bluev8

That is a fair point, and I agree that trying to log “everything in the truck” would be the wrong direction. There are a lot of modules and a lot of traffic. If the product became a full-truck datalogger, the amount of data would get huge very quickly, and most owners would never use it. I think the first useful version would need to be narrow: - powertrain-side event evidence - selected high-value parameters - communication / voltage / reset events - pre/post event window - short report first, raw log only as backup One distinction I should make is between active OBD/PID polling and passive bus capture. If you are polling PIDs through OBD, then yes: the more parameters you request, the lower the effective sample rate becomes, and you are adding diagnostic traffic to a vehicle that is already busy running itself. With passive CAN capture, the recorder is not asking all the modules for data. It is listening to traffic that is already on the bus. So it does not consume vehicle bus bandwidth in the same way that a scan tool polling hundreds of PIDs would. But your point still applies in a different way. Even if passive capture does not add bus traffic, the recorder still has limits: - processing rate - storage rate - timestamp accuracy - decoder workload - event filtering - report size - user attention span So the answer cannot be “log everything and let the user figure it out.” The product would need to store enough raw evidence to be useful, but only decode, graph, and present the important parts around the event. A practical report should probably show: - what triggered the capture - how much pre/post data was preserved - which selected parameters changed - how those values compared to baseline - whether the same pattern happened before - whether any voltage, reset, bus-off, lost-message, or communication fault occurred - selected graphs around the event - raw data only as supporting evidence So I agree with you. More data is not automatically better. The real product is the reduction from raw data into a useful event report.

That makes sense, and I agree with most of that. I think the product would need both: 1. a default powertrain template, so it is useful out of the box; 2. user-selected priority parameters, so the owner or shop can choose what they want to see first. Different users are going to care about different things. One owner may care about oil pressure and voltage. Another may care about misfire trend, AFM/DFM behavior, or U-codes. A shop may want communication events and repeatability first. Your baseline point is probably the most important one. Raw data is not very useful unless the report can show what normal looked like for that vehicle under similar conditions. The way I would think about it is: - start with a basic known-good baseline - learn normal behavior for that specific vehicle over time - allow the event to be overlaid against baseline - show whether the event was a one-time spike or a repeatable pattern - provide a simple severity level, but with clear limits on what that severity means For example, early severity could be something like: - Info: event captured, no obvious abnormal pattern - Watch: value moved outside baseline, but not repeated - Warning: repeatable abnormal pattern under similar conditions - Critical: communication loss, voltage drop, bus-off, reset, or severe repeated event I would not want the first version to say “replace this part.” That would be overclaiming unless there is repair-confirmed data behind it. It would be more honest to say “this pattern deserves inspection.” On the OBD port question, I think OBD absolutely has a role. OBD is probably the right place for: - DTCs - freeze frame - VIN - calibration information - normal scan-tool parameters - Mode 6 / enhanced diagnostic data if available The reason I am still looking at an ECM-side recorder is that the failure may happen before anyone connects a scan tool. If the owner plugs in a scanner after the event, the pre-event evidence may already be gone unless the ECU happened to save it. So I do not see this as “OBD versus ECM-side.” I see it more like: - ECM-side recorder: always armed, rolling buffer, event evidence - OBD/DLC companion: DTCs, freeze frame, VIN, calibration, normal scan data - phone/cloud: status, notes, upload, report generation, notifications I agree that phone connection and push notifications would be useful. I just would not want the phone or cloud connection to be required for capture. The recorder should save the event locally even if the phone is not connected. The phone should help with event marking, download, notes, upload, alerts, and report viewing. For a default GM V8 event report, would this list make sense? - RPM - calculated load / MAP - throttle position - vehicle speed - gear / torque converter state if available - coolant temperature - oil pressure - oil temperature if available - battery voltage - commanded AFM/DFM state if available - actual AFM/DFM state if available - misfire counters / roughness by cylinder if available - fuel trims - relevant U-codes / communication events - bus-off / lost periodic message / module reset / voltage drop events Which of those would you remove, and what would you add?

