Ultra Precision Reference LTZ1000

[SeanB]

To ensure a very clean connection of course.......

[Andreas]

Hey folks.
How much time does properly designed and constructed LTZ1000(A) ref takes to get stable output, counting from power on?

Also what is usual current consumption of reference, given that temp set to 12K/1K resistors and powered from +12 or +15VDC?
Is 10.5mA sounds corrent?

Hello,

I have around 20 mA / 21 mA @ 23 deg C environment with 14V and LTZ1000A and setpoint 12K5 / 1K with good thermal isolation.
so 10.5 mA is rather low.
You either have a perfect isolation of your LTZ or something went wrong.
The zener alone needs already 4-5mA the Op-Amp also 1mA and the Setpoint divider also 1mA. So there is almost no heater current.
I also think that 12V is too low (will need a long startup-time).
I have not measured the startup time (since my devices are on 24/7) but it should be around 15-30 minutes.

with best regards

Andreas

[TiN]

I see, then I f'up somewhere again, as voltage still bounces around way more than few ppm.   
With +15V current is 12 mA, which is still seem too low.

As of connection, I just stuffed something proper size to keep copper UTP5 wire in banana connector fixed and not have big thermal mass Cleaning sticks are just perfect size for bananas 

[Andreas]

Hello,

I guess you have interchanged Collector and Emitter of your heater transistor.

how high is the heater voltage?
If I have calculated right you should have around 6V+0.7V (Diode) at Pin 1 of the LTZ1000.
Which would give around 20mA at 28 deg room temp / 40 deg setpoint for the heater (300 Ohms).
The Op-Amp output should be around 0.7V(+0.2V resistor) higher (at 7.6V)

With best regards

Andreas

[TiN]

No, that was fixed already. It was inverting input resistor connected incorrectly to VREF out, so heater was not heating
Now it's running. I'll do measurements when get home.

[TiN]

Thanks!

Initial data with working reference. Spikes and sags are due to dirty lab environment with lots SMPS's turn on/off around

[TiN]

Had LTZ1000 + LTC2057 ref running over couple weeks, discovered wierd low-frequency oscillation on output.
I think it's heater circuitry control, as consumption current is following trend, going from ~28.5 to 30.5mA.
Probably temperature setpoint is too low, this ref using 12K + 1K resistors.

Will test other units with different opamps and LTZ1000A, too see if it happens there too.

REF was powered from Keithley 2400 SMU, +13.000VDC set.
Sampling speed is every 2000ms, settings are:

Code: [Select]

:SENSE:FUNC 'VOLT:DC';
:SENS:VOLT:DC:DIG 8.5;
:SENS:VOLT:DC:NPLC 10;
:SENS:VOLT:DC:RANG:UPP 20;
:SENS:VOLT:DC:AVER:COUN 10;
:SENS:VOLT:DC:AVER:STAT OFF;
:READ?
8.5digit, 10 NPLC, Autozero syncronous.

 

Reference was enclosed in foam box and taped around, which was put into small plastic cage.
Tried various ambient temperatures, +24c to +30C, was same result.

Started study on LabView to setup nice logging setup. I already got Keithley 2001-TCSCAN card for my 2002, which is 10-channel scan card.

Have idea to use Keithley 7001 as a box for multiple LTZ1000's references. It have nice shielded compartment, can have one scan/digital card to mux outputs/ratios.

[Andreas]

And you are shure that the 2002 is stable?
I also had around 2uV unstability on a K2002 with my references.

I would try the following:

- Measure the difference between 2 LTZ1000 (higher resolution)
- install a 10-100nF Cap between base + emitter of the temperature sensing transistor of the LTZ (Pin 6+7).
- how does the LTZ temperature (your sensor) behave during shifts?
- does the heater voltage change also?

With best regards

Andreas

[Dr. Frank]

Well,

12k/1k is perfectly fine and stable for 45°C on a LTZ1000, but environmental temperature is limited, and it's not suited for the A version (??).
(The photo is not sharp at all, but I think you have an ACH type here, and then, 45°C is too low.)

Your measurement seems to be 17 min time span, I guess, if your time scale is seconds.
Your output shows about 1µVpp (=0.15ppm) short term noise plus possible superimposed 3.5 µV (=0.5ppm) of longer termed oscillation, period of ca. 2.5 minutes.

I also use 12k over 1k , LTZ1000, LT standard circuitry, and get less than 0.15ppm = 1µVpp noise, see again a typical measurement on my LTZ1000, NPLC 100.
That is totally normal, covered in the datasheet.

If your superimposed oscillation of about 0.5ppm is real, something else is fishy, either in the Keithley instrument, or as the period seems to be about 2.5 minutes, some thermal issue like a regular air draught or so.. Maybe there's a fan going on and off..
Hope all of the solder junction are also thermally isolated.

