Best solution to generate DC 1 milli volt 1 micro volt with precision ?

[Overspeed]

Hello

If solutions exist to generate DC from 1 Volt to 10V as precision component exist .
Does there is a solution to generate low voltage with precision ? at a level comparable to a Ref54 for example so in the ppm ..?

high ratio resistive divider don't seems to be the best solution to generate milli ou micro volt.

Regards
OS

[Conrad Hoffman]

What's wrong with a precision divider? When you drop a lot of voltage the lower resistor can be low, so the output impedance is low. Look at KVDs and the old General Radio 1346 Microvolter for inspiration.

[Gyro]

...
high ratio resistive divider don't seems to be the best solution to generate milli ou micro volt.

Why not? It is the best way to achieve a low source resistance, and any active component noise and offsets are attenuated. You could use a low level current source into a low value resistor, but it would have similar properties.


Edit: ^ What he said.

[Overspeed]

What's wrong with a precision divider? When you drop a lot of voltage the lower resistor can be low, so the output impedance is low. Look at KVDs and the old General Radio 1346 Microvolter for inspiration.

Hello
I don't know what wrong as I have asked a question on the MI 1340 and its seem that not the best

I link a GR 546 manual with schematic and BOM , doesn't look as rocket science which is a good point

Regards
OS

[BILLPOD]

Good Morning Overspeed, I've got the LB06 calibrator and It is GREAT for this type of application.  It costs a couple hundred bucks, but I have found lots of uses for it, and I'm sure you will too.  It would work BEAUTIFULLY for what you want to do.  Here is a link for
one on Amazon:  https://www.amazon.com/Multifunction-Process-Calibrator-Frequency-4-20mA/dp/B0CZHSXXHV?gQT=1&th=1

[Victorman222]

I would second the resistive divider suggestion. Check out this article about Keithley low thermal divider, it is made exactly for the purpose of getting microvolts with low noise. https://xdevs.com/fix/kei262/

If you are after great linearity and stability an inductive voltage divider based decade calibrator might work (for millivolts definitely). But such solutions complexity is infinitely higher than resistor divider.

[Gyro]

@Overspeed:

That's the switch with PCB wafer and sprung Copper 'rivet' contacts. I knew I'd seen it somewhere when you asked a few weeks back (the Marco Reps nV scanner video). An ideal candidate for DIY low thermal emf switching.

[David Hess]

high ratio resistive divider don't seems to be the best solution to generate milli ou micro volt.

Resistive attenuator sections can be calibrated individually, and then stacked to get high levels of precision attenuation.

If a zero output impedance is required, then the low level output from the attenuator can be buffered with a precision operational amplifier.

[tszaboo]

If you want a novel way of generating microvolts, you could intentionally heat a thermocouple. If you get an oven like used in the LTZ1000 for both ends, which gets to 1/1000K accuracy, then a thermocouple would be super accurate, and you can buffer it with an already heated (thus accurate) opamp.

[Overspeed]

Good Morning Overspeed, I've got the LB06 calibrator and It is GREAT for this type of application.  It costs a couple hundred bucks, but I have found lots of uses for it, and I'm sure you will too.  It would work BEAUTIFULLY for what you want to do.  Here is a link for
one on Amazon:  https://www.amazon.com/Multifunction-Process-Calibrator-Frequency-4-20mA/dp/B0CZHSXXHV?gQT=1&th=1

Hello
Thanks , I have a Fluke 724 and a Time Electronic 1006 but I search a solution to generate fixed value in uV range with lab precision ( able to be validated ) , problem look to be ( perhaps ) than most voltage reference have a limited current output level .

Regards
OS

[Gyro]

Most 10V IC references (thinking REF102 / AD857 style) are capable of 1-10mA -  obviously the higher the current, the greater the need for Kelvin sensing lead compensation.

