HP41: accuracy of 13-digit routines

[Dieter]

(09-02-2015 04:45 AM)Thomas Klemm Wrote:  Thus the internal 13-digit result isn't rounded correctly but cut off.
...
From this only the first 10 digits are used.

(09-02-2015 06:11 AM)Joe Horn Wrote:  Thomas is correct.
...
This "truncate, then round" procedure was not only used by the old 10-digit models, but is still used by the newer 12-digit HP calculators, which calculate internally to 15 *truncated* digits, then round that result to 12 digits.

Thank you very much. So the 13-digit routines truncate after 13 digits. I assume this also applies to the transcendental functions (log, exp, trig etc.). Now if I set a WP34s to RM 4 (truncate) and apply a RSD 13 to every result, this should yield the same result as the 41's 13-digit routines?

The reason I'm asking this is testing a Lanczos Gamma approximation with an optimized coefficient set, cf. the recent thread in the HP41 software library.

Dieter