Video: Analog/Mixed-Signal (AMS) Safety Verification Flow for Automotive IC Designs | Duration: 3820s | Summary: Analog/Mixed-Signal (AMS) Safety Verification Flow for Automotive IC Designs | Chapters: Webinar Introduction (29.95s), FMEDA Driven Verification (448.045s), MIDAS Safety Platform (1510.345s), ADE Artist Demo (1901.6951s), Safety Analysis Demonstration (2999.8599s), Fault Sweep Explained (3440.76s), Visualizing Fault Injection (3523.155s), FUSA Export Clarification (3654.0698s), DCLS Implementation Query (3685.85s), Test Bench Transformation (3724.325s), Webinar Conclusion (3773.9001s)
Transcript for "Analog/Mixed-Signal (AMS) Safety Verification Flow for Automotive IC Designs": Welcome, everyone, to today's webinar, analog mixed signal safety verification flow for automotive IC designs. We will begin in a few minutes. Welcome, everyone, to today's webinar, analog mixed signal AMS safety verification flow for automotive IC designs. We will begin in a few minutes. On behalf of Cadence Design, I would like to welcome you to today's webinar, analog signal safety verification flow for automotive IC designs. My name is Kale Hawkins. I'm hello? If you could, mute your mics, I'll we'll finish with the introduction, then we'll we'll go from there. K? Again, my name is Kale Hawkins. I'm the field marketing manager here at Cadence Design. Before I introduce today's presenter, I'd like to go over a few housekeeping items. All attendee lines are muted to prevent background noise. There will be a question and answer segment at the conclusion of today's webinar. If you have any questions during the presentation, navigate to the educate engagement panel on the left hand side of the screen and locate and open the q and a icon and type your question into the text box provided. We will attempt to answer all questions at the end of the webinar. However, if we run out of time, questions may be answered via email within the next few days. Please ask your questions early so that we can get them queued up and ready to go for the q and a session. Today's webinar is being recorded, and a link to the recording will be emailed to you in the coming days. Final note, we will control the movement of the slides. Some slides include graphics, may take some time to advance, depending on the speed of your connection. At this time, I'll turn the event over to our speakers joining us from the Munich office, Jens Nave, product engineering architect at the Cadence Design custom product group, and Farhan Rasheed, the principal product engineer in the systems verification group. Jens, take it away. Thanks, Kale. Also from my side, welcome to the Cadence Design Automotive webinar series. Our topic for today is the FMEDA driven analog signal safety verification. In the next hour, we will discuss how full cadence flow can be applied to the safety verification of automotive mixed signal designs. I'm starting off with a short introduction, just two or three slides that will be really quick. Then I will present the functional safety flow architecture. Next, I will cover the false simulation as the foundation technology. And after that, Farhan, will take over and guide you through the MITRE safety platform for EPBA. In the last section of the session, there will be a short live demo of the safety verification flow with ADE artists and MITRE. This is the motivation slides. So the ISO two six two six two standard is a comprehensive framework, for achieving functional safety in automotive systems. It's there to ensure that electric components perform safely through the vehicle life cycle even in case of hardware failures. Hardware failures are generally unavoidable. Our focus here is the chapter five of the ISO two six two framework, the product development at hardware level. Essential methodology is FMEDA. This stands for failure modes, effects, and diagnostic analysis. FMEDA is used to to systematically investigate and assess failure modes, quantifying failure rates, and guiding the design of robust safety measures. Fault simulations, based on fault injection techniques, is an important aspect of functional safety verification. The ISO 26,262 standard mandates, a rigorous assessment to ensure that the system performs safely even in the presence of faults or hardware defects. It is recommended to apply fault simulations for safety critical components such as designs targeting ASIL c or ASIL d certification. So you can see on the table here, there's certain safety metrics required for, for the respective automotive certification level. Fault injections, mimic hardware defects, allowing to assess the system behavior under abnormal conditions with with a hardware defect in place. This webinar this webinar, we only we will only present a small subset of the overall cable's functional safety solution. We are focusing on the MITRE safety platform product here, and and the virtual also ADE spectral legato part on the lower side of the of the slide, for the IMS fault campaigns. But to paint the bigger picture, the Cadence Design functional safety solution also addresses, digital design and verification here on the right hand side with products like, Verisium, Xelium, Chester, Palladium for simulation, formal and and emulation, as well as the digital implement implementation part to allow safety aware digital implementation, plus our design IP and methodology consulting offerings, which are not even mentioned here on that slide. Let us focus on the AMS safety verification flow for now on a more technical level. From looking at our customer base, we figured that the environment which the analog and mix signal design engineers are most comfortable with is the virtual also platform, the ADE simulation environment, and the spectra and AMS designer simulators. To account for that, we have centered our solution around virtual ADE rather than a digital verification cockpit. The FMEDA is managed through the MITRE safety platform product, which handles both analog and