Ultra Precision Reference LTZ1000

[Dr. Frank]

Hm, do I glimpse a bit of inspiration from my box?
How did you wire the ferrite toroid - common mode choke? source material?

BR
Hendi

Hello Hendi ,
yes of course 
it's wound common mode, like a CAN bus choke, and it's the cheapest lacquered TDK ferrite (B64290L38X830) I got from voelkner/Conrad, about 1.8µH, or so.

best wishes to Aschebersch!

Frank

[Dr. Frank]

All my references use a 12k/1k divider.

Hello Frank,

You simply had luck that the voltage of the LTZ (7.1V) is more at the lower end of the tolerance.
With 7.3V zener voltage the oven would fall out of regulation already at 30 deg C.

with best regards

Andreas

Hello Andreas,
I figure out what you're referring to, i.e. higher reference voltage gives lower oven temperature, 200mV more give about 8°C less.. theoretically.. as I can't discover this systematicity in the data I've summarized. The U_BE of the temperature sensing transistor will also vary, and therefore, it's not out-and-out that 7.3V reference voltage really gives about 42°C.

And it's no luck at all, because, I determine the final oven temperature for each of my references. 
If I'd find a too low oven set point, I could assemble a 12.5k/1k divider as well.. a simple, variable, fine tuning circuit for the oven temperature, as well for the T.C. would be great.

Have you done that measurement on your references as well, especially on the one with 7.3V?
Your 34401A can also do the job! 

Frank

[openloop]

You guys, it seems, have a lot of references sitting around...   

Have you tried to do a Battle Royale between them? 

Like take 3 of those, power them from separate batteries, tie 0 terminals together so they form a Benz-like star, let them warmup, of course.

Now voltages between tips should be less than 100mV (1V, if you're unlucky). In that range even my humble 34401As are precise enough (like 60nV stddev).
Take two DMMs (3 for sanity checks) and log the voltages between neighboring tips (with simultaneous triggering) for several hours.

Now you have 3 rows of data (third is the difference between first two, if only 2 DMMs used). 
Now, estimate the variances. Each of those 3 variances is a sum of 2 corresponding variances of your references.
Now you have 3 unknowns and 3 equations so you can solve for individual variances!
Maybe rotate the setup and redo to see if repeatable.

Thus you can find the quietest reference of all.

[dietert1]

Many of those self built references have issues with their outputs: Since they have OpAmp output pins directly or almost directly connected to the binding posts, eventually connected with some meters of cable, they are sensitive to RF and to capacitive loads. Often hobbyists found it difficult to wire more than one reference and more than one voltmeter in one circuit. With a more robust reference output circuit that would not be a problem at all and the proposed difference measurement would be the way to go.

When drawing the Keithley 213 4x DAC schematic i saw how to implement a stable output circuit that includes and drives a 100 uF buffer cap.

Regards, Dieter

[MegaVolt]

When drawing the Keithley 213 4x DAC schematic i saw how to implement a stable output circuit that includes and drives a 100 uF buffer cap.

Will you share your find with us?

[openloop]

Dieter,

Often hobbyists found it difficult to wire more than one reference and more than one voltmeter in one circuit.

Because references are independently battery powered and connected in only one point there are no ground loops or some such.
In my case, two DMMs are well grounded and were connected to the circuit with two 15cm long twisted pairs.

Worked fine.

[dietert1]

OK, here is the Keithley DAC output section i mentioned. A similar buffer with output filter including a large buffer cap can be used for a lab voltage reference with binding post interface.

Regards, Dieter

[openloop]

After thinking about it some more, I must admit, there is asymmetry in the connection of 2 DMMs. That's one of the reasons why I mentioned rotating the setup.
In my case though, outputs are driven by LTC2057s and they seem can handle capacitive load of 2 DMMs just fine.

