Spehro Pefhany a écrit :
100M isn't that high, well, depending on what kind of accuracy you're
looking for.
How much instability are you seeing? How much above 100M?
LMC6042 has 2fA (typical) input bias current and costs a couple
dollars one-off.
A 100M 1% resistor with 50ppm/K tempco runs around five dollars.
A 1G 1% resistor with similar tempco is maybe double that.
Most inexpensive DMMs on the lower ranges have very high input
impedance so the LMC buffer might not be necessary.
High value resistors are (depends on the model) somewhat unstable.
I currently have a batch of Dale 1G/1%.
They measure fine with 100V bias which is the datasheet measuring
conditions. When measuring them at low voltage, they're all over the
place, from +3ish% to +7ish%. Yep, not even grouped...
On the contrary I've some Caddok rated <0.02ppm/V!
A friend of mine worked at, IIRC, Vishay/sfernice on that specific high
value resistors 'feature' and went auditing some of the 'production'
lines (the quotes are his). He said the voltage dependency was mostly a
prod issue. High value Rs is a very small niche market and as such,
production inherited some highest tech tools, like hand lapping and the
likes. Being labor intensive it was relocated in low labor cost
countries, with poor buildings (he said, almost backyard
and less
than ideal handling cleanliness.
That could explain a lot some of the strange behaviors...