digital. The flow is as follows. The design information is in the Cadence OA open access database. The virtual also site collects the design information for a device under test in the design hierarchy extraction step. So this is collected and, made available to to MIDAS for design mapping step. MIDAS then issues so called fault campaign orders. The fault campaigns are fault simulation sweeps being managed by artist and are simulated with Spectre or AMS designer. The FUSA results are then collected by by NDE artists and provided back to MIDAS as a FUSA report here. So this is the back notation step of functional safety results. Going a bit more into details. So this is the next level. The fault campaign setup is done in ADE artist. So that means the test bench and default in analysis setup here. It's done on the AD artist side. The fault expressions are then calculator measurements that monitor the functional correctness and the check output in ISO terms, the observation and the detection points. The heavy lifting is done by Spectra or AMS designer. That's the fault simulation sweeps, down below. These can be run-in in many parallel jobs at the same time. Then fault sweeps, the results are collected, and FUSR summary reports are then read back into MIDAS. That's the big notation step here. For each failure mode, it is recommended to use a separate ADE artist session. So in case of two failure modes, it's recommended to have a second ADE artist session, and there can be many imperatives, many failure modes that you want to account for. ADE can run fault campaigns in a highly parallelized fashion. So the fault campaign run time is strongly dependent on on your IT resources and the available licenses. Going into the topic of the legato fault simulation, which is the technology foundation for Adanuk fault campaigns, There are three steps involved in the defect analysis flow. The first step is the defect identification. So based on the DOT circuit topology, there are possible hardware defects extracted, and the user controls the step through fault rules, explaining where to expect possible bridge and open faults and what weights to assign for these faults. The second step is then, the faults get injected into the DOT one by one, and the fault observation and detection is collected. Either all faults can be run-in the fault sweep covering the anti so called fault universe or a reduced number of faults through sampling. In the third step, the fourth sweep results are used for coverage analysis, calculating the fault detection matrix and generating FUSAR summaries. Going a bit into the details of of the defect extraction, there's a number of possibilities where bridge or open faults can occur in a transistor level circuit. As an example, for a MOS transistor, a hardware defect can create a short between two terminals. So these are the possibilities here between drain the drain and gate, drain and source, and so on. Or there can be opens on on terminal spins. And not all defects have the same likelihood. So in the fault rule setup, GUI, the user can define equations for assigning fault rates. The default is to extract all faults with an identical rate of one, but you can also define area dependent rate equations like we have it here, the weight expression w times l, which is an area type of base code. The collection of all of the possible faults of a d e u t is called the fault universe. What is done automatically for the users is then the fault collapsing, the reduction of the total number of faults. So there are faults which have identical effects, like, if you consider here the drain the drain source, if here, the short gets created, and that would have the same effect as a short created between the source and the gate here. So, a gate a bridge fault for this PMOS structure can be simulated once with just one fault being inserted with, with the sum of the weights. Other faults can be removed entirely. So for instance, bridge faults for terminals that are shorted in the design anyway like this. In this current mirror structure, the design intent to to short, drain with gate. So there's no point in trying to introduce an additional short here. And there can be structures like spare devices in the in the design that have short terminals anyway, so there there also wouldn't be a point to to introduce shorts here because it's already sorted by design. For the fault ejection simulation, there are, in general, two different algorithms available in spectrum. The so called direct fault analysis, DFA, and the transient fault analysis, DFA. In the DFA, the fault is injected at the beginning at time zero, and then there's the transient simulation run, and the test bench is simulated with that particular fault. At the stroke point, it is evaluated whether there is the fault observable or detected, and this is done by each and every fault separately. So the next fault gets inserted at time equals zero, and then the same procedure is recalculated. TFA, the transient fault analysis runs a much shorter loop. So it injects the fault at a certain so called lead time before the screw point, allowing to have a bit much shorter loop here in order to simulate less less time steps. And at the scope point, the detection is evaluated. TFA can be much faster than TFA, so some customers mentioned up to a 100 times faster. But it needs more user experience, so I would recommend, in general, to start with DFA. That's the golden reference, and then experiment with DFA and a suitable lead time. Sorry. Let me explain the fault classification process. We distinguish between functional outputs, the so called observation points, and check outputs, the detection points. So consider this example here. We have a functional module device under test with, safety mechanism, a diagnostic module in in vision mode, and that's the direct output, the functional output, and that's the checker output signal and the detection point. After injecting a hardware fault here