Gauge vs absolute pressure: the 14.7 psi that breaks bubble-point matching

If your lab quotes a bubble point in psig and you type it into a model that works in psia, your bubble point is one atmosphere too low. That's 14.696 psi. At a 2,750 psia bubble point the error is 0.5 %, and you'd shrug it off. At a 360 psia bubble point it's 4.1 %, and it carries through into solution GOR, any regressed gas gravity, and the oil you forecast before gas comes out of solution.

#Gauge and absolute

Absolute pressure is measured from a vacuum. Gauge pressure is measured from the local atmosphere:

p_abs = p_gauge + p_atm

A standard atmosphere is 101,325 Pa exactly. One psi is 6,894.757 Pa (NIST SP 811, Appendix B.8). So:

p_atm = 101,325 / 6,894.757 = 14.696 psia

Every PVT correlation, every equation of state and every material-balance equation works in absolute pressure. Gas volume goes as 1/p, and that relation has no meaning on a gauge scale. A separator or wellhead gauge reads relative to the air around it.

NIST's style guide (SP 811, section 7.4) says information shouldn't be attached to the unit symbol at all. It belongs on the quantity: "gauge pressure, 2,435 psi", not "2,435 psig". The industry writes psig anyway, and lab reports mix the two, so read every suffix.

#The worked example

Physics regime: a black oil at reservoir temperature, at and below its bubble point. Units are oilfield throughout (psia, °F, scf/STB, rb/STB).

#Standing's bubble point

Standing (1947) correlated 105 measured bubble points on 22 California oil-gas mixtures (API Drilling and Production Practice, pp. 275–287):

pb = 18.2 × [ (Rs / γg)^0.83 × 10^a − 1.4 ]      psia
a  = 0.00091 × T − 0.0125 × API                   T in °F

The data range is pb 130–7,000 psia, T 100–258 °F, Rs 20–1,425 scf/STB, API 16.5–63.8 and gas gravity 0.59–0.95. Both cases below sit inside it.

Take the fluid from Converge's sample field: 35 °API, gas gravity 0.75, 200 °F. Then

a    = 0.00091 × 200 − 0.0125 × 35 = 0.182 − 0.4375 = −0.2555
10^a = 0.5553

High bubble point, GOR 650 scf/STB:

Rs/γg        = 650 / 0.75 = 866.67
866.67^0.83  = 274.42
× 0.5553     = 152.38
− 1.4        = 150.98
× 18.2       = 2,747.8 psia   (2,733.1 psig)

Low bubble point, GOR 60 scf/STB:

Rs/γg        = 60 / 0.75 = 80.00
80^0.83      = 37.98
× 0.5553     = 21.09
− 1.4        = 19.69
× 18.2       = 358.3 psia     (343.7 psig)

The 2,747.8 psia result is Converge's benchmark case V1 (2,748 psia by hand). It's on the benchmarks page.

#The 14.7 psi slip

Now say the lab reports the bubble point in psig and someone types it in as psia. The offset doesn't change, but its share of pb does:

The same 14.696 psi is eight times worse at the low bubble point. Shallow reservoirs, heavy and low-GOR oils, and depleted fields with a low current pressure all live at that end.

#Where the error lands

I computed every row below with Standing's equations and checked them by running Converge's black-oil engine on the same inputs. Pressures are psia.

#Rs and Bo

When you honour a measured bubble point, the saturated Rs curve is stretched so that Rs equals the GOR at that pressure. A pb that's too low pulls the whole curve to the left. At every pressure below pb, the model then holds more gas in solution than the oil really does. That's +5.1 % at 300 psia in the low case.

Bo hardly moves, about 0.1 %. If you only check Bo, you won't see the problem.

#Regression

Suppose you tune the correlation to the lab bubble point instead of honouring it directly, and you leave gas gravity free. The fit pushes gas gravity from 0.75 to 0.786. The regression reports a perfect match, so nothing looks wrong. But gas gravity feeds the pseudo-critical properties, the Z-factor and the gas density. You've traded a pressure-label error for a fluid-description error.

#Material balance

Above the bubble point, an undersaturated oil produces by expansion alone. The Havlena-Odeh form reduces to F = N·Boi·ce·Δp, so the oil recovered before pb scales with pi − pb. A pb 14.7 psi too low gives you 14.7 psi of extra single-phase depletion. That's +6.1 % on the recovery to the bubble point when pi is 600 psia, and +3.2 % when pi is 3,200 psia.

The relative error depends on pi − pb, not on pb. A high-pressure reservoir sitting just above its bubble point is as exposed as a shallow one.

The same slip in a pressure history hurts more below pb. Gas FVF, Bg = 0.02827·z·T/p (ft³/scf, T in °R, p in psia), goes as 1/p. A reading of 285.3 psig typed as psia gives a Bg 5.2 % too high at constant z. That error goes straight into the (Rp − Rs)·Bg term of the underground withdrawal.

#Inflow

Vogel's (1968) composite IPR is the least sensitive result here: +1.5 % on AOF in the low case and +0.3 % in the high case.

#What Converge does about it

Engines, saved projects and the API hold pressure in psia only. The interface converts to bara at the display boundary, and pressure labels always carry the basis. Here is what's built:

• Measured pb input. The PVT page's "Measured pb" field is labelled absolute. Its hint reads: "Add atmospheric pressure to a gauge reading (psig + 14.7)." The engine rejects a measured bubble point at or below 14.7 psia.
• Range and mismatch warnings. The black-oil model warns when inputs, including the bubble point, fall outside the selected correlation's data range. It also warns when a measured pb differs from the correlation's own value by more than 20 %.
• Imports. Production and material-balance history headers such as "p (psig)" or "p (barg)" are read and converted with the atmospheric offset. The Data import tab has an "Atmospheric pressure" setting for gauge readings. A bare "psi" header is read as gauge on a wellhead pressure and as absolute on a downhole pressure, and the import preview says which it chose so you can change it.
• AI Engineer. Every AI Engineer tool that takes a pressure needs the unit stated: psia, psig, bara, barg, kPa or MPa. Values are converted to psia before any validation or comparison, and the reply lists each conversion. This came out of an audit case where a lab bubble point of 2,435 psig was compared with a model value of 2,435.7 psia and reported as a +0.03 % match. The lab value is 2,449.7 psia, so the real error was −0.57 %.

#A short checklist

1. Read the suffix on every lab pressure: bubble point, CCE, differential liberation and separator stages. Separator pressures are very often gauge.
2. Convert at the boundary, once. Never convert halfway through a calculation.
3. Be most careful when pb is below about 1,000 psia, or when pi − pb is a few hundred psi.

Read the black-oil PVT guide for the inputs and correlation ranges, and units for how Converge handles field and metric display.