Technically, variance measurement does not require simultaneous acquisition.  One can collect variance data for for each pair separately and then solve the equations.
But that makes the problem of how long to run the experiment even more acute. Because it's not clear at what time scale LTZ1000 can be considered ergodic enough.
You know, running into a "Black Swan" problem.
In my case, one of those happened when on one of the references buffer LTC2057 went bad: suddenly started drawing non-trivial amount of bias current...

[e61_phil]

You have to keep in mind, that the inputs of the DMM aren't perfectly the same. The HI input is really high-impedant, but the LO input is normally connected to the whole GND of the floating electronics (higher leakage and much more capacitance).

A couple of weeks ago, I made some experiments with my SR1010-1k and the 3458A. I applied 10V to 10 resistors of the SR1010-1k in series (4W connection sourced by Fluke 5440B). In the first experiment I connected the 3458A LO to the lowest resistor (also LO from the 5440B) and the 3458A HI to the middle of the chain. After that I reversed the connectors on the 3458A and I measured a difference of a couple of µV. Unfortunately, that made it impossible to make the linearity test I was intended to do.
I also played around with the Guard, but that didn't improve much.

What I just wanted to say: You really need proper low impedant output for such difference measurements.

[rigrunner]

I would like to summarize about the newly built two LTZ1000As, which are the first ones to me, after 7 non-A types. At first, these references will be powered from a common linear 12V supply, that are about 11.5V for the circuit.

Nice to see the boards in use 

[Andreas]

it's wound common mode, like a CAN bus choke, and it's the cheapest lacquered TDK ferrite (B64290L38X830) I got from voelkner/Conrad, about 1.8µH, or so.

Hello Frank,

so 1 winding on each side? (cannot see it exactly on the photo).

with best regards

Andreas

[doktor pyta]

https://www.changpuak.ch/electronics/Arduino-Shield-BHUMI.php

[Dr. Frank]

it's wound common mode, like a CAN bus choke, and it's the cheapest lacquered TDK ferrite (B64290L38X830) I got from voelkner/Conrad, about 1.8µH, or so.

Hello Frank,

so 1 winding on each side? (cannot see it exactly on the photo).

with best regards

Andreas

yes, quite randomly.

[Andreas]

https://www.changpuak.ch/electronics/Arduino-Shield-BHUMI.php

Its a pity.
All those costly parts and then the temperature setpoint with 11:1 (below room temperature).
And a switchmode supply.

with best regards

Andreas

[openloop]

All those costly parts and then the temperature setpoint with 11:1 (below room temperature).

I've crunched some numbers, for the fun of it.

And, according to two dudes named Ebers and Moll, to keep temperature the same as for 12:1 ratio, for 11:1 R3 needs to be 13K7 

I do not know what consequences of increasing collector current to .5mA will be. Will the whole thing stay in regulation? Will somewhat increased self-heating of the temperature sensing transistor screw things? Will the additional voltage drop on pin7 be noticeable?

On another hand, dividers can be made out of them cool LT5400s ... 

[uski]

Hi,

Did anyone hear about a group buy of LTZ1000A ?
It is out of stock anywhere but AD sells directly for $54.50 each for a pack of 20.

Any other way to get my hands on one ?
If someone wants to make a group buy I'd happily order one.

Or if multiple are interested maybe we could get a group buy going ?

Thanks

[Andreas]

Hello,

I finally have adjusted the T.C. of my LTZ#9 (LTZ1000 non A version)

first measurement on 03.11.2020 without R9: gives -128 ppb /K
from my rule of thumb (derrived from the LTZ1000A measurements) that each 1 Meg resistor compensates  +40 ppb/K
So I tried a 330K resistor.

measurement on 06.11.2020 shows clearly some over-compensation giving +31 ppb/K
So it seems on the LTZ1000 (non A) I have to calculate with 53 ppb/K for each 1 Meg resistor.