or injecting the hardware faults one by one, this can cause a functional problem in the circuit or lead to a checker detection of that fault or both or none of them. So the the different combinations that can occur is, unobserved, undetected through the classification c one, which means a fault doesn't propagate to the output. It's not visible or not observed, and it's not detected by by the safety mechanism. So that is unobserved, undetected. C two means, the fault is not propagating to the outside, so it's not observed. But, anyway, the checker detects a problem, and that is unobserved detected. The third combination is, the fault is propagating, to the outside system, so we see a functional problem here at the observation point. And the diagnostic module, the safety mechanism is able to catch it, and that means it's dangerous detected. And the last combination here is dangerous undetected where, there is a functional problem, but it's missed by the checker. So the checker doesn't see it, which means there is no way for the system to, mitigate or recover. The diagnostic coverage number tells how the how well the checker is performing, detecting dangerous faults. So this goes into the FMEDA safety metric calculation. It's not yet the safety metric, but this is used then in for FMEDA to calculate something like failure rates. There is new infrastructure in the new Cadence which also '6, '25 dot one release. It's called the FUSA functional safety results view in ADR test. So the the view directly shows the functional safety related fault detection matrix as a result at fault simulation sweep run time. So the user sees the detection matrix filling up with results, including the safety related fault classifications, UU, UD, DU, and DD. It's not not a post processing step that really happens or is up is being updated once, fault injection simulation completes. All the user needs to do is to provide fault expressions, that are calculator expressions and add an additional tech or property to it to distinguish between functional and checker detection for expressions. Then this distinction is known by the ADE artist at runtime, not only doing post processing. So I'm going to show this in a demo so it's it's very easy then to to just put an additional check onto that. And then this runtime classification, is possible in that AD artist, results view. After the simulation completes, the sweep, there's a typical FUSA report that the user can create. And so you see the defect coverage being listed in the text report or the diagnostic coverage or the unobserved, undetected, unobserved, detected, dangerous detected, dangerous undetected percentages. So these are weighted, numbers, that are given as part of the function safety report. The diagnostic coverage number here, this is to be used in FMEBA for the single point fault metric calculation. There's one step missing, which I mentioned in the beginning. So the last step that is needed for the MIDAS FMEDA flow is the DHE, the design hierarchy extraction. MIDAS needs the design information for for the design mapping step. There's a new icon in ADE artist in the false setup assistant. If you click on that, the safe DHE database GUI is prefilled with the DOT information and the file name. So by clicking on okay, the DHE database file gets created. It's really that simple. So we will show that during the demo. Let me cover that topic really quickly. So analog centric simulations in general can be time consuming, especially with DFA. A good indicator of the runtimes, the nominal run time without faults in the size of the fault universe. There is a number of tech techniques to manage the fault campaign run time. So the first one is consolidation of the fault list, trying to reduce the number of faults that need to be simulated. So this can be done by fault collapsing, merging, waiting. The second option is to, wisely choose between SpectreX and FX solvers. So the FastSpice solver FX can significantly speed up the simulation. In general, the the rule is only run with the accuracy level, that you really need in the end. Then the fourth algorithm, as I mentioned, the TFA can be much faster than the TFA. So this is something to consider. Fault sampling means rather than simulating the entire fault, even worse, all possible faults, you can apply sampling like random weighted sampling. This takes the forward weights into account, and the coverage result will be reported with confidence interval. So this may be accurate enough, especially for early investigations with a largely reduced number of of fault runs. And, of course, the fastest approach approach is to run a fault sweep massively parallelized on powerful machines. It depends on the IT infrastructure and the license availability. We have received good feedback from customers having adopted our SpectraFX FastSpy solver, the next generation FastSpy solver. On the highest precision side, our golden reference product is SpectraX. Both X and FX offer a variety of accuracy and preset modes, and this allows trading of speed versus accuracy and optimizing the simulation run time. Okay. And with this, I would like to hand over to Farhan for the MITRE safety part. Hi, Jens. Thanks thanks for the for the all the details. So I will talk about MIDAS safety platform, which is FMEDA based function safety design and verification tool. At this point, you might be wondering that if what you saw are analog design environment, you can do the design and, fault simulation, then why do you need something as an additional tool or as a safety platform? So this is answered through this slide. So if you see the functional safety concept, if you see the sub safety analysis and, verification flow, you start from functional safety concept, which is which typically contains your safety requirements. Then you move to FMEDA, defining your failure modes, error and effect analysis, and then moving to the end, which is the safety metrics. So in this in in this flow, this