At any frequency, regardless of the mode of operation, a transmission line can always be shortened to a length below which the line operates not in a distributed fashion, but in a mode reminiscent of a simple lumpedelement circuit (Section 3.2).
The mathematical extent of the lumpedelement region includes all combinations of w and l for which the magnitude of the propagation coefficient l g ( w ) remains less than D .
Equation 3.22
where 
D is an arbitrary constant typically set to about 1/4. 
l is the length of the transmission line, m, and 

g ( w ) is the propagation coefficient of the transmission line (complex neper/m) at frequency w (rad/s). 
For typical digital transmission applications the quantity g ( w ) increases monotonically from DC so that inequality [3.22] need be checked only at the maximum length and maximum anticipated frequency of operation. If the transmission line satisfies [3.22] at that maximum point, it will similarly satisfy [3.22] at all shorter lengths and lower frequencies.
Beware the fallacy that a short transmission line, even one short enough to fall into the lumpedelement region, never requires termination. Such is not the case. Even a perfect zerolength transmission line may resonate horribly if used to interconnect a ferociously reactive combination of source and load (see Section 3.4.6, "Step Response (LumpedElement Region)").
The classification of a transmission line in the lumpedelement region does not determine how the line is going to act. It determines merely how the line may be analyzed .
POINT TO REMEMBER
3.4.1 Boundary of LumpedElement Region
A exact physical interpretation of equation [3.22] is fairly difficult to comprehend; however, by making a few reasonable assumptions you may approximate the boundary of the lumpedelement region in the following way. First assume that the propagation coefficient for a transmission line is given by [3.13], and that parameters R, L , and C are constants that do not much vary with frequency. Substituting [3.13] for the propagation coefficient in [3.22], you may derive the following equation relating w and l LE .
Equation 3.23
Since the boundaries of the lumpedelement region is by definition a rather fuzzy concept in the first place, you needn't bother with precise calculation of [3.23], substituting instead two asymptotic approximations that handle the cases where j w L is either much smaller than, or much larger than, R . These are the boundaries drawn in Figure 3.2.
Equation 3.24
Equation 3.25
where 
w is frequency of operation, rad/s, 
R DC is the series DC resistance of the transmission line (signal and return resistances added together), W /m, 

L is the transmissionline series inductance per meter, H/m, 

C is the transmissionline shunt capacitance per meter, F/m, 

l LE is the upper boundary of the lumpedelement region, m, and 

D is an arbitrary constant customarily set to 1/4. 

NOTE: 
For differential configurations, define R DC as the sum of the resistances of the outbound and returning conductors, and L and C as the inductance and capacitance respectively of the differential transmission line thus formed , L = Z DIFFERENTIAL / v DIFFERENTIAL and C = 1/( Z DIFFERENTIAL · v DIFFERENTIAL ). 
If the length l is known and you wish to determine the maximum frequency of lumpedelement operation, the constraints in [3.24] and [3.25] may be inverted to produce
Equation 3.26
Equation 3.27
Here you can see the value of the approximations used in [3.24] and [3.25], as the direct inversion of any equation utilizing the full, frequencyvarying form of g ( w ) would indeed be a formidable undertaking.
In physical terms, constraint [3.24] asks that the RC time constant l 2 R DC C formed by the structure's DC resistance and capacitance remain far smaller than the time constant 1/ w associated with the highest frequency of operation. Constraint [3.25] asks that the LC delay of the transmission structure remain much shorter than 1/ w . Working together, these two constraints ensure that the delay internal to the transmission structure itself remains far smaller than the signal rise or fall time. Under these conditions the transmission line enjoys a peculiarly tight coupling between the source and load impedances. When connecting obnoxiously reactive components , the system can still exhibit vigorous amounts of simple harmonic resonance , but what it cannot do is create the sort of lingering, unexpectedly late reflections that happen on a truly distributed circuit.
Transmission lines short enough to operate in the lumpedelement region rarely require termination except in unusual situations involving very lowimpedance drivers coupled either through exorbitant amounts of packaging inductance or connected through transmission lines to heavily reactive loads.
POINT TO REMEMBER
3.4.2 Pi Model
The pimodel circuit approximates the behavior of a short transmission line (Figure 3.4). This circuit is equivalent to a secondorder Taylor series solution of the transmission equations (see Appendix D, "Accuracy of Pi Model").
Figure 3.4. The pi model applies to any transmission line electrically short compared to the signal wavelength, and where the time constant l 2 R DC C remains small compared to the signal period.
The name pi model derives from the schematic configuration of the three circuit elements in the model whose positions resemble the three strokes of the Greek letter p .
Over the domain prescribed by [3.22], the fractional accuracy E of predictions made by the pi model is approximately bounded by
Equation 3.28
where 
Z S , Z C , and Z L represent the source impedance of the driver, the characteristic impedance of the transmission line, and the impedance of the load respectively. 
In typical cases where the ratios Z S / Z C and Z C / Z L each remain less than two and 

