Target: shunt regulator, 300V 0.1% from 20uA to 500uA.

Used LM285, J175, and trimmed divider for 25V 0.1% regulator.
Used 1% resistors, added one calculated fixed trimming resistor to
get 25.000V for 0.1% result at 20uA.
Made twelve of these, in series to achieve 300V for presumed 0.1%
precision.

Well, the results are OK, but not quite "right".
Firstly, it is high by 0.3%.
Then there is a seeming discrepancy: a Tektronix 7CT1N curve tracer
on six of the 25V trimmed regulators reads 150.24V (a bit above the
0.1%), but minimum current is 40uA (??).
Direct measurement of the twelve is 300.9V from 30uA to 470uA which
is high by 0.3%.

I do not have an adjustable supply to try 20uA "zener" current so
must assume the current displayed on the 7CT1N is in error, but that is
a factor of two and not a mere 1% or so.

Any idea as to why the error is so high (0.3%) and why the apparent
minimum operating current is way above the individually measured 20uA
used for setting each 25V regulator?

Robert Baer wrote:
Target: shunt regulator, 300V 0.1% from 20uA to 500uA.

Used LM285, J175, and trimmed divider for 25V 0.1% regulator.
Used 1% resistors, added one calculated fixed trimming resistor to get
25.000V for 0.1% result at 20uA.
Made twelve of these, in series to achieve 300V for presumed 0.1%
precision.

Well, the results are OK, but not quite "right".
Firstly, it is high by 0.3%.
Then there is a seeming discrepancy: a Tektronix 7CT1N curve tracer on
six of the 25V trimmed regulators reads 150.24V (a bit above the 0.1%),
but minimum current is 40uA (??).
* Basically resolves with more careful measurement using no added metering.

Direct measurement of the twelve is 300.9V from 30uA to 470uA which is
high by 0.3%.

I do not have an adjustable supply to try 20uA "zener" current so must
assume the current displayed on the 7CT1N is in error, but that is a
factor of two and not a mere 1% or so.
* Roughly resolved, by separate, added voltage measurements with a
forgotten 200K/V analog meter, at currents near the "knee".

Any idea as to why the error is so high (0.3%)?