Regarding the discussion about null tests as a measure of accuracy of cloning.....
Null test is neither completely meaningless nor the ultimate measurement. It is simply one of many possible performance metrics. In my opinion, it is also a valid scientific measurement method. For example 'sum of error' is very standard way of measuring in signal processing field, I guess.
In a very specific case, where a clone reproduces the reference gear with 100% accuracy, the null test result would be negative infinity. There is no debate there.
If a clone is not infinitely perfect, then there are several possible outcomes. For example, clone X may reproduce transients better but be less accurate in the overall frequency spectrum. Clone Y may be better at mimicking note decay and cleaning things up. Clone Z may not be the best in either transients or spectrum, but may offer the best overall balance.
To those who argue that a single number cannot summarize the complex behavior of a cloned amp, I would ask: what is the alternative? Endless debate about what sounds better? And if you already have a well-established personal taste, then you probably do not need advice or guidance from others in the first place.
To follow the car analogy, imagine you are talking to two car dealers.
Car dealer 1:
“This car feels good, sounds nice, goes fast, and is fuel-efficient. That’s what I think, and other people agree.”
Car dealer 2:
“A number cannot represent the whole driving experience, but this car does 0–100 km/h in 5.3 seconds and gets XYZ MPG.”
Personally, I would choose dealer 2. Dealer 1 is not necessarily bad, but only if I can test the all available cars myself which takes much time and effort.
Coming back to modeling or cloning, a null test result cannot represent every aspect of the sound, but it can still be a useful guideline.
