The bark fix

In a recent post, The Bark Syndrome, I identified a problem that is difficult to pinpoint its source. Barking voices. You've no doubt experienced on multiple recordings the bark of a loud voice. The signer gets going and suddenly you wince. Your face scrunches. Where in the chain does this change in quality occur? If we look at the raw waveform for signs of clipping or dynamic compression, we see nothing alarming. I first ran into this issue back in the early 70s when Stan and I started PS Audio. At the time, we were building only phono preamplifiers. Those preamps started out life based on an IC op-amp, the venerable 709. It wasn't long after those first circuits that we became dissatisfied with the limitations of an integrated circuit operational amplifier. Among them was an annoying tendency to sound musical only on softer passages of simple music. Whenever the levels got too high or the music became complex, we winced. The problem with integrated circuits is the designer's inability to change anything internal to the chip. The chip does what the chip does. Our first cure for facial wincing came in the form of a discrete op amp. By using our own discrete transistors and resistors to mirror the structure of an operational amplifier (differential input pair feeding a gain stage whose output fed a buffer), we could solve one of the IC's biggest problems, that of sounding muddled when the music became complex. We theorized our discrete amplifier was free of this problem because we were finally able to control the IC's over-the-top high open-loop gain. (What this means is that when an amplifier has extremely high input to output gain to the point of instability without resorting to feedback to lower that gain to something useful, we tend to suffer from the problem I described earlier of complex music becoming cluttered. By ensuring the amplifier's gain structure and speed are capable of stable operation at high gain without the necessity of feedback, much of this problem is no longer heard by the listener). Now, with the clutter problem solved by the use of a discrete circuit in which we could control the circuit's gain parameters, our next challenge was the Bark Syndrome. That had unfortunately remained and was for a long period of time a real head scratcher. Why, when the maximum output signal the preamplifier had to deal with was a mere 2 volts (and our circuit could easily output 7 volts without any increase in distortion), did signals approaching that level tend to make our faces scrunch? On a whim, we decided to see if it might be something we could not yet measure. Headroom or more accurately, exceeding the linear region of the circuit. Because we had moved away from fixed integrated circuits to a discrete version we were no longer bound by the power supply limitations of the chip. (Integrated circuit operational amplifiers are typically limited to +/-15 volts on their power supply rails). Easy enough for us to rejigger the power supply rails to double what we had been using. Now, with +/- 30 volt rails we played again the same passage of music that had formerly been barking. Voila! Less bark and by a rather noticeable margin. Just a bit of history for you to consume.