针对采用蓝光激发荧光粉产生白光的YAG型白光LED，通过分析其光谱波谷特性，采用常规可见光光谱仪和温控系统设计了一套基于光谱特征参量的LED结温测试系统.测量方法分为定标函数的测定和任意状态下的测量两部分.首先采用光谱仪测量在给定的多个不同结温和正常驱动电流下的相对发光光谱数据，再分析其光谱波谷处的相对光谱强度.从实用性和降低成本的角度考虑，采用正常工作电流驱动，但以正常工作电流驱动下的LED在光谱仪的固定反应时间内其自加热效应不可忽略.因此采用选定基准状态法，将各温度下的相对发光光谱强度与基准状态下的逐点作差得到相应的发光光谱强度差，同时为了减少温控系统引入的温度偏差，同样将各温度与基准温度作差得到相应的结温差.实验表明高低色温大功率LED的结温差和发光光谱强度差经过一定的函数拟合形成的定标函数其线性度都较高，R²达到0.99以上；利用定标函数，可以测量出在任意状态下的LED结温.最后将采用本方法得出的高低色温LED在不同条件下的结温数据与通过Mentor Graphics公司的T3Ster仪器的测量结果进行了比较，最大偏离度为2.82%，在可接受的误差范围内，表明此方法完全具备可行性，具有一定的实用价值.

Aiming at YAG white LED generated by blue-light excitation phosphor, a set of junction temperature measurement system based on characteristic spectral parameters is designed by using conventional visible spectroscopy and temperature control system through analyzing spectral trough characteristics. The measurement process is divided into two parts:the measurement of calibration function and the measurement of arbitrary state. Firstly, the relative luminescence spectra at different junction temperature and normal driving current are measured by spectrometer, and then the relative spectral intensity at the spectral trough is analyzed. Considering practicability and cost reduction, normal working current drive is adopted, but the self-heating effect of LED driven by normal working current can not be neglected in the fixed reaction time of spectrometer. Therefore, the selected reference state method is used to make the difference between the relative luminescence spectral intensity at each temperature and the reference state point by point to get the corresponding difference value. At the same time, in order to reduce the temperature deviation introduced by the temperature control system, the corresponding junction temperature difference is also obtained by making the difference between each temperature and the reference temperature. Experiments show that the calibration functions of junction temperature difference and luminescence spectral intensity difference of LED with high and low color temperature have high linearity, and R² is above 0.99. Using the calibration function, the junction temperature of LED can be measured at any state. Finally, the junction temperature data of high and low color temperature LED obtained by the proposed method under different conditions are compared with the measurement results by T3Ster instrument of Mentor Graphics Company. The maximum deviation is 2.82%. Within the acceptable error range, it shows that the proposed method is completely feasible and has certain practical value.