Reading Noise Specs & Decibels
Most published noise specs are useless as they do not give a proper reference. When they do it is often to full output, which still requires some recalculation if we use the amp less than full output. For instance, if an amp's noise is -80 dB referenced to full output of 100 watts it is only -60 dB at one watt and -40 dB at 0.1 watt where folks with quiet rooms, sensitive speakers, and sensitive ears may likely be listening. Those are not quiet amps.
I have lots of suggestions for the industry and one is that we should stop using full output as a reference because it allows noisy high-power amps to have good noise specs. Instead, we should be referencing it to 1 watt. Since a speaker’s sensitivity is rated at 1 watt/1 meter we can then easily calculate the noise at one meter. If your speaker sensitivity is rated 100 dB at 1 watt/1meter and the amp is -60 dB below 1 watt, then your signal to noise will be 60 dB at 1 meter and at the listening position wherever that may be. I would love to rate my amps this way and I might include this handy spec. Unfortunately, it does not tell the whole story of what will happen in the system because we still must take the line section of the preamp into account.
Line sections typically have their gain stage after the volume control and put the full noise of the line tube into the power amp. This is why in most instances the noise does not go down with the volume control setting. If the noise does go down with volume setting, then the noise is ahead of your volume control which puts it in your signal source. This is important information for those trying to locate the noise source.
I have gone to a lot of effort to make very quiet power amps. When I get a note from an owner who thinks it is the power amp making the noise, I always tell them to try shorting plugs in the input. 99% of the time the noise disappears, and it turns out the preamp is making the noise. Hum is often cable and ground related and floating the preamp and signal sources fixes this.
Here is what happens with a noisy preamp. Very few preamps are below 20 uV of output noise and the exceptionally good ones are 5 uV. It is hard to do much better than that. You cannot even get near those numbers with a 6SN7, a truly inappropriate tube to use for a line amp. If you have 20 uV of noise and a typical power amp gain of 30 dB (which also happens to be 30x) the noise at the speaker will be 600 uV or 0.6 mV which is a lot higher than my amp's resident noise. So, you see, just because the amp is connected to the speaker, it is not the amp that is at fault, it is just doing its job which is to amplify what it is given.
Do all of us who make power amps a favor; try shorting plugs first. You can make them by taking any old RCA cable, cut it off at 2 inches from the plug, then strip and twist the hot and ground together. If you are using an XLR, twist all three wires together and plug it into the input.
Another way is to rate noise in absolute values as I do in my amps. Simply state it in millivolts or microvolts if your amp is that good. Then knowing the gain of the power amp and the noise of the preamp you can figure what the noise at the speaker will be and which component will be the limiting factor.
For comparison the early amplifiers that David Manley sold often had hum of 2mV at the speaker terminals. My RM-9 was typically 0.5 mV and the RM-10 typically 0.15 mV and these are achieved with AC heaters. The RM-10 noise at that level is very hard to beat. In my amps hiss is even lower.
If you have a speaker with 100 dB sensitivity, 2.83 mV will produce noise at 40 dB SPL at one meter. Since we have 2 speakers add 6 dB to that, and more if your chair happens to be in a standing wave peak at 60 Hz or 120 Hz. Every time we cut the output noise in half, we get 6 dB less noise (or increase in S/N if you prefer). So, reducing the noise in an amp from 2 mV to 0.25 mV reduces the noise by nearly 20 dB!
[Source: circa 2011]