There are many applications which are based on the nuclear magnetic resonance with microcoils detection, such as MR imaging, spectroscopy and well logging. But due to the low Zeeman splitting difference in populations， usually the specific signal-to-noise ratio associated with the micro solenoid coils is very low. The geometry of the detection coil has numerous influences on the detected MR signals. Under a given duty ratio of the sample volume to the coil size, as well as a certain static magnetic field and so on, it is concluded that the signal-to-noise ratio is in proportion to the radio-frequency magnetic field by unit current flowing through the RF coil, and is inversely proportional to the squared root of the resistance of the coil under certain frequency. In the paper, for a 039 Tesla main field the relationship between the optimization on coil geometry parameters and the skin-depth effect of bulk conductor is addressed. Compared with sophisticated processes in MEMS technologies, the micro solenoid coils was winded with lacquered wires on polytetrafluoroethylene holding. After electronic measurement, the signal-to-noise ratio on a certain frequency for the coils made is benefit from the proper number of turns of coil. On the other hand, this fabrication is quite simplely relative to a couple of masks, lithography and electroplating. In the content of 169 MHz, i.e. the operation condition, the quality factor of a modeled micro solenoid coil is scanned in a frequency span. It is in good agreement that the simulation predicted a maximum SNR of 1998 when the number of turns is 11, under the condition of wire and coil diameter 05 and 55 mm, respectively. So the investigated principle is well verified from the practical point of view. In the future, this method can be further used in the proton density MR imaging, relaxation spectrum analysis on rock fluidics.