请问一般移动通信中仿真能达到多大的信噪比？

40dB太高了。一般也就10dB，15dB这个量级。小一些也有可能。

In simulations, we usually simulate from 0 dB to 20 dB.
Sometimes, 30 or higher SNR is also tested for asympototic performance.
I am also wondering how to relate the simulated SNR to the real SNR.
One possible way may be first find the energy of background noise. Usually on the frequence of 2G, it is -95 to -100 dBw in Beijing. (But I don't know how much is the themal noise in the receiver).  Then, according to the fading channel model, we can get strength of the received signal, which is related with 1/d^r, where d is the distance and r >= 2, according to the environment.

仿真用的信噪比通常都是接收端的信噪比吧，然后在根据这个作链路预算，你所说的
relay SNR是什么东东，没有听过

Typo /shy
It should be "Real SNR"
I guess you are from MAC layer or Network layer. While in physical layer, both transmitted SNR and received SNR is okay for simulation. For myself, I use transmitted SNR, due to a fair comparison.

what a fair comparison indicates? the same tx power?

yes. In some of my cases, the receiver is moving around and the received SNR is not easy to compare.

是bit error, symbol error还是packet loss？
这个还是有很大区别的。
另外还要看是code signal还是uncoded signal。

coded signal的 BER 应该很快就降下来吧。

功率谱密度不应该是dBm/Hz吗

不给出仿真条件，你这样的问题是没有任何意义的。加不加编码，有没有各种干扰，你的发射端有没有预处理，你的接收机是什么，这些都是影响BER的因素。