If you take a 10V reference at 1mA max, a 10k (OK, 9.999k)/1R divider will give you a nice low source impedance 1mV. Decreasing the 1R to 1mR (or increasing the 10k to 10M) will give you 1uV.

[Kleinstein]

If needed one can support the reference with an extra current source to provide most of the known current to the divider. The actual reference would than only deliver the smaller rest of the current. The difficulty can be the thermal EMF at the lower resistor. Not all resistors a equal in this respect and the specs may not tell you much. For a build it is also about getting a small thermal gradient at the resistor, so not just a good resistor.

Chances are one would want some way to switch the voltage off for offset measurements.

[Overspeed]

If needed one can support the reference with an extra current source to provide most of the known current to the divider. The actual reference would than only deliver the smaller rest of the current. The difficulty can be the thermal EMF at the lower resistor. Not all resistors a equal in this respect and the specs may not tell you much. For a build it is also about getting a small thermal gradient at the resistor, so not just a good resistor.

Chances are one would want some way to switch the voltage off for offset measurements.

Hello
Does this circuit design is a solution ? that boost the current of a reference , this circuit is similar for some references .
Other concern is to draw too much current as I suppose that a Ref IC shall be used at a reasonable level of current even if the MAX current could be 10 mA

Regards
OS

[Kleinstein]

The more normal way to provide most of the current would be to amplify the voltage and than use a resistor to provide the approximate calculated current, or a little less. If one wants to include switching the divider drive side off for an offset / zero reading, one would likely more go to an amplifier to drive the current anyway. If really needed one could still use 2 mechnical switch poles (for the rough current and the fine correction).

At 1 mV for the output it is anyway not so much about the stability / accuracy of the reference or amplifier, but more the divider and terminal thermal EMF.

[David Hess]

The more normal way to provide most of the current would be to amplify the voltage and than use a resistor to provide the approximate calculated current, or a little less. If one wants to include switching the divider drive side off for an offset / zero reading, one would likely more go to an amplifier to drive the current anyway. If really needed one could still use 2 mechnical switch poles (for the rough current and the fine correction).

I have done it that way before when the reference had to travel across a board.  I used the reference to generate a precision current which travels across the board to a single resistor at the load.  In this way, a 4-wire Kelvin connection is not required to deliver a precision reference voltage.

This sort of circuit also works to transfer audio across a noisy ground without ground loop isolation like a transformer or instrumentation amplifier.

[coppercone2]

I think the question is about low voltage references. intrinsically low voltage references.


The zener effect is poor until a number of volts. Usually low voltage references are only used for sensing temperature (BE junction) etc. They are not stable.

A true low voltage reference would be a discovery. I assume this has something to do with current effects and the impedance of the source material.

While interesting, the power loss from using a low voltage reference combined with a high stability divider, makes the low voltage reference not very appealing to develop.

[Kleinstein]

The Josephson junctions are low voltage to start with. For just a single junction there are a few reports getting away with 77 K. This not getting the same very high accuracy as the 4 K version, but still pretty good.

The more standard way is a divider, likely a resistive divider and than a Zener or maybe even band gap reference as a start. Here it does not make a difference if one starts at 10 V, 7 V, 5 V or 1.2 V: the divider does not get much easier with a lower voltage to start with.
The power loss is not that bad. One can get away with some 0.1 to 1 mA for the divider, so usually less than the power at the reference.

For me the question is, if it is worth adding a way to turn off the drive side to also have 0 V from the same output. So not just a DC reference, but the extra switching to allow correction of the thermal EMF from the cables and connectors (not just at the ref. side but also the meter side).

[Overspeed]

Most 10V IC references (thinking REF102 / AD857 style) are capable of 1-10mA -  obviously the higher the current, the greater the need for Kelvin sensing lead compensation.

If you take a 10V reference at 1mA max, a 10k (OK, 9.999k)/1R divider will give you a nice low source impedance 1mV. Decreasing the 1R to 1mR (or increasing the 10k to 10M) will give you 1uV.

