Dynamometer Data Acquisition System: Channels and Timing

01 / TEST DATAWhat a dynamometer data acquisition system records

Separate the measurement, the alignment and the record

A dynamometer DAQ performs three jobs that are easy to confuse. It converts sensor signals into engineering values, it aligns those values in time, and it retains them as evidence that another engineer can re-examine. A system that only displays live numbers has done the first job and part of the second.

The alignment job is the one most often under-designed. Torque arrives as an analogue bridge signal, speed as a pulse train, temperatures as slow analogue channels and engine data as bus messages with their own transmission rhythm. If the record does not say when each value was valid, a fast event and a slow one can appear to agree when they never occurred together.

Understand the layers before comparing catalogues

Dynamometer control and data acquisition are often supplied together, as with the Hance HD-3000 control and data acquisition system, and that integration is convenient for the operator. It does not remove the need to confirm, channel by channel, what is measured, how often and against which clock.

Treat the channel list as a procurement decision

The channel list should be written from the test tasks the facility must answer, not copied from a previous cell. An engine acceptance test, a durability cycle and a combustion study need different channel sets and very different timing. Writing the list first also exposes the instruments that must supply each channel and the calibration evidence each one needs.

02 / CORE CHANNELSCore channels: torque, speed and the power that follows

Torque is measured; power is calculated

Torque and speed are measured quantities. Power is arithmetic performed on them, so an error in either input propagates into every power figure in the report:

That relationship is worth stating in the report method, because a reader who receives only a power curve cannot tell which torque and speed inputs produced it. If speed is briefly misread, the power trace will show it as a power event.

Torque channels differ in response and in what they sense

Torque sensors are not interchangeable. A reaction sensor measures the torque the absorber's housing reacts against, which is convenient on a trunnion-mounted dynamometer but filters fast events through the machine's mass. In-line sensors measure the shaft directly and differ further by signal transfer method. HBK's rotating torque sensor guidance puts rotary-transformer response at around 300 Hz while digital telemetry can reach the kilohertz range, and notes that reaction sensors are less accurate than in-line types for the same reason.

Whichever arrangement is installed, record the sensor identification and its calibration status with the test. A torque channel is only as traceable as the certificate behind it, which is why torque calibration is treated as its own subject rather than a line item in a test plan.

Speed channels fail differently at low speed

Speed usually arrives as a pulse train from a toothed wheel or an encoder. The acquisition system counts pulses over an interval, so the reported resolution depends on that interval and on the pulse count per revolution. At high speed this is rarely an issue; at idle or during a slow sweep, a short update interval can produce a visibly stepped speed trace and a correspondingly stepped power trace.

Record the pulse arrangement and the update interval with the test. Also confirm how zero speed is represented, because a system that reports a nonzero floor can make a stationary engine look like it is turning, and a sweep test starting from rest will contain a false segment.

Keep the derived values labelled as derived

Brake mean effective pressure, specific fuel consumption and corrected power are all calculated. Their inputs, correction basis and constants belong in the method section, not only in the spreadsheet that produced them. When a report is re-examined months later, a derived value without its inputs cannot be defended. The engine dyno test report guide covers how those derived values are usually presented.

03 / AUXILIARY CHANNELSEngine and fluid channels around the dynamometer

Every supporting channel has its own timescale

Temperatures move slowly; pressures during a transient move quickly; fuel consumption is a rate accumulated over a window. Assigning one global sample rate to all of them either oversamples the slow channels or undersamples the fast ones. A per-channel decision, recorded in the configuration, is easier to defend than a single number in a specification table.

Hance's HD-3000 product information lists an accessory set that includes smoke measurement, emissions analysis, engine pressure and temperature instrumentation units, a heavy-duty diesel ECM channel and fuel measurement integration. That list describes what can be connected. It does not state the channel count, resolution or bandwidth of any of them, and those figures must be confirmed for the actual configuration before they are used in a test method.

Fuel measurement is a rate, not an instantaneous value

Fuel consumption instruments fall into two practical families. Gravimetric systems weigh fuel removed from a supply over a window and suit steady-state work well; Coriolis meters measure mass flow continuously and suit transients. Sierra CP's fuel system range describes a combined arrangement in which a two-way valve holds the gravimetric tank level constant while a Coriolis meter measures flow, so the two techniques cover normal and low flow rates and can verify each other.

The consequence for acquisition is that a fuel channel's meaningful update interval is set by the instrument and by the length of the consumption window, not by the DAQ's fastest rate. Reporting a fuel number computed from a window that overlapped a speed change will mislead, even if every individual sample was valid.

Pressure and temperature channels need a stated conditioning

Pressure instrumentation often carries a fast transient and therefore needs a stated filter, or the trace will contain electrical noise that a reader interprets as a physical event. Temperature channels are usually slow enough that filtering is harmless, but the thermocouple or RTD type, its range and its calibration must be recorded, because two nominally identical temperature channels on different instruments are not automatically comparable.