D = 1/4 the accuracy is better than 1% (see Appendix D). 
Those experienced in analog design may be concerned that the inductive and capacitive components of the pi model could under some circumstances resonate severely. If you have made this discovery I should like to congratulate you on your astute observation and also address your discomfort by pointing out that condition [3.27] precludes the realization of such resonance by restricting the range of applicability of the model to only those frequencies well below the resonant frequency.
POINT TO REMEMBER
3.4.3 TaylorSeries Approximation of H (LumpedElement Region)
Operation within the lumpedelement region assumes a line delay short compared to the rise and fall time of the externally applied signals. Both source and load therefore exert an almost instantaneous influence over the system behavior, creating the illusion of a direct connection between the two. Keep in mind, though, that even on a short line two fully independent modes of propagation still exist (out and back). Equations [3.13] through [3.17] still apply (see Section 3.2.1, "A Transmission Line Is Always a Transmission Line").
Within the lumpedelement region the limited magnitude of the propagation coefficient l g renders possible the use of Taylorseries expansions for H and H “1 in this region. These expansions greatly simplify the analysis.
Equation 3.29
Equation 3.30
where 
H is the oneway propagation function of the transmission line (a complex, dimensionless quantity), 
l is the transmissionline length (m), and 

g is the perunitlength propagation coefficient (complex neper/m). 
Calculations of circuit behavior made using the Taylorseries approximation depend only upon ordinary differential equations as opposed to the partial differential equations required to construct a fully distributed model. Therein lies one advantage of operation in the lumpedelement region ”it's relatively easy to determine in this region how a transmission line will respond.
This text uses only the first four terms of the Taylorseries expansion to generate the following approximations used in calculation of lumpedelement input impedance and system gain.
Equation 3.31
Equation 3.32
where 
H is the oneway propagation function of the transmission line (a complex, dimensionless quantity), 
l is the transmissionline length (m), and 

