Sounding harsh and compressed

Yesterday we pointed out that the vast majority of IC op amps were not designed to be high performance audio amplification devices, but rather general purpose functional amplifier blocks with a nearly infinite amount of gain. The problem we find with these is that when we feedback all that gain to something usable the audio quality is compromised. The "infinite" gain we are referring to isn't, of course, infinite but it's large enough to stop the op amp from working should you remove the feedback altogether. This state of no feedback is called open loop gain (where the feedback loop is open and not connected) and closed loop gain is what we get when we use feedback to set the gain. So what does one hear when you use an IC op amp with all this gain? It really depends on the op amp. Bipolar based op amps like the classic NE5532, used by so many designers of audio equipment, can sound a bit brittle on the top end - transistory and slightly compressed overall. The best way I can describe it is impressive and powerful sounding always - but the harmonics and overtones are slightly etched sounding and overall dynamics sound a bit like trying to squeeze the music through cheesecloth. I know, this description is going to freak out the engineers reading this. Sorry. I just have spent too many hours listening to all these variations to really care. Another classic favorite of designers is the TL072, a JFET input device that has relatively low open loop gain and the softer sound of FETs on its input. While it doesn't have this harsh transistory etched sound its softer presentation is at the expense of the overtones, upper harmonics and extended ringing of notes of acoustic instruments - it also can be far more compressed sounding than the more powerful 5532. This is a deceptive little device because it always sound pleasant, musical and good - just on extended listening it loses too much of the life of the music. These characteristics seem to amplify themselves with increased feedback. So, for example, if you use these devices as what we call followers - 100% feedback and a gain of 1 vs. using them at 20dB or even 40dB, the sound closes down and becomes more compressed as the feedback increases and the gain decreases. As you remove feedback and increase the closed loop gain the life of the music returns. Why does this happen? There are many theories about this including the famous work ofEero Leinonen, Matti Otala, and John Curl in their 1976 AES paper on Transient Intermodulation Distortion and later Slewing Induced Distortionwritten by Walt Jung, to black magic, the weather and so on. The flat earth crowd (those who believe if we can't measure it then it doesn't exist) to the "I'll believe anything" crowd and everyone in between, I have yet to hear a perfectly explainable argument one way or the other. What I can tell you is that after years of listening to these op amps I can almost always spot the sound of them in designs without even knowing they were based on them - so distinct are their signatures - and I am not alone. So when it comes to open loop gain, how much is the right amount? Here's the rule of thumb: if an amplifier design can function reasonably well open loop (no feedback) then its chances of sounding open, dynamic and musical alive are excellent. If it cannot function without feedback (as most op amps can't) then its sound is dictated mostly by the type of semiconductors used and the amount of open loop gain it has. So here's problem number 1 spelled out for you: there are few, if any, IC op amps with reasonably low open loop gain that are available to audio designers. This means pretty much all IC op amps have massive amounts of gain that has to be fedback to set the closed loop gain and designers then have to be careful with limiting the input bandwidth so you don't get nasty ringing and harsh artifacts from fast transients and out of band noise and other tricks of the trade that help these high gain devices sound reasonable. If you don't use an IC op amp, but rather design your own discrete version, you can control exactly how much open loop gain you want, and what sounds and measures the best for your particular application. With an IC, you don't have a chance. Tomorrow, we cover strike 2 - power supply voltage.