Created on 2008-11-21 20:54:00
Question
We need to conduct distortion tests of power amplifiers during design and development, before heat sinks have been installed. Without heat sinks, the DUT can only withstand a duty cycle of approximately 0.2 seconds on and 0.8 seconds off. In addition, the amplifiers may have an onboard DSP, so the test needs to account for a delay in the DUT. How can we do this test with the AP2700?
Resolution
To conduct these measurements with a very short duty cycle, we recommend using the Arbitrary Waveform generator in the AP2700 with the FFT spectrum analyzer.
First, we construct a waveform file that has the appropriate duty cycle. Please see the attached macro file "Burst THD_Mod1.apb" and the accompanying test file "Burst THD_Mod1.at27". For this sample test, we have set the output sample rate at 16,384 Hz. Given the generator buffer length of 16,384 samples, this provides a generator buffer 1.0 seconds long. In this sample project, the waveform loaded into the generator buffer consists of six cycles of a 32.0 Hz sine wave with the rest of the 1.0 second buffer padded with zeroes. This results in a duty cycle of 187.5 msec on and 812.5 msec off (see Figure 1). In the example, the Auto On feature of the Analog Generator is used. As a result, the system has plenty of time to turn the generator off, preventing the DUT from being exposed to successive bursts of signal.
Figure 1. Plot of the waveform used in the Burst THD_Mod1.at27 test
In the test file, note that the Input sample rate is set to HIRes A/D @65536 and the FFT size is 4096. As a result, the FFT transform buffer length is 62.5 msec, or exactly 2.0 cycles of the 32 Hz sine wave. The frequency of 32 Hz was chosen as a synchronous frequency – an integer number of cycles fits exactly into the FFT buffer, allowing a leakage-free FFT without windowing.
Note also that the FFT start time has been set to 93.75 msec. This allows for a delay time in the DUT. This value was arbitrarily chosen as being one half the length of the signal-on time.
In the test file, the sweep table for the FFT spectrum has been set up to have 10 points (the fundamental frequency and the first 9 harmonics. The macro runs the test and then computes the THD as the RSS (root sum of squares) summation of the 9 harmonics relative to the fundamental. The THD number is shown below the spectrum graph when the macro finishes (Figure 2).
Figure 2. Spectrum from the Burst THD_Mod1.at27 test and the calculated THD.
A second set of files named "Burst THD_Mod2.apb" and the accompanying test file "Burst THD_Mod2.at27", provide the same measurement, but this time with a signal at a frequency of 1024 Hz.
For the IMD measurement, please see the attached macro file "Burst IMD.apb" and the accompanying test file "Burst IMD.at27". This test uses the same principles as the THD measurement above. However, in this case the IMD test uses sine waves with frequencies of 18 and 20 kHz. To accommodate these higher frequencies, a higher sample rate must be used in the signal generator. Unfortunately, the higher the sample rate is, the shorter the time length of the generator waveform buffer. In this case, we selected a sample rate of 51.2 kHz. This provides nice whole number frequencies that fit within the generator and FFT transform buffers, and it provides a generator buffer length of 320 msec.
Figure 3. Signal used for Burst IMD measurement in test "Burst IMD.at27"
Figure 4. Spectrum from burst IMD measurement in test "Burst IMD.at27"
Based on the measured FFT spectrum, this macro computes the second and third order IMD difference products and their RSS sum. It then calculates the Total RMS IMD as the ratio of the above distortion products to the overall RMS level from 50 Hz to 20 kHz. The results are listed in the Immediate window of the macro panel (Figure 5).
Figure 5. Results of IMD calculations in macro "Burst IMD.apb"
RELATED DOWNLOADS
Series of macros written to support KB article https://www.ap.com/kb/87 on THD & IMD Measurements of Power Amplifiers with Short Duration Signals