Partial Response / Maximum Likelihood
Partial Response / Maximum Likelihood (PRML) is a reading method for reading data from magnetic data carriers such as hard disks with a higher write density.
Procedure
In order to write data on a magnetic data carrier, this data is first encoded using a line code . This coding is based, for example, on the group of RLL codes , which map the data in the form of pulse trains of different lengths on the storage medium. When reading this data , the height of the individual pulses can vary due to fluctuating material properties of the magnetic medium, high write densities or even random noise , which means that the usual, fixed threshold value decisions for generating a digital signal have higher error rates than PRML.
With a fixed threshold value decision, for example, analog signal values of the read head with over 50% are interpreted as logic-1 , signal values with less than 50% as logic-0 . With PRML, on the other hand, the threshold value decision is shifted dynamically as a function of the signal curve and thus made dependent on the temporal environment within the framework of a maximum likelihood method . In addition to better interference suppression on magnetic media, this enables higher write densities to be achieved, since the individual write pulses can flow “into one another” to a certain extent and are estimated based on their proximity .
For example, it is an analog signal curve with the following percentage curve
- 70%, 60%, 55%, 60%, 70%
read by the read head, PRML interprets this as a logical bit sequence of 101 , while with a fixed threshold value decision it is read as a constant sequence 111 . The situation is similar for signal values such as:
- 30%, 35%, 45%, 35%, 30%
which in PRML is interpreted as a bit sequence of 010 , whereas in the case of a fixed threshold value decision it is read as a constant 000 .
Independent of PRML, the digital data sequence obtained in this way is then decoded by the line decoder (RLL), which is usually followed by further stages such as error correction processes .
PRML is remotely comparable to the Viterbi algorithm , although no convolutional codes have to be used with PRML .
literature
- Todd K. Moon: Error Correction Coding. Mathematical Methods and Algorithms . Wiley-Interscience, Hoboken NJ 2005, ISBN 0-471-64800-0 .