Is the Hazen-Williams resistance constant 4.727 exact as implemented, or a rounded published value?

[hawstom]

I maintain a small suite of free web-based hydraulic calculators (https://hawsedc.com/engcalcs/
GPL, twelve calculators, twenty-seven languages), including a looped-network page that solves by the
global gradient method. I am currently aligning its Hazen-Williams head loss with EPANET's so that
the two agree, and I would rather get the constant right than approximately right.

My pages have been using the SI form:

Sf = 7.8828 / d^4.8704 * (Q / (0.849 * C))^1.852

EPANET's documented US form is:

hL = 4.727 * L * Q^1.852 / (C^1.852 * d^4.871)

Converted to SI these differ by up to about 0.1% over a 50 mm to 2 m diameter range. The diameter
exponents also differ (4.8704 against 4.871), so the discrepancy is diameter-dependent rather than a
constant offset — it changes sign near 300 mm.

Two questions for anyone who knows the source well:

1. Is 4.727 the exact value used in the code, or a rounded form of something derived at higher
   precision?
2. Since EPANET converts SI input to US units internally, is there any accumulated conversion
   difference I should expect between a network entered in SI and the same network entered in US
   units?

I am planning to adopt EPANET's constants outright rather than carry both sets, on the grounds that
EPANET is what my users will check me against. Happy to be told that is the wrong call.

Thomas Gail Haws
Mesa, Arizona