this intent of safety analysis is not part of the IC design tools. So there is no formal way to propagate this concept of functional safety and, technical safety concept inside IC design environment. Hence, to close the gap between FMEDA and safety verification, we introduced this detailed FMEDA, which allows the designer to map their FMEDA hierarchy to associate design. And in in this presentation, actually, our focus is on analog mixing design where the abstraction layer is complex compared to digital design. So you have a transistor level design, and, you have to do safety for safety verification, you run on analog fault simulations. Could you please go to the next slide, Jens Nave? Yeah. A quick overview of, Midas safety platform. So with Midas safety platform, you can run FMEDA based functional safety analysis for analog and analog mixed signal and digital design flows. If you do not have your, design data available, you can also do architectural exploration. So you can put estimated values for the design data and generate FMDA reports only based on estimations. And when you have once you have your design data available, you can map your FMDA to your design and then run the detailed, FMED analysis. It also supports unified safety format. So with USF format, you can actually define your safety goals. You can define your safety objects. Next slide, please. Yeah. So to, to make it easy for you to follow the demo, so this is the slide I, prepared to give you an overview of how Midas application look like. So in the blue in the blue box, you see the standalone GUI, application my Midas application. It also has a command line interface, which is shown in the screen box. So and the back end is USF, so it's a TQL environment. You can you can automate your safety verification flow through it. And, this pink box is the shows the summary of your FME data. And these matrices, the safety matrices are calculated through the Midas backend, which is functional safety engine. It supports ISO two six two six two and IEC six one five zero eight standards. Yep. On this slide, I I just showed the key key information required to follow the, the demo. So on the top left, you see this safety case, which we call as FMEDIA project. So you create an FMEDIA project. Under this, this there is a green portion, green highlighted part, which shows your FMEDIA hierarchy view. So this is how you actually do the design partitioning, and then you map your design information. And then you have the pink overlay part where you have your design hierarchy view. So this design hierarchy is what Jens just showed that you can export your design in a in a hierarchical view, and then you can import the same design inside Midas. And below that area is the shared library view where you can define your technology, your fit rate for the technology, also the safety mechanisms. And on the right side, you have this green green portion showing the failure modes, how you can define your failure modes. Next to that is, the design instances which are mapped to your failure mode copied from the design hierarchy view. And then, IC design parameters, they are automatically calculated from your design hierarchy. This is the last slide before moving towards the demo, just to show the difference between the architectural FMED and, detailed FMED. So once you start from architecture of FMED, you do not have any design data available. You define your FMED hierarchy, and then you you can calculate your failure rates and distribution based on estimation. And when you have your design data available, then you you are in the detailed FM media mode where you have you can define your, parts and subparts of your, from the actual design, and then you can map these parts and subparts to your failure modes. So at at this point, you have failure rates estimated or calculated based on the design instances area, but still, this does not include the safety metrics as, calculated from, your simulations. And for that, you can run fault simulations and do detailed FME data analysis based on annotations. So in this part, even once you get the safety reports from spectry or analog design environment, you can bring these reports inside MIDAS environment and annotate your safety results to your failure modes. Yes. As a collateral, we have a rapid adoption kit available, which is based on what Jens explained. So starting from, what you saw analog design environment from design and simulation. And it also contains the second, module, which is about MITA's safety verification, where you can define your FMED hierarchy and moving to towards the safety analysis and verification. And now for the demo. I would like to hand over to you, Jens. Thanks, Farhan. Yeah. Coming to the demo part, yeah, I just wanted to add, the demo database or example that we are showing. This is here packaged as the rapid adoption kits. That's something you can pick up on the ASK portal and try yourself so we are not using a different example for the demo. The demo example is, band cap device under test with the safety mechanism. So the safety mechanism is, in this case, it's just the Verilog a module as a, simple checker, and we prepare two failure modes for that device under test. So failure mode one, is means that the band gap output voltage is out of the specification range, and we defined, the spec range for the for the band gap out voltage to be between one dot three eight five volts and one four one dot four one five volts. So whatever this is outside of the range, it's considered fading the spec. The second fader mode is that, a defect is causing, the band gap output to become high ohmic. And, the yeah, definition we used for that is if the output impedance of the band gap is more than 500 kiloohms, then we consider this, the failure. So with this example, we want to demo the ADE artist step. So the fault campaign setup running the DHE extraction, generating FUSA results reports, and we are mimicking the flow in a