# Comparing Op-Amp & Video Amp Topologies, Part 4

In my previous blog in this series, we looked at the video amplifier used in a piece of professional video equipment, the Sony BE-3000 video switcher/effects board. Figure 1 shows a video amplifier similar to one in the Ampex AVR-3 video tape recorder's demodulator board. In the original circuit, Q4's collector was connected to the feedback resistor directly, and the variable capacitor CF was not needed. A separate buffer amplifier then provided drive via the collector of Q4.

Figure 1

A video amplifier similar to one used in an Ampex AVR-3 video recorder with a gain of 4 V/V.

This amplifier has a slightly different second voltage gain stage that allows more symmetrical output voltage swings by tying the collector of Q4's load resistor to the minus rail. In this example, the amplifier is set for a gain of four instead of two. And for optimal flatness in video frequency response, a variable capacitor (CF) is adjusted. Amplifiers similar to this have been seen in many audio devices.

The measured open loop gain and -3 dB open loop frequency were about 514 V/V and 560 kHz, respectively. The gain bandwidth product is about 288 MHz.

For a quick calculation of the open loop frequency response, first we determine the input resistance into Q4's base. The internal base emitter resistance is r•πQ4 ~175Ω given the DC collector current of Q4 is about 15 mA and with β = 100. Therefore, the input resistance at the base of Q4 is about

This input resistance is in parallel to RL_Q1 (402Ω), and the paralleled resistance is 1185‖402Ω = 300Ω.

To determine the open loop pole, we calculate the input capacitance to the base of Q4 via the voltage gain of Q4 and the Miller multiplier capacitor Cc_Miller.

The voltage gain (A) of Q4 is about 820/[(1/gmQ4 ) + 10] ≈ 820/[1.73 + 10] = 70 V/V. There is an internal collector-base capacitance CcbQ4 of about 2.5 pf.

Thus the equivalent capacitance at the base of Q4 is approximately:

(1 + A)(Cc_Miller + CcbQ4 ) = (1 + 70 )(10 + 2.5) = 887.5pf. The calculated open loop pole then 1/[2π(300Ω)(887.5pf)] = 598 kHz.

With a closed loop gain of +4 V/V, this amplifier's frequency response was well within 0.2 dB out to 4.2 MHz. The differential phase and gain were about 0.15° and 0.1 percent, respectively.

One may ask about comparing it with op-amps of the 1970s or 1980s. For the comparison, an LM318 was chosen, since its high slew rate (50 V/μs) and gain bandwidth product (50 MHz) would be suitable for video use.

The LM318 was set for a closed loop gain of 2 V/V by using 1000Ω resistors for the feedback network. Unfortunately, the differential phase and gain measured about 10° and 5 percent, respectively.

See Figures 2-5 for a comparison between the input and output signals of the LM318 and the two op-amps we described in detail in our previous blog. As seen in Figure 3, the LM318 has an uneven frequency response (e.g., a peak of nearly 1 dB) that also causes overshoot.

Figure 2

Input Mulitburst signal with sine wave packets at 0.5 MHz, 1 MHz, 2 MHz, 3 MHz, 3.58 MHz, and 4.2 MHz.

Figure 3

LM318 video frequency response.

Note a slightly peaked frequency response for this 50 MHz gain bandwidth product op-amp.

Figure 4

Amplifier circuit from Figure 2 of our previous blog.

The amplifier circuit from our previous blog's Figure 2 also had a 50 MHz gain bandwidth product and exhibited much flatter frequency response than the LM318.

Figure 5

Amplifier circuit from Figure 1.

With a gain of four (instead of two), this amplifier from Figure 1 still provides a very flat frequency response.

By the 1990s, current mode feedback amplifiers such as the EL2120 and EL5261 provided superior video frequency response and very low differential phase and differential gain distortion. Likewise, voltage feedback amplifiers such as the ISL55002 will perform superbly in terms of video specifications.

Related posts:

## 3 comments on “Comparing Op-Amp & Video Amp Topologies, Part 4”

1. D Feucht
November 19, 2013

Some possible ways of refining the amplifier:

1. The Q1 collector node suffers from the Miller effect to its base. Q2 does not and has a Miller effect only from the feedback path. It would be a higher-bandwidth path to use instead, and with the same gain.

2. Q4 also has (as you point out) the Miller effect and might be made instead into a current mirror, where the Q4 collector is connected to the Q2 collector instead. Then the gain of Q2 is not lost (increasing gain-bandwidth) and the mirror provides a low-speed gain path for low-speed precision. The RL_Q4 resistor, which had to be made larger to be returned to VEE instead of ground, does increase gain but also reduces bandwidth.

3. By making RE1, RE2 somewhat larger than zero, the Cbe effects of Q1 and Q2 are reduced and the amplifier is made faster with a slight loss in gain. The input Z is also made closer to resistive.

4. The Q3 base resistor (47 ohm) will form a pole with Cbc(Q4), but if it is connected to the collectors of Q2, Q4 – a high-impedance node – then it is not needed anyway.

The subject of active-circuit dynamics is complicated. You might take a look at my EDN series on Circuit Dynamics for (much) more on all of this. Last year I built some circuits like this (at lower speed, but same principles) using the CA3906 BJT array. These circuits can be both enjoyable and surprising.

2. rnquan
November 19, 2013

Thank you for your comments. The 47 ohm base series resistor was put in to ensure that the output emitter follower did not oscillate.

3. D Feucht
November 22, 2013

Ronald,

There is a way to compensate an emitter follower (common-collector) stage that is capacitively loaded without the bandwidth-reducing effect of the base resistor. See more on this in the following three articles: