Why An Amplifier’s Sound Changes When You Change Tubes
By Roger A. Modjeski
Over the years, it has been noted by many astute listeners that changing tubes often produces a change in the sound of their equipment. While these changes are often attributed to the tubes alone, it is almost always a case of tube-and-unit interaction. Therefore, it is incorrect to say that a certain tube sounds a certain way. In this article we shall look at the ways in which tubes affect the equipment they go into. First, let us look at some of the problems involved in manufacturing tubes. Given a batch of tubes from a factory there will be a spread of parameters. Each manufacturers noise, microphonics, gain, and operating point will fall into a bell curve due to the nature of the machines that manufacture the tubes, the materials for the batch, and control over processes. A bell curve of gain is a good example.
In an ideal world, all the 6DJ8 makers would adjust their grid lathes to get the same target value of 30. They do try, but they have a specification of ą1 or ą2 or ą3 depending on how important this parameter is to them. It is very hard to wind that little 6DJ8 frame grid to get a voltage gain (Mu) of 30 ą1. I encourage you to break open an old 6DJ8, pull off the top mica, cut the lower supports, and carefully pull out the two posts on either side of the hollow cathode sleeve in the center. Look at the fineness of the grid wire wound on the posts and the spacing of those hundreds of turns. The Mu of 30, which we want to control to ą3% is the ratio of the diameter of that invisibly fine wire to the spacing between turns. It is amazing they can do it at all! Tear apart a 12AX7 and you will see how much easier it is to wind that grid.
However, controlling gain by Mu alone is not sufficient. Mu is the product of transconductance (gm) and plate resistance (Rp). Although most triode circuits are designed to let Mu pre-dominate, sometimes gm is more in control of circuit operation. This is typically the case in cascode circuits. Here we have the opportunity for a larger variation since two tubes with the same Mu may not have the same gm as Rp differs. Now add in the variation in operating point (bias) and it is easy to see why different manufacturers tubes sound different in a circuit, even to the point of not functioning in some overly sensitive circuits. As an extreme example, we can reference the 12AX7 in the Moscode Minuette. George Kay and I found that the Tungsram 12AX7 worked extremely well in the front end of his unit, but the smooth plate Telefunken did not. The latter caused the gain to fall 10-20 dB and the distortion to increase 10-100 times. In this case, anyone would certainly hear differences between the two. However, the main thing they would be hearing is the difference between a functioning and malfunctioning preamplifier.
A less dramatic situation is found in feedback-type RIAA phono circuits. These include Audio Research, Modulus, and others. The problem stems from the fact that there is not enough loop gain for the feedback to provide consistent equalization at low frequencies. This means that the shape of the low end, and low-to-high tonal balance will vary with the Mu of the tubes. I remember magazines of the late '70's encouraging manufacturers to get this RIAA EQ accurate to some 10 millibel (mB). It is impossible to get ą1 dB when tube variations cause ą2 dB errors. I believe this RIAA EQ shift to be the major effect one hears when changing phono tubes in active RIAA circuits. Fortunately, passive EQ circuits are free from this effect, and therefore RIAA is consistent over tube life.
Moving on to noise, there are four areas to investigate:
AC hum is rarely a problem as most designers use DC regulated filament supplies. However, for those with some ripple, the filament construction becomes a factor. The 6DJ8 and 12AX7 are made with folded or coiled filaments depending on the manufacturer. It is not part of the specification. A coiled filament will cancel hum whereas a folded one will not.
Noise, however, is the big battle. All tubes have thermal noise and in a perfectly made one there is only the thermal noise, which cannot be improved upon. A 6DJ8 has a signal-to-noise ratio (s/n) that is 8 dB better than a 12AX7. No wonder it is the tube of choice for phono inputs.
In addition to this thermal noise, there is excess noise which can be large. This is the noise we grade for at RAM Tubes. A RAM Tubes "A" grade has only 3 dB excess noise. After listening to and measuring noise over the years, I determined that two tubes could measure the same total noise but one would be more annoying than another. I designed a device to quantify the annoyance factor in the noise and that became the RAM Factor.
The last area of noise performance is susceptibility to microphonics. We all know the sound emitted by tapping on a tube or the chassis. The sound is caused by the relative motion of the grid and cathode. Many tubes use all sorts of fancy mica insulators to hold things steady, but large variations occur from tube to tube and between manufacturers. Low microphonics is important to imaging and detail in that a microphonic tube will receive sound from the speakers and put it back in the chain in a very non-musical way.
Finally, I am often asked if X tube is bright or Y tube is bloated, etc. This paper is the answer to that question. I cannot ascribe a particular sonic quality to a particular tube because tubes themselves do not have a sound. This information was gathered over my last 25 years of working with tubes and tube circuits. It is a combination of many books, experiments, and visits to tube manufacturers.