Ultra Precision Reference LTZ1000

[branadic]

I now have more data of my LTZ1000(non-A)#1. Obviously the TC is 0.28255ppm/K or 2.02µV/K in the observed change in ambient temperature. I have a 402k compensation resistor installed. What is your suggestion, increasing or decreasing the resistor value and by what amount? Is there some rule of thumbs?

-branadic-

[Andreas]

Hello,

for the non-A I do not really have a receipe.
But I think the principal direction should be the same as here for the A-Version:

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg999857/#msg999857

So to decrease the T.C. by 80 ppm you will have to leave out the 400 K resistor.
Probably on the non-A version the effect from heater is larger than on the A-Version.
A estimate could be sqrt(400 K/W / 80 K/W) so a factor 2.3 more than on the A-Version.

But: 0.8 deg C room temperature change seems to be too little for me to really differ between LTZ T.C. and instrument T.C.
I would do at least a 20 deg C temperature change for the LTZ at constant conditions of the instrument.

With best regards

Andreas

[branadic]

Thanks Andreas,

so I need to go into the climate chamber with my LM399 and my LTZ1000 as soon as it is available. Don't have a working temperature box at home

-branadic-

[Andreas]

Conrad has:

https://www.conrad.de/de/kuehlbox-thermoelektrisch-g30-dc-12-v-rot-grau-29-l-eeknrel-mobicool-1423217.html

With best regards

Andreas

[Andreas]

Sorry was a 12V only version:

So at home this one is better if you do not want to outsource the beer from the fridge.

https://www.conrad.de/de/kuehlbox-thermoelektrisch-v26-12230-12-v-230-v-rot-25-l-eeka-mobicool-1423216.html

with best regards

Andreas

[branadic]

What temperature slope do you suggest? Up to 5K/min are possible with the chamber we have at work.

-branadic-

[Andreas]

Hello branadic,

for practical reasons I usually use 0.12K/minute.
-> one measurement cycle over 30 deg C with rest time at minimum and maximum needs and some overlap around 13-15 hrs. = 1 day

I have already used 0.03 or 0.05K/minute for final measurements especially when the thermal mass is larger.
Then the measurement usually is splitted into a warm and a cold phase on different days.

Maximum rate I used is around 0.3K/minute but only for smaller temperature changes (+/- 5 to +/-7 K).

If you have a too large slope then you will have additional "measurement dependant hysteresis" because the temperature sensor has not the same temperature as your zener/resistors.

Consider also: maximum slope in my "lab" is around 2K per hour (or 0.03K per minute).
So if you have the time you should choose that what you expect in later use.
For the same reason I put the reference into a shoe box so that I do not have large air drafts (similar to later use).

with best regards

Andreas

[Kleinstein]

5 K / min is already relatively fast for testing a precision reference. A realistic speed would be more like less than 1 K / min.

The effect of the compensation resistor is approximately linear in 1/R.  As the correction is coming from the heater, it depends on the thermal setup. Due to the resistive heater the change in heater voltage also depends on the heater power and  thus the set temperature.

For the A version the thermal coupling is thermal resistance is larger, but the higher set temperature and lower heater power tend to have an opposite effect. Due to the large number of influences it is hard to predict how strong the effect of the compensating resistor is. The best bet is more like a measurement of the TC with and without the 400K. If needed one can then calculate the suitable resistor.

[kj7e]

Regarding the TC resistor and the LTZ1000A, all three of mine showed about 0.1 ppm / K negative TC without the TC resistor.  The 400K value is just a ballpark starting point it seems.  I found one needed 330K, one 390K and one 470K to achieve similar end results.  All three show the same behavior with the TC resistor in place, the TC slides downward slightly from 10deg C to about the 24deg C point, then it will level off or show a slight positive TC around about 28deg C and onward. The TC dropped from the 0.1 ppm / K to less than 0.05 ppm / K.  I have been testing from 10deg C to 50deg C with a very crude manual slope control (but it works).  I have seen too aggressive temp slope gives false results as not all of the parts on the reference board have had time to thermally stabilize.  If the temp is increased from 25 to 35 deg C, in say 10 min (1 K / min), my references will show a more positive TC while warming, then settle down in value after the temp has stabilized.  I have seen much better/stable results with a less aggressive slope, around 10 K / hour or more.

I use a small portable cooler as like in the link above but the air movement inside is too disruptive.  I found I can avoid this by using some frozen cold packs, place everything inside and let it cool to about 8 deg C, then remove the ice packs and start taking measurements as it slowly warms from 10deg C to near room temp (takes about 2 hours).  Then I use some large wirewound resistors in the cooler and my lab supply to gently warm from there, starting off at about 5 watts which gets me to about 30 deg C, then increase the wattage by 5 watts every 5 deg C.  Its crude, but it works.

[TheSteve]

All kinds of boring and unoriginal - I just boxed an HP 3458A A9 ref board.

[branadic]

Oh wow, the 7480T is the first multimeter with the input jacks for high and low arranged in a horizontal way instead of vertical that I see.