way that I'm handing over, files of FUSA results in DHE database files to MIDAS to navigate, FMEDA and MIDAS based on these results. Okay. So that's not the demo part. We are using the band gap lip, test bench band clip. And as I said, for each failure mode, it's recommended to have a separate ADE artist session. And the first Maestro session is, the the one associated with failure mode one. So if we open that by the way, I'm using, the most recent Cadence releases. It's 51 20 what 25 dot 1 ISR two on both the the Spectre and the virtual ASO side. And what is new with 25 dot one is the ADE artist tier. So you see in the banner, it's ADE artist. It looks similar to ADE similar, but it's new ADE tier. And, I have switched to the fault simulation workspace. There is the fault setup assistant on the right hand side. I'm coming to that. And, here's the setup. So what I wanna wanted to show is, in the fault setup, there are two things to be done. One is the fault rules. This is done here in the fault setup assistant. So if I'm clicking on that, you see the fault rules extraction form, and I'm having one rule to extract bridge faults for the device under test. I'm extracting I've had a simple setup here. It's, the one as a weight expression. I'm using the IEEE two four two seven mode and enable fault collapsing. So this is the RB one rule defining which faults. And I also have a parallel open rule, which, is also run-in I I triple e two four two seven mode. So that's the rule setup for the fault, extraction. What is what needs to be done is to define fault expressions on the output side. So I'm going here to the output setup in ADE artist. I have three relevant expressions here. One is the bandgap voltage at fifty milliseconds. It's a fault expression. But the new, feature or the new property, which is added is the tech, telling this fault expression to be considered a functional and an observation point. So here's the calculator expression and the combination of fault expression type and the tech functional means it's an observation point. The second functional expression here, which is the difference between, the band gap voltage at hundred milliseconds and fifty milliseconds, it's also it also has this additional tech functional. The check if you, look into the into into the environment the first time, the check column may not be visible, but you can enable this through the right mouse button here. So this please make sure that you are enabling, the display of the text column. Then there is a checker for expression, which is the, output signal of the, simple checker. So it's a fault expression type with the additional tech checker. And this simply monitors the output of the simple checker output. It must remain below 0.5 volt. If it goes high, then this means that the fault is detected. So here, it's the normal feature of defining spec ranges for the bandgap voltage. That's the voltage range considered within spec. And for the, OP settling, I have a spec defined. It should be less than seven milli millivolts difference. And the safety mechanism output should remain at zero if it goes above, 0.5 volt and the checker detected default. Okay. The device under test or the test bench, let me open that. So we only have one test here. Open design and tap opens the test bench schematic. So that's, the simple setup here. We have a disabled signal going into the band gap. So once the disabled signal is released, then the band gap work starts to work, and it's producing the band gap output voltage here to note, bandgap out. And the safety mechanism has a simple check out signal here. If we go down one level of hierarchy, that's what you saw in the screenshot. It's the core bandgap circuit here, yeah, producing this reference voltage plus a safety mechanism in mission mode, which is, in this case, just the dialogue a, behavior module. Okay. So that's all in terms of the setup. For false simulations, we need to run-in the run mode false simulation. In the interest of time, I'm not starting the simulation. Only one word, for the job policy setup. False simulations or, these sweeps can be time time consuming, so please make sure that you are running and distributed with as many jobs as you can can afford in order to reduce run time, can be 100 or more. That really depends on on the resource that you have. But here in this example, I'm I have been using, a SunGrid engine distributed setup, but we also have the command mode for LSF and and other options available. So I'm distributing to Red Hat eight machines, and I was running on eight parallel jobs. In the interest of time, I'm opening the results directly. So that's the false simulation dot two, history item view results. And that's the FUSA result view that I mentioned. So there are, it's a new view in parallel to the to the ones that you know from previous releases, like detail detail transpose and so on. And it has a special FUSA, setting here. And once this is enabled to see the fault expressions being evaluated for each and every fault, so the the number of, you see here represents one of the faults being applied to the device in a test. And then after the simulation run, the expressions were evaluated. And color coded means, that's the simple checker output here. If it's green zero, that means the check output remained on zero. No detection. If it went up to one, it's outside spec and detected the fault. And similarly here for the, functional fault expression, so that's the band gap at fifty milliseconds and the difference between hundred and fifty milliseconds. And this combination is then combined into the safety category, DDUUDD UD and DU. So just searching for one example here. So for instance, here we have dangerous undetected because the check output remained on zero, so it didn't detect anything. But the functional fault expression detected that the that the bandgap voltage is slightly outside of the of the spec range. So that's the result summary. You still have, the all