
System
+5.0V
CMOS/TTL
GROUND PLANE
System
Ground
PECL
+5.0V PLANE
*
**
ECL
Sub System
PECL
GROUND PLANE
*
**
Low frequency bypass at the board input
High frequency bypass at the individual device level
Figure 5. Power Plane Isolation in Mixed Logic Systems
5.0V
VCC = 0V
Driver
5.0V
VEE = GND
VEE = GND
Figure 6. Generic Driver/Receiver Pair
ON Semiconductor and
are trademarks of Semiconductor Components Industries, LLC (SCILLC). SCILLC reserves the right to make changes
without further notice to any products herein. SCILLC makes no warranty, representation or guarantee regarding the suitability of its products for any particular
purpose, nor does SCILLC assume any liability arising out of the application or use of any product or circuit, and specifically disclaims any and all liability,
including without limitation special, consequential or incidental damages. “Typical” parameters which may be provided in SCILLC data sheets and/or
specifications can and do vary in different applications and actual performance may vary over time. All operating parameters, including “Typicals” must be
validated for each customer application by customer’s technical experts. SCILLC does not convey any license under its patent rights nor the rights of others.
SCILLC products are not designed, intended, or authorized for use as components in systems intended for surgical implant into the body, or other applications
intended to support or sustain life, or for any other application in which the failure of the SCILLC product could create a situation where personal injury or
death may occur. Should Buyer purchase or use SCILLC products for any such unintended or unauthorized application, Buyer shall indemnify and hold
SCILLC and its officers, employees, subsidiaries, affiliates, and distributors harmless against all claims, costs, damages, and expenses, and reasonable
attorney fees arising out of, directly or indirectly, any claim of personal injury or death associated with such unintended or unauthorized use, even if such claim
alleges that SCILLC was negligent regarding the design or manufacture of the part. SCILLC is an Equal Opportunity/Affirmative Action Employer.
PUBLICATION ORDERING INFORMATION
USA/EUROPE Literature Fulfillment:
Literature Distribution Center for ON Semiconductor
P.O. Box 5163, Denver, Colorado 80217 USA
Phone: 303–675–2175 or 800–344–3860 Toll Free USA/Canada
Fax: 303–675–2176 or 800–344–3867 Toll Free USA/Canada
Email: [email protected]
ASIA/PACIFIC: LDC for ON Semiconductor – Asia Support
Phone: 303–675–2121 (Tue–Fri 9:00am to 1:00pm, Hong Kong Time)
Email: ONlit–[email protected]
JAPAN: ON Semiconductor, Japan Customer Focus Center
4–32–1 Nishi–Gotanda, Shinagawa–ku, Tokyo, Japan 141–8549
Phone: 81–3–5487–8345
Email: [email protected]
Fax Response Line*: 303–675–2167
800–344–3810 Toll Free USA/Canada
ON Semiconductor Website: http://onsemi.com
*To receive a Fax of our publications
For additional information, please contact your local Sales Representative.
N. America Technical Support: 800–282–9855 Toll Free USA/Canada
http://onsemi.com
8
AN1406/D