Ultra Precision Reference LTZ1000

[iMo]

FYI - a simulation of my specific circuit - 4 wires force/sense.
Example wires 20cm long, 0.3mm dia copper to the two OUT posts.
With a typical Rload 10k - 10G wired to the posts (x axes).
The V_OUT voltage spread at those loads is 72uV at aprox 10V.
Not usable below, say, an 1Meg Rload (-1uV)..
 

[grx]

Hmm so I have to think about that...

Are you able to do the same simulation with a LTC2057 opamp as the source? Whats the maximum current then? How does it depend on the output series resistor?

Based on what I understand of four wire connections, the sense wire is a feedback line with no current that allows to servo the output of the force wire so the voltage is correct at the end of the wire, where sense and force meet together.

I can do that super easily with the positive side by using the opamp feedback path as sense wire.

However, for the negative side, there is no similar circuit that can do a remote sense and adjust the ground wire potential so the ground voltage is zero at the end of the wire.

If I really want to achieve that, I see no other solution that a secondary buffer to drive the negative side, and that would require a negative supply.

However, I dont want to do that on this board. I want this board to use an unipolar supply, in order to keep the ability to drive it with a battery in a simple way.

The question that remains is: is it worth doing on the positive side and not on the negative side? That would be a three wire connection that only cancels the voltage drop on the positive wire. Given that this board is only meant to be connected to a proper 7v-10v conversion stage with a much more complex output buffer, I think I can keep it simple. I do not intend my board as a generic standalone calibrator with no other hardware around, it's only the core part of a more complex device I want to build next.

I updated my schematic to reflect these changes.

[dietert1]

Sometimes one can use a silicon diode in the negative supply connection to derive a 0.7 V negative supply. That may be sufficient to implement the negative sense buffer.

Regards, Dieter

[iMo]

Hmm so I have to think about that...

Are you able to do the same simulation with a LTC2057 opamp as the source? Whats the maximum current then? How does it depend on the output series resistor?

Here it is with your simple buffer..
It is working a little bit better because you are not loading the sense with the divider I need in 7->10V buffer.
So your sense+ draws something like XXpA.
At Rload 100k the drop is 3.4uV..

Max current is the max current of the 2057 and the simulation shows 43mA (1ohm load).

[grx]

Oh, thanks a lot for this, thats great news. Three wire it is then.

BTW, what are the 4.7 ohms resistors for? (R8,R9 in your sim)

[iMo]

The small 4R7 resistors there (those are real and you should have them there) are for closing the opamp's loop when the 4 wires are floating (the load is not connected via 4 wires), thus the opamp/buffer will not behave wildly. Their value does not matter much as they are bypassed by the 4 wires when the load is connected with those 4 wires. Without proper 4 wire connection you will get slightly different voltage between F+ and F-, of course..

[grx]

I understand. These did not seem important to me before, as of course the opamp loop will be open, but we're not expected to have it working with the loop open.

I see how that's still a valuable precaution, thanks.

[iMo]

This looks a bit better, at least in the simulation..
A better opamp in the sense-/force- would be required with say -2V negative Vee created by the 3 diodes (example only).
Like OPA388 ??

[Kleinstein]

One can largely avoid the extra noise and drift from a low side buffer. In stead of classic drive an sense, one can connect the sense part directly the the reference ground as sense terminal and mirror the current from the high side drive to use it to drive the same but opposite sign current for the low side drive. It does not provide a full compensation of the terminal / wire resistance but can reduce the effect by a factor like 1000 depending on the accuracy of current mirroring.
One would still need negative supply though.

[grx]

@iMo: So the whole reference voltage generation is now referenced to this floating virtual ground.

What happens if this virtual ground potential moves around under thermal drift of the diode forward voltages? does it affect the output?

[iMo]

One can largely avoid the extra noise and drift from a low side buffer. In stead of classic drive an sense, one can connect the sense part directly the the reference ground as sense terminal and mirror the current from the high side drive to use it to drive the same but opposite sign current for the low side drive. It does not provide a full compensation of the terminal / wire resistance but can reduce the effect by a factor like 1000 depending on the accuracy of current mirroring.
One would still need negative supply though.

