Is the noise of balanced structure twice that of the single ended counterpart?

The balanced structure have twice the noise sources. Thus it seems that if I connect the previous stage deferentially, its noise performance will be worse than if I connect the it single ended (because the differential circuit contributes twice the noise). It is true that if I do so without changing the previous stage, the source resistance seen by the half circuit is halved in balanced structure compared with the single ended circuit, and the gain is changed, but it is unlikely that this change will results in a large enough drop in gain to compensate for the increase in output noise.

So can I conclude from this that the single ended structure has better noise performance?  And when noise performance is critical, like in the case of LNA,  I should use single ended structure?

Thanks!

Chen

thanks for your reply. But I think if we work with small signal, and keep the half circuit the same as in the case of the single ended circuit, the differential circuit have the same gain (for differential circuit, it is differential to differential gain) as the signal ended circuit. Indeed, since the balanced circuit has more headroom, we could make it in such a way that it have higher gain. But in cases like LNA design I think the headroom is not the limiting factor of the gain.

Stated differently, noise does not double with a differential structure; it increases by sqrt(2). This is because the noise on each side is uncorrelated. On the other hand, the signal can be twice as large if it is differential.  

So, with differential:
1) Noise goes up by sqrt(2)
2) Signal goes up by 2
Therefore, signal-to-noise increases sqrt(2) (equal to 3dB).

However, depending on what you are trying to do, differential may take twice the power so the noise advantage is lost. You still get cancellation of even ordered harmonics which is a good thing.

Dear all,
Thanks for your replys. I understand the benefits of differential circuit over single ended one, such as common mode rejection. The problem is whether the noise is larger or smaller. My argument is that the noise power is doubled. the signal stay the same(say, a differential CS has a voltage gain of gmXro, which is the same as a single ended CS). I don't understand why everyone here thinks that the gain is doubled.

Thanks!

Chen