This paper examines various sources of error in MIT's improved top oil temperature rise over ambient temperature model and estimation process. The sources of error are the current parameter estimation technique, quantization noise, and post-processing of the transformer data. Results from this paper will show that an output error parameter estimation technique should be selected to replace the current least squares estimation technique. The output error technique obtained accurate predictions of transformer behavior, revealed the best error covariance, obtained consistent parameter estimates, and provided for valid and sensible parameters. This paper will also show that the output error technique should be used to minimize errors attributed to post-processing (decimation) of the transformer data. Models used in this paper are validated using data from a large transformer in service.Morris, A. TerryLangley Research CenterPARAMETER IDENTIFICATION; NOISE (SOUND); TRANSFORMERS; AMBIENT TEMPERATURE; MEASUREMENT; MATHEMATICAL MODELS; PARAMETERIZATION; ESTIMATING; ERROR ANALYSIS...