[volt-nuts] Low noise reference

[Randy Evans]

I have a question for the group.  I was looking at an article for building
an ultra-low noise voltage reference by Walt Jung, published in Electronic
Design June 24, 1993 and a URL to the article is below.  I want to filter
the output of an LTZ1000 based 10V reference I am building and this circuit
has a very low freq corner of 1.6 Hz.  I was concerned about the leakage
through R1-C1.  If C1 had as little as 1ua leakage, it would drop the
voltage through R1 by 1 mV.  The spec on 100 uF electrolytic and tantalum
capacitors show a leakage of 20 ua  at rated voltage so this could be of
great concern.  However, at the low few tenths of a volt that should be
across C1, the capacitor should have a much lower leakage amount, which is
the theme of the article.

To get a better appreciation of the issue, I connected a precision 0 to 10
V source (100uV resolution steps) to a series combination of a 1 Kohm
resistor and a 100 uF electrolytic and, later, another 47uF tantalum and a
47 uF electrolytic capacitor.  In all cases the leakage, as measured with a
Keithley 414 picoammmeter, showed a leakage or around 0.08 uA at 10V and
varying 0.04 to 0.12 uA, around .1uA at 1V and varying , and around 1 pA at
0.1V, but with widely varying leakage current of 0.5 to 1.5 pA, with
occasional peaks of -0.5 to 2 pA.  This would equate to about +/- 2 uV
voltage variation across R1, making a 10 V 0.1ppm stable voltage reference
of questionable value.

I also tried a 0.68 uF polystyrene capacitor and also saw leakage current
variations, although much less than the electrolytic and tantalum
capacitors, as one would expect.

Thinking the problem might be the the picoammeter, I put a 100 megohm 0.1%
precision resistor in place of the capacitor across the precision voltage
source set for 0.1 V and measured the current through the resistor at a
very stable 0.9 pA on the Keithley 414 (sb 1pA but accurate enough for my
measurements - the resistor shielded box likely has some sub pA leakage
also).  Note that I used shielded cables for all measurements, and the
resistor and capacitor were in a shielded box, as well as the 100 Mohm
calibration resistor.  Touching the cables or boxes did not change the
picoammeter reading at all, indicating to me that the shielding was
reasonable.

I suppose the best approach is to build it and characterize it, but it's
not fruitful if someone has already done this. So my question is: has
anyone built this circuit and characterized it, particularly over
temperature for stability at the sub ppm level?

Thanks,

Randall Evans







http://waltjung.org/PDFs/Build_Ultra_Low_Noise_Voltage_Reference.pdf