Current balancing

Suppose I have a voltage reference and some associated components connected between a supply and ground. Now imagine I have a resistor connected between ground and a negative supply to minimise the current actually flowing from the reference circuit through the rest of the ground. (I know that HP use or used to use this arrangement).

I was wondering what a better way might be to obtain the same or better result and not depend on precise resistor selection. A current mirror was the first thing to spring to mind, but then the output side has to be going negative.

Thoughts, anybody?

voltage references usually don't like a lot of current through them. however i'm really having trouble visualizing the circuit you are describing. if you are trying to minimize the current through the ground, then you should use a single point ground, such as a "star" ground.

Theoretically the reference could be supplied from a separate supply , and the ground current from it only determined by the loading of the circuit referencing it. That so a current source / sink could be used or a higher voltage hence larger resistor value. It could also have positive and negative supplies, but it would obviously have to have a ground contact of some form. My first question would be, has the reference a precision current source in it anyway?

Tracy

I didn't phrase first post very well, did I...

Ok, the idea comes from a reverse-engineer someone posted on volt-nuts forum (I think that's what it was) some years ago, of the reference circuit from an HP DMM - might have been 3458 but I'm going from memory - it was a really good one anyway.
The board had a precision reference, op-amp and supporting components. Lets assume a shunt reference for argument's sake. Might have been LTZ1000.

What I found really interesting was a resistor going from ground to negative supply, the only purpose of which must be to match the supply current so there is no (or very little) current flowing through this circuit's ground connection to the supply.

I imagine this is to eliminate errors due to the resistance of the ground conductor. I suppose they would have used an ADC with a reference ground and connected it to this point, but I'm wildly speculating here.

So it's creating a virtual ground which is connected to real ground anyway.

It's fascinated me ever since and I started wondering if I can make use of the idea but adapted to tolerate DIY-grade parts by using some kind of mixed-polarity current mirror in place of the resistor.

Like this:

So it basically simulates a kelvin connection? You could just actually use a kelvin connection (or star ground as previously suggested).

Are you sure this is the entirety of what you saw? For example...maybe there wasn't actually a Supply+ and a Supply- but just a single supply that had split rails to produce the GND in an inexpensive way that could not handle a lot of current? If that was the design objective, it seems like such a shoddy way to go about things.

This would have been from a 7 or 8 digit (or 6 at minimum) DMM from 20 years ago, so I doubt very much anything would have been shoddy, and I think it's very likely it did have a real V- supply.
I'll try to find the thread so I can show you properly

Ok I was mistaken it's not a forum thread, it's here, and it's not a reverse-engineer it's an actual manual schematic!

http://www.maxmcarter.com/vref/

The resistor of interest is R418

Well, not just R418 seems useless. The entire -15V supply is there for no other purpose than to power R418 so maybe that really is the case.

You could simulate in LTSpice and measure the currents with and without to verify if you're really curious

I found a photo of the back of the board. That resistor actually goes directly to pin 7 of the ref.