[volt-nuts] Ultra low noise reference

[Dave M]

I've been following the "Traveling Standards" thread (and its aberrations).
It took me a while to find it, but the postings reminded me of an article
that I had read in Electronics Design magazine some years ago. The article
by Walt Jung
(http://www.waltjung.org/PDFs/Build_Ultra_Low_Noise_Voltage_Reference.pdf)
describes a technique using a 10V reference IC and an AD797 opamp to build
an exceptionally low-noise voltage reference (in the neighborhood of
1.25nV/root Hz).

The circuit, as drawn, uses an AD587 10V reference IC, buffered by an AD797.
After looking the circuit over again (after having lost it for many years),
I'm thinking about getting some parts and building a few copies of it, using
a variety of voltage references, such as the LM399 and 1N829 (I have a few
left over from days long past). Almost any type of voltage reference could
be used in front of the opamp, and a variety of output voltages provided
with a bit of gain in the opamp. Possibilities are intriguing. In terms of
voltage stability and drift, the AD587 doesn't look particularly attractive,
considering today's available references that's why I'm thinking about other
types of references.

I'm not an engineer, but spent a number of years as a calibration tech, and
later, as an engineering tech. Precision voltage and frequency equipment
continue to hold my interest, even though I've been away from it for several
years. I'd be interested in hearing a discussion of the pros and cons of
the circuit described in the article.

One last thought; I have a couple of old proportional crystal ovens
(salvaged from old military revivers, about 75C internal temp, controlled to
approx. 0.1C) that I'd like to use to house the assemblies. What are some
thoughts on enclosing the entire voltage reference circuit inside an oven?
Would the heat inside the oven adversely affect the noise and voltage drift
of the opamp? Better to put only the voltage reference IC or diode inside
the oven?

Thanks for comments,
David
dgminala at mediacombb dot net

[ws]

David
Your posting brings up several points.

As far as the oven question, need to first ask Why use one?
The oven has many disadvantages with only one main advantage, that being
it's great if you're making something that has to have a very low TC over a
wide temperature range. If that is the case, then put everything that drifts
due to temperature changes inside of it, and keep its temperature as low as
you can.

As for the adding the low noise filter / buffer, Why?
It's only needed if you're driving something such as a ADC or DAC that has a
changing load and/or needs low High freq noise.

As far as the AD587. Right its not too good, but they can be selected and
tweaked to be as good or better than most anything you're going to be able
to build at home.

If you need resistors for gain or attenuation, they are likely going to be
some of the biggest problems when making a really high end reference.

ws

**************************
[volt-nuts] Ultra low noise reference
Dave M dgminala at mediacombb.net

I've been following the "Traveling Standards" thread (and its aberrations).
It took me a while to find it, but the postings reminded me of an article
that I had read in Electronics Design magazine some years ago. The article
by Walt Jung
(http://www.waltjung.org/PDFs/Build_Ultra_Low_Noise_Voltage_Reference.pdf)
describes a technique using a 10V reference IC and an AD797 opamp to build
an exceptionally low-noise voltage reference (in the neighborhood of
1.25nV/root Hz).

The circuit, as drawn, uses an AD587 10V reference IC, buffered by an AD797.
After looking the circuit over again (after having lost it for many years),
I'm thinking about getting some parts and building a few copies of it, using
a variety of voltage references, such as the LM399 and 1N829 (I have a few
left over from days long past). Almost any type of voltage reference could
be used in front of the opamp, and a variety of output voltages provided
with a bit of gain in the opamp. Possibilities are intriguing. In terms of
voltage stability and drift, the AD587 doesn't look particularly attractive,
considering today's available references that's why I'm thinking about other
types of references.

I'm not an engineer, but spent a number of years as a calibration tech, and
later, as an engineering tech. Precision voltage and frequency equipment
continue to hold my interest, even though I've been away from it for several
years. I'd be interested in hearing a discussion of the pros and cons of
the circuit described in the article.

One last thought; I have a couple of old proportional crystal ovens
(salvaged from old military revivers, about 75C internal temp, controlled to
approx. 0.1C) that I'd like to use to house the assemblies. What are some
thoughts on enclosing the entire voltage reference circuit inside an oven?
Would the heat inside the oven adversely affect the noise and voltage drift
of the opamp? Better to put only the voltage reference IC or diode inside
the oven?

Thanks for comments,
David
dgminala at mediacombb dot net

[Andreas Jahn]

Post by Dave M
an exceptionally low-noise voltage reference (in the neighborhood of
1.25nV/root Hz).

This is the broadband noise which is only of interest if using a high speed
A/D converter. For integrating A/D converters the 0.1-10 Hz noise
(or even lower frequencies as I have learned now) are of interest.
The around 2-3 Hz corner frequency of the setup does not give much
advantage on precision measurements where integration times start from
2 seconds or more.

Post by Dave M
One last thought; I have a couple of old proportional crystal ovens
(salvaged from old military revivers, about 75C internal temp, controlled to
approx. 0.1C) that I'd like to use to house the assemblies. What are some
thoughts on enclosing the entire voltage reference circuit inside an oven?
Would the heat inside the oven adversely affect the noise and voltage drift
of the opamp? Better to put only the voltage reference IC or diode inside
the oven?

I would put all within the oven minimizing the number of thermocouple
junctions (soldering points) with different temperatures.
But note: with higher temperature the ageing is accelerated
and the noise increases. So for some applications it might be better
to characterize the temperature drift and compensate it by measuring
the temperature.

(see diagram with a 3rd Order polynominal for approximating the
temperature error curve of a LT1027CCN8-5 reference voltage
the resulting error is in a 3uV band / 3V = around 1 ppm).

With best regards

Andreas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ADC8_TC.PNG
Type: image/png
Size: 42751 bytes
Desc: not available
URL: <Loading Image...>