[time-nuts] Voltage standards

[Mike Monett]

  Warren wrote:

  > If you  want to reduce the measurement noise of a system  you need
  > to do one or more of the following:

  > Lower the  source  impedance, by reducing  the  resistance  of the
  > thing you are measuring

  > Lower the Bandwidth, by filtering over a longer time period

  > Lower the temperature, by making it colder.

  > Doing even  any one of these things enough will in theory  let the
  > noise approach  zero, When you do two (or all three)  at  once the
  > noise will approach zero sooner.

  > see "Johnson-Nyquist noise" for the details

  >> From a  practical  standpoint, 1 nV of  resolution  is  doable by
  >> comparing the difference between two voltage sources if  one uses
  >> a lot of care and applies some form of extra filtering.

  > A 1nV (1e-9) is way below the noise level of any  voltage standard
  > that puts out volts.

  > This means  reference  measurements are not limited  by  the noise
  > level when using a good but simple setup until the references gets
  > to be in the 3e-10 precision range.

  > The answer  to  "What can an amateur do to get  a  good  low noise
  > reference for less than, say, $500" IS shop at the US eBay site.

  >WarrenS

  The Keithley  Low  Level Measurements Handbook,  Sixth  Edition, has
  some nice graphs that illustrate the limit due to noise. Figure 1-2,
  on page  1-4, shows the theoretical limits  of  voltage measurements
  due to  noise. The next figure, 1-3 on page 1-5,  shows  the typical
  limits of measurement with various source resistances:

  https://www.keithley.com/servlet/Data?id=9538

  The EM  Electronics D.C. Picovoltmeter, Model P12, has  a resolution
  better than 5 picovolts, dependent on response time. From  their web
  page:

  "The noise performance is such that, using a source impedance  of 30
  milliohms or  higher,  the detection level will  be  limited  by the
  Johnson noise of the source. The equivalent noise resistance  of the
  P12 is only 30 milliohms. To put this figure into  perspective, this
  is the  noise  generated by 30cm of 0.5mm diameter  copper  wire! To
  make full  use of the noise performance, it is desirable  to  use it
  with very  low source impedance. While the P12 can be  operated with
  source impedance  of  ten ohms or more, its  characteristic  will be
  overwhelmed by  the  source  noise if the  source  impedance  is too
  high."

  http://www.emelectronics.co.uk/spec/P12.html

  The EM DC Picovoltmeter Model P13 only goes down to 20 picovolts:

  http://www.emelectronics.co.uk/spec/P13.html

  For the  creme de la creme, the Keithley Model 1801  Nanovolt Preamp
  gives 1 picovolt resolution when used with the Model 2001 DMM.

  This extends  the  range by a factor of 1,000, so  the  20 millivolt
  range becomes 20 microvolts. The Model 2001 DMM is 7 1/2  digits, so
  the least significant digit becomes 1 picovolt (1e-12 volts.)  It is
  interesting to  note  the Model 1801  input  connection  uses copper
  nuts.

  From the brochure:

  "The 1801's  chopper-based  design  offers  several  advantages over
  traditional preamps  designed for high-frequency  amplification. For
  example, most DMMs and lock-in amplifiers are designed to filter out
  random noise by averaging."

  "However, low-level  measurements are also subject  to  (1/f) noise,
  which cannot be filtered out in this way. The chopper design cancels
  out (1/f)  noise  completely, which lock-in  preamps  and  most DMMs
  can't. By  continuously  inverting the DC input  and  amplifying the
  resulting AC  waveform,  the 1801's chopper  measures  the  DC input
  amplitude precisely, exclusive of voltage offsets and (1/f) noise in
  the amplifier circuitry. Carefully designed feedback circuitry makes
  the 1801  capable  of rise times as fast as 500us 10  -  90%  and DC
  input resistance  of  up  to  1GOhm.  A  one-eighth-inch-thick steel
  shield and an internal mu-metal shield prevent magnetic  coupling of
  noise to the amplifier."

  http://www.teknetelectronics.com/DataSheet/KEITHLEY/WEBKEITH1801.pdf

  The 1801 web page is at:

  http://www.keithley.com/products/lovoltloresist/?mn=1801

  For those  interested in making your own  nanovoltmeter,  the Analog
  Devices AD8571  offers  Rail-to-Rail Input and Output,  20  pA input
  bias current, 1uV input offset, and 5 nanovolt/C input offset drift:

 
http://www.analog.com/static/imported-files/data_sheets/AD8571_8572_8574.pdf

  Regards,

  Mike Monett