MIL and SIL

Both modes run the same plant in the same diagram layout. They differ only in the controller block, so a difference in the results points at the controller.

What runs in each

	SIL	MIL
Block	STEPPER_CTRL_IO	STEPPER_CTRL_MIL
Controller	The Stepper-C library, called by sil/c/sstep_xcos.c	ctrl_mil_block(), which calls ctrl_period()
Numeric form	Q15 or float, as on the MCU	Scilab doubles, with the integer motion generator of the library
Plant	sil/c/sstep_plant.c	The same
Input limits	sstep_xcos.c	ctrl_mil_limit(), the same limits
Speed	Used for everyday runs	15 s per simulated second on a 0.15 s run
Use it to	Test the code that ships	Study or change the algorithm without compiling it

The compiled library is needed in both modes, because the plant and the monitor are C blocks in MIL too. lab_run() and Open in Xcos link it by themselves.

Backends

Stepper-C has two numeric backends. sil_build(backend) compiles the chosen one into sil/build/fixed or sil/build/float, once, and again when a source file changes.

Backend	Numbers	For
FIXED	Q15	MCUs without an FPU
FLOAT	Single precision	MCUs with an FPU

A Scilab session links one backend. To change it, select the other one in the window and restart Scilab. tests/run_all.sce does this by starting a second Scilab process for FLOAT.

How close they are

Tests M01 and S01 to S06 compare the modes sample by sample. These are the limits they enforce and the values measured on the final run:

Comparison	Limit	Measured
MIL against SIL, current step, phase currents	5 mA	1.02 mA
MIL against SIL, move of 640 microsteps, phase currents	5 mA	1.28 mA
MIL against SIL, commanded position	1e-6 microsteps	0
MIL against SIL, load step above pull-out	Same fault within 1 ms	OVERCURRENT and STALL at 204.9 ms in both

After the rotor slips, the ADC quantization is amplified and the samples diverge, so a comparison of samples after a pull-out means little. The stall tests compare the fault and its time there.

Compare MIL / SIL shows this for any scenario, and reports differences only up to the first fault. From a script:

rs = lab_run("SIL", 1);
rm = lab_run("MIL", 1);
c = lab_compare_data(rs, rm);
mprintf("%s\n", c.panels(1).title);

A configuration the library refuses

Stepper-C checks its configuration in sstep_init().

• In SIL the simulation stops with the message of the library.
• In MIL the block asks the library the same question at the start. If the answer is no, or if the configuration asks for voltage drive (ctl.drive = 1), which the prototype does not have, the outputs stay off, the state is 3, the fault flags are 16, and the reason is printed. lab_run() raises the reason as an error, so the window shows it. In Xcos no dialog appears and the run reaches T_end, because an error raised in this kind of block at its start ends Scilab 2027.

The stepper_sil diagram

xcos/stepper_sil.zcos, opened by start_sil.sce, is the first SIL diagram of the project and is kept as it was. It has a flat layout, and its input is the command table.

Context variable	Meaning	Example
cmds	Commands [t kind value]: 1 MOVE to, 2 RUN, 3 STOP	[0 2 12800; 0.15 2 -12800]
loads	Load torque steps [t TL]	[0 0; 0.15 0.25]
brd.topology	"shared" or "full"	"full"
brd.stage	"avg" or "switching"	"switching"
mp, ctl	Motion and control settings	mp.v_max = 25600
locked	1: rotor locked	1
T_end	Simulated time, s	1.0

Its results have the first 19 columns of sil.values.