Hello
OK sound realistic , perhaps the 1 Ohm tolerance can be a problem as low value high precision tolerance are very hard to locate even from Vishay but a trial is possible without too much $$

Regards
OS

[coppercone2]

the JJ effect I ignored because while low voltage, it needs a microwave signal to work. hardly neat low power electronics , even at room temperature

Thought some kind of microwave oscillator coupled with a room temperature JJ plate? is a chip waiting to happen one day. Not much more complicated then AD voltage refernece chips

I don't think it super stable without a atomic clock though? What would happen if you use a crappy oscillator to power a joseph junnction?

[Gyro]

Most 10V IC references (thinking REF102 / AD857 style) are capable of 1-10mA -  obviously the higher the current, the greater the need for Kelvin sensing lead compensation.

If you take a 10V reference at 1mA max, a 10k (OK, 9.999k)/1R divider will give you a nice low source impedance 1mV. Decreasing the 1R to 1mR (or increasing the 10k to 10M) will give you 1uV.

Hello
OK sound realistic , perhaps the 1 Ohm tolerance can be a problem as low value high precision tolerance are very hard to locate even from Vishay but a trial is possible without too much $$

Regards
O

... but easy to trim with parallel resistors (assuming that you can't trim the reference at the divider input). Kleinstein is right, thermal EMFs are going to be your biggest enemy (whatever method you use. You would want to put the 10k resistor lsome distance from the 1R and run tightly twisted leads to the 1R to minimize thermal difference. Much less of a problem with the 10M resistor (almost vanishingly so) but the thermal emfs are a far bigger influence at 1uV. Of course it depends on required accuracy - at 1uV, you're only likely to be able to reliably resolve 100nV, so 10%.

[TimFox]

The classic Keithley 260 nanovolt voltage source is a precision voltage supply followed by decade-switched resistor divider networks.
The final divider (to get sub-microvolt DC voltage output) is made from copper to minimize thermal emf problems.
Although the resistance tempco for copper is crummy, the ratio stability is adequate for the application.
The voltage outputs are low-emf Leeds & Northrup binding posts, hidden behind a front-panel door to shield them from drafts.
An upside-down pdf file:  https://xdevs.com/doc/Keithley/260/KEI_260_Instruction_Manual.pdf

[Overspeed]

The classic Keithley 260 nanovolt voltage source is a precision voltage supply followed by decade-switched resistor divider networks.
The final divider (to get sub-microvolt DC voltage output) is made from copper to minimize thermal emf problems.
Although the resistance tempco for copper is crummy, the ratio stability is adequate for the application.
The voltage outputs are low-emf Leeds & Northrup binding posts, hidden behind a front-panel door to shield them from drafts.
An upside-down pdf file:  https://xdevs.com/doc/Keithley/260/KEI_260_Instruction_Manual.pdf

Hello
Interesting instrument some internal pictures in
https://www.eevblog.com/forum/metrology/keithley-260-teardown/
all copper built

Regards
OS

[David Hess]

A true low voltage reference would be a discovery. I assume this has something to do with current effects and the impedance of the source material.

A bandgap reference can be configured to deliver 0.2 volts instead of 1.2 volts.

[coppercone2]

A true low voltage reference would be a discovery. I assume this has something to do with current effects and the impedance of the source material.

A bandgap reference can be configured to deliver 0.2 volts instead of 1.2 volts.

I thought this is the configuration they use for a temperature sensor, because it has a high but linear tempco?

[David Hess]

A true low voltage reference would be a discovery. I assume this has something to do with current effects and the impedance of the source material.

A bandgap reference can be configured to deliver 0.2 volts instead of 1.2 volts.

I thought this is the configuration they use for a temperature sensor, because it has a high but linear tempco?

No, the normal bandgap configuration also produces an signal which is proportional to absolute temperature, but this signal is not normally brought out.  Some bandgap references do make this signal available.