Recommended, but hard to set up
Standards ask for back-to-back testing, yet a working harness for model and generated code takes real effort to stand up.
BTC TestStack
Back-to-back test.
Show that the generated production code behaves like the model — automatically, for Simulink, dSPACE TargetLink and MathWorks Embedded Coder.
MILSILPILSimulinkdSPACE TargetLinkEmbedded Coder
Back-to-back · Controller_Rev4MIL vs SIL
Model · MIL
Controller.slx
Code · SIL / PIL
/* Controller_step */
void Controller_step(void) {
real32_T err = ref - fb;
real32_T u = Kp * err;
state = plant(state, u);
out = state;
fb = out;
}
dSPACE TargetLink & Embedded Coder
Compare outputs
torque_cmd142.8 ≡ 142.8
mode_outRUN ≡ RUN
1 284 samples comparedModel ≡ code
The challenge
Showing that model and code match shouldn't be a manual project.
Back-to-back testing is recommended by ISO 26262 and related functional-safety standards. The question it answers is simple — does the generated code behave like the model? Getting to that answer by hand is not.
Test data by hand, everywhere
Generating enough stimulus and comparing outputs across many subsystems is slow, repetitive, manual work.
Every model change re-runs everything
Change the model and you re-generate, re-run and re-compare both sides — a full cycle, on every iteration.
The solution
100% completeness, 0% manual effort.
BTC TestStack runs the back-to-back test between model (MIL) and production code (SIL/PIL) automatically within one single test project. It extracts the architecture from dSPACE TargetLink and MathWorks Embedded Coder, generates the test cases, runs both sides and reports the comparison — no manual harness, no hand-written test data.
Architecture extracted for you
BTC TestStack reads the model and code architecture straight from TargetLink and Embedded Coder — no harness to wire up by hand.
Harness models on-the-fly
For large models, test harness models are generated automatically, so even deep hierarchies run without setup.
CI/CD
Runs headless in your CI, with a versatile REST API orchestrating BTC TestStack either running in a virtual machine, a Linux Docker container or BTC TestStack × Cloud in AWS.
The back-to-back testing workflow
Automatic
Architecture extraction
Interfaces and subsystem hierarchy read directly from the model.
Automatic
Test generation
Automatic test generation for 100% structural coverage incl. MC/DC.
Run both
Run both sides
Run on model · MIL
Run on generated code · SIL / PIL
Automatic
Compare results
Outputs matched sample-by-sample within defined tolerances.
Output
Report & differences
HTML report highlighting any sample where code diverges from model.
Automatic test generation
Complete coverage, generated by model checking technology.
The test cases that drive a back-to-back run aren't guessed at random. BTC TestStack uses model checking to generate the smallest set that reaches every coverage objective — and to show which objectives can't be reached.
Generated, not random
Model checking derives test cases from the structure itself, so 100% coverage is reached by construction rather than by chance.
Smallest, shortest set
The fewest and shortest test cases that still meet every objective — less to run, less to review.
Unreachable, identified
A dynamic analysis distinguishes an unreachable objective from one that is simply untested — so a gap is explained, not just flagged.
Inputs & calibrations reached
Direct access to inputs and calibrations surfaces behaviour that requirement-driven functional tests can miss.
Back-to-back report — Controller_Rev4 · build #482MIL vs SIL
Test cases
24
generated · model checking
Model ≡ Code
23 / 24
matching within tolerance
Differences
1
flagged for review
MC/DC coverage
100%
| Test case | Model (MIL) | Code (SIL) | Max Δ | Result |
|---|---|---|---|---|
| TC_01 | 142.80 | 142.80 | 0.0e0 | Match |
| TC_02 | RUN | RUN | — | Match |
| TC_03 | 0.0491 | 0.0491 | 2.1e-7 | Match |
| TC_04 | 88.20 | 88.21 | 1.0e-2 | Difference |
| TC_05 | -1.000 | -1.000 | 0.0e0 | Match |
Keep exploring
Same engine, different question.
Back-to-back, regression and migration testing all run on the same BTC TestStack engine — automatic model-checking test generation plus structural comparison.
Regression test
“Did a change break anything?”
Shows the software still behaves as intended after a change — at model level (MIL) or code level (SIL/PIL). Push-button execution, HTML reporting.
MIL vs. MILSIL vs. SILnon-AUTOSAR vs. AUTOSAR
Migration test
“Did a tool upgrade change behaviour?”
Shows a new tool version — code generator, compiler, modeling tool or OS — leaves behaviour unchanged. Jenkins runs the workflow across tool versions and operating systems.
New Simulink versionNew TargetLink versionNew OS
Certified
BTC TestStack
is ISO 26262 certified
The certificate addresses functional-safety standards across multiple industries:
ISO 26262
Automotive
IEC 61508-3:2010
Industrial / cross-industry functional safety
ISO 25119
Agricultural & forestry machinery
IEC 62304
Medical device software
EN 50716
Railway / rail applications
For ISO 26262, BTC TestStack is certified with the highest Tool Confidence Level (TCL), valid for all ASIL levels including ASIL D. We provide the certificate and report to customers free of charge on request — removing most tool-qualification effort on the customer side.
Related use cases
Where BTC TestStack goes next.
Use case
Requirements-Based Testing
Requirements-based testing for Simulink, TargetLink, Embedded Coder, handwritten and AI-generated code.
Use case
Code Coverage
In BTC TestStack, code coverage is automatically calculated and updated in the background.
Use case
Formal Methods
The power of proof for safety-critical systems.
Feel the Quality
Talk to our engineering team.
Tell us where your engineering problem actually is. We'll tell you whether an evaluation license, a workshop, or a conversation is the right next step.