During the first power amplifier shootout it became evident quickly that the new power amplifier we are proposing to go to market with was great in many respects, but lacking in a few key places, specifically in the upper treble regions of frequency response, a lack of openness and a propensity to fall apart when the music gets very loud.
Between shootout one and two we've made great progress in improving these problems but haven't yet fixed them to our satisfaction and this may necessitate a major redesign to get where we need to go before this product is ready to ship. I promised yesterday I'd detail the areas where we made those improvements that we've achieved to date.
At first thought a lack of openness and extension in the upper octaves would seem simple to fix: just extend the bandwidth of the amplifier, improve the transient response, and bingo - you're in. Unfortunately this rarely is the case as evidenced by the number of power amplifiers who seem to have great openness and extension but less than great transient and top end measured responses. In fact, there's little to correlate what we're hearing with what we're measuring in these areas. Point of fact, our amplifier is flat to 20kHz and has great looking square wave and transient response characteristics but from a sonic standpoint seems to have the "top octave" lopped off. Of course it technically does not, but that's what it sounds like.
So, how did we make the major strides we have accomplished so far? By lowering the feedback of the gain stage feeding the amplifier. The gain stage we started with had between 80dB and 90dB of global feedback (depending on frequency) and we reduced this to about 27dB flat up to 100kHz. This opened up the top end and extended the perceived upper octave energy a lot - an awful lot - but it still doesn't get the magnitude of change we need.
The power amplifier stage that precedes the input gain stage we're using needs an extremely low impedance output up to about 1mHz to work properly and the lower we make the feedback the higher this impedance becomes - so it's a somewhat tough game we're playing here.
I am convinced that there will be multiple steps to get where we're going, the first of which will be to completely redesign the input stage to have low distortion and output impedance with little or no global feedback - that's quite a challenge but on the trip back from shootout number two the circuit popped into my head.
We're on the trail! Tomorrow, we'll get back to cables. If there's been any amp breakthroughs I'll let you know.