Besides that, the amplitude of the basic noise of your circuitry is very acceptable.


Frank

PS: Added the 10min stability figure from a "quiet" time area, the noise is even less: +/- 0.05 ppm = 0.7 µVpp.
PPS: I use NPLC 100 , but 4 sec measurement time, with Auto Zero On which doubles the measurement time, and I take data as fast as possible.

Check whether your Keithley is set up properly, e.g. if it also does Auto Zero on each measurement (synchronous AZ doesn't tell me anything).

And your 2000ms measurement time does not fit to NPLC 10 (which is 166 or 200ms). Or do you average 10 samples for one data point?

PPPS: Be aware, that your self built reference is or might probably be more stable than your Keithley instrument!
At least, as you measure the "naked" output voltage, whereupon your instrument (and any other 8.5 digit instrument ) includes additional circuitry or measurement modes which in sum are less stable.

Therefore, distrust your Keithly more than your circuitry, in first instance..   

4PS: One last idea.. the other guys mull something about mains beat frequency and AC disturbance, .. naah wouldn't explain directly oscillation with 2.5minutes.. But on the other hand, if you operate that circuitry unshielded, it is extremely susceptible to AC noise signals... those spikes you see in your measurement, and in mine, are evidences for that.

I also observed longer termed changes (dips & more) due to AC noise pickup, especially when I turned on my LED light, which has an switching mode PSU..

Afterwards I have thrown out all switched modes PSUs of my analog lab, and I am using light bulbs again, which run on 50Hz, here.

Problem solved. Maybe you check shielding and turn off ALL switch mode PSUs in your lab..

Is your environmental temperature stable during measurement??

I can achieve a few tenths of °C over 35h..w/o air conditioning.

[Andreas]

If your superimposed oscillation is real, something else is fishy, either in the Keithley instrument, or as the period seems to be several minutes, some thermal issue as regular air draught or so..

Perhaps wrong mains frequency? or a beat frequency with mains?

With best regards

Andreas

[MK]

perhaps your thermal set point circuit has an rf oscillation that is affecting the setpoint? or pehaps with a low demand temperarure you need to slow down the max rate of heat input so that the heatup and cooldown timeconstants are not too far apart, put a resistor in series with the heater for instance.

[TiN]

I have quite wonky setup right now to collect data by LabView, so time scales and sampling speed is longer than actual measurement.

I gathered data from unaged another LTZ1000A-based reference, but now added EDC MV106 output (measured by cal'd K2001)
and added current output data from K2400 (which is powering LTZ ref to +15.000VDC)

Here's overnight data with airconditioned room, where all gear reside, with temperature variance shown on graph.
LTZ1000A definately looks way more stable than non-heated zener-based MV106.

Autozero on both K2002 and K2001 set to syncronous, which means:

Synchronous (the default mode) is the most accurate,
but slowest mode. In this mode, each trigger causes
three A/D conversions: one for input signal, one for an
internal zero, and one for an internal gain. This mode
also yields a constant reading rate.

And all instruments powered from 60Hz 110VAC.
There are few other gear working around, so I cannot say it's pure clean environment.

I will try with capacitor between pin 6 and 7 later.

Next plan is to make an assembly board with 3 x references powered from Linear LT3081ER LDO and powered from separate linear power supply, or maybe even car two car batteries

[Andreas]

Hello,

whats that:
the LTZ1000A (heater-) current increases when temperature is rising?
Are you shure that the op-amp pins are not interchanged?

With best regards

Andreas

[TiN]

Sorry, i missed label Heater works correctly, current goes down with temperature raise.

[TiN]

More data, LTZ1000A raw output overnight.

Upper chart - EDC MV106 +7.13663 VDC output to Keithley 2001
Lower chart - KX LTZ1000A ref output to Keithley 2002

Both run NPLC 10, 8.5 digit mode.

Current "spikes" - I manually turned off +15V source, to "sync" graphs, as right now all data is collected by two separate labview modules,
will work later to use one labview app to get data from both SMU and multimeters.

Next step, run group test on all five LTZ references with Keithley 2002 + scan card as logger. Also will gather MAX6610's temperatures.

[JesT]

This is my first post on this forum.

A while ago I stumbled over Dave Jones’ videos and that revived my longtime slumbering interest in electronics (actually, I started building stuff over 60 years ago, mostly audio and test instruments, including an oscilloscope).

I have always had an interest in accurate measurements but have never owned any really good instruments, simply because they were too expensive. I was therefore intrigued to see in DJ´s videos all the quality vintage stuff you can buy at reasonable prices (until DJ promotes them, that is!). I had no idea these treasures existed.

