A LITTLE THEORY
Resistors, next to wire, are the simplest of the electronic devices. They resist the flow of current, both A.C. and D.C. Theoretically, the resistance to A.C. is linear. In either case, the resistance is proportional to the ratio of voltage to current. R=E/I. This equation is derived from Ohms law, E=IR. This states that the voltage across the resistor is the resistance times the current through it. Ohm most likely measured it before he formulated it. This is one rule of thumb, namely "Why calculate what you can meaure?" What he did was put one volt across a resistance. His finding on the ammeter was that one amp flowed. He therefore, I guess wanting to make a baseline standard, declared it one unit of resistance. We call it after him, an Ohm. If one put two of these exact same resistances together and put one volt across them, one would read 0.5 amp on the ammeter Why is voltage the deciding factor? In this experiment it is the constant. So the change in resistance made a change in current, while voltage remained the same.
So he could logically conclude that two units of resistance made the current lower by 1/2, but voltage remained the same. So if I multiply 2 by 0.5 I get one, just like the voltage. E=IR or 1 = 2 X 0.5!
Resistors are generally made with carbon. The molecular structure of carbon makes it conductive but with a little resistance. The less carbon electricity must flow through, the higher the resistance. It really doesn't take much carbon, that is why resistors are so small. A copper wire with the resistance of a 1000 ohms would likely have to be miles long and very thin. However, carbon makes for noisy audio components. The noise is due to the electrons having to bounce around those nasty little carbon molecules.
Metal film is the higher quality version and the resistor of choice for audio amps. There is less noise because the metal film is more conducive to conduction. More free electrons, therefore less noise. However, there is a floor to how much lower one can go before one can't eliminate any more noise. This is known as Johnson noise. It will be the same level for any component one uses.
Metal film resistors are accused of making an amp sound "metallic". I believe that this has nothing to do with the resistor being metal and more to do with the noise level. A friend of mine told me that with amps of a curve where frequencies higher than 15 kilohertz are about 1 to 2 decibels higher than flat tend to sound mettalic. The noise content of carbon resistors tends to mask the high frequency boost, hence sounding less mettalic. It's a psychoacoustic thing.
Resistors come in a variety of tolerance values. The tolerance in this case is how much from the "advertised" value the resistor is off. If I have a resistor of 1000 ohms and the tolerance is 10 percent, then the value could range from 900 ohms to 1100 ohms. Resistors used to be had with 20% tolerance. I doubt that anyone could even find a ten percenter nowadays, let alone 20. The tolerance can go as low as 0.1%. They are pretty expensive though. 1%ers go for about 25 cents US each, where the 0.1%er will go for $2.50 or more! I would be satisfied with the 1% ones.
How do they get such tight tolerances? "Modern technology, Ted." No, laser trimming. ;-)
But as for the psychoacoustic effects, does replacing resistors in an amplifier make a difference? Absolutely. If an engineer designs an amplifier, in this case a tube amplifier, he designs it with the specifications of the tube in mind. Any variation from the design will result in, even if subtle, a difference in the overall response of the amplifier.
According to "The Psychology of Music", our ears can detect from five to seven harmonics in a tone. If there is a variation of amplitude by a small amount of any of the harmonics, we can detect it. This effect is more sensitivve than phase changes in harmonics. According to the book, a room is likely to smear the phase relationship of the harmonics, yet we know the difference between a guitar and a violin playing the same note. This is based on the timber which is dictated by the amplitude relationship of the harmonics.
There are some who actually have compared the "sound" of different resistors. This to me is somewhat absurd. There are claims that tantalum resistors (no longer manufactured except in Japan. Why? They cost up to 10 bucks apiece!) offer the warmth of carbon composition with the low noise and precision of metal film. I think that there is either something in the amp that is amiss or the person listening has super-duper absolutely perfect wide range hearing ears and can discern even the most minutest variation in response. I can't (yet I can hear a difference in the sound of wire. Well, jumping from 28 gauge to 20 there is a difference anyway). I have used both types in the exact same amp and they both sound equal. The only reason I would go with metal film over carbon is the precision values one can get. In high gain applications the low noise is a plus. But that is it. There is no reactive component in a resistor that causes it to affect the frequency response of an amplifier (except maybe at RF frequencies where skin effect is the greatest). Therefore they are linear.
In a tube amplifer, the tube has several different response curves for different settings of bias. I have noticed that the closer to zero the tube is biased the more linear the curve. This changes dramatically as the tube's curve gets closer to cut-off.
If the cathode resistor that determines the bias setting in most triode applications was to vary by 10 percent, the voltage drop across the resistor will also change. This is because more or less current will flow. If I have a 1000 ohm resistor dropping 1 volt, and it changes or actually has 900 ohms resistance, then the voltage will be .9 volts for the same current, or less for a larger current. In a tube, the closer one gets to a bias of zero volts, the more current flows, so the voltage drop across the resistor of 900 ohms could actually be .75 volts, depending on the tube. But that is the subject for the tube psych page. This would, if you take a look at a triode curve sheet, make enough of a difference that it will sound different to us.
A similar effect would be had by changing the type or make of tube. Again, the subject for the tube page.
So, in conclusion, if you just bought your first or fifth ST-70 or Conrad-Johnson or whatever, along with replacing the capacitors, invest the five or six bucks and extra hours and replace those resistors too.
My first experience with this was an old 1961 Magnavox console stereo. I replaced the capacitors, the tubes, the wires and even the tube sockets, but the most dramatic change I got was when I replaced all of the resistors. The sound was fat and lower midrange-upper bass strong. Afterwards the response "flattened" to give less fatness and lots of deeeep bass. My father ran out of his room and asked "What did you DO!?! It hasn't sounded like that since it was new!" Well, at first I didn't like it, but then I got used to it.
I suspect that most advocates against negative feedback usually like the fat sound, even if it is a coloration from the real sound that purists like me like. Negative feedback tends to flatten the curve and remove the fatness to a degree, while extending the frequency response in both directions. Subject for the NFB page.