Q: I get CA&E in South Africa and it's an excellent magazine. I have two questions, which I know you can answer. I'm busy installing a Zapco DC650.6 amplifier in an Opel Corsa and we want to compete in IASCA SQ events. The component system consists of Morel Elate 5" and Morel Supremo tweeters. When I use the Creos FFT real-time analyzer with a pink noise CD, I notice a dip from 19,200Hz and above. It's almost like the system cannot play 20,000Hz, but when I use the frequency generator the system is capable of close to 23,000Hz. We've seen the same problem in my friend's car with a full-on Brax system. What would the possible cause of that be?
My second question is that a lot of professional SQ installers and competitors say that they don't use any time alignment at all - only equalization. If they say they use only equalization, how can they possibly get the left and right speakers of the front stage to sound like they're equally spaced from you (as you know, SA is right-hand drive so the right mid and tweeter would sound much closer to you). I can get a perfect focus and image using time alignment, but you can still hear that the right mid and tweeter are closer than the left. How can I improve this?
Thanks,
Abrie Rautenbach
A: You have been afflicted with a case of Nyquist's Criterion. Before you run off to the local doctor for some industrial-strength antibiotics, let me explain this.
Digital audio is a slave to a rule of technology named after Nyquist-Shannon sampling theorem. Although it's accompanied with a pile of formulae, the concept is pretty simple to grasp. Essentially, the rule is that if you want to digitally sample a sound to store it and play it back later from a digital file, you need to sample at two times the highest frequency you want to hear to prevent sonic trash from occurring. When the CD player and its digital characteristics were first invented, engineers and scientists knew we want to hear at least 20,000Hz, so they rounded it up to 22,050Hz and doubled it to get the Nyquist number of 44,100Hz sampling rate (commonly seen as 44.1kHz on product spec sheets).
The A/D (analog to digital) converter used in the encoding process takes a series of "snapshots" at the rate of 44,100 per second and stores each one. If you can imagine the sound of a camera clicking away that fast, you begin to wonder "is there any way I can keep that high-pitched whining out of my music?" By using a "brick wall filter" you can keep over 90 percent of the noise away from the musical information and the remaining bit of hash is covered over by "masking."
Most filters, like a crossover for a tweeter in your speaker system, are first or second order, meaning that every time the frequency is cut in half (an octave), the sound passing through the filter is 6dB or 12dB quieter. Cutting a huge amount of noise that started at 44.1kHz by 6dB before it hits 22kHz isn't nearly enough, so we use steeper filters. In fact, by trying to wipe out 90 percent, we're really looking for a filter that can stop over 90dB in the one octave that exists between 22 and 44kHz. This filter, when viewed on a graph, isn't a gentle slope but is almost vertical like the "brick wall" name it earned.
Your FFT analyzer (not many car audio installers have one of those) shows you that the output from your CD player isn't as accurate as we had hoped, and instead of allowing your pink noise test disc (hopefully a precision piece of software) to play the full bandwidth up to 20kHz, it chops off the last 800Hz. It's also possible that the pink noise happens to be recorded with that drop-off.
So, is this a real problem? First off, you're missing 800Hz of an octave that's 10,000Hz wide, which equates to 8 percent of the octave. If this was in the middle of your most sensitive hearing band (about 500 to 3,000Hz), you might actually notice something if you use really good drugs. Second, if you are over 9 years old, you can't hear over 19,000Hz anyway. If you're older than 20, then I'd bet you couldn't hear pure tones over 17,000Hz. Sucks to get old, doesn't it? Finally, I mentioned a pure tone, because that's how my last hearing tests were performed when I was being custom-fitted for Ultimate Ears UE-10pro in-ear stage monitors. If you think I'd stoop so low as to buy a set of Bose noise-canceling headphones, you need to check out these $900 ear buds!
When you're dealing with music, or even pink noise, you have a new psycho-acoustic phenomenon called "masking" that will be sure to hide the fact that there is anything missing. Yet, when you plug your signal generator into the system downstream of the CD player, your system can now deliver 23kHz at the top end. Whether you're introducing a tone or pink noise from your generator, it doesn't matter because the rest of the system can pass it through since you're now using a piece of test equipment that isn't limited by Nyquist.
This means that your system is great and your CD player is probably typical. If you plugged in a DAT player using a tape recorded at 48kHz instead of a CD player, the Nyquist limit is 24kHz. After filter problems and other errors, your entire system would still be good for at least 3,000Hz above the limit of perfect hearing in a small child.