Later I dug into the eevblog forums - especially this LTZ1000 forum and other fora on voltage references, precision resistors, power supplies and DMMs.

I have learned a lot during the last 4-5 months and I´m so impressed by all the people in here - professionals and amateurs - who readily share their knowledge, academical as well as practical, make experiments and discuss possible solutions to problems, all in a very friendly tone. I have been through these 62 pages (including links) several times and learned something new each time.

I believe I´m now ready for action. My plan is to build 4 units with LTZ1000 and 4 units with LM399. I have got the components, the references are burning in and I´m going to make PCBs (Eagle is new to me, too, I´ve only made PCBs with photoresist over 40 years ago).

Why on earth am I doing this and what do I expect to get from it?

Well, I´m doing it mostly for the fun of it. I think the fun comes from the fact that it is really challenging to do these things well. If I didn´t know beforehand, I know it now, thanks to all the good people who contributed.

Hopefully, the result will be that I have some references that will give me confidence in my (presently 8 ) DMMs without the need to formally calibrate all the time.

Regards

JesT 

[Andreas]

Hello JesT,

welcome to the club.

4 references of each is a good number to start ageing comparisons.
I hope you will publish your design here.
And dont forget to use a good thermal isolation for the references.

Which resistors did you chose for the LTZ cirquit?

With best regards

Andreas

[wiss]

I am currently working on a design for a "poor man's thermal converter" that can be used to measure AC waveforms very accurately.

Please start a thread on this subject!

[bingo600]

if you are like me you will be compelled to make or buy some ultra high precision and/or stable resistors.  It just gets more crazy after that...

Ohh yes to the CRAZY part

I started with a HP34401A , and a Geller SVR/SVR-T
Then i build a LM299 Ref (Branadic version) , and i now have all the parts for a LTZ1000 ref (PCB sponsored by quarks).
Note that the LM299 & LTZ1000 are the cheap parts ... The resistors (Outchh)

Then i got a new problem ... How to accurately measure a LTZ1000.

Solution (Big Outchh) ... But i got my dream meter


/Bingo

Ohh , on the picts i have't done the ACAL ALL yet (my first power on)

[Andreas]

The resistors (Outchh)

Then i got a new problem ... How to accurately measure a LTZ1000.

Solution (Big Outchh) ... But i got my dream meter

Hello Bingo,

Which resistors did you choose?

Your problem is still not solved: LTZ1000 accuracy is around 1-2 ppm/year.
HP3485A is specced with 8 ppm/year.

So you will at least need one (up to 4) Fluke 732B to calibrate the HP each time before you measure the LTZ.

With best regards

Andreas

[bingo600]

The resistors (Outchh)

Then i got a new problem ... How to accurately measure a LTZ1000.

Solution (Big Outchh) ... But i got my dream meter

Hello Bingo,

Which resistors did you choose?

Your problem is still not solved: LTZ1000 accuracy is around 1-2 ppm/year.
HP3485A is specced with 8 ppm/year.

So you will at least need one (up to 4) Fluke 732B to calibrate the HP each time before you measure the LTZ.

With best regards

Andreas

Andeas i suppose you meant 3458A .... It's an OPT-002 (4 ppm/yr) from early 90's so well aged reference.

And i know ... Was thinking about a 732A , but have spend my "toy budget for a while".

If i get the 732A , the next i need a ESI SR104  ... It never stops.

I got econistors , but was thinking about getting a "set from edwin" , didn't you get some from him ?

Next project is to replace the 3458A calram , and prob. the storage rams also , and then while i have it open ...
Upgrade with OPT-001 (only 10$) in ram chips.

/Bingo

[HighVoltage]

Solution (Big Outchh) ... But i got my dream meter

Congratulations on fulfilling your dream.
Your 3458A looks very clean, even inside the binding post copper area.
Now you need a Fluke 8508A for comparison
It seems like a never ending spiral for just one more digit of precision and accuracy.

[Andreas]

Andeas i suppose you meant 3458A .... It's an OPT-002 (4 ppm/yr) from early 90's so well aged reference.

And i know ... Was thinking about a 732A , but have spend my "toy budget for a while".

If i get the 732A , the next i need a ESI SR104  ... It never stops.

I got econistors , but was thinking about getting a "set from edwin" , didn't you get some from him ?

/Bingo

Hello,

yes of course the 3458A.

Caution the Fluke 732A  is specced with 6 ppm/year.
I meant the Fluke 732B which is specced with 2 ppm/year for the DMM.

For the resistors: I still have no price indication from Edwin for his resistors.

With best regards

Andreas

[JesT]

Hello JesT,

welcome to the club.

4 references of each is a good number to start ageing comparisons.
I hope you will publish your design here.
And dont forget to use a good thermal isolation for the references.

