Tonight I breadboarded my first "serious" headphone amp, measured it & did some listening. OK, but not spectacular on the subjective, into the limits of the measuring instruments on the objective. My full-sized system beats this initial amp w/Senn 600s readily. Obviously, there is much more to do on the amp, but I knew that before power-up. (Hmmm - is that why I'm not dazzled?)
Much of what we hear can be measured, with the right techniques, of which 1930s THD is just one of an arsenal of methods needed (some of which haven't been discovered yet). Years ago, I found that perception of distortion is much more sensitive to IMD than THD. Ordinary amplitude clipping produces much more IM than HD in complex-waveform simulations. Yet so many measurement-oriented folks remain fixated on THD (this is less so in Europe). And the vast majority of published data involves THD or 2-tone IMD, nothing more complex.
This gets more interesting when feedback is considered. When I listened to a bandwidth-limited, measureable delay, low-distortion amp with and without FB, the FB worsened the sound. The FB was obviously trying to correct a condition that had already passed and, therefore, adding a stale corrective factor into a later input signal. Does this mean that FB is always bad? Or that this particular set of propagation delays and bandwidth limitations just wasn't compatible with the particular FB used?
Phase distortion and phase modulation in particular aren't considered in traditional audio measurements, yet we know they're audible.
The popular steady-state tests reveal few of these phenomena, as feedback aligns itself to correct the signal within the time frame of the measurement process. Could other, transient-oriented tests reveal them? Maybe not all phenomena, but certainly more than what we obtain from the traditional methods.
So, where are the techniques, and why haven't they been developed? First, they don't illustrate theoretical principles being taught at the undergraduate level, so there's little market in the university laboratory environment for test equipment & techniques that don't compliment the traditional educational sequence. Second, some of the folks with the best potential ability to develop correlational techniques simply reject objective measurements out-of-hand. If half the effort expended in arguing the merits of subjective vs. objective went into the development of better-correlated measurement techniques, objective measurement and subjective result would both take a leap forward, I would think.
We have computational and simulation tools available now that were unimaginable when I was an undergraduate 30 years ago. Hopefully, some bright young minds will step outside the box and develop DSP applications that reveal more of the anomalies we hear, thereby providing the means for more rapid sonic improvement of products.
Appropriate measurements help us understand what we hear. Improving measured results is a tool that's often helpful along the road to better sound. But the goal cannot be solely the measured data, as that's not what satisfies the customer in the end. That doesn't invalidate the usefulness of measurement, it merely establishes some limits on it.
The measurement techniques I have available on my bench won't, by themselves, help me improve my amp's initial sound. Steady-state tests won't reveal a dynamic regulation issue in a power supply, for example. But they'll come in handy every now and then during the evolution of this project, which I will know is complete when it matches or beats the sound of my full-sized system.
|