Excellent question Andrio, and as luck would have it you chose the exact same site I was talking about. You will find a thread in the Lounge I started where I described this site in particular as it is setting out to PROVE that there are differences but in fact their own results show that there are none of any note. If you can wade through the snake oil of course.
Firstly ask yourself "why have they posted that site at all"? If you look closely you will find this at the bottom of the page:
Amplified Parts wrote:By Kurt Prange (BSEE), Sales Engineer for Antique Electronic Supply - based in Tempe, AZ. Kurt began playing guitar at the age of nine in Kalamazoo, Michigan. He is a guitar DIY'er and tube amplifier designer who enjoys helping other musicians along in the endless pursuit of tone. [My Bold, BB]
Notice Kurt Prange is the "Sales Engineer". That highlights the fact that they are selling valves (and other things) so this is actually a piece of marketing copy. It's to their advantage to spread this sort of belief as it increases sales.
Now, ignore the first 95% of the page with its pretty pictures of 12AX7s and the Red/Grey/Yellow bargraphs representing the responses of each model and go to the ONLY piece of relevant information on there. That's the single graph right at the base of the page which shows the frequency response of all of the graphs laid over each other. Did you spot that one? If you did, were your eyes quickly drawn away to the pretty descriptions of the valves above it with their captivating but meaningless diagrams because it was a "bit techy"? I have attached a slightly more informative version of it. It is a common thing to only show part of a graph as it expands and shows up the fine detail better but it also distorts the relevance of that fine detail to the overall picture. When the full graph is shown the differences are a bit less significant. I won't criticise them for that point but it is relevant to our discussion. Click on that graph at the bottom of their page and it will expand to a better size. Is minimising the only bit of real info, perhaps to make it less clear and a little more insignificant, a marketing strategy or not? I'm not sure on that one either.
Look at the individual plots on that graph. You should notice a couple of things. Firstly each one is sitting at a different level. That level shows the gain of the circuit they have been tested in. Putting the valves into a Fender style preamp and measuring each of them within that circuit is a reasonable way to approach things but it would have been better to show some measurements of the valves' real characteristics in a bland test bed too. Anyway, you should understand that the plots are related to a valve's characteristics but are not a direct measurement of them.
Look at the vertical axis and see that it is graduated in Decibels representing the voltage gain of the preamp with each valve in place. Do you really know what the difference in decibels means to the human ear? The maximum difference between all of those models is about 2.8dB and that includes a rogue valve (the lowest yellow plot) which is an odd amount below all of the rest. In truth, statistically speaking, that one should be given less significance as should the top one. Every other plot lies within about 1.6dB of each other. That is a ridiculously small spread and your ear would be hard pressed to even detect it in the real world. For a bit of background try going to this site,
Harbeth Sound Level Test, and IMMEDIATELY go to post #11 and take the test. Listen to the 4 samples and see what order of volume you put them into.
Now to the REAL stuff! The levels of those plots are really irrelevant. Ok, so each valve has a very slightly different gain within the circuit. (Valves do not have a stated "voltage gain" of their own.) Don't we have Volume controls to trim that if it becomes noticeable? The REAL point becomes obvious when you slide each plot over each other. The SHAPE of each plot is the only measurement of their "tonal signature". They are all the same! Exactly the same!!! The slightest less than 0.1dB differences at any particular frequency are totally irrelevant in audible terms and are within the spread we could expect from even different valves from a single manufacturer's output, and the test equipment even though it was good quality, and variations in the test conditions like the warm up time of each valve or mains variations. The manufacturers spread of characteristics is around +-40% for valve transconductance, anode current is allowed to vary around +-20%, what is the relevance of those comparisons when even valves in each manufacturer's acceptable range can vary by far more than those results show? (
"Getting The Most Out Of Vaccuum Tubes" - Page 64" The whole thing is a complete fudge of the reality of the results.
So why the pretty Red/Grey/Yellow bargraphs? Surely they show there are differences? Let me show you what that is about with an example. Have a look at the Dummy Valve Comparison I set up. I made up some data similar to the frequency plots and chose a value for Bass, Mid and Treble just as they did. I then took the values at those points and plotted them on a barchart just as they did. That is what it looked like, just like theirs. Now from that you could say that "the blue valve has a well balanced frequency response with good tight dry bass, bell like natural mids and a smooth transparent treble, while the orange valve has strong bass and mids with a slightly recessed delicate treble." The graphs show that don't they? However if you look at the horizontal axis you will see that, even though the length of the bars looks impressively different it actually represents only a range of 0-0.05 MAX. A tiny difference of
less than 0.05% in the real original measurement values, way within measurement error or natural characteristics spread, has been stretched out to look highly significant. Now take a look at the individual barcharts on that site and you will see that they are designed to take advantage of the same effect. They even recalibrate the tiny decibel differences to a fake scale of 1-10 to disguise the truth. If we accept that the maximum difference between all of the valves is less than 3dB at any frequency, each step in value from 0-10 represents a difference of less than 0.3dB which is inaudible. Those charts are nothing more than a marketing ploy. In common with the latest rash of TV adverts you could say that they utilise "Visuo Lookyseewhatwewantyoutosee Technology".
The interesting thing is they have done a pretty good job of the measurements but then if they fudged those "real" results it would really be too obvious and perhaps open them up to accusations of malpractice. Keep on questioning, it's the only way to the truth. No salesman is ever going to tell you that. Nor is any self styled guru who doesn't understand too much of this anyway. They just "know what they know what everyone knows and says".