Low noise input?

[MikeW]

Why is it so important to have a low noise at the input of an amplifier? It
seems to me that the Pre-amp and the source would be a bigger factor for low
noise in an audio system.
http://www.rfcafe.com/references/ele...ise_figure.htm

Would it not be better to have a higher gain pre-amp and run it into an amp
like the PLH?

[peranders]

This is just mathematics.

Example:

You have two gain stages and the gain is 10. First stage has 1 uV noise, the second has 5 uV. Which stage contributes with noise?

First stage feeds the second with 10 uV and then 5 uV is added. The sum is not 15 uV but 11.18 uV. So this 5 uV is hardly noticable. If the gain is even more even less is it with low noise in the rest of the chain.

It's always most important to put effort in the first stage, both in noise and distortion.

You are totally right when you suggest lower gain in the power amp but you must also make the amp fit for a standard audio level.

[GRollins]

I confess to chuckling when I saw this. I got fried in a thread such as this one a couple of years ago when someone asked essentially the same question. I said something similar to what peranders said, but kept it as a purely straight ratio thing. X amount of noise from the front multiplied by the gain of the second stage. And here's the crux: I noted that it was oversimplified.
Well, by the time I got back to the thread the next day, about thirty-seven people had piled on, subjecting me to a veritable blizzard of mathematical formulas, derision, and condemnation.
Guess they didn't notice the single word 'oversimplified.' Interesting test of people's ability to read.
Technically, they were all quite correct. The noise isn't purely additive (or perhaps I should say multiplicative) from one stage to the next. There's some reinforcement and some cancellation, but the upshot is that there's somewhat less noise than you might expect. The formulas? Hell, I don't remember. I don't keep but about a dozen formulas in my head; the rest I look up on an as-needs basis. (There's an anecdote about Henry Ford I could throw in here, but it would be a diversion and those self-same pendants would probably jump on me about tiny details in the story, so we'll skip it in the interest of peace and quiet.)
Anyway, Mike, you're right. High gain in the front is better than at the end. But reality comes knocking and all the things you do to get high gain generally bring along baggage in the form of higher distortion, decreased bandwidth, and other such nonsense. As with most things in life, it's the art of trading off gracefully. Folks like Nelson and Charles Hansen have had plenty of practice at this and are nimble on the dance floor. Newbies seize one concept and develop a fixation--ignoring all other parameters.
As far as stereo gear goes, there are rough standards to be adhered to. Amplifiers generally have about 26dB gain, although they range from less than 0dB for followers to over 30dB for some of the beastly ones. Phono preamps used to be clustered around 40dB, although there were some that were down in the 30's. Line stages were more variable, but were generally in the 20's, at least until the rise of CDs, when they fell to upper single digits on up into the teens. Phono stages rose at the same time, partially to work better with moving coil cartridges and partly to compensate for the loss of gain in the line stages. To the extent that you intend to use your designs with other peoples', it's a good idea to stay reasonably close to those norms. If you're going to stay home, you can do anything you want.
Me? I just wandered in here by mistake. The bouncer will be along shortly to whisk me out the door.
Quietly, of course.

Grey

[tlf9999]

I confess to not understanding the math behind it.

If the two "noises" are really noises: signals without any correlation, 100% independant of each other, it seems to me the "expected" value of the resulting noise at the output is really 15uv.

However, the "loudness" of that noise, due to the lack of inter-dependance, is equivalent to that of a 11.18uv signal (=sqrt(10^2+5^2)).

so depends on your perspective, both answers (simple arthismatic summing and RMS summing) are both correct.

[peranders]

Originally posted by tlf9999
I confess to not understanding the math behind it.

If the two "noises" are really noises: signals without any correlation, 100% independant of each other, it seems to me the "expected" value of the resulting noise at the output is really 15uv.

Are you really sure about that?

[tlf9999]

Originally posted by peranders

Are you really sure about that?

the mean of the sum of two uncorrelated variables is the sum of the means of those uncorrelated variables.

Stat 101. It can also be fairly easy to prove as well.

[MikeW]

Grey, how is that new front end coming along? Can you send me a schematic so I can make a PLH intergrated amp? I'll send it to you when I am finished listening to it.

[ctong]

Originally posted by tlf9999


the mean of the sum of two uncorrelated variables is the sum of the means of those uncorrelated variables.

Stat 101. It can also be fairly easy to prove as well.

For independent variables the variance of the sum is the sum of the variances. Therefore, the sum of two independent random noise sources with identical variances equals a noise with twice the variance or sqrt(2) times the rms values (1V+1V=1.41V). So 10uV+5uV=11.8V is correct.

[tlf9999]

But why are we summing variances?

[ctong]

It is the ac part (fluctuations) of the noise that counts. The dc part is actually not noise.