True gravity of a fermenting or finished sample from two refractometer
readings.
Alcohol refracts far more strongly than sugar, so a refractometer reading
taken after fermentation has started is meaningless unless the correction
also knows the original reading. Sean Terrill's linear regression
(seanterrill.com, "Refractometer FG Results") against measured hydrometer
data:
OB and FB are wort-correction-factor-adjusted Brix, which is how
Brewer's Friend and Brewfather apply the model. Use
refractometerFinalGravityRaw to pass uncorrected instrument readings.
Validation
A model has to give physically sensible results across the brewing range,
and this one does. Given OB = FB (a sample that hasn't fermented at all)
it returns within 0.002 SG of the true gravity of that Brix from 8 to
20 °Bx, and over realistic fermentations it implies apparent attenuations
of 69–75 %. We tried a second, widely circulated "Terrill cubic" and
rejected it. As transcribed it returns SG below 1.000 and implied
attenuations above 100 % for strong beers, so its coefficients can't be
right.
True gravity of a fermenting or finished sample from two refractometer readings.
Alcohol refracts far more strongly than sugar, so a refractometer reading taken after fermentation has started is meaningless unless the correction also knows the original reading. Sean Terrill's linear regression (seanterrill.com, "Refractometer FG Results") against measured hydrometer data:
OBandFBare wort-correction-factor-adjusted Brix, which is how Brewer's Friend and Brewfather apply the model. Use refractometerFinalGravityRaw to pass uncorrected instrument readings.Validation
A model has to give physically sensible results across the brewing range, and this one does. Given
OB = FB(a sample that hasn't fermented at all) it returns within 0.002 SG of the true gravity of that Brix from 8 to 20 °Bx, and over realistic fermentations it implies apparent attenuations of 69–75 %. We tried a second, widely circulated "Terrill cubic" and rejected it. As transcribed it returns SG below 1.000 and implied attenuations above 100 % for strong beers, so its coefficients can't be right.