Sunday, August 18, 2013
Audio System Signal Validation
My goal today was to couple my signal generator, Jensen isolator, and MiniDSP in a circuit and measure the output with my oscilloscope at locations throughout the circuit to determine the loss through each device, the “final” voltage that would be sent to the amplifiers, any clipping or other distortion, and, to a lesser extent, the frequency response of the circuit. I began by connecting my laptop to the MiniDSP and defeating all filters since I knew I'd eventually be changing the frequency to sweep the audio band and did not want to affect the output levels.
Test 1: Look for clipping at maximum signal level
Earlier testing with the DEH-80PRS headunit in the circuit revealed a maximum output of 4.81 Vrms. When I plugged that number into the signal generator I noticed a very slight clipping on the negative half of the waveform on the output from the MiniDSP. Once I reduced the signal to 4.7 Vrms the clipping went away. It is highly unlikely I will operate the system at that volume so I am not immediately concerned about this.
Test 2: Verify the differential output from the isolator and MiniDSP
This test was designed to verify the circuit would accept a single ended signal and produce a balanced signal suitable for use with the MiniDSP and ultimately the Arc amps. As you can see in the following oscilloscope screenshot the yellow trace (channel 1) is displaying the "positive" (in-phase) signal, while the blue trace (channel 2) is showing the "negative" (180 degrees out of phase) signal. Viewed together they form an “eye” pattern. When the first signal is subtracted from the other, the amplitude of the resulting signal at any point in time is the vertical distance between the two signal traces.
Normally, the way to measure differential signals is with a differential probe. I don't own a differential probe (yet) so I used two probes in combination with the scope's math function to subtract the voltages received and thus calculate and plot the differential voltage shown in the purple trace. You may note that the vertical scale of both channels and the math function is set to 2 volts per division. I did this intentionally to highlight the fact that the differential signal amplitude is twice that of the individual input signals, as expected. The real-time voltage measurements are shown at the bottom of the screen color-coded appropriately. Notice that the blue signal voltage is 2.49V RMS while the differential signal is 4.96V.
This should explain that while it's possible to connect a single-ended / ground-referenced signal like that from the headunit to a differential input like that on the “Balanced” version of the MiniDSP, doing so effectively discards half of the signal amplitude. Proper conversion between single-ended and differential interfaces is thus required. As I have indicated previously I think the best way to do that is with a transformer. It is not, however, the least expensive solution, which explains why all modern mobile amplifiers that support balanced interfaces do it with active components.
You may note the output voltage is higher than measured the other day, despite provisioning my signal generator to be slightly lower than that measured at the output of the headunit. I believe that is due to some high frequency interference that appears to be getting into the MiniDSP or generated by it somehow. That may explain the "spikes" seen on the peaks of the purple waveform. I know this not an artifact of the math function because when I zoom in I see the same spikes on the individual signals.
While researching the problem in the MiniDSP forums I found several older threads in which people complained about high frequency noise (hiss) with these units. Most of the issues seemed confined to the mobile environment and their tech support people suggested it was due to grounding issues. These problems were reportedly solved by using the MiniDC DC-DC converter / isolator specifically built to address these issues. Considering how inexpensive they are ($12) and the fact that they also provide a remote turn-on delay circuit that may come in handy I'll probably wind up buying one. The MiniDC documentation indicates the unit's delay circuit can source a maximum of 100mA so I naturally asked Arc how much the KS125.4 sinks via its remote turn-on circuit. They responded with 1.5mA per amp or, in other words, well within the capabilities of a single MiniDC.
Before I gave up on this test I decided to run the output of the MiniDSP through the FFT function of my scope. In the above screenshot I've zoomed in to show a frequency range from DC to 35Khz. The peak seen to the left is the fundamental frequency (1Khz test signal) and the decay that occurs to the right is typical. There are some artifacts in this plot due to the limitations of the scope (it's not a spectrum analyzer after all) but the general plot seems to indicate no spurious signals or other undesirable content. That leaves only one thing left to do: connect an amp and speaker and listen to the output.
Test 3: Frequency Response
Using the jog wheel of my signal generator I manually swept the generated frequency from 1KHz to 25KHz in 1KHz steps looking for changes in the characteristic output from the MiniDSP. I confirmed that with the MiniDSP filters disabled the circuit has effectively flat frequency response from 1KHz to 21Khz.
At 22Khz, however, the signal drops off a bit but otherwise appears distortion free. At 23Khz the signal is reduced considerably and an odd and unexpected oscillation in amplitude occurs as shown in the screenshot. I reduced the signal level a bit (as shown in the measurements at the bottom of the display) in attempt to fix the problem but that had no effect. This may have something to do with the brick wall filter designed to attenuate output beyond the Nyquist frequency of 24KHz and may very well be normal but I'm not sure. This is one of the many reasons I wanted the MiniDSP to sample at 192Khz but unfortunately I cannot change it.