Measurement

Sampling rate and HPPC: what 10 Hz hides

The difference between 10 Hz, 100 Hz and 1000 Hz is not file size. It is which part of a pulse, and which dQ/dV feature, survives the record.

· 5 min read · All notes

Sampling rate is usually discussed as a data-volume setting. In HPPC and dQ/dV work it is a measurement setting first. The rate decides which part of a pulse you can see, how much a derivative amplifies noise, and whether a short pulse appears in the record at all. A 10 Hz trace and a 1000 Hz trace of the same HPPC pulse can support different conclusions about the same cell.

What the rate means in samples

At 10 Hz the instrument records one sample every 100 ms. At 100 Hz it records every 10 ms, and at 1000 Hz every 1 ms. The cell-testing platform publishes selectable 10 / 100 / 1000 Hz sampling on the relevant model classes, with current response down to ≤100 µs on 9-series channels. The channel datasheets give the concrete values for individual models: the BTS-6008A-5V300A-H records at 100 Hz (10 Hz when an auxiliary channel is connected), and the BTS-4008Q-5V100mA-124 records at 10 Hz.

The sample interval is a lower bound on what a pulse record can show. A feature shorter than one interval may be missed entirely, and a feature only two or three intervals long is represented by two or three points. Nyquist puts the usable bandwidth at half the sampling rate, but that is a statement about reconstructing a continuous signal, not about resolving a step edge. For pulse work, the practical rule is to keep at least ten samples across the feature you intend to fit or report.

RateIntervalSamples in a 10 s pulseSamples in a 50 ms pulseWhat it resolves
10 Hz100 ms1000–1Endpoint values, slow relaxation, cycle trends
100 Hz10 ms1,0005Pulse shape, 10 ms ohmic point, early relaxation
1000 Hz1 ms10,00050Fast transients, overshoot, dQ/dV detail

A 50 ms pulse is the published minimum pulse width on the BTS-6008A-5V300A-H; at 10 Hz it can fall between samples.

HPPC: the endpoint is forgiving, the transient is not

An HPPC pulse is often defined as a 10 s discharge or charge pulse followed by a rest period, repeated at a series of states of charge. The resistance at the end of the 10 s pulse is a slow quantity: at 10 Hz there are about 100 samples, and averaging a short window around the endpoint gives a stable value. For that specific number, 10 Hz is adequate.

The same 10 Hz record cannot separate the ohmic drop from the polarisation that follows it. The first usable sample is 100 ms after the step, by which time the current has settled and the fast interfacial response has already been absorbed into the voltage. If the test needs an ohmic resistance, a 10 ms point, or a relaxation time constant, the record has to come from a 100 Hz or 1000 Hz channel. The model matters: the BTS-6008A at 100 Hz gives a 10 ms grid; the BTS-4008Q at 10 Hz does not.

There is a second timing issue. The published current response and the recording rate are independent. A channel with a ≤100 µs response can still be recorded at 10 Hz. The response tells you how fast the current arrives; the rate tells you how often the result is stored. Do not infer one from the other.

dQ/dV: differentiation amplifies the sampling decision

dQ/dV is a derivative, so it amplifies noise in both capacity and voltage. At 10 Hz, a slow charge curve may advance only a few microvolts between samples, and the voltage quantisation of the channel becomes a visible staircase in the derivative. At 1000 Hz the same curve has a hundred times more points, which reduces the interval between samples but also increases the number of correlated, noisy differences. More samples are not automatically a cleaner derivative; the differentiation window and the smoothing have to be chosen for the rate.

A practical approach is to record fast and analyse slow. Keep the 1000 Hz record for the pulse and the fast transients, then resample or average to a uniform grid before computing dQ/dV. Averaging adjacent samples before differentiating lowers the noise without changing the underlying capacity. Report the window used. Two dQ/dV curves computed with different windows are not directly comparable, even if they came from the same file.

Capacity integration is less sensitive to the rate than the derivative, provided the current and voltage samples are time-stamped and the integration uses the actual interval. A dropped sample or a variable interval shows up as a step in capacity. Check the time column before trusting the capacity column.

Data volume is real but manageable

Eight channels at 1000 Hz for one hour produce 28.8 million samples. The same eight channels at 10 Hz produce 288,000. On a multi-channel rack running for weeks, that difference decides whether the database stays responsive and whether an export is practical. The BTS 9.1 workflow allows the rate to be selected per test, so a long cycle-life run does not have to carry the sampling rate needed for a short pulse characterisation.

Use the high rate where the question is fast — pulse shape, overshoot, ohmic drop, dQ/dV near a phase transition — and the low rate where the question is slow — capacity fade, coulombic efficiency, end-of-cycle trends. Record the rate in the method and in the file name. A trace without its rate is ambiguous.

Takeaways

  • Sampling rate sets what a pulse record can resolve: 10 Hz gives one point per 100 ms, so a 50 ms pulse can disappear between samples.
  • For HPPC endpoint resistance, 10 Hz is usually enough; for ohmic drop, early relaxation or dQ/dV detail, use 100 Hz or 1000 Hz and state the rate.
  • A fast current response does not imply a fast recording rate — check both specifications, and choose the rate per test rather than for the whole campaign.

Request a configuration review

Tell us the channel count, current range and protocol. We will check the measurement class against the note and return a configuration, not a generic datasheet.

Request a configuration review

References

← All technical notes

Talk to an engineer

Request a quote