One of the biggest schisms among audio system builders in the last two or three decades has been that between the so-called "subjectivists" and "rationalists". In case you don't know what I'm talking about, these two groups are split as follows. The subjectivists say that if you can hear a difference in the sound due to a change in some part of the system or environment, then that perception is valid, even if they can't come up with any scientific explanation. In fact, many say that scientific explanation is not even necessary. The rationalists, on the whole, say that things that can't be explained by current state of scientific understanding, can't be measured, and can't be demonstrated by double-blind comparison tests, are figments of imagination.
There is a common meeting ground, which rationalists usually accept, but many subjectivists don't. This is the double-blind comparison test. In this sort of test, scientific explanations and conformance to accepted theory are kept aside, and two systems are compared in actual listening tests under controlled conditions. Rationalists are usually willing to accept that if something repeatedly makes a difference to the sound of an audio system, as demonstrated in double-blind tests, then there must be some basis for the claim. Many subjectivists too want to go with this approach, but some subjectivists have rejected this as too rigorous. This extreme subset of subjectivists inquire about the change in the sound of an audio system due to replacing one short, well-made interconnect cable with another. They comment about the possible change in the sound of an amplifier because the power-amp chip is fixed too tightly to the heatsink.
Anyway, these two groups have worked from opposite ends and have contributed to our understanding of how audio systems work and how we relate to them.
It is usually believed that engineers who build chips are in the rationalist camp. They live in the world of measured parameters and datasheets, not in the world of subjective tweaking of the sound of a CD player by putting maple-wood cones below the player's feet. These engineers are the people who have, even now, continued to use THD+N as the most important indicator of the quality of sound of an amplifier. These august notables claim that two amplifiers with "vanishingly low THD+N" will sound identical, unless proved otherwise in double-blind tests. Subjectivists find these remarks so odious that they refuse to invite such notables to their children's birthday parties.
It is therefore quite interesting, and revealing, to see that engineers too listen to music, keeping aside cherished technical beliefs. Read the National application note on the LM4702 audio amplifier driver IC. This application note says things which I have never read in an engineering datasheet or article about an amplifier chip before.
The first surprise was this passage:
The LM4702 Demonstration Stereo Power Amplifier achieves THD+N as low as 0.0006% with only 15 to 20mA of bias current. Increasing the bias to drive the class AB Darlington output stage into a more Class A region of operation was tested, but there was no improvement in THD+N or any other audio measurements. There was also no improvement in the critical listening tests performed at National Semiconductor's in-house dedicated sound room.
I have never seen any article from any amp chip manufacturer saying that they tried Alternative B after trying Alternative A, even though Alternative A was giving them a THD+N of 0.0006%. Till now, I have always seen an implicit belief that THD+N figures this low are well below audible thresholds anyway, hence any further experimentation is superfluous.
The second thing that caught me by surprise was the reference to "critical listening tests". These chaps don't just listen --- they have an "in-house dedicated sound room." They begin to appear almost human --- they probably even enjoy listening to music, perhaps without even hooking up the distortion analyser. In a lighter vein, one wonders whether this dedicated sound room has been accounted for in their books of accounts under the heading of Staff Welfare or Product R&D?
The next surprise was en passant: just a line at the start of a section:
Sound quality can be a very heated and subjective debate.
Interesting, coming from a community which till now seemed to believe that the only debate about sound quality was whether the THD measurement had been done accurately enough. But the most unkindest cut of them all came a few lines down the page:
In listening tests at National's sound room evaluating different circuit components used in the LM4702 demo amplifier, there was one part whose negative effect on audible signal quality was undeniable. A DC blocking capacitor on the input of the LM4702 degraded sound quality. In multiple listening tests, with different participants and at various locations around the country, the negative effects of even the best film and foil polystyrene DC blocking input capacitors in the audio signal path was confirmed. It is therefore recommended that DC blocking capacitors not be used in the signal path for mid to high-end audio equipment. Where DC offset from another signal source may be a problem then the use of a DC servo circuit that keeps DC offset from appearing at the output of the amplifier is recommended.