-branadic-

[TheSteve]

Oh wow, the 7480T is the first multimeter with the input jacks for high and low arranged in a horizontal way instead of vertical that I see.

-branadic-

Yep, just to force you to really focus when making connections so you get them right. They are also not in a .75 x .75 arrangement so I can't use a shorting block that would fit many of the Keithley and Keysight meters.

[2N3055]

What is it, 20x20mm ?

[Andreas]

Oh wow, the 7480T is the first multimeter with the input jacks for high and low arranged in a horizontal way instead of vertical that I see.

-branadic-

Yep, just to force you to really focus when making connections so you get them right.

but at least they are more likely at the same temperature.
I ever wondered how one could make a precision instrument with connectors at different temperature levels.

with best regards

Andreas

[TheSteve]

What is it, 20x20mm ?

It fits a standard dual banana just fine for measuring voltage or 2 wire resistance. It looks to be 19.05x25mm.

but at least they are more likely at the same temperature.
I ever wondered how one could make a precision instrument with connectors at different temperature levels.

with best regards

Andreas

I had never considered that before - maybe that was what they had in mind.

[Insatman]

I recently built a battery backed voltage standard.  Four references are included.  The primary reference is a LTZ1000 scaled to 10V.   There are three secondary references.  First sec ref is based on three LM199s which are averaged and then scaled to 10V.  2nd sec ref is an LT1021BCN8-10.  This was the best Single chip 10V reference type I have tested so far.   Third sec reference is the infamous VRE305AD.  This 5V reference has stellar specification which unfortunately it does not quite meet.  But it is still a pretty good reference, just quite pricey for the performance level.

The chassis uses two 12V, 7ah batteries in series to power the references.   A LM319 pre-regulator drops the voltage from ~24 to 19VDC.  It is then filtered to remove battery charger noise.  The 19 volts are then fed to the two reference modules where a LT3080ET regulator drops the voltage to 15V.   The primary regulator module contains the LTZ1000 reference on a custom circuit board.  Originally I planned on using the KX reference board.  But...after two failed attempts I designed my own board.   

The secondary module contains the LM199x3, LT1021 and VRE305 references, all sharing the same 15V rail.   Both modules are enclosed in a cast Aluminum enclosure within the steel enclosure of the voltage reference chassis.   

Schematics, photos and data are attached.  I will need more than one post for all the attachments...sorry.

[Insatman]

More PDFs for my prior post on DIY LTZ1000 Reference

Note the data also includes an modified MV-106 and a DIY LM399 reference

[Insatman]

Final PDF for DIY LTZ1000 Vref

[borghese]

I think that OPA177 is not specified for input voltages of about 0.5 Volts; it is better to use the LT1013 or LT1006
Best regards

[branadic]

My triple set is almost ready. Board is the LTZ1048B including BMON (battery monitor and battery charger) and battery pack (12x AA NiMH) by Andreas. It currently delivers only the buffered zener voltage, but VSMP 10k is already installed and VSMP 25k is on my desk to boost the voltage to 10V after I do have an idea about its drift and aging. This is probably the most professional complete solution for the LTZ I know. Thanks Andreas. 

-branadic-

[Dr. Frank]

Very cool screws!!

[branadic]

Only a few stainless steel screws. The panels were designed with Schaeffer Frontpanel Designer and laser cutted by a friend of mine. The back panel does contain Pomona 3770 binding posts. Back panel will be finished soon, but I need some spacer as D-Sub25 connector is not flush with the circuit board edge.

-branadic-

[Andreas]

Hello,

Only a few stainless steel screws. The panels were designed with Schaeffer Frontpanel Designer and laser cutted by a friend of mine. The back panel does contain Pomona 3770 binding posts.

Really good looking panels against my "hand crafted" ones.

but I need some spacer as D-Sub25 connector is not flush with the circuit board edge.

Why don´t you leave the D-Sub connector off (like Frank) or replace the connector with a different one?
There are only 4 pins of interest used on the D-Sub connector:
2 for the internal temperature sensor.
2 for the unbuffered LTZ-output.
But I think for the later that it harms more (I had many accidents with unbuffered outputs) than is useful to have it outside the reference. Except if you use the 7->10V transfer option on the buffered output.

with best regards

Andreas

[Andreas]

Back panel will be finished soon, but I need some spacer as D-Sub25 connector is not flush with the circuit board edge.

Hello,

the intention was that the D-Sub front is in flush with the front plate.
But that works only with the original thickness of the front plate from the housing.
I also did not want to fix the D-Sub to the front plate to avoid mechanical stress to the PCB.
The PCB is already fixed from the BMON-side (back) so I wanted to leave the front side floating to avoid stress on the PCB.
So using Pomona connectors on the front side is a step in the right direction.

With best regards

Andreas

[branadic]

Well, I could replace the D-Sub25 angled pcb type by a panel mounted one for wire soldering to be compatible to your solution. This way the existing backpanels can be used without modification, there is no additional board stress and spacers are obsolete.

-branadic-