If you’ve played with or read about tubes for longer than fifteen seconds, you’ve probably come across the concept of ‘tube rolling’ and the amazeballs things it does for sound. Proponents may claim transformative improvements to be had by rolling a tube or two in an amp and some will swear by d-getters/round mica spacers/black plates/red bases/etc.
But WTF is tube rolling?
1. the act of swapping one fully-functional vacuum tube in an audio device for another of a different brand and/or tube type
2. use of the inherent manufacturing variation in vacuum tubes to change the character of an audio device through trial and error
3. an act not-unlike navel gazing whereby the actor derives masturbatory satisfaction from perceived improvements that may or may not be founded wholly on expectation bias
Some define tube rolling less generously than others. Like most things in life though, the truth is probably somewhere in the middle.
A tube’s suitability for a circuit is largely defined by its characteristics Mu, Gm, and Rp. These are the tube’s theoretical voltage gain, transconductance, and plate resistance respectively. Two tubes of the same type are assumed to have the same characteristics. Two tubes of different types can also have the same characteristics quoted on the datasheet. Thing is, tubes are slippery little shits. The characteristics (which we assume to be immutable) can change drastically depending on the operating point you choose and what you ask the tube to do.
Here are the plate curves for a 6SN7 (Mu 20, Gm 3 mA/V, Rp 6,700 ohms):
Here are the plate curves for a 12AU7 (Mu 20, Gm 3.1 mA/V, Rp 6,500 ohms):
Setting aside the difference in socket and heater configuration for a moment, these tubes look completely interchangeable if you only take into account their quoted characteristics. When you look at the plate curves, on the other hand, you see that the 6SN7’s grid lines below 4mA and above 250V are much more evenly spaced.
This is important because as the input signal pushes the grid up and down in voltage along the load line, inconsistently spaced grid lines mean the anode (output signal measured on the x axis) will swing more in one direction than the other. That creates distortion. A 6SN7 with a large input signal (say 2Vrms or 5.7Vptp and up) will probably produce significantly less distortion than a 12AU7 in many cases.
However, if the bias point is between 150V and 200V at 7.5mA+, and the load is appropriate (15k-20k), and the signal is 2Vrms or less, the audible difference between tubes is probably debatable. Both tubes are pretty linear (i.e. evenly spaced grid lines) in these regions. That said, dropping one tube type into a design for another may change the bias point (anode voltage and current through the tube) and this change, rather than the quality of the tube itself, is likely to produce an audible difference in the character of the amp.
Outside of the effects of operating point choices, tube construction has a definitive effect on performance. The characteristics outlined above are governed by the physical geometry of the electrodes and the chemical properties of the materials used. For example, Child’s Law states that current across a vacuum is limited by three-halves power of the voltage potential divided by the square of the distance between electrodes. Taking this into account, manufacturing tolerances over the century that vacuum tubes have existed have been all over the place and so small differences in geometry and materials can have effects on the performance of any given batch of tubes. Whether these variations are actually audible depends on the tubes, circuit, ancillary equipment, and ears.
TL;DR: Even tubes with the same or very similar ratings may behave differently in a circuit. Tube rolling has at least as much to do with the circuit as the tube itself, making generalizations (e.g. 12AU7s suck because 6SN7s always sound better) about as useful as a screen door on a battleship.