I couldn't believe my eyes. This issue of including versus excluding the coupling capacitor at the input to an amplifier has been a constant subject of debate, with proponents of both camps heaping derision on each other. I must admit that I have seen more derision on this issue from the rationalists to the subjectivists, often in the form of taunts referring to the subjectivists' ability to hear DC. After all, in the textbooks, the purpose of this capacitor was to block accidental input of DC into the power amplifier --- if this ever happens, it destroys the speaker that the amp is connected to. The rationalists could not believe that anyone could object to this capacitor, because after all, its presence would block only DC which no one can hear anyway, right? Subjectivists have timidly tried to suggest that a capacitor in the signal path is a distorting addition, irrespective of what its stated purpose is. Rationalists have laughed at these claims in rather unsubtle ways. Many such unsubtle examples are in evidence in online fora on audio system design.
Now, for the first time in my limited knowledge, a company known for solid left-brained engineering and with no business need to pander to audiophile fads has gone on record that they don't like the effect of the DC blocking capacitor on the sound of the amplifier. They also claim that they have confirmed this in multiple listening tests, we presume of the double-blind variety. The same claim coming from an audiophile amp manufacturer would be treated with skepticism by the engineering and rationalist community, because amp manufacturers have a business need to pander to current audiophile beliefs in order to sell their products to well-heeled and fashion-conscious audiophiles. But when a company like National with no such exposure and investment in the audiophile equipment market goes on record with this sort of observation virtually uninvited, the credibility of this statement is many times higher.
National didn't just stop at the remark on the need to get rid of the capacitor. They went a step further and made veiled references to differences between capacitors. They have referred to "the negative effects of even the best film and foil polystyrene DC blocking input capacitors", thus hinting darkly that other types of capacitors may be worse. Verily, oh verily, those engineers have gone over to the Dark Side.
I find these small remarks in one corner of one application note a serious watershed in audio system design debates.
On the one side of this schism were people like the great Seigfried Linkwitz, of the Linkwitz-Riley filter fame. He has broken new ground in speaker design, virtually changed the rules of the game, and has designed open-baffle speakers, where the speaker drivers sit on a plain sheet instead of on the wall of a box. His speakers, described on his Website are, by all accounts, absolutely among the best in sound quality, and both engineers and subjectivists appear to agree on this. But Linkwitz says the choice of amp to drive his speakers is not important. His opinions on amplifiers are well explained in his design notes on both the Orion and Phoenix speakers, both of which are active speakers and need one power amp per driver. Will someone like Linkwitz want to consider the mere presence of a good quality DC blocking capacitor a factor affecting the sound of his excellent Orion speakers? There are others from this school; G. Randy Slone, the amp designer, has been reported to claim that two well constructed amps with very low measured THD will be indistinguishable in double-blind tests.
On the other side is someone like Lynn Olson, a gifted audio system designer and talented writer. He has had a long career as a design engineer where he used all the analytical instruments and tools engineers normally use to design and measure speakers and amplifiers; his writings on amp design clearly indicate his familiarity with the quantitative and objective measurement approach. Yet, today he is one of the most vocal critics of using THD as any useful indicator of subjective sound quality, and an energetic evangelist drawing attention to the sound of passive components like capacitors. Till recently, Lynn Olson and Seigfried Linkwitz would never find much common ground to agree on these contentious issues.
I have DIY amp and speaker builder friends who have always claimed that the sound of things like capacitors is almost impossible to verify in listening tests. They have always said that they can't hear any difference between a physical piece of copper and, say, a large-value series capacitor of good construction (in their opinion, metallised polypropylene). Many have tried and failed to hear any difference, and will therefore laugh in abovementioned unsubtle ways when they hear Lynn Olson and others talk of the difference the Hovland Musicap in the Ariel's crossover made to its sound. From now on, Lynn's attempts to build a totally transformer-coupled valve amp may be taken a bit more seriously. Can we hope that some of our ratioinalist brethren will actually begin to wonder about the pros and cons of GNFB (global negative feedback) too?
From now on, such discussions will, or at least should, happen differently. Let us hope they become less unsubtle, at the very least. :)