the detailed analysis capabilities that you are used to to use for, in ADE. So for instance, you can use the detail view and see results and do very detailed post processing and debugging. But the standard way for functional safety sweeps is to have have this fault FUSA results view where you see the FUSA categories at a glance. After the sweep, you generate FUSA reports with this icon here. You are selecting fault expressions to fall into the functional expression or checker expression category. So I'm doing this for failure mode one. And then the functional safety report is created from the results database. And you see in the functional safety report, like the for instance, the diagnostic coverage number, the fault tolerance, and these are the c one, c two, c three, c four results, unobserved, undetected, unobserved, detected, dangerous detected, dangerous undetected. So that's the FUSA result summary for for particular tests. Then you see all the detection details for all of the different, faults that were that were run. And that's then the summary for all of the tests down below. So here, in our case, it's simple. We just have one test in ADE artist enabled, but there can be multiple tests. So there can be more than one section in the FUSA report. As an alternative, because, in MIDAS, we can also import or read, JSON results for the functional safety report. You can also generate a JSON, result structure for that functional safety summary. So here you see the defect coverage number, the diagnostic coverage number, the detection percentages, and so on. And the structure also supports confidence intervals. I mean, here, we were running the entire fall sweep. So we have the upper and lower bound of the confidence interval is null. But let me show, this can be quite handy if you run a sample sweep. So this is an example here. We ran a sample sweep with auto auto stop. That means, once a certain confidence interval for the criteria for the diagnostic coverage was reached, the sweep was stopped automatically because it was seeing, there's no more faults needed in order to close that on this accuracy goal. Then you see in the FUSA report also the confidence interval information. So here we see, for instance, the defect coverage was measured, but because we were running a sampling sweep, there's some uncertainty, and that's the upper lower bound of the confidence interval with the global Pearson, calculation. Okay. So that's the information given to Midas. The second information that needs is needed for Midas is the design hierarchy extraction. So if you, go back to the design, we want to we want to summarize all the design internals, the hierarchy structure inside of this device under test. And for this, there's a new icon in the fault setup GUI. Because the DOT is already defined here, we can just click on that, run design hierarchy extraction icon, and then it's automatically prefilled. The DOT is the band gap with safety mechanism, which is the DOT definition here. It defines a a file name. So if I click on okay, it extracts the design hierarchy, writes into a database file, and that's that's the file here that is read into MIDAS for the for the design mapping. Okay. So that was failure mode one. And, similarly, for failure mode two, it's a similar exercise. So that's the high z, high omic type of failure mode. We have a separate Maestro session for ADE artist. It's quite similar to the the one we just discussed, but here the test bench looks slightly different. So I'm opening the test bench schematic. We have the same DOT, but some, yeah, different circuitry here for detecting that failure mode. And, also, the fault expression is slightly different, so we are measuring the output, resistance the effective output resistance of the device on a test, and that's the functional expression here. After running the fault sweep, we are getting let me open the results that are produced. We have a similar detection matrix, but now for the failure mode two for all of the faults of the device and the test that were simulated. And then we are creating the functional safety report through this icon here, through the screen icon. The fault expression functional the functional fault expression here in this for this failure mode is the r out effective. The checker expression is still simple checker out. It's the failure mode two we want to consider, and then the functional safety report gets created. So here we have slightly different numbers. This is related to the failure mode ID number two, and this is the second file or information in terms of FUSA, summary that is handed over to MIDAS. Alternatively, as I mentioned, the JSON results file is created anyway by Cadence Design, so we can also extract that. So that's the JSON file, JSON structure, and this can be handed over to to MIDAS for the for the back end notation step. Okay. Did I forget anything? One thing I wanted to mention just for the detailed debugging that might be interesting. So, I mean, it's very handy to have a FUSA summary here to see the statistics, whether you, you are in the in the correct ballpark for the diagnostic coverage and and so on. But what I wanted to show is that you can still do all the detailed debugging work. So if you want to drill down into the results, I want just want to show one simple example for the first failure mode. So I'm opening the results again, view results. You can still change to the different detail view, for instance, inside of the, you see for each of the design point, for each of the fault, all the signals that were saved. And, for instance, you can now plot all the band gap output voltages for all of the faults with the spec definition, seeing what failed, this particular spec here. So let me do that quickly, because it's then nicely grouping all of the bandgap voltages for all of the faults over, the spec criteria. So we see all the waveforms that were passing that particular spec, all the failure all the faults that we