LOSING THE HUM BAR
Jul 1, 2002 12:00 PM, BILL WHITLOCK
▪ IN THE JUNE 2002 ISSUE I SHOWED HOW ENTIRELY proper power connections to equipment can still cause hum bars. That article described some simple measurement techniques and closed with a step-by-step troubleshooting procedure to locate the offending interface. Normal voltage differences between system grounding points cause small AC power currents to flow through signal cables. At power frequencies, the shield impedance (AC resistance) of a typical coaxial video cable is about 3 ž per 1000 feet (310 m). As ground loop current flows in this impedance, a small voltage drop is created. Because the interface is unbalanced, common-impedance coupling adds half this voltage to the video signal. A voltage drop of only 7mV peak can cause a visible hum bar.
The objective is to reduce that voltage drop. Of course, it would make sense to reduce system ground voltage differences as much as possible, but this generally requires utility power modifications or rewiring. Likewise, using shorter cables or types with less shield resistance will reduce the impedance that causes the coupling. Although these are certainly important design considerations for new systems, they are rarely practical in existing systems. For an existing ground voltage difference and cable shield resistance, it's preferred to reduce shield current by inserting a high-impedance device somewhere in the ground loop circuit.
To be effective, the device's common-mode impedance must be high enough to significantly reduce ground loop current (Common-mode means applying to both signal and shield connections between input and output ports). Any voltage drop in the cable shield due to residual loop current will simply become part of the video signal (a hum bar) seen at the input port. Because voltage drops in a series circuit are proportional to impedance, nearly all of the system ground voltage difference will appear between the input and output ports of the device itself when it is installed. Therefore, the common-mode rejection of the device must also be high enough that this voltage has no significant effect on the video signal at the output port. CMR is a measure of output signal while a ground (common-mode) voltage of specified level and frequency is applied between input and output ports. The results of the measurement can also be calculated as a ratio of output response to normal-mode (signal) and common-mode (ground voltage) inputs of the same voltage. This ratio is known as common-mode rejection ratio, or CMRR, and is normally expressed in decibels, in which larger numbers indicate more rejection.
Here's a look at how the three basic design strategies work and a comparison of a few products.
▪ The Video Isolation Transformer. A transformer uses its primary winding to convert the signal itself into a fluctuating magnetic field, which then induces a replica signal in the secondary winding. Because the two windings are electrically insulated, they have only one residual capacitance between them. For the Jensen ISO-MAX isolator, this 2 nF capacitance makes its common-mode impedance more than 1 Mž at 60 Hz, and the insulation easily withstands 350 V RMS common-mode voltages. Its CMRR is typically 120 dB. Therefore, the biggest advantage of a transformer isolator is that loop current remains negligible, and rejection is high even when system ground voltage differences are extreme or cables are very long. Transformers also require no power, and they are bidirectional.
The chief disadvantage of a video transformer is limited bandwidth. Usable transformers cannot have response to 0 Hz (DC) and are limited to a frequencies spanning about a 1,000,000 to 1 range. For NTSC video, poor low-frequency response can cause black level to vary from image top to bottom, an effect called field-rate tilt. Good performance in this respect requires low-frequency response extending to about 10 Hz. Therefore, high-frequency response is limited to about 10 MHz in a practical video transformer.
▪ The Video Common-Mode Choke. The video CM choke device goes by many names, including hum eliminator, hum suppressor, humbucker, ground loop inhibitor and ground loop isolator. Sometimes it is mistakenly called a transformer. The CM choke is the most widely used solution for hum bars. Although these devices use windings and core material like a transformer, they operate in a fundamentally different way. Video signal current flows from device A to device B through one winding and returns through a different one. Because the two windings have the same number of turns, and the same current flows through them in opposite directions, their magnetic fields cancel and there is no signal-related field in the core. In fact, the inner conductor and shield of a single piece of coaxial cable are the windings or coil of these CM chokes, so the only effect on the signal is that of the same length of unwound cable.
However, ground loop current flows through the shield and inner conductors in the same direction. Therefore, ground loop current does produce a magnetic field in the core that reacts with the coil to create an inductor, or choke, that determines common-mode impedance. Inductance of the tested units ranged from 25 milli-Henry for the Extron GLI-1000 to 250 mH for the Allen Avionics HEC-2000, making the impedance of the ground loop about 10 ž to 100 ž respectively at 60 Hz. As explained, because the choke is the highest impedance in the ground loop, most of the ground voltage difference will appear between its in input and output shield connections. Because the two windings are a single length of coax, the number of turns in each is precisely the same, and transformer action induces precisely the same voltage into the center-conductor signal path. Thus most of the hum is effectively removed from the signal at the output port. However, the cable used to wind the choke itself has DC resistance that ultimately limits rejection even when short external cables are used. For the tested units, this resistance ranged from 0.04 ž for the Extron to 0.368 ž for the Allen Avionics.