The first measurements with 402K  (e.g. 10.11.2020) where somewhat drifting so I had to wait until the measurements stabilized.
(perhaps I shouldnt have used a cheap 1% metal film resistor)

Now on 21.11.2020 the final result is 1 ppb/K in average. But of course due to the non linearity introduced by the temperature compensation there is some "Box T.C." of around 45 ppb/K over the temperature range of 12-38 deg C measured on top of the housing of the LTZ.

with best regards

Andreas

[Kleinstein]

The drift only looks bad because the scaling of the graph changes. In absolute numbers the curve without the resistor looks similar. One can also see that the resistor only has a really small effect on the voltage. So drift of the compensation resistor (even with cheap 1% resistors) should not be an issue.
The visible drift and hysteresis is more from the other resistors. Chances are this is still in the early hours for the parts and thus still in the initial drift after soldering / cleaning.

The compensation is nonlinear. So the right size of the compensation resistor would depend on the heater power / set temperature and likely also the thermal design.  With a higher set temperature and thus more heater power the heater voltage will change less with temperature and thus a smaller compensation resistor is expected. Similar more isolation could prefer a smaller resistor.
I would expect the right compensation resistor to be about inverse proportional to the heater voltage at a fixed temperature.

[openloop]

Andreas,

What is the time scale of your temperature sweeping graphs?
I mean, what is the total elapsed time?

[Andreas]

Hello,

Setpoint temperature ramp speed is 0.12 deg C / minute.
(Actual temperature at the top of the aluminium housing of the PCB follows more ore less).
I usually start from 25 deg C down to cold, ramp up to hot and do another cold phase until minimum reaching 25 deg again.
In this case I did another hot phase and switched off the cooler during last phase.

Attached a different graph over time from first and last measurement.
So total time takes 1500 - 1900 minutes (so 25 - 30 hours) in this case.

with best regards

Andreas

[branadic]

Hello Andreas,

we recently had a discussion about the temperature slope the other day. Even though I use 0.1K/min for the slope, which is the minimum that my Arroyo 5305 can do automatically, this is still to large to reach thermal equilibrium across the board and components. TiN uses a maximum slope of 0.02K/min and a minimum slope of about 0.0084K/min, so we are talking about a factor of 10 smaller value. This is achieved by a software controlled temperature profile and available here:

https://xdevs.com/guide/teckit

GitHub repository:
https://github.com/tin-/teckit

My explaination of the "open eye" (or hystersis if you want to call it like that) in your but also my measurements is, that thermal resistance and capacitance come into play if the slope is too fast. Thus, run up and run down are different. Solution, decrease the slope.

One thing that I don't like about the standard datasheet circuit of LTZ is, that tc-compensation and amplification to 10V are correlated and not independent from each other. This (let's call it) issue was solved in a commercial product of a voltage reference though.

By the way: I found two documents related to pre-aging of LTZ1000 provided by Cern:

1101699_V1_Burn_in_of_LTZ1000.pdf

1070495_V1_HPM_equipment_description_DS22bit.pdf

-branadic-

[Andreas]

Hello branadic,

you are right (at least for a normal lab).

But in my case in summer I have up to 2 deg C / hour temperature rising. (from 22 deg to 33 deg C in a few hours).
So testing at 7.2 deg C / hour is only testing something like the "worst case".

On the other side: my battery pack within the LTZs limits the test time to a maximum of ~2 days.
So setting the ramp speed to below 0.05 deg C / minute is not really a option.

with best regards

Andreas

[niner_007]

Hey folks, I finished the layout of my LTZ1000 reference module. It is built around an 3458A reference form factor. 6 layers, uses mostly through hole parts, the divider is VHD200, the other sensitive resistors are VHP202Z, I also use MKP capacitors, and plan on using the free space for a few LMT70 temperature sensors, a USB interface and microcontroller for tracing and calibration.

I appreciate any feedback on the layout, specially around grounding.

P.S. each cell in the picture is 1mm, so that makes the thickest trace 1mm, an the thinnest trace 0.4mm.

[Andreas]

Hello,

for me it looks like you have to populate the LTZ on the bottom side of the PCB.
Is this intended?

with best regards

Andreas

[niner_007]

it is not meant to be populated on the bottom, perhaps orientation is not visible well