are failing and all that that that are near, that means still failing, but only for little bit. So these are you can still go into all of these separate results if you save these and and debug each and every single fault if you want to. Okay. With this, I would like to, hand over to to Farhan again to follow-up on the on how to import these results into into MIDAS for for the detailed SMEDA with back end notation. Thanks, Joanne. Thank you, Jens. I will share my screen. Okay. To save some time, I have already prepared the environment. I will quickly go through the environment so that, you can follow. So, I created a project, the call bandgap FMEDEA, and then I created a FMEDEA hierarchy, which is nothing but just defy defining your parts, subparts. And so it's a divide design partitioning. For example, what Jens showed as a design, I changed it to I transformed it to two parts, functional block and checker block. So checker blocks, is actually the safety mechanism here, and then also map the design instances. So the design instances are mapped from the design hierarchy. So this is the design hierarchy view, and this design hierarchy is also extracted from the analog design environment. And once these are map mapped, you can further, define the subparts of of your of your design. For example, here, we have only defined into two subparts. Then you can start defining your failure modes. What showed, there were two failure modes about, band gap output and high pretenses output. So we added some details that this failure mode is relevant to this part, this subpart, and some description about the failure mode. And the design instances are then, copied to this failure modes. Then there is, another information. So you also need to map your, failure mode to the technology. So which technology are are you using? For example, here, under under the shared library path, you can define your technology. For example, for this example, we took generic PDK. It's a analog type pathology, and we define the fit rate for this technology. And as a safety mechanism, we have only one hardware block of safety mechanism. For that, we define some default diagnostic of these values. Okay. This is all about failure mode definition. And then if if you have to do safety verification, then we go into the safety verification tab. We define our campaign. So we we say that, okay, for these two failure modes, we would like to, run or execute the strategy of safety verification. So we define our metric for which metric and then fault injection type and then some other parameters that you can define here. And also for annotation in the same menu, you can also define the annotation strategy that if you would like to use, expert standard expert judgment or some other methodology for results annotation back in MIDAS. With this, you can actually generate the collaterals. So if you press this generate button, you will generate the collaterals. I will show you the collaterals. So you will generate a campaign order directly something like this. It will have the simulation setup. For example, if this strobing point will have your, the the equations not the equations, but just just the signals, and then you have to transform them in into equations in in any environment. But it will have some information that you have defined here inside MIDAS environment to set up your false simulation. Here, I have already copied the, simulation results that Jens already executed. So these are the the results in JSON format. For for both of the failure modes, I have already copied these results here. And then once you have your, results simulation results finished, your results copied back to the failure mode directories, you can use this annotate option to annotate your results. So what happens when you annotate is I will go here just to show you where is it used. Yeah. For example, I have my failure modes defined, and I have some default diagnostic coverage. So this is the diagnostic coverage default values. And, currently, this is 0%. I have, based on estimation, there are 0% safe faults. But if I if I run the false simulations, export back the results into MIDAS environment, and run the annotation strategy, I can annotate the results. And now I can accurately estimate or accurately calculate based on simulation results how what is the coverage for the safe faults and what is the diagnostic coverage. So and, also, if you have your if you use your estimated values, you can still calculate it, and then you can also export your summary of the of the of your safety analysis. So you have many parameters from from your safety matrices. So you can you can and, also, some details on your FMEDEA, how many parts and failure modes are there. So but this so these are all without simulation results. And if you would like to see based on your simulation results, you move back to your show annotated values, and then you recalculate it. Press the summary button, and then you have your FMED results now assisted through, fault simulation results. So here you are seeing this using annotated values. That concludes our demo. Now I would like to hand over to Kale again. Thanks. Thank you so much, gentlemen. We will go ahead and take some questions, now. As a reminder, please be sure to type your questions in the q and a panel off to the left. You can click on the icon on the left, and you can type it in there. Or to the right, I think you may have an option to type it into there as well. We'll go ahead. And, Jens, I think you can go ahead and start answering some of the questions now. Okay. Let me connect. There was one question regarding the fault rates. Is it recommended to use weight equations? So what do these rates account for, and where do do I get these equations? So this goes back to the to the question how likely, a device is failing, with a with a defect. So in reality, it's, highly geometry dependent. So you can see, for instance, if if a device has has a very small area, then it's much