g is the perunitlength propagation coefficient (complex neper/m). 
POINT TO REMEMBER
3.4.4 Input Impedance (LumpedElement Region)
On a lumpedelement line the load exerts an almost instantaneous influence over the input impedance of the structure (see Section 3.2.1, "A Transmission Line Is Always a Transmission Line").
This happens because, by definition, the transit delay of a lumpedelement structure is limited to a short fraction of the signal risetime. In the time the signal progresses only partway through a transition, information about the changing input propagates to the far end of the line, interacts with the load, and reflects back to the source. The source therefore receives almost instantaneous feedback about the conditions at the load.
On very long lines the same general scenario applies with the exception that the longer transit delay effectively disconnects the source and load in a temporal sense. Information about the load reflects back to the source too late to affect the progress of an individual rising or falling edge. The reflection in such a case, rather than being considered part of the "input impedance" of the line, may be considered separately as a distinct "reflection." From a mathematical viewpoint, whether you consider the reflection as a perturbation in the input impedance of the structure or as a structurally distinct reflection from a distant load makes no difference, provided that you properly account for the reflectedwave effect somehow in your calculations.
Let's next examine the input impedance of a lumpedelement structure under various conditions of loading. To begin, apply the lumpedelement Taylorseries approximations [3.31] and [3.32] to the general equation for input impedance [3.16], neglecting all but the constant and linear terms.
Equation 3.33
Assuming the line to be very lightly loaded ( Z L >> Z C ) causes the righthand terms in the numerator and denominator to vanish , leaving you with a very simple expression for the input impedance.
Equation 3.34
Plugging in the definitions of g [3.13] and Z C [3.15] reveals a classic result ”that the input impedance of a short, unloaded transmission line looks entirely capacitive.
Equation 3.35
The amount of capacitance equals the total distributed capacitance ( lC ) of the line. The approximately equal sign in [3.35] exists to remind you that this equation applies only under very special circumstances:
Violation of either constraint invalidates [3.35]. I mention this because one of the most widely held misconceptions about transmission lines is the belief that the input impedance of a transmission line looks capacitive. It holds this appearance only when short enough to qualify as a lumpedelement structure and when lightly loaded. Under other conditions the input impedance varies considerably.
For example, let's see what happens if the line is shortcircuited to ground at the far end. You may be asking yourself, What good is a transmission line shorted at the far end? The application for such a line is the grounding of an integrated circuit (IC) pin on a PC layout that does not permit room for a ground via contiguous with the IC pin pad. In congested situations it is common for (inexperienced) layout persons to connect the IC ground pin with a short trace to a ground via (see Figure 3.5). The following discussion computes the effective input impedance of the trace leading to ground, as viewed from the perspective of the IC ground ball.
Figure 3.5. This ground via lies too far from the ballgrid array (BGA) ground ball.
Assuming the line to be shorted to ground at the ground via implies that ( Z L << Z C ). This assumption inflates the righthand terms in both numerator and denominator of [3.33], causing them to dominate the equation and leaving you with this simple expression for the input impedance.
Equation 3.36
Again plugging in definitions [3.13] and [3.15] reveals that the input impedance of a short, grounded transmission line looks either inductive or resistive, depending on the ratio of j w L to R .
Equation 3.37
Equation [3.37] tells you that the input impedance of a short line shorted to ground at the far end looks like nothing more or less than the total series impedance, inductive plus resistive, of the structure. Formula [3.37] provides the basis for making calculations of the groundbounce (also called simultaneous switching noise ) generated when a current i ( t ) passes through a short trace leading to ground.
At the frequencies associated with most highspeed digital grounding problems, the inductance of a pcb trace is almost always much more significant than the resistance, leading to this simplified conclusion:
The input impedance of a short, grounded pcb trace looks entirely inductive .
The amount of inductance equals the total distributed inductance ( lL ) of the transmission line formed by the pcb trace and its associated return path . The approximately equal sign in [3.37] exists to remind you that this equation applies only under very special circumstances:
You should check [3.33] to verify that when the line is properly terminated ( Z L = Z C ) the numerator and denominator return the same result so that the effective input impedance of the structure equals precisely Z C . This is the situation on an endterminated transmission line as typically implemented on a pcb.
I am compelled to point out that for lines operated at frequencies below the onset of the LC region, the input impedance Z C ( w ) is not at all constant but a strongly frequencyvarying quantity with a phase angle of 45 degrees. Accurately matching the impedance Z C in this region is not a trivial exercise. Fortunately, lines short enough to qualify for lumpedelement analysis hardly require termination.
POINTS TO REMEMBER
3.4.5 Transfer Function (LumpedElement Region)
Substitute expressions [3.31] and [3.32] into [3.17] to compute the overall circuit gain G .
Equation 3.38
From [3.38] you can (given sufficient mental effort) deduce the conditions necessary to achieve gain flatness . To begin your investigation of this equation, observe that as the term l g approaches zero it renders negligible all the terms associated with various powers of l g . The propagation function G in that case asymptotically approaches Z L /( Z S + Z L ), precisely the transfer gain you would expect if the source and load were directly connected. This simple mathematical deduction indicates that as the length l of the line is foreshortened, eventually you come to a length so short that the transmission line exerts no observable influence over the outcome ”the response at that point depends only on the configuration of source and load. In other words,
Any transmission line can be shortened to the point where it acts as a perfect connection .
Another conclusion you might draw from the analysis so far is that the gain Z L /( Z S + Z L ) is in some sense the best you can do given the source and load. No directly attached configuration of transmission lines will be able to improve on this response. A corollary to this principle is simply stated:
