Vehicle testing
A 400 kS/s card will not show you a 50 ns edge — and that is not the problem it sounds like
Vehicle work asks two things of a measurement system that a general-purpose DAQ does not obviously offer: a front end that survives a chassis which is not at 0 V, and enough rate to be useful around events that last microseconds. Knowing which measurements need which is most of the engineering.
Isolation, sample-rate budgeting, adjacent-channel skew, pulse counting and IEPE excitation on a vehicle bench — including where a DAQ card genuinely beats a scope and where it does not.
There is a version of this article that claims a USB module replaces an oscilloscope on a vehicle bench. That version is not useful. A general-purpose DAQ converts a few hundred thousand samples per second; a switching edge on an injector driver happens in tens of nanoseconds, and no amount of cleverness recovers information that was never captured. What the same module does better than any scope is record thirty channels for twenty minutes at a time, count pulses without aliasing them, and stay synchronised to a laptop that is also logging temperature.
The chassis is not at 0 V
A laptop on a USB cable is referenced to mains earth through its power supply. A vehicle chassis is not. Put a single-ended input between them and the measurement includes whatever potential difference exists between the two grounds — which, with a starter motor turning or an injector firing, is neither small nor constant.
The USB-6215, USB-6216 and USB-6218 answer this with 60 V of channel-to-ground isolation. Note precisely what that is and is not. It is a common-mode barrier between each analog input and the module ground, which is what stops a ground loop from turning into a noise floor problem. It is not galvanic isolation between the channels themselves, and it is not a CAT-rated barrier for probing a traction battery. For the second of those, the answer is a differential probe with its own rating, not a DAQ card.
Budget the sample rate before you wire anything
As on every multiplexed card, the headline rate is shared between all enabled channels. The number that matters for a vehicle bench is not the per-channel rate, though — it is the skew between two channels that are supposed to have been measured at the same instant.
A multiplexed front end converts one channel at a time. So if the converter runs at R conversions per second, two channels scanned one after the other are separated by 1/R seconds, regardless of how many channels are in the scan. Adding channels reduces how often each channel is revisited; it does not change the gap between neighbours in a single scan.
| Total conversion rate | Skew between adjacent channels | Phase error at 50 Hz | Phase error at 10 kHz |
|---|---|---|---|
| 400 kS/s | 2.5 µs | 0.045° | 9° |
| 250 kS/s | 4 µs | 0.072° | 14.4° |
| 100 kS/s | 10 µs | 0.18° | 36° |
| 1.25 MS/s | 0.8 µs | 0.014° | 2.9° |
Swipe the table sideways to see all columns
The table is the whole argument for choosing a tier. At line frequency, every row is irrelevant — 0.18° of phase error on a 50 Hz waveform is nothing, and a 250 kS/s card measures battery current and pack voltage perfectly well. At switching frequency the same card is 36° out, which turns a power calculation into fiction. The rate you need is set by the fastest signal whose phase relationship you actually care about, not by the fastest thing on the vehicle.
What the card can say about an event, and what it cannot
Take an injector driver. The interesting content is a decade of time scales at once: pulse widths that change from about 2 ms at idle to around 15 ms at full load, a diagnostic knee point where a peak-and-hold driver drops its current, and a turn-off edge whose ringing is measured in sub-microseconds. No single acquisition setting covers all three, and that is the actual engineering problem — not the card's sample rate.
- Pulse width, dwell time, repetition rate and duty cycle — measurable at tens of kS/s, and best measured on a counter or with a hardware-timed edge-detection task rather than by finding edges in a sampled record.
- Slow context — battery voltage, supply droop during crank, coolant temperature, average current — where bandwidth is irrelevant and channel count is what you are short of.
- Event statistics over a long record — how often a misfire occurs in twenty minutes, and what the supply was doing at the time. This is where a DAQ beats a scope outright, because the scope is not going to sit there for twenty minutes.
- Edge shape and ringing frequency. This is the scope's job. If a channel needs it, take a scope to that channel and let the DAQ do everything else.
Wheel speed, encoders and the double-count trap
Pulse trains belong on the counter. Every 62xx module here carries two 32-bit counters clocked at 80 MHz, which is enough for a 40-tooth tone wheel on a wheel spinning far faster than any vehicle will manage. Counting in hardware means the count is exact — it does not miss a tooth between samples and it does not alias.
The classic error with a variable-reluctance (passive) sensor is counting extra zero crossings at low speed. A VR sensor's output amplitude falls with speed, and the signal conditioner's own threshold then produces spurious crossings around the true ones. The symptom is a speed reading that jumps to roughly double the truth at low rpm and settles down as speed rises. The check is to log amplitude alongside frequency: on a healthy VR setup the two track each other proportionally.
Knock, vibration and the IEC 61000 immunity question
A piezoelectric knock sensor or an accelerometer is an IEPE device: it needs a constant-current excitation of a couple of milliamps to power its internal amplifier, and its output sits on a DC bias of several volts that has to be stripped before conversion. Only the USB-4431 and USB-4432 provide that excitation on the card, at 2.1 mA per channel, with AC coupling to remove the bias. Everything else in the range needs an external IEPE conditioner between the sensor and the input.
As for immunity: a bench sits next to an ignition system, which is an EFT generator with a spark plug on the end. The practical defences are the boring ones — keep the sensor cabling away from the coil harness, use shielded twisted pair with the shield grounded at one end, and prefer an isolated front end so that the common-mode energy has somewhere to go that is not your signal path.
Choosing between the four plausible parts
| The job | Part | Configuration | Why this one |
|---|---|---|---|
| Mixed slow and medium-speed signals on a running vehicle | USB-6216 | 16 SE / 8 DI, 400 kS/s, 32 digital lines, 60 V channel-to-ground isolation | Isolated, and enough digital lines to drive the indicators and relays on the rig |
| Fast transients or high channel count at speed | USB-6259 | 32 SE / 16 DI at 1.25 MS/s, four 16-bit outputs at 2.86 MS/s, 48 digital lines | The fastest tier, with enough outputs to drive actuators while measuring |
| Voltage and current together, for power and efficiency | USB-6229 | 32 SE / 16 DI at 250 kS/s, four simultaneous 16-bit outputs at 833 kS/s | Four analog outputs is what makes a programmable load or a driver emulator possible |
| Accelerometers, microphones, knock sensors | USB-4431 | 4 differential 24-bit IEPE inputs at 102.4 kS/s, plus one 24-bit output for shaker or exciter drive | Built-in 2.1 mA excitation and automatic anti-alias filtering; no external conditioner on the rack |
Swipe the table sideways to see all columns
Mistakes that cost a day each
- Grounding the DAQ to the chassis at the same point the sensor return uses, creating a loop that carries starter current.
- Assuming a card with isolated inputs is safe to connect across a traction battery. It is not a CAT-rated barrier.
- Reading a pulse train by sampling it, then reporting a speed that is quantised by the sample rate rather than by the encoder.
- Expecting channel-to-channel phase from a multiplexed card, then explaining the result as sensor error.
- Driving a solenoid or relay coil from a digital line without a flyback path, and replacing the module a week later.
- Trusting a 1.25 MS/s record to show the shape of an injection event. It shows the envelope, not the edge.
Every figure in this note is either arithmetic from published resolution and sample-rate specifications, or a value stated by the standard it cites. Where a number depends on a particular module, confirm it against that module datasheet before design freeze.