With the aim to realize the miniaturization of absolute optical encoders with the characteristics of greater accuracy and more reliability, the new coding mode and structure of a reading head are investigated. Firstly, this paper describes the principle of eight-matrix coding used in absolute optical encoders. Different from traditional four-matrix coding,the eight-matrix coding only needs two grating tracks to realize a 10-bit binary.The first grating track realizes the 3-bit coding of Gray code; and the second one realizes the 7-bit coding of Gray code. Then it can achieve the 10-bit binary coding through the relation between Gray code and binary. The Moiré fringe obtained by using phase compensation of incremental slit is another new technique in this paper.The single reading-head is also good for reducing the number of LEDs, therefore the dimension of encoder is smaller. Finally, the method of signal pick-up is discussed as well. Experimental results indicate that the proposed eight-matrix encoder has a smaller size of Φ25 mm×16 mm and a lighter weight of 28 g, while his resolution is 16-bit and accuracy is 30″. It is concluded that the encoder can normally work in very hard conditions with more reliability for long time and can be widely applied to the field of modern military and aerospace.