Space-Saving Techniques for Driving Video

Three options head the list when trying to reduce all-important board space—a SAG network, charge pump, or dc-restore circuit.

More and more mobile devices feature videooutput ports, and these ports must drive video signals conforming to display standards. For example, NTSC video is a 1V peak signal driven into 75W lines. If the circuit uses termination at both cable ends, the signal from the video-driver output must be 2V. Often, the video ASICs (encoders) are built on lowvoltage platforms, running off of 1.8V or lower supplies.

In such cases, a video buffer is needed to raise the signal to proper driving levels. In addition, the video signal must be ac-coupled to prevent undesired current into the load and unwanted draining of the portable battery.

The ac output capacitance, typically 220mF, needs to be large to preserve the video signal’s lowfrequency contents. Physically, a 200mF capacitor is very large— too large for mobile devices like cell phones or MP3 players.

In this article, we’ll discuss the traditional ac-coupling technique for transmitting video signals, with the associated design tradeoffs. Then we’ll move into the various improvements that can be made. For instance, a SAG feedback network can be applied to minimise the output capacitance. Or a charge pump can be added. Or a dc-restore circuit can be applied at the input of the video driver. In the latter, two of the three original coupling capacitors are eliminated and the third is drastically reduced.

Since design choices in these cases are made for similar performance, the circuits can be compared in terms of size of coupling capacitors, power-supply noise, and board area.


The traditional technique for transmitting signals is accomplished with a single supply (Fig. 1). Resistors R1 and R2 set the bias voltage at the amplifier’s input, placing it in the linear operating region. The parallel combination of R1 and R2 form a lowfrequency pole with C1. R1’s and R2’s values are set by the video driver’s input bias current. If they’re too large, the offset voltage at the input will be unacceptably large as well. Because their sizes are limited, C1 must be sufficiently sized to guarantee that the input pole is lower than the minimum video signal frequency.

C2 is required to keep the dc gain at unity. The size of C2 is determined by the R3 and R4 values. For a current-feedback amplifier, feedback resistor R3 is predetermined by the amplifier design. For a voltage-feedback amplifier, the feedback resistor value is limited by the interconnecting parasitic capacitance at the amplifier’s inverting input. A large feedback resistor and excessive parasitic capacitance will lead to instability problems. Parasitic capacitance can be minimised to approximately 3pF with short traces and good layout.

C3 is needed to avoid putting unnecessary dc bias voltage into the load. The size of this capacitor is determined by the line impedance, typically 75Ω, and the minimum signal frequency. These three poles occur in close proximity, so their effect is additive. Therefore, all three poles must be placed well below the desired cutoff frequency. In a video system, for example, the lowest frequency of interest is the vertical sync at 60Hz.


Figure 2 shows the effects of pole placement on the low-side cutoff frequency of a video driver system. The curves represent two values of C1—0.1μF and 6.5μF—from the Figure 1 circuit. With the 0.1μF capacitor, the pole appears at 318Hz, and a 60Hz signal is attenuated by 11dB. When using a 6.5μF capacitor, the pole is lowered to 6Hz. This makes it possible for a 60Hz signal to pass with less than 1dB of attenuation.

The effect in the time domain is drastic (Fig. 3). The input to the video driver is plotted in yellow. The output with C1 = 6.5μF, a respectable duplicate, is plotted in pink. Finally, the output with C1 = 0.1μF is plotted in blue. The capacitance is small enough so that the voltage drifts up during the sync pulse, throwing off the average value once the video information returns. This offset corrupts the intensity information of the video signal.

Thus far, we’ve argued that larger capacitance is better for video systems, since it reduces the lowfrequency pole and preserves the low-frequency contents of the video signal. How large is reasonable? Size and cost provide an upper bound for capacitance value at 220μF.


In many instances, 220μF is physically too large to be included on a board. A feedback network can be added to create an effectively large output coupling capacitance (Fig. 4). The primary tradeoff when employing this setup is capacitive size versus load on the driver. Simplistically, if the capacitor is three times smaller, the driver must drive three times harder to deliver the same signal to the load. The additional output swing may suffer from linearity issues.

The area that’s consumed by this feedback network is relatively small, despite the number of components. R1, R2, and R3 can be very small and placed close to the inverting input. The size and proximity will ultimately reduce the parasitic capacitance.

The gain versus frequency for CSAG (the feedback capacitance) varies from 1μF to 220μF (Fig. 5). The lower cutoff is extended through peaking. When the feedback capacitance is 1μF, that extension is not enough to pass the vertical sync information. For all plotted values greater than 1μF, the lower cutoff is sufficient. Larger values can be used, but they will affect settling time.

The most appealing aspect of the circuit in Figure 4 is the dual function of CSAG. In this configuration, C2 isn’t needed. CSAG provides the feedback to scale the output capacitance, as well as the unity-gain dc characteristic formerly provided by C2. CCOMP is added instead of C2, but since it utilises the Miller effect, it’s three orders of magnitude smaller than the original C2.


A primary motivation for including coupling capacitors regards the migration of designs from dual supply to single supply. A charge pump is a separate IC that creates a negative supply. By adding a charge pump (Fig. 6), the designer can eliminate the output capacitance, but with added cost, increased noise, and larger power dissipation. The input coupling capacitance, C1, is still needed to remove the dc content of the video signal.

The limitations of this option depend on the charge pump’s characteristics. Because a switching circuit is used to create the opposite polarity of voltage, the frequency of this switching will appear as noise on our negative power supply and in our circuit. This conversion is imperfect, so a positive input of 5V can be used to create about 3.5V with a diode-based charge pump. A charge pump with integrated FET switches can create 4.5V.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.