Smoke and emissions channels need their own timing treatment

Smoke and exhaust gas instruments are separate analysers with their own internal response behaviour. Their outputs can be recorded by the same DAQ, but the analyser's response time belongs to the channel's meaning. An emissions value aligned to the wrong second of a transient is not a slightly late value; it can be attributed to the wrong operating point entirely.

04 / BUS AND TIMINGECM data, CAN and the synchronization problem

Bus data arrives on the ECU's schedule, not yours

Modern engines publish data on a CAN bus, and heavy-duty engines commonly use SAE J1939. Vector's J1939 overview describes the standardised baud rates of 250 kbit/s and 500 kbit/s, the 29-bit identifier scheme, parameter groups with defined cycle times and a transport protocol needed for messages longer than eight bytes, with a 10–200 ms inter-packet interval in the broadcast case.

The practical point is that each parameter group is broadcast on its own cycle. A torque value from the dynamometer and a fuelling value from the ECM are therefore sampled at different moments by design, and any comparison between them inherits that offset unless the acquisition system resolves it.

Put both signal families on one clock

There are two broad approaches. The stronger one gives the acquisition hardware a common time base and stamps every sample in hardware; the weaker one lets each source keep its own clock and aligns the streams afterwards in software. HBK's PTP explainer describes IEEE 1588 PTPv2 as hardware-based stamping with sub-microsecond accuracy between participants, and contrasts it with NTP and IRIG-B approaches and with the separate question of real-time control latency.

Dewesoft's engine test bench case study shows the same principle in practice: analogue channels and CAN-FD data are acquired in one piece of hardware with hardware-stamped analogue data, while FlexRay is soft-synchronised using the arrival time of its messages. The distinction is honest, and it is the distinction a test report should carry too.

Say which channels are hardware-synchronised

If some channels are hardware-stamped and others are aligned in software, the report should say so. A reader who assumes everything is on one clock will over-trust a correlation between a fast combustion event and a bus parameter that arrived on a slower cycle. Recording the synchronization method per channel is a small documentation burden that protects the conclusion.

Cross-check the alignment with a physical event

A useful sanity check is an event that all channels can see: a speed step, a load change or a deliberate input. If torque, speed and the ECM's reported speed respond in the expected order and with plausible intervals, the alignment is working. If the ECM value appears to lead the measured speed, the timestamps are worth investigating before more tests are added to the record.

05 / ACQUISITION PRACTICESampling rates, alarms and reporting

Apply the sampling theorem per channel

Kistler's sampling rate explanation states the Nyquist–Shannon requirement plainly: the sampling rate must be at least twice the highest frequency of interest, and if it is not, higher frequencies appear as lower ones and the digitised signal is wrong. That is aliasing, and it produces a plausible-looking trace rather than an obvious failure.

So the sampling decision is not a global preference. It follows the fastest phenomenon each channel must resolve. A steady-state fuel measurement and a combustion pressure channel can live in the same system and legitimately need very different rates; HBK notes analysis-oriented data rates of 100 kS/s per channel and above in test applications, and Dewesoft's combustion channels run to 1 MS/s while its NVH channels run to 200 kS/s. Those are other suppliers' configurations, quoted to show the range, not Hance specifications.

Decide what the record is for before choosing a rate

A high rate is not free. It consumes storage, it can make overview screens unreadable and it can bury the steady-state trend the test was actually about. A practical pattern is to record fast channels at the rate the phenomenon needs and slow channels at the rate the decision needs, then compute the report values from the stored raw data rather than from a downsampled display curve.

Step and sweep tests illustrate the same point. Dewesoft's case study notes that in a step test the inertia of engine and dynamometer does not affect the torque reading, because there is no acceleration, whereas a sweep test's acceleration rate does affect torque and must be applied consistently. The acquisition side must therefore preserve the acceleration context, not only the torque samples.

Separate alarm logic from protective function

An acquisition system can raise an alarm on a channel condition, and that is useful for the operator. It is not automatically a safety function. The distinction matters when the alarm is computed from a filtered value, from a channel with its own response delay, or on a system that shares a PC with other software. The protective function, its response time and its independence belong to the approved safety design, not to the logging configuration.

Make repeatability visible in the record

Two runs can agree closely and both be wrong, so repeatability is not proof of accuracy. What a record can do is make the conditions comparable: the installed configuration, the instrument identifications, the sampling decisions, the synchronization method and the events that occurred. That is the same discipline the engine dyno testing procedure applies to run setup, and the same logic that test cell component planning applies to the instruments themselves.

Where the test cell's water and cooling arrangements affect the measurement, the water supply requirements for a water brake dynamometer and the maintenance checklist define the conditions the data depends on. A torque trace from a cell with an unstable cooling supply is not comparable with one from a stable cell, even at the same set point.

Share your test tasks, channel list and instrumentation with Hance to define a control and data acquisition configuration that matches the way your cell actually runs.
Discuss your data acquisition configuration