The major advantage of a CM choke is high bandwidth. As stated on the Pelco data sheet, video characteristics are “equivalent in all respects to approximately 200 feet (60.96 m) of RG59/U coax.” Thus, low-frequency response extends to DC, and high-frequency response is limited only by the length and type of cable used in the winding. Typical bandwidths of 100 MHz or more make these units suitable for high-definition TV systems. As with transformers, CM chokes are also passive and bidirectional. The core material and number of turns used will determine the common-mode voltage at which the core will become magnetically saturated, causing the choke's impedance to plummet. Therefore, ground voltage differences over a certain level will cause hum rejection to deteriorate or vanish, an obvious disadvantage. The tested units varied widely in that respect.
▪ The Video Differential or Isolation Amplifier. A differential amplifier can null its response to common-mode voltage. Of course, at video frequencies, a very wideband amplifier is a requirement, and the values of the resistors immediately surrounding it must be kept low to prevent stray capacitances from degrading bandwidth. In most designs, these resistors determine the common-mode impedance for the device and, for the VAC unit tested, it was about 1 kž. A problem with differential amplifiers having low common-mode impedance is that their rejection depends on the driving source impedances. In a video system, these impedances are inherently unbalanced, near zero for the shield and 37.5 ž nominal for the terminated signal line. They also vary with cable resistances (length) and terminator accuracy. Therefore, most differential amplifier devices require a trim adjustment to achieve maximum common-mode rejection. That could be a disadvantage in portable applications in which cables are frequently rerouted. Although differential amplifiers do require power and contain potentially vulnerable circuitry, features such as multiple outputs and adjustable gain add little incremental cost.
PERFORMANCE IN A REAL SYSTEM
Figure 1 plots the output of six devices using a 100-foot length of typical RG59/U coaxial cable (0.38ž shield resistance). The system ground voltage difference was held at 100 mV peak as frequency was varied. Some observations:
- Because the common-mode impedance for these devices is inductive, making it increase with frequency, the plots for all the CM chokes have a general downward slope.
- Likewise, the capacitive impedance of the transformer decreases with frequency, giving its plot an upward slope.
- The differential amplifier, with its resistive impedance, has an essentially flat plot. The insignificant bumps at 120 and 240 Hz are due to its power supply ripple.
Under the test conditions, all the units provided enough common-mode rejection to keep their outputs below the 7mV peak visibility threshold. However, the rejection of any device will decrease with cable length. For example, if a 1000-foot length of the same cable were used, output voltage would be ten times (20 dB) greater and all plots would move up two divisions. That would make the Extron and Inline devices marginal, but it wouldn't make sense to use either on such a long cable because of bandwidth considerations.
Figure 2 plots the output of the same devices, cable and configuration, but this time the frequency was held at 60 Hz as the system ground voltage difference was varied from 10mV to 10V peak. If the output of a device remains proportional to the common-mode input (hence, constant rejection ratio), its plot will be an upward sloping straight line over the full voltage range.
- The rejection ratio of the Allen Avionics and Pelco CM chokes is fairly constant over the range. Core saturation occurs (off the graph) at 28V peak (20V RMS) for the Allen Avionics unit and at 24V peak (17V RMS) for the Pelco. The smooth curvature is due to core properties that vary somewhat with magnetic field intensity.
- Core saturation is evident for the Inline and Extron CM chokes, occurring at 1.4V peak (1V RMS) for the Inline unit and only 170mV peak (120mV RMS) for the Extron.
- At levels below about 1 V, the output of the Jensen unit is below the instrument noise floor. Rejection remains constant until internal insulation breaks down (off the graph, at least 500V peak or 350V RMS).
- At levels under about 100 mV, the output of the VAC unit is below its own noise floor. Rejection remains constant up to electronic overload that occurs off the graph at 16.2V peak (11.5V RMS).
▪ How Much Bandwidth is Enough? It depends — but you knew that. Remember that coaxial cables limit bandwidth, too. For example, the — 3 dB bandwidth for 100 feet (30 m) of cable is about 150 MHz for RG59 types and about 300 MHz for RG6 or RG11 types. It is only 5 MHz for 1000 feet of RG11 and response is — 6.5 dB for the same length of RG59. Table 1 may help you make a decision (computer formats are included for comparison).
▪ Which One to Use? Again, it depends on many factors: how much bandwidth is necessary, how much ground voltage difference is present and how long and what type of cable is used. Often, information critical to the decision is absent from manufacturer's data sheets, and in other cases a specification, such as rejection or CMRR, is offered without a hint as to how the test was performed. Hopefully, the information presented here, and in Part 1, will help lead you to the data you need and then guide you to a successful decision.
Bill Whitlock is president of Jensen Transformers Inc. He has designed audio and video circuits and systems for 30 years. He can be contacted by e-mail at firstname.lastname@example.org.
An obvious way to raise the impedance of a ground loop is to disconnect a safety ground, most commonly done with a 3-to-2-prong adapter. This practice is dangerous, illegal and makes you legally liable for subsequent shocks, injuries, electrocutions or fires!