AD8146/AD8147/AD8148

Rev. 0 | Page 20 of 24

20

Under no circumstances should capacitance be intentionally

added to an output to introduce frequency domain peaking.

Figure 36 and Figure 37 illustrate how adding just 5 pF of

excessive load capacitance influences time and frequency

domain responses.

2.0

–2.0

0

TIME (ns)

1.5

1.0

0.5

0

–0.5

–1.0

–1.5

2

4

6

8

10121416

18

Figure 36. Large Signal Transient Responses at Various Capacitive Loads

12

2

10

1000

FREQUENCY (MHz)

100

11

10

9

8

7

6

5

4

3

Figure 37. Large Signal Frequency Responses at Various Capacitive Loads

While high frequency peaking is desirable in some cable

equalization applications, it should be implemented using

methods that do not compromise the stability of the driver and

that do not depend on amplifier parasitic elements. The parasitic

elements are affected by process variations and cannot be

depended upon for circuit designs. The amplifier may break

into oscillation when excess load capacitance is intentionally

added. For more information on this topic, see the Adding Pre-

Emphasis to the AD8148 section for a description on how to

introduce a controlled amount of pre-emphasis for 30 meters of

UTP using the AD8148.

ADDING PRE-EMPHASIS TO THE AD8148

UTP cables exhibit loss characteristics that are low pass in

nature and are exponential functions of the square root of the

frequency. Over wideband video bandwidths, the losses are

predominantly due to the skin effect, which causes the resistance of

the cable to increase with frequency. Even though the loss

characteristics are nonlinear, suitable linear networks can be

designed to approximately compensate for the losses.

Placing the compensation network at the transmitting end of

the cable is referred to as pre-emphasis, because the higher

frequencies are emphasized, or boosted, before they are sent, to

compensate for the low-pass response of the cable. Because the

higher frequencies experience more loss than the lower frequencies

as they pass through the cable, the high and low frequencies

arrive at approximately the same level and at the end of the cable

when a properly designed pre-emphasis network is used at the

transmitter. The ideal cascaded frequency response of the pre-

emphasis network and the cable is therefore nominally flat.

Because the AD8148 has an internally set, closed-loop gain of 4

(12 dB), it is possible to reduce the gain at low frequencies using

external frequency selective components, then use these

components to provide increasing gain with increasing

frequency, back to a value close to 12 dB. These components, along

with the AD8148, form the pre-emphasis network. When properly

designed, the combined frequency response of the pre-emphasis

network and cable is approximately flat with a gain of 2 (6 dB).

Figure 38 illustrates how to construct a pre-emphasis network

using the AD8148 that compensates for 30 meters of UTP cable.

The network in the lower leg is required to match the transfer

function of the two feedback loops.

At dc, the capacitors are open circuits, and the network has a

gain of approximately 6.5 dB. (The additional 0.5 dB is added to

compensate for the cable flat loss that occurs at frequencies

below where the skin effect begins to take effect.) Moving up in

frequency, the 30 pF capacitor begins to take effect and introduces

a zero into the frequency response, causing the gain to increase

with frequency. Continuing to move up in frequency, the 30 pF

capacitor becomes an effective short, and the 487 Ω resistor

goes in parallel with the 442 Ω resistor, forming a pole in the

response. Continuing to move up in frequency, the 15 pF

capacitor takes effect, introducing another zero, and causes

the gain to further increase with frequency until it becomes

an effective short, and the gain starts to flatten out until the

amplifier response begins to roll off. The gain does not reach

12 dB before the amplifier begins to roll off because the 12 dB

value is a high frequency asymptote. The pole and zero locations

cited in the previous discussion are qualitative, but the

discussion describes the basic principles involved with the

operation of the pre-emphasis network.