FPRE004 typically uses a Qm.n format:
Example: Q3.4 (8 bits total: 1 sign, 2 integer, 4 fractional)
| Format | Range | Resolution (LSB) | |--------|-------|------------------| | Q2.5 | -2 to +1.96875 | (2^-5 = 0.03125) | | Q3.4 | -4 to +3.9375 | (2^-4 = 0.0625) | fpre004
The prefix "fpre" is heavily used by Frontiers journals to denote articles that are in the "Frontiers Preview" stage (an uncorrected proof prior to final formatting).
When a real number ( x ) is quantized to a fixed-point value ( \hatx ), the error ( e = \hatx - x ) depends on the rounding method: FPRE004 typically uses a Qm
However, based on common academic and scientific naming conventions, here is what "fpre004" most likely refers to and how you can find the proper article:
For a signal uniformly distributed over ([-A, +A]), the signal power ( P_s = A^2/3 ).
Quantization noise power ( P_n = \frac\textLSB^212 ).
Thus:
[
\textSQNR = \fracP_sP_n = \fracA^2/3\textLSB^2/12 = 4 \cdot \fracA^2\textLSB^2
]
In dB:
[
\textSQNR\textdB = 6.02n + 10\log10\left( \frac4A^2(\textrange)^2 \right)
]
For full-scale sine wave ((A = \textrange/2)), this simplifies to ( \approx 6.02n + 1.76) dB. Example: Q3
Functional:
Non-functional:
Compliance: