Why Analogue Does Not Mean Obsolete

Why this lesson matters

There is a persistent assumption that better data requires better instruments. It produces capital proposals to replace serviceable equipment purely to obtain a digital output, and monitoring programmes that stall because the replacement budget never arrives. This lesson separates two questions that get collapsed into one.

Two independent questions

Question A: is the measurement fit for purpose? Accuracy class, measuring range, turndown, installation conditions, calibration status, physical condition.

Question B: is the data path fit for purpose? Reading frequency, acceptable latency, who needs the value, what evidence the value must carry, what it feeds into.

A gauge can pass A and fail B. That is the normal case, and it does not call for replacement. It calls for a capture layer.

Where analogue instruments genuinely outperform

Reliability and life. A Bourdon tube gauge has no power supply, no firmware, no network stack and no configuration. Service life of a decade or more in ordinary conditions is unremarkable. A transmitter adds a power dependency, an addressable configuration, a firmware revision and at least one more thing that can go offline without anyone noticing that it has.

Failure visibility. When a mechanical gauge fails it usually fails obviously: a pegged needle, a broken pointer, a fogged glass. When a digital sensor fails it can keep transmitting a plausible last value, or drift quietly. Silent plausible failure is considerably more dangerous than visible failure.

Immunity. No EMC susceptibility, no supply transients, no cybersecurity surface, nothing to patch.

Hazardous area simplicity. In a classified area, a mechanical gauge requires no Ex certification, no intrinsically safe barrier, no certified gland and no additional inspection regime.

Metrological standing. A pattern approved, sealed revenue meter is the legally recognised measurement. A secondary sensor beside it is an estimate of that measurement, and for billing or regulatory reporting the seal usually wins.

Independence. During a BMS outage, a commissioning exercise or a network fault, the local gauge still shows the truth. Local indication is a safety and operability feature, not a legacy inconvenience.

Where analogue instruments genuinely fall short

Be equally honest about the limits:

• Sampling frequency. A weekly manual round cannot see a twenty minute pressure excursion or a night time leak signature. If the phenomenon you care about is faster than your round, manual capture will not find it.
• Latency. Manual reading cannot trigger a response in minutes. Anything requiring rapid intervention, such as a tank overfill, a cold room excursion or a burst main, needs a continuous point.
• Coverage cost at scale. A hundred instruments read weekly is a substantial recurring labour commitment. Automation economics improve steeply with point count and required frequency.
• Drift without calibration. An uncalibrated mechanical meter can under register progressively and give no external sign of it. Analogue is not self validating.

The decision framework

For each instrument, work through these in order:

1. Is the measurement itself sound? Check accuracy class against what you need, that the normal operating point sits in the middle third of the scale (a gauge operating at 10 percent of full scale is a poor measurement regardless of class), the calibration or verification status, and physical condition. If this fails, replace it. Not for data reasons, for measurement reasons.

2. What frequency does the decision actually need? Match frequency to the time constant of the thing you are monitoring. Building water consumption trends usefully at daily resolution. Boiler pressure during a fault investigation needs minutes. Compressor hours need monthly.

3. What latency does the consequence require? If undetected failure for a week is tolerable, manual capture is adequate. If it is not, you need a continuous point and an alarm path.

4. What are the constraints? Hazardous area classification, metrological sealing, network reach, power availability, lease or landlord restrictions, and who owns the asset.

5. Only then choose the data path. Structured manual capture, non invasive retrofit, or full automation.

Retrofit options that keep the instrument

Between "read it by hand" and "replace it" sits a useful middle ground:

• Pulse output heads on water and gas meters, where the meter body supports a retrofit register. Gives you volume per pulse without disturbing the measuring element.
• Optical and inductive register readers that observe the existing mechanical index without breaking a seal.
• Camera based register capture for indices that cannot be touched at all.
• Clamp on ultrasonic flow meters for liquid, non invasive and removable, though installation straight lengths and pipe condition matter.
• Non invasive current transformers for electrical loads.
• Structured photo capture, which is the zero infrastructure option and works on literally any instrument with a visible face.

Each of these preserves the existing measurement and only changes the data path, which is exactly what the two question framework tells you to do.

Calibration still applies

Digitising the record does nothing for measurement accuracy. Keep the verification regime running:

• Record calibration or verification dates against the asset, and treat an overdue instrument as a data quality flag on every reading it produces.
• Where practical, cross check against an independent source. Utility invoice volume against site meter totals is the easiest and most frequently skipped check available.
• Re baseline after any meter exchange, and record closing and opening indices explicitly so the consumption series does not break.

Common mistakes at this stage

• Replacing instruments to obtain data, before knowing the required frequency. This is the most expensive mistake in the whole subject.
• Assuming a new sensor is more accurate than an old gauge. Compare accuracy class and installation conditions, not age.
• Breaking a revenue meter seal to fit a pulse head. Check who owns the meter and what is permitted first.
• Fitting continuous monitoring with no alarm path. A point that nobody watches has the same operational value as a gauge nobody reads, at considerably higher cost.
• Dropping the calibration programme once readings look tidy on a screen. Structured wrong data is still wrong data.