Which resistors did you chose for the LTZ cirquit?

With best regards

Andreas

Hello Andreas,

Thank you for your welcome!

I chose 4 units from the following reasoning:
1 unit: You feel happy, but really, you know nothing
2 units: Maximum uncertainty, if different, which one is right?
3 units: Better, the two agreeing are (probably) right
4 units: Even better, three units agreeing can’t be wrong (or?)
You could go on forever...but you have to stop at some point.

I will in due time present my project in detail. For the time being i will just say that there will be 4 units with a buffer to unify to 7 V for each unit, a separate unit to boost the individual unit or the average to 10 V and 5 isolated external linear power supplies.

In my humble view the most critical parts are the PCB (avoiding EMFs from temperature imbalance) and the resistors in the boost unit. I see resistors for the LTZ1000 circuit itself as a lesser problem when it comes to their contribution to the total error budget.

For the LTZ1000 circuit I have bought 5 sets of used Vishay S102 0.01% resistors. They have pretty short leads so they will, no doubt, be influenced somewhat by soldering.

I will first make a prototype with one of the 5 sets in order to check the working points, stability (additional compensation?), temperatures and TC in the circuit with all four LTZ1000. This may lead to changes in the final version. The prototype will NOT be able to determine whether the resistors are good enough.

The booster will be a copy of dr. Franks (first post on page 13), only with 10K wirewound resistors instead of 50K (yes, 50K would have been better). I’ve bought 20 old new ones, unknown manufacturer, 0.01%, all of them still within 0.0035% with a TC of 3-5 ppm/?C by a rough test. I will match these resistors by value/temperature compensating each resistor until “perfect” tracking (0.2-0.4 ppm?). It can be repeated anytime in order to conserve accuracy, so ageing is out of the equation.

I have comtemplated using LTC1043 for dividing/transferring voltages up and down. On paper the specifications look very promising, but according to your own experiments it is not that easy. Anyway, I have bought some and will play with them - someday.

And, no, I won’t forget isolation! Who could after reading Bob Dobkins comments? (Thanks, DiligentMinds)

Best regards

JesT

[JesT]

<snip>
Why on earth am I doing this and what do I expect to get from it?

Well, I´m doing it mostly for the fun of it. I think the fun comes from the fact that it is really challenging to do these things well. If I didn´t know beforehand, I know it now, thanks to all the good people who contributed.

Hopefully, the result will be that I have some references that will give me confidence in my (presently 8 ) DMMs without the need to formally calibrate all the time.

Regards

JesT

Well, it sounds like you have been bitten by the "precision bug".  Many of us here in this thread and many on the "Volt Nuts" and "Time Nuts" mailing list also have this affliction.  It can be a rewarding hobby, but at least for me has also been a very expensive one-- always striving for the best possible absolute accurracy.  In real life [as in engineering and technician work] this kind of precision is rarely [if ever] needed.  But that hasn't stopped me from buying a lot of ultra high precision gear.  {...sigh...}  It seems that there is no end in sight.  After building and calibrating [which is the hard part] some precision references, if you are like me you will be compelled to make or buy some ultra high precision and/or stable resistors.  It just gets more crazy after that...  I am currently working on a design for a "poor man's thermal converter" that can be used to measure AC waveforms very accurately.

Hello DiligentMinds,

Yes, I've been bitten. But, officially, I'm not a real "volt nut" - yet - as I do not have a 8 1/2 digit DVM. I have restricted myself to buying 6 1/2 and 7 1/2 meters because the improvement in drift from 6 or 7 digits to 8 digits comes with a very high price tag, both initially and later on (calibration). I really do not need absolute accuracy to perform my experiments with LTZ1000 references, the differences (or lack of such) they will show will be sufficient. Resolution and short term stability, on the other hand, are essential and here, I think, 7 1/2 digit is enough. Furthermore, in my environment, I'm afraid, a 8 1/2 digit meter will not fully come to its right.

BTW, my three HP 3456A (three different series, not the first one with a lot of problems), one of which has been calibrated in 2014, have shown remarkable stability. The two uncalibrated ones differ less than 20 ppm from the calibrated one - after more than 20 years since last calibration. And they have best TC of all my meters. Amazing.

So far I have been able to restrain myself, maybe because I have set up the rules of the game for myself (no 8 1/2 DVM, no high end calibrators, no SR104). Keep it at a reaonable level and get the most fun possible out of it. Only time can show whether "craziness" really catches up with me!

And, yes, so far it has been rather expensive, because I have also bought some other gear (vintage scopes, decade resistors, KV divider, precision voltage supplies and quite a lot of electronic components). I'm looking forward to bringing these beautiful instrument back to their former glory, as far as I can (I do not possess Robrenz' skills, helas!).

Best regards

JesT