Balgalvis: I agree. As it happens, I happen to have the same Lazarus amps that Dick reviewed, plus an additional pair that Lazarus sent me because the guy felt that I should drive them balanced into the Divas, two per side. Again, just to emphasize what was heard, I have to say I heard not much difference on the Divas from what Dick heard on the speakers that he used, which were the Quads. But the point is that you have to play with an amplifier with cables. And we get into an area that for me is really like a snake-infested area and has nothing to do with what Dick did with his review, which was an admirable job. But an amplifier is very difficult to review because you don't know what you're going to put in front and in back of it. I mean cables, speakers, etc. It becomes very very tricky, and I think that I agree with Dick: if you knew that there's a half-an-ohm output impedance for an amplifier, then you can start compensating for it, use a Celestion SL600 let's say, which has a high impedance.

Atkinson: Except in this specific case, I asked Bob to measure the Lazarus's output impedance because Dick had already found by listening that the sound of the amplifier was significantly loudspeaker-dependent. Dick had discovered without any help from measurements that the amplifier did have a serious interaction problem with loudspeakers, and the measurement was purely just to see if indeed it did have a high output impedance. The end result was right. Dick heard something; Dick investigated it subjectively; Bob's measurement just said "Well, here's why."

Holt: What is going to be the approach if, for instance, the reviewer says "This preamp has thin low end," and the measurement shows that it's perfectly flat down to 5Hz. Someone should interject something in there to suggest at least a possible reason for such a discrepancy.

Galo: I think it's necessary to go beyond the measurements into the circuitry to try to explain just what it is that's going on. If an amplifier, for example, appears to be lacking in dynamics, well, what are the output impedances of the circuitry? How is the power supply designed?

Holt: How would you even define this?

Balgalvis: Having had conversations with a number of manufacturers who said confidential things, I tell you, there's almost a consensus out there, they say, "We did this circuit, we do certain things, and it comes out sounding a certain way." I say, "Well, do you know exactly why it sounds like this?" and they say "Nope."

Norton: They don't know either.

Balgalvis: They don't know. The designer, in many cases, says something like "We tried this new circuit, it sounds better now, and this is how we're going to build it." It's really sometimes frightening to hear this type of honesty, but it's out there. So for us to go into a circuit that they designed and find out why it sounds a certain way, seems to me to be a very difficult task...It's very dangerous and we may end up in the situation where we start making comments that are not technically valid.

Norton: If we think we're going to be able to explain why the measurements, why everything sounds the way it does, in some cases, such as the Lazarus, we may be able to. But if we think we're going to be able to do it consistently, we're fooling ourselves.

Holt: Over a period of time I'm willing to bet a lot of things are going to start falling into place.

Norton: Over the long haul, we may come up with some statistical evidence that gives us some leads on individual products.

Greenhill: There's one other element. In reviewing using measurements, a great deal depends on the person doing the measurements. Even with a system like the Audio Precision. Over a period of time, I think, Bob is going to have a chance to learn the craft of measuring which will enable him to rapidly get to the heart of the matter. After you take every measurement exactly the same way a number of times, you begin to look for anomalies that appear and you begin to track them down...I think that if Bob begins to test more and more equipment, he's going to get a feel for when something doesn't seem to be quite right. Maybe he'll start off with ten tests of an amplifier, but as time goes on he'll have another five optional ones as the software is upgraded. He'll begin to use those other optional ones and begin to find interesting things where the component begins to stop performing well. And that's very helpful because he can then comment on what may have happened in a specific situation.

Atkinson: There's two points raised here: I think the first was very important that, though people in general regard measurements as objective, in fact they're not. The whole decision on what measurements to make and how to carry them out is a learned experience. Anybody doing it will go through quite a long learning process about how to carry out the measurements and reject spuriae. What the instrument tells you is not necessarily the truth and you have to learn how to distinguish between the truth and something due to a bug...

Greenhill: You can't get the measurement to work, you can't get it to work, you can't get it to work, you blame your equipment, and all of a sudden you realize it's the connector or something. There was one amp I had from a very well-known manufacturer, the terminals were wired backwards! It became very clear that you couldn't get the thing to work and there was a reason for it.

Atkinson: The second thing is that I see carrying out measurements on a component as being analogous to chemical analysis. You have a tree structure of tests, each one giving you a yes/no answer. You carry out your first test; did it do this, yes or no? And that will lead you in two paths, and then you have more tests which lead you in further bifurcations until you end up down one tiny little twig which is where the truth lies.

Greenhill: As long as you don't print all the standard things that begin at the root. Just show the reader the ones that you found significant.