Could you show a simple schematics?

[iMo]

@iMo: So the whole reference voltage generation is now referenced to this floating virtual ground.

What happens if this virtual ground potential moves around under thermal drift of the diode forward voltages? does it affect the output?

Yep, there is a virtual ground for the reference stuff and for the Vref posts.
The drift of the diodes may mess up a bit when the opamps would be too sensitive to the Vcc/Vee changes.
So it depends. I recommended above to remove the serial protection diode, now I inserted there 3..
You may use a PSU with 2 voltages to mitigate the drift of the diodes, of course..

[grx]

Last opamp datasheet I read specified that asymmetric supplies were supported, but with a penalty of 500 uV of input offset per volt of asymmetry.

So yeah I think we really need a symmetric supply, thats why I keep thinking that the output buffer is a complete project of its own.

[iMo]

.. and even a better version below..

[grx]

It feels like it's just the symmetric of the positive buffer but reversed for the ground line. Do you think it would be useful to add the same current limiter as the positive side, but in the ground line?

[iMo]

You may put there the limiter, sure.

But my above V3 has still an issue - it does not eliminate the dependency on the 4 wires lengths.
So - with constant L (like 20cm from the board to the posts) it somehow works.

But it will not work when you will have 4 posts on the box and then 30cm or 100cm or 150cm long 4 kelvin wires to the load.

For that the sense+ input has to be highZ (so an opamp buffer there may help), and the bottom of the divider has to be connected to sense- somehow (not loading it and tracking it)..
Not an easy stuff, indeed..
 

PS: below a quick idea with opamps only for simplicity - V4..
Wires in 3 lengths  - 20cm, 50cm, 100cm - all fits within 0.65ppm..

[iMo]

Interestingly with 100k "shorting" R8/R9 (because of highZ sense+/- ) and with the kelvin wires lengths 0.2m to 5 meters it now fits inside 0.38ppm (RLoad 1k..10Meg)..
 

Now, what opamps to use there??

[iMo]

..and the complete sim V5..
Works nice in the sim..
All fits into 0.16ppm.
 

[grx]

very cool!

--moved reply to new thread--

[grx]

I'm not dead, I just managed to get components for the board: film caps, LT1013, transils, components to simulate a ltz1000 instead of burning one, and also a nice tin box of this model: https://fr.farnell.com/perancea/cfl2t/boitier-rfi-emi-cadre-ferme/dp/1181387

PCB could not be ordered yet, hopefully soon.

[Andreas]

For that price you get 3 TEKO boxes:
https://www.reichelt.de/de/de/shop/produkt/stahlblechgehaeuse_83_x_68_x_28_mm_silber-34031

with best regards

Andreas

[grx]

Yes but when life gives you farnell you make farnellonade...

PCB have been finally ordered!

[grx]

PCB pakidge have arrived!

Let me present you a cursed thing: the ltz999, to avoid destroying expensive components while I'm testing my design! 6v8 zener, BC847 transistors, 300 ohms resistor.

And assembled in my v1 pcb with no output buffer and crap components! Which might also be, the worse ltz-ish reference in the world!

Results so far: I get voltage regulation! Output is 7.4 volts, which is a nice 6.8 zener + 0.6 Vbe.
I have not checked """stability""" yet.

But I get no temperature control of any kind: The base gets about 0.56 volts from vref, but the collector of the temperature control transistor is stuck at zero volts. Obviously a problem with my extremely experimental design that I will fix soon. the 12.5k/1k divider ratio might not be a good match for the BC847.

Anyways, as bad as it could be, it is invaluable to tinker with the design without risking damage to anything!

I also have received the v2 pcbs with all the bell and whistles suggested by you all including the output buffer. I'll work on that a bit later.

[GigaJoe]

lt1013  -  Drift: 2μV/°C Max - ok ... your wish ...

in overall noise budget for such output somewhere around 3 microvolts p-p , like realistically look at ..

[grx]

thats only what I have on hand right now... this board will get a TL082 as soon as I'm back home.
This board is not made for any kind of accuracy.

Anyways, did anyone test the LT1013 made by Texas Instruments and compared it to the one made by Analog Devices? They should be identical, and yet...