What is the storage capacity of dataloggers?

Turn a memory specification into a recording time

A datalogger's storage capacity depends on the model and its recording setup. A specification may quote individual measurements, complete records, or bytes of memory. The useful question is how long your chosen configuration can run before you must collect the data. Start with what a datalogger does and how it builds a record, then compare devices using the same channel count and logging interval.

What actual specifications look like

The Lascar EL-USB-1 lists storage for 16,382 temperature readings. The Onset MX1101 lists 128 KB, with a maximum of 84,650 measurements. Documentation for the Campbell Scientific CR1000X identifies 4 MB SRAM plus 72 MB flash for certain automatically allocated data tables, with microSD expansion up to 16 GB. These are examples of different designs, rather than minimum and maximum limits for all dataloggers.

Check what the manufacturer counts before comparing those figures. Recording Pt100 temperature, relative humidity and pressure together may consume storage differently from recording temperature alone. Your sensor configuration and the enabled output fields matter as much as the headline memory figure.

Does “100,000 records” mean values or rows?

One timestamped row can contain readings from four channels. Alternatively, each channel value may count as one measurement against a shared limit. Ask whether the rating refers to total values, complete rows, or capacity per channel. Choosing a thermocouple or comparing thermocouples with Pt100 sensors addresses the input; the logger's storage format determines how that input uses memory.

For a hypothetical device with room for 100,000 individual channel values shared across all channels, use: Recording time = value capacity ÷ active channels × logging interval. This assumes all channels log at the same interval and no additional fields consume that quota. The result uses the same time unit as the interval. Do not divide a complete-row capacity or a per-channel capacity by the channel count in the same way.

How long a shared allowance of 100,000 values lasts
Active channels Logging interval Values per day Approximate duration
1 1 minute 1,440 69.4 days
4 1 minute 5,760 17.4 days
4 10 seconds 34,560 2.9 days

This is a worked example, not a runtime specification for the products above. The first-record timing and integer record count can slightly change the timestamp span. Define the duration needed for your application and industry before programming the logging schedule.

When the specification gives bytes

Estimate complete rows = usable data memory ÷ bytes per row. Suppose four values each occupy 4 bytes and a shared timestamp occupies 8 bytes. That hypothetical format uses 24 bytes per row. If all of 1 MiB, or 1,048,576 bytes, were available for those rows, it would hold 43,690 complete rows: roughly 30.3 days at one row per minute.

Real formats may add headers, status information and event records, or use different encodings. 1 MB and 1 MiB are different quantities, and an exported CSV file need not occupy the same space as the internal archive. An average calculated later during data analysis does not use additional logger storage; separately logged minimum, maximum and average fields can. Verify the storage rating against the software's estimated fill time for your final configuration.

Sampling rate is not necessarily the logging rate

A logger might measure once per second but save only one average each minute. Memory usage follows what is retained; the archive will not contain every sampled value. For a brief excursion in temperature control or pressure control, a slow average may hide the event you need to investigate. Likewise, the refresh rate of a live monitoring display does not establish the archive's logging interval.

Some models support event-triggered logging or faster logging above a threshold. Adding variables from a PLC over Modbus or Modbus RTU can also increase the stored field count. Include these variables in the calculation even though they are not separate physical sensors attached to the logger.

What happens at the storage limit?

With stop-when-full operation, the archive keeps its existing records and accepts no new ones. With circular or wraparound storage, new records replace the oldest. Available modes and defaults vary by model. A rolling archive can keep accepting data, but its historical window remains finite. Check that transfers into your SCADA archive succeed before old records disappear. Downloading data does not necessarily clear the logger's memory.

Separate local storage from cards and cloud archives

SD or microSD expansion requires explicit support. Check the permitted card type, capacity, file system and write performance. A USB port alone does not establish that the logger can record to a USB flash drive.

Remote access and wireless operation provide ways to move data. Ethernet, Wi-Fi and Bluetooth connectivity do not, by themselves, increase local capacity. A cloud archive has its own retention terms, while the logger still needs enough storage for an outage. In an edge computing design, size the local buffer for the longest expected interruption. For IIoT deployments, verify whether the specific equipment forwards missed records when the connection returns.

Specify the recording job before buying

Write down the capacity definition, enabled fields, logging interval, full-memory behavior and collection schedule. Include memory cards, transfer equipment and archive services in the total system cost. Add storage checks to the maintenance plan, choose suitable environmental durability and maintain measurement accuracy. More memory cannot repair an incorrect measurement. Manage archive access and deletion permissions as part of industrial cybersecurity.

Technical sources — example capacities checked on 8 October 2026: Lascar EL-USB-1 · Onset MX1101 · Campbell Scientific CR1000X