less likely to happen to have a defect there in comparison with a with a, very big device. Also, if there are metal tracks over long distance, very close to each other, then, shorts are more likely. In terms of the equations, a good starting point typically is to have, geometry or area based weight equations, but better would it be still to get some foundry numbers regarding, manufacturing or reliability data. Or sometimes you you have some in house data, available about the device reliability. Okay. So, I mean, it's it's a good starting point to start with a weight of one. The second best option or the more accurate option is to, represent the area, and the best more most accurate option is the the one getting found with data about reliability. Okay. Do I have another another question? I don't currently don't see any additional Yeah. If you click on the q and a button on the left just to get questions, one one is what is a fault sweep? A fault sweep means, you're running all of the different faults one by one. So, typically, you extract the fault universe, and then you, let Spectra automatically sweep over all the different faults one by one and and collect the the coverage information. Another question is could you explain what the safety mechanism Verloc was representative of? In the demo, you mean probably the demo example. So that's something where, it's just a Verloc a module. We didn't have a an actual circuit implementation or a circuit design of a safety mechanism. And in Verloc a, so the analog, flavor of Verilog, we coded, a checker that is monitoring the voltage. And if the voltage is outside of the spec range given by instance parameters, then this was generating a high output. So it's a behavioral language that you can, use to easily model analog hardware without having a transistor level schematic. Request to share this PowerPoint after the webinar. So the recording is is shared with all the attendees. That's correct. If you needed a specific slide or you needed a specific, item gone into further detail, just please contact your account team, and they can organize to get that for you. How can we visualize digital and I look forward injection campaign report together? So that's probably something we could do in in Midas. Right? Typically, I mean, the environments for collecting this data is quite different. I mean, the digital campaigns are typically run-in digital environments like the Resium, whereas, and that's also possible for analog, by the way. So we had this flow before, but we came up with this ADE artist graphical flow because that's really the the genuine environment the analog mixed signal engineers are working in. So that means the FUSA report is rolled up into FMEDA. And, Farhan, maybe you can elaborate a bit how you can combine this report with a spot or see the combined effect in terms of FMVDA. Yes. So this is actually relevant to how you define your failure modes in Midas. For example, if you do if you have your failure modes, which some of so failure mode map to the digital technology, so then, of course, it it has to be annotated based on the digital fault injection campaign. And if your failure mode is based on an analog technology, then the other is it it will be annotated based on this factory fault injection simulation report. Then if you create one campaign that has both analog and digital failure modes, there you can annotate the results, for for to both failure modes in in in one campaign. But here, I see the question on mixed signal fault injection because the question is digital analog fault injection campaign. So here, I assume it the question is relevant to analog signal fault injection campaign where you have both analog and digital design and then cut the the digital fault injection results merged with analog fault injection results and how to visualize this. I understood it this way. Did you also Jens, do you also understand it this way? Yep. Then then this is actually supported in, ADE environment. You can you you can view the AMS fault injection simulation through this FUSA ex, export in in ADE. Yes? Okay. Okay. I'm just Okay. Yeah. They had a follow-up clarification there. They said, yes. My question was about that. Yes. Okay. Okay. Yeah. So Okay. Great. That's one question. I'm just picking up any recommendations or app note for DCLS dual core lockstep USF implementation if shadow core is already present in processor IP third party. So that's something I I can't answer. So I roughly know what that that is, but we don't have the the, ideal colleagues in the call. That's something for us to follow-up and and let you know. So that's something where we, I know who to contact within Cadence, but I cannot answer this right away. We can follow back up with you, Adam, specifically on that one. Yep. Okay. There is one question on, the test benches. So does the test benches in ADE represent failure modes in functional safety domain? If yes, how does the transformation of test benches from ADE environment to failure mode in Midas happens? The first, the cost answer to your the first part is yes. They are actually failure modes. In AD, they are treated as test benches. The transformation can happen, by manually. Manually, you define your failure modes in Midas, or we also have a feature where you can directly import your test benches from ADE environment to Midas by mapping your design hierarchy. Yeah. So mapping is manual, but you can import it. Okay. Okay. I don't see any additional questions. Yep. I think that might be it. Conclude the q and a section, giving back to Kale. Okay. Well, thank you so much. That was a fantastic, webinar and presentation. Ladies and gentlemen, this concludes today's webinar. Please take a minute to complete a short survey I'm about to launch, and it'll be on your screen. It's only four questions. It'll take you ten seconds. On behalf of Cadence, I'd like to thank you for joining us, and, we will definitely see you next time. Thank you so much.