That makes sense, and I think you are describing the real product problem. Capturing data is the easy part. If the owner or technician has to manually dig through five minutes of millisecond-level logs, the product has already failed. The device would be at the ECM harness, not at the OBD port, so I agree that data retrieval and event marking need to be thought through carefully. The way I am thinking about the architecture is: The recorder itself should not depend on a phone, app, Bluetooth, Wi-Fi, or cloud connection to capture the event. It should always keep a local rolling buffer and lock the event locally. A button, phone app, or small cabin device would only act as an event marker. If the driver feels a stumble and presses the button 10–30 seconds later, the pre-buffer has to already contain the useful data. For data retrieval, the practical options would be a sealed service USB lead, Wi-Fi download, or a phone/cabin companion device. I would not expect the owner to remove the ECM-side module or work with raw files directly. The cloud or AI side would be for interpretation, not for capturing the event. The truck may have no connection when the issue happens, so the evidence has to be saved locally first. After that, cloud processing could help decode the data, compare it against baselines, and generate a readable report. For the first version, I would keep the automatic triggers conservative and objective: driver event marker bus-off error passive voltage drop / brownout device reset FIFO or queue overflow a normally periodic message disappearing side-to-side communication mismatch, if the topology supports that For “learning normal,” I agree with your point, but I would not want to overclaim it as automatic root-cause diagnosis at first. A realistic first step would be learned baseline comparison for that specific vehicle and operating condition. For example, a value would only be compared against similar conditions: RPM range load / MAP throttle position gear / vehicle speed coolant and oil temperature battery voltage AFM/DFM state, if decoded and validated Then the report could flag things like: this periodic message disappeared compared with its normal timing this value deviated from this vehicle’s normal range under similar conditions the same abnormal pattern repeated after the same type of event the anomaly occurred together with voltage, oil-pressure, misfire, or communication changes But I would still call that “abnormal pattern detected,” not “replace this part,” unless there is enough validated repair data behind it. So the intended product would not be “here is a huge log.” It would need to be an event package: what triggered the capture how much pre/post data was preserved what changed before and after the event whether the device itself reset, overflowed, or saw a bus error selected graphs around the event raw data only as supporting evidence From your perspective, what would make this kind of report useful instead of just another datalog? For example: What are the top 5 parameters or events you would want highlighted first? Would you trust a learned baseline for that specific vehicle, or would you prefer fixed thresholds? How much false-positive flagging would be acceptable before you stopped looking at the reports? What would a one-page report need to show for an independent shop to take it seriously? For misfire, AFM/DFM, oil pressure, or U-code complaints, what would you want the tool to flag automatically?

New member here. I am researching a read-first event-recorder concept for late-model GM V8 trucks. This is not a sales post. There is no product link, price, preorder, or mailing list. I am trying to determine whether the underlying problem is real before building anything. Has your truck ever had a brief problem such as: - rough running or a momentary misfire - an oil-pressure warning or unusual pressure event - reduced power or a brief stall - a U-code or lost-communication problem - a symptom that disappeared before the dealer or independent shop could reproduce it If so, I would appreciate the following details: - year, model, engine, and mileage - what happened and under what conditions - whether a DTC and useful freeze-frame data were stored - whether the shop was able to reproduce it - what the eventual confirmed repair was, if known - what additional information would have helped the diagnosis The concept being evaluated is a removable leave-in recorder that continuously retains a rolling window from before and after an event. It would not tune, reflash, clear codes, or change the vehicle calibration. I am also not claiming that it could predict lifter failure or see every internal ECM variable. The real question is whether continuous event history would add enough useful evidence beyond freeze frame, GDS2, and existing scan tools — or whether it would simply be another unnecessary gadget. For owners and technicians, which problem would make something like this genuinely useful: 1. intermittent misfire or AFM/DFM-related behavior 2. oil-pressure events 3. lost communication or electrical faults 4. none of the above Please be blunt. Negative feedback is just as useful as positive feedback.