If the source can't drive the load in the first place, then hooking the source and load together with a transmission line isn't likely to make things better .
Within the lumped element region you may assume the extent x of the line is set to a value such that the magnitude of the coefficient l g remains less than D = 1/4 at all frequencies of interest. Under such an assumption you might reasonably expect to ignore the second and thirdorder terms in [3.38], but you could hardly ignore the firstorder term, particularly if either of the ratios Z S / Z C or Z C / Z L exceeds unity. This observation may be cast into a statement of the conditions under which the wiring exerts negligible influence over the connection of source and load.
Equation 3.39
Equation 3.40
Equation 3.41
Conditions [3.39] through [3.41] may be simplified and perhaps made more recognizable by inserting the definitions of g [3.13] and Z C [3.15].
Equation 3.42
Equation 3.43
Formulas [3.42] and [3.43] stipulate the two conditions that are required above and beyond [3.41] such that the line not exert any deleterious influence over signal quality, namely,
If either condition is not satisfied, then the presence of the transmission line may substantially affect the signal amplitude and/or quality.
In the event the line is driven by an ideal voltage source ( Z S = 0) and coupled to an ideal load at the endpoint ( Z L = ), equation [3.38] reduces to this form:
Equation 3.44
Equation [3.44] provides a direct relationship between the definition of the lumpedelement boundary l g < D and the circuit performance expected under ideal conditions. Given D = 0.25, the unloaded gain of the transmission line at the lumpedelement boundary can differ from unity by no more than D 2 /2 = .032. Beware that the appearance of good performance with ideal loading does not imply that the transmission line would achieve satisfactory performance under other conditions of loading.
It is usually sufficient to check the conditions [3.39] through [3.41] at the highest anticipated frequency of operation. For digital circuits, that frequency corresponds to w knee . If the circuit performs satisfactorily at w knee , then conditions will generally improve at lower frequencies, provided that the following conditions are met:
Taken together, the preceding three conditions suggest that any ordinary transmission line that is satisfactorily terminated for operation at frequencies at the lumpedelement boundary will also work satisfactorily at frequencies within the lumpedelement region. This principle should relieve you of certain worries about the peculiar way in which the transmissionline impedance spikes to infinity near DC. Even when Z C goes to infinity, as long as conditions [3.19] through [3.41] are met, the circuit will still function.
POINTS TO REMEMBER
3.4.6 Step Response (LumpedElement Region)
The best way to illustrate the general effect of a lumpedelement transmission line is by example (Figure 3.6).
Figure 3.6. This short trace meets the conditions for a lumpedelement circuit, yet when loaded with 10 pF it rings horribly.
Driver
Transmission line
Load
From the highfrequency values of Z and v you can determine the transmission line parameters L and C :
Equation 3.45
Presuming that the inductive effect of the line far outweighs the resistance (in this case it does), you may approximate the magnitude of the propagation coefficient using only the inductive term.
Equation 3.46
This value of the propagation coefficient puts you just outside the official boundary of the lumpedelement region at a point where you should expect the pimodel approximations (and indeed equation [3.38]) to be accurate only to about 3%.
Next check conditions [3.42] and [3.43] to see whether the line significantly influences signal quality. Condition [3.42] appears to be satisfied.
Equation 3.47
Checking [3.43] for the case of no load capacitance (infinite Z L ), you can see that it is clearly true as well ”leading to the important conclusion that for rise and fall times no faster than 1 ns, this microstrip in the absence of a load induces practically no distortion in the transmitted waveform.
Rechecking [3.43] for the case of a 10pF load capacitance leads to an opposite conclusion. In this case Z L exceeds the series impedance of the line by but a small amount (a factor of only 2:1) implying that the transmission line will have a noticeable effect on the shape of the received signal.
Equation 3.48
Indeed, when the received waveform is plotted under the two conditions, showing the outcome with no load and with a 10pF load, the unloaded condition shows an undistorted outcome, while the 10pF load displays prodigious amounts of ringing (Figure 3.7).
Figure 3.7. Even a short transmission line displays prodigious amounts of ringing when combined with a lowimpedance driver and a heavy capacitive load.
Those readers steeped in the art of analog design may find some satisfaction in comparing this result with the simple pimodel approximation (Figure 3.4). The values for inductance and capacitance in that figure are
Equation 3.49
Equation 3.50
When driven by a lowimpedance source, the capacitor on the left in Figure 3.4 plays only a small role in determining the result. The main effect is an RLC seriesresonant circuit formed by the output resistance of the driver, the 10.6 nH series inductance, and the 2.5 pF capacitor on the right, which connects in parallel with the load .
In an unloaded state the resonant frequency w res associated with the inductive and capacitive components lies far above the bandwidth of the driver.
Equation 3.51
An RLC resonance certainly exists at 6.1 Grad/s, but since this is well above the spectral center of gravity of the rising and falling edges coming out of the driver (2.2 Grad/s), you never see it. Condition [3.41] ensures that under lightly loaded conditions the line is always sufficiently short to guarantee the resonance falls above the frequency of operation.
With a 10pF load applied, the situation changes. The new load capacitance swells the total amount of capacitance on the right side of the pi model, reducing the resonant frequency to within a stone's throw of the bandwidth of the driver. In this example the new resonant frequency is a function of the series inductance of the line (10.6 nH) and the parallel combination of 2.5 pF and 10 pF.
Equation 3.52
Such a resonance should have a period of 2 p /(2.7 ·10 9 ) = 2.3 ns, corresponding nicely to the ringing period displayed in Figure 3.7.
POINT TO REMEMBER
Fundamentals
Transmission Line Parameters
Performance Regions
FrequencyDomain Modeling
Pcb (printedcircuit board) Traces
Differential Signaling
Generic BuildingCabling Standards
100Ohm Balanced TwistedPair Cabling
150Ohm STPA Cabling
Coaxial Cabling
FiberOptic Cabling
Clock Distribution
TimeDomain Simulation Tools and Methods
Points to Remember
Appendix A. Building a Signal Integrity Department
Appendix B. Calculation of Loss Slope
Appendix C. TwoPort Analysis
Appendix D. Accuracy of Pi Model
Appendix E. erf( )
Notes