Going against the grain

I promised in yesterday's post I'd tell you what I preparing to implement to improve the high frequency performance of the proposed PerfectWave power amplifier, and here it is: I removed all the feedback. Simple as that may sound there are some serious problems doing that and I'll try and detail them here without getting too technical. First and foremost it is 100% counter to what Bruno Putzey, the brilliant designer of the Hypex output stage we're probably going to use wants. So adamant is Bruno that you simply cannot have too much feedback, he's written an article called "The F-Word" and why no amount of negative feedback is too much. If you're interested you can read a copy of it here. Bruno is a lot smarter than I am. Read his paper if you want to be dazzled. Me, I am just a designer who knows what he hears. What I hear is a far more open transparent sound signature when there is less feedback in the input stage of an amplifier. So bucking the numbers guy is my first hurdle. Ok, I am over it. Second hurdle is the Hypex module itself. The Hypex modules require the input stage to have low output impedance. In fact, the closer to zero Ohms the better the measured response of the amplifier. Negative feedback helps you get low output impedance. It helps a lot. If one were to focus all energies at achieving near zero output impedance (at frequencies as high as 2mHz) to properly feed the Hypex module, then the first thing you'd do is apply gobs of feedback. That is, unless you listen to the results. That's something I do. The third hurdle is distortion vs. output swing. For a proper amplifier you need low distortion and high output swing so you can have unlimited dynamic range within the constraints of the amplifier and the system. As we've discussed many times in these posts, dynamic range is a big deal and modern digital audio has gobs of it. The last thing you want in a power amplifier is the sound changing when the music gets loud, We all want an effortless, transparent sound to our systems. To achieve both low distortion and high output swing, designers typically employ feedback. That's something tough to do if your goal is zero feedback. There are plenty of other challenges as well but this post will get too long. So with those challenges in mind, I set out to design a zero feedback input stage with extremely low output impedance, low distortion and large output swing. To manage this challenge the circuit board would take a dramatic shift in layout. The original board was filled with mostly amplification electronics, accounting for 80% of the board real estate, the remaining 20% afforded to power supply. The new board is just about the opposite. Here's a picture of the new PCB. The area closest to the technician's right hand, behind the 4 big yellow film capacitors is the gain stage, consisting of two MOSFETS, one for each side of the balanced stage. Everything else on this board is either output stage or power supply. I'll try and let you know a bit more in the next few days. While technically a great input stage and one I am certain will sound great, it still has to prove itself in the environment we need it to work in. That's what I meant yesterday in Rolling the Dice.