Liquid bath versus dry-block calibrator

Both devices do the same job: hold a volume at a stable, uniform temperature so a reference thermometer and a unit under test can be compared in it. They differ in how well they manage the two things that matter, uniformity across the working volume and stability over time, and that difference propagates straight into the budget.

The same calibration, in both

Four components, identical except for the source's gradient and stability term and the immersion term that the source's geometry drives. Reference certificate at U = 0.030 °C with k = 2, repeatability of 0.010 °C from ten readings.

The headline is the 3.6-fold difference in expanded uncertainty, but the more useful observation is the change in which term dominates. In the bath, the source has become small enough that the reference thermometer's own certificate is the limiting factor, which is where you want to be: buying a better reference now buys you something. In the dry block, the reference is almost irrelevant. Fitting a 0.01 °C reference to a block with a 0.1 °C gradient is spending money on the wrong term.

Where the dry block loses it

• Radial gradient across the block: wells at different radii sit at different temperatures, so the reference and the unit under test are not at the same temperature even when both are fully inserted.
• Axial gradient up the well: the temperature falls towards the mouth, so immersion depth becomes a term rather than a detail.
• Well fit: an air gap between sensor and wall is an insulator. A 6 mm sensor in an 8 mm well is a different measurement from the same sensor in a 6.5 mm well.
• Stem conduction: heat travelling up the sensor sheath and out of the block. Worse for short sensors, thick sheaths and large differences from ambient.
• Loading: inserting a second, cold sensor disturbs the block and it needs time to recover. A bath barely notices.

Every one of those is manageable and none of them is free. EURAMET cg-13 exists to characterise them, and a block that has been characterised against its own wells and insertion depths is a far better instrument than the same block used on its specification sheet.

Where the bath loses

The bath is not simply better. It is slower to change temperature, heavier, not portable, needs the right fluid for the range and needs that fluid maintained. Silicone oil above about 200 °C smells, degrades and eventually has to be replaced; water below 0 °C is not an option at all. A bath in a van is a bad idea, and a bath for a single point at 250 °C on a customer's site is a worse one.

Choosing

• Target uncertainty below about 0.05 °C: stirred bath, with the reference as your limiting term.
• Target between 0.05 and 0.3 °C: a characterised dry block is usually adequate, and much faster per point.
• On-site work, many points, wide range, portability required: dry block, budgeted honestly.
• Short sensors, thick sheaths or awkward geometry: bath, because immersion is the term that will otherwise dominate.
• Either way, calibrate in the condition of use. A sensor calibrated in a stirred bath and used in still air carries a self-heating difference nobody corrected.

The honest failure mode here is a budget copied from a bath calibration and used for a block. The reference, resolution and repeatability lines look identical; the source and immersion lines are the whole difference, and they are the ones that get carried over unchanged.