OUTPUT NOISE MEASUREMENT

Output Noise Measurement

In DC/DC Converters

In the application of DC/DC converters, conflicts often arise between the end user's output noise measurement and the manufacturer's specification. Typically, the end user will connect an oscilloscope across the converter's output with a 3 or 5 inch long ground lead on the scope probe, set the input to AC coupling, increase the gain and measure the peak-to-peak noise on a 100 MHZ bandwidth oscilloscope. The measurement may or may not agree with the manufacturer's published specification. Usually, it does not. What went wrong?

What is being measured are two noise signals - the conducted noise actually on the converter's output pins, and radiated noise picked up on the long ground lead of the scope probe. The conducted output noise has two components - a "low" frequency ripple, and a high frequency burst. The low frequency ripple present at the output of converters is at the primary switching frequency. The high frequency noise is a burst resembling a heavily damped ringing sinusoid at 40 to 50 MHZ, and is due to reverse recovery of the rectifying diodes. This ringing occurs at the time the switching transistor changes state from saturation to cutoff. It is this high frequency noise that radiates over a few inches and couples into the loop antenna formed by the scope probe and its associated ground lead.

The peak-to-peak amplitude of the low frequency ripple and high frequency burst noise increase with increased output load current. As load current rises, the output filter capacitor's ripple current rises, causing a larger drop across its ESR (Effective Series Resistance) and, consequently, a larger output ripple voltage. The rectifying diode's recovery time and the energy dissipated during turn off both increase, which raises the amplitude of the high frequency burst.

Factors affecting noise measurements:

Several factors affect measurements of noise amplitude, including oscilloscope bandwidth, scope probe characteristics, the type of load, the number and length of attached leads and the method of grounding the scope probe.

The output noise of most converters (such as the 10A48R5 or 15A/A48R5) is specified for a bandwidth of 20 Hz to 20 MHZ. The oscilloscope's bandwidth must be limited to 20MHz when testing them since a 100 MHZ bandwidth will measure noise amplitude several times higher.

Oscilloscope probes of different types do not give equal measurements of noise amplitude. The bandwidth, input capacitance, inductance, attenuation and grounding rings differ from one type to another. Only 10X probes should be used for measuring noise because 1X probes do not have a wide enough bandwidth.

While Tektronix P6103 probes are good for general measurements, they are generally poor for noise measurement and show high noise amplitude because their ground ring makes a poor connection at the oscilloscope. Tektronix P6105A or P6131 probes of 3-foot (1 meter) length are best for noise measurement because they make good ground connections.

Leads attached to the converter act as antennas which pick up radiated noise and add this to the measurement. The longer the lead, the more noise is added. Digital voltmeters generate switching noise internally and this can appear on their input leads. For these reasons, remove all unnecessary leads from the converter. Only the input power leads and output load leads should be attached to the converter when measuring noise.

A converter under test is generally loaded either with a resistor or with an electronic load such as a Kikusui PLZ-72W. A resistor used to load a converter should be non-inductive or have low inductance because inductance presents a high impedance to high frequencies and results in a larger amplitude of the ringing noise. Measuring noise when using an electronic load is not as simple as with a resistor. The difference is related to the input characteristics of the load. An electronic load can give either higher or lower noise measurements than a resistor load, depending on the type of converter under test. A converter such as the 15A/A48R5 will show lower output noise when loaded with an electronic load than when loaded by a resistor. The 15A/A48R15-15 will show higher noise when an electronic load is used.

Measurement technique:

The last and most important consideration in measuring noise is the method of using the oscilloscope probe. Contact leads must be as short as possible. Remove all unnecessary leads which are attached to the converter.

It is easy to make a 10-to-1 difference in the apparent amplitude of noise by changing the technique of measurement. A P6105A probe with its spring-loaded tip in place and with a 3 inch ground lead can show switching noise amplitudes of 500mV or greater on a 15A/A48R5 converter. With exactly the same setup (and not even turning off power to the converter), removing the ground lead and spring tip from the probe and using the exposed ground ring and tip to make contact will show noise amplitude near 50mV. The spring-loaded tip adds about 1 1/4" of unshielded wire to the internal tip and acts as an antenna. The ground lead adds another 3" of antenna.

Converter output noise is measured without a ground lead or push-on spring-loaded tip on the scope probe. The ground connection is made by using the ring behind the probe tip to touch the converter's output ground pin while the probe tip touches the output pin. A Tektronix 206-114-00 hook tip adapter may be necessary to make contact between pins on some "A" pinout converters. This gives a true measurement of conducted output noise. Figure 1 illustrates this technique.