Ultra Precision Reference LTZ1000

[TiN]

Hm, I have a questions to ones who have LTZ's running.
Any one used "bad" resistors with tempco 25-50ppm/C ?

Since I'm still waiting on proper resistors, I built one LTZ ref to try, and seeing
big voltage/temperature drifts over time.

Like in attached log.

[Mickle T.]

All of my old LTZ references were based on the "bad" wirewound MRH resistors with 15 ppm/C tempco (max) and metal-film S2-29V with up to 50-100 ppm/C (max). All have a typical datasheets specs without any problem.

[CaptnYellowShirt]

What are the two temperature columns?

[TiN]

There was Honeywell Pt1000 RTD HEL-705 and MAX6610 in deadbug position superglued to top of LTZ can.
So 111 is RTD temp and 902 is calculated MAX6610 temp (T = Vout - 0.75V / 10mV).
Whole board was put into small cardboard package box and wrapped with 2 layers of packaging bubble-wrap.



Also what is current consumption of your references?

[CaptnYellowShirt]

Ah I see now. Thanks.

[Andreas]

Since I'm still waiting on proper resistors, I built one LTZ ref to try, and seeing
big voltage/temperature drifts over time.

Since the resistor tempco is divided down by a factor of around 100 to the 7V output it is evident that you should have at maximum around  some 10uV change.
Do you use a switchmode power supply?
Try to use batteries.

With best regards

Andreas

[TiN]

I used KI 2400, which is kinda switchmode.
Need to find battery which can hold 50mA 15V for quite a time first...

[Andreas]

I used KI 2400, which is kinda switchmode.

Does the output voltage change when you put your hand over/on the reference or
the power supply or the DMM?

Perhaps 2*100nF capacitors between base + emitter (short!) of the LTZ-Transistors will give a little improvement.

With best regards

Andreas

[CaptnYellowShirt]

6) After telling him that I wanted a voltage transfer device, he said a better way might be to use [at least 6] LM399's in parallel [like the Bob Pease idea].  He said that the burn-in procedure would be to operate these in an oven set to 125C for 2 weeks, which would be equivalent to 1000's of hours of normal operation.  Any LM399's that are drifting too much after that can be replaced [i.e., you burn-in more than you need, and select the best units for the array].  The LM399 is much more sensitive to board stress than the LTZ [because the LTZ has a special mechanical arrangement in the die mount]-- so the LM399 should be mounted off of the PCB a little bit to allow for this.  The long term stability of the array of LM399's will be directly related to the power required to run the heater-- and this can be minimized with insulation-- the more the better!  The LM399's do not age when they are turned off, and have almost no hysteresis-- so keeping the reference *off* until a few hours before you need to use it [and/or calibrate it] is the best way to keep the long-term drift minimized.

7) Note that for LM399-based designs, the slots in the PC board [plus a lot of insulation top and bottom] make sense-- Bob said that the less power the heater requires, then the more stable the output voltage will be.  So, in this case, the slots [plus insulation] are helping with this.  Oh-- and he also said that the LM399 should be run at about 1mA of Zener current for best stability.  The more stable you can make the Zener current, the more stable will be the output voltage.  He said that there is about 1uV of voltage change for 1uA of current change.

I'm running a small experiment to test these two ideas in another thread. I'd love to get some peer review on my thoughts (especially the math section).

https://www.eevblog.com/forum/projects/long-term-lm399-stability/msg419452

[CaptnYellowShirt]

A few weeks ago there was a discussion on this thread about the effects of humidity on the Fluke 732a/b.

I'm reading a paper (http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=377817) that discusses these humidity effects on a set of 732a's. The paper claims that humidity effects the 1.018 v output -- around  - 1uV /  + 10% change in relative humidity and time constant in the 20-30 day range.

However, the paper claims there is no detectable change on the 10 v output.

The authors seem to imply that the change could be a product of the 1.018 v voltage divider or internal leakage paths.

For those of you with a 732 (or similar), what can you say about this? Seems weird to me that one would change and the other wouldnt?

[CaptnYellowShirt]

The cited paper on the physics of moisture effects is here:


"Moisture solubility and diffusion in epoxy and epoxy-glass composites"
http://domino.research.ibm.com/tchjr/journalindex.nsf/0b9bc46ed06cbac1852565e6006fe1a0/0521bda5a94455f085256bfa0067f5f7!OpenDocument

 
Anybody here work for IBM?

[acbern]

the problem is that although I have a 732a and a 4910, I cannot really tell which one is drifting when, unless i would correlate to an external standard. too expensive.
the only other way to test would really be humidity cycling, one against teh other. have no chamber either.

however, I have done some general tests on precision resistors (z-foil, highest grade, not hermetic) agains an hermetic one (said to drift by less than 0.5ppm/a by vishay). within a year, a 1k drifted by 10ppm, a 100k drifted by 3.4ppm (assuming a stable 10k hermetic). both reduced their resistance values. that also is in line what vishay states and mostly related to humidity influence, hence the hermetic versions. oxigen seems oi be of minor importance. vishay also states the effect is reversible.
not having disassembled my 732a, maybe the anser is to check the different resistors used. now probably they are not vishay, but may be a first indication.

[ManateeMafia]

I have a couple of pics of the reference board inside one of my newer rev 732A. I replaced the 10V pot and thought I should take some pics. These were taken a year ago.

I apologize for keeping these to myself for so long.

These were taken as a visual aid for reassembly of the oven and to note some of the design details not shown in the schematics.

The second picture shows the four resistors used in the 1V and 1.018V dividers. I believe these were a rev to the original design and I don't know what prompted the change.
Perhaps the sensitivity to humidity may have been a factor?

[branadic]

I want to refresh the discussion about the substrate. I already mentioned that LTCC could be interesting, but is that really the case? I reviewed a few datasheets. As always, units seem to be a real problem for some guys

Isola IS410
Thermal Conductivity: 0.5 W/mK
Coefficient of Thermal Expansion X, Y: 13 ppm/°C (Post-TG)
Coefficient of Thermal Expansion Z: 250 ppm/°C

RO4003C
Thermal Conductivity: 0.71 W/m/°K @ 80°C
Coefficient of Thermal Expansion X: 11 ppm/°C (-55 to 288°C)
Coefficient of Thermal Expansion Y: 14 ppm/°C
Coefficient of Thermal Expansion Z: 46 ppm/°C

9K7 GreenTape
Thermal Conductivity: 4.6 W/m-K
TCE, (23° - 300°C): 4.4

951 GreenTape
Thermal Conductivity: 3.3 W/mK
TCE, (25° - 300°C), ppm/°C: 5.8

A6M
Thermal Conductivity: 2 W/mK
Thermal Coefficient of Expansion: 7 ppm/°C

L8
Thermal Conductivity: >3 W/mK
Thermal Coefficient of Expansion (25-300°C): 6ppm/°C

So for what I see, the use of more available RO4003C instead of expensive LTCC could be worth a try. It has better thermal conductivity (good for all pins on same temperature) and smaller TCE. LTCC would be