Notes from the Test Bench
By Bruce Hofer, Chairman & Co-Founder, Audio Precision
I’ve been in Hong Kong this week visiting our Chinese distributor AP Technology to recognize their accomplishment of being 2010 AP Partner of the Year—their third such award in the past four years! It has been over five years since I was last in Hong Kong, and there have been some significant changes. In addition to a number of new 70 to 80 floor buildings in the Kow Loon skyline, one thing that struck me is the almost complete changeover from neon sign technology to super bright, full color spectrum LEDs. You can still see some older neon signs, but they really look aged in comparison to the new generation. In the same time period here at AP, we’ve come out with a new generation of analyzers (the APx500 Series), eight analog channels, HDMI connectivity, high-bandwidth 24-bit FFTs, a new friendlier graphic interface, and more. And our mains power LED has advanced also over the years from yellow, to blue, to white. So, perhaps I shouldn’t be so surprised.
AP Chairman and Co-Founder Bruce Hofer giving the Partner of the Year award to AP Technology's Man Li in China.
Output: Designing Switching Power Supplies with APx
Although primarily intended for audio, designers of switching (switch-mode) power supplies are finding the APx525 Series with the BW52 High Bandwidth option to be superior to their scope-based FFTs and spectrum analyzers for circuit analysis. The BW52 option extends the measurement range to over 1 MHz.
Switching supplies can be tricky to design. Because of the sharp switching cut-off, they can generate significant harmonic and intermodulation content. Analysis in the frequency domain is necessary in order to optimize input/output filtering, efficiency, feedback loop stability, input/output regulation, reliability, and other design elements.
APx500 display of a switching power supply, linear x and y axes. Any combination of linear or log scales may be selected, and multiple windows with different scales can be open simultaneously.
APx500 display of the same switching power supply, log x and y axes.
Two major advantages of the APx instruments with BW52 are the high bit depth and high frequency resolution. The 24-bit extremely low distortion AD converters in the APx instruments provide a much wider dynamic range and lower noise floor than the 8-bit converters in scopes or spectrum analyzers. Selectable averaging (up to 1000 averages) also helps to reduce random noise from obscuring low level detail. Additionally, most spectrum analyzers only offer a 50 Ω input impedance and a very low maximum input voltage tolerance. An external attenuator can be used to raise the input impedance and multiply the voltage tolerance, but the consequence is to degrade the noise floor even further.
Frequency resolution at maximum FFT length is 2.38 Hz from DC to 1 MHz. This makes it possible to see 50/60 Hz power line noise and intermodulation, switching frequency drift and instability, and other artifacts that would be obscured or be impossible to identify with the much lower resolution typical of scopes and spectrum analyzers.
FFT of the same supply, using a scope.
For switching power supplies used in audio equipment, the APx525 Series with BW52 has the additional advantage of being able to make both power supply and audio measurements with a single instrument. Since leakage of switching power supply noise into audio circuits is a major concern in these devices, it is often necessary to have a high-bandwidth view of the audio output. Likewise, by generating an audio signal with the APx instrument and observing the power supply, any resulting intermodulation products on the supply rails can be seen.
APx500 FFT of a Class D amplifier with 10 kHz audio stimulus. Note the intermodulation products centered around the switching frequency and its harmonics.
Being completely software controlled, many advanced measurement, automation, and reporting capabilities are built into the APx500 software that can be advantageous in switching power supply design. Measurement Recorder, for example, can make a continuous real-time plot of switching frequency, DC voltage, and audio THD+N simultaneously over a specified time period. Built-in automation with GPI control, as well as external automation via LabVIEW drivers and a .NET API, are included. And, the automated reporting allows you to save results to a Word or PDF document, so that you can compare FFTs and other measurements to past results when you make design or production changes.
Sound Advice: Measuring High Impedance Sources
Measuring high impedance sources, such as some high impedance audio transformers, can result in poorer high-frequency performance than expected. This is due to an input shunt capacitance of around 190–300 pF that is present in almost all audio analyzers. AP’s APx500 Series has a shunt capacitance of 220–230pF, while the 2700 Series has a shunt capacitance of 95 pF in balanced mode (differential) and 185 pF in unbalanced mode. Most of this capacitance comes from an input RC filter that removes unwanted RF interference. Defeating this filter is possible, but it can result in nasty problems with demodulation by the input stages, since audio analyzers employ relatively slow, low noise and low distortion op-amps. We do NOT recommend this approach!
To see the effect of the shunt capacitance on frequency response, take a 47 kΩ high impedance source as an example. If the effective source resistance is shunted by the input resistance, you get
If you include an additional 15 pF for the stray capacitance of the test leads, the predicted bandwidth (–3 dB point) is
which will cause significant measurement roll-off in the audio band.
Rolled-off response without input attenuator.
The best solution to this measurement challenge is to build an external attenuator that connects directly to the input of the audio analyzer. For example, a series 900 kΩ resistor shunted by approximately 25.6 pF should give a relatively flat response, with a 1 MΩ total input impedance and an effective shunt capacitance of about 23 pF. This is the same principle used in a 10:1 oscilloscope probe, except for the obvious value differences. We strongly suggest building this attenuator in a shielded box or enclosure to prevent unwanted pickup of hum and other interference. Provision should also be made for trimming the value of the shunt capacitance to achieve the best flatness.
Flat frequency response using input attenuator.
Test Results: AP News & Events
©2011 Audio Precision, Inc.