This library contains bittrue implementations in VHDL (for synthesis) and Python (for fast simulations) of standard signal processing components. For the fixed-point arithmetics, the library en_cl_fix provided by Enclustra GmbH is used.
Radoslaw Rybaniec [radoslaw.rybaniec@psi.ch]
- Oliver Bründler [oli.bruendler@gmx.ch]
- Radoslaw Rybaniec [radoslaw.rybaniec@psi.ch]
- Benoît Stef [benoit.stef@psi.ch]
This library is published under PSI HDL Library License, which is LGPL plus some additional exceptions to clarify the LGPL terms in the context of firmware development.
See Changelog
See Documentation
This library contains fixed point processing logic that is not very application specific and can be reused easily. Code must be written with reuse in mind. All important parameters as well as number formats shall be implemented as Generics.
One of the main ideas behind en_fix is to have bittrue Python models of each VHDL component in the library. Therefore only components that have a bittrue model shall be added to this library. If some VHDL components don't have a bittrue model, it may be better to open a separate library for "non-bittrue fixed-point code".
It is suggested to use one .vhd file per Package or Entity.
Examples for things that belong into this library:
- Processing Entities such as FIR filters
- Code generators for en_fix (e.g. for Approximations)
- Packages that help with fixed-point processing
Stable releases are tagged in the form major.minor.bugfix.
- Whenever a change is not fully backward compatible, the major version number is incremented
- Whenever new features are added, the minor version number is incremented
- If only bugs are fixed (i.e. no functional changes are applied), the bugfix version is incremented
The required folder structure looks as given below (folder names must be matched exactly).
Alternatively the repository psi_fpga_all can be used. This repo contains all FPGA related repositories as submodules in the correct folder structure.
- TCL
- PsiSim (2.1.0 or higher)
- VHDL
- psi_fix
- psi_common (2.15.0 or higher)
- psi_tb (2.7.0 or higher)
- en_cl_fix (1.2.0 or higher) - fork of a a library provided by Enclustra GmbH
Original Location
Dependencies can also be checked out using the python script scripts/dependencies.py. For details, refer to the help of the script:
python dependencies.py -help
Note that the dependencies package must be installed in order to run the script.
- Python 3.x (for executing the bittrue models)
- Python Packages
- SciPy (pip install scipy)
- NumPy (pip install numpy)
Note: On Linux Python 3.x and Python 2.x can be called explicitly by using python3 and python2. However, this does not work out of the box for Windows but explicit calling is required since Python 2.x may be present. To enable the python3 command for Windows, follow the steps below:
- Add the path to your Python 3.0 installation to the PATH environment variable
- Create a copy of python.exe and rename it to python3.exe
The python models can be called from MATLAB. Not having separate MATLAB models allows maintaining only one code base.
Some helper functions as well as an example about how to use python models from MATLAB can be found here.
Python is the main development tool for the models, so minor problems due to limited Python support of MATLAB can occur when using a model the first time from MATLAB. If Python changes are necessary, they shall be implemented fully backward compatible.
For everything that is non-trivial, self-checking testbenches shall be provided to allow easy and safe reuse of the library elements. Testbenches shall check bittrueness by comparing outputs from the python model to VHDL outputs.
A regression test script for Modelsim is present. New Testbenches must therefore be added to the configuration of the regression test script sim/config.tcl.
To run the regression test, execute the following command in modelsim from within the directory sim
source ./run.tcl
The fixed point number format used in this library is defined as follows:
[s, i, f]
s: 1 = Signed number (two's complement), 0 = Unsigned number i: Number of integer bits f: Number of fractional bits
The total number of bits required is s+i+f.
The value of each bit depending on its position relative to the binary point (i-bits left, f-bits right) is given below.
... [4][2][1].[0.5][0.25][0.125] ...
Some examples are given below:
| Number Format | Range | Bit Pattern | Example Int | Example Bits |
|---|---|---|---|---|
| [1,2,1] | -4 ... +3.5 | sii.f | -2.5 | 101.1 |
| [1,2,2] | -4 ... +3.75 | sii.ff | -2.5 | 101.10 |
| [0,4,0] | 0 ... 15 | iiii. | 5 | 0101. |
| [0,4,2] | 0 ... 15.75 | iiii.ff | 5.25 | 0101.01 |
| [1,4,-2] | -16 ... 12 | sii--. | -8 | 110--. |
| [1,-2,4] | -0.25 ... +0.1875 | s.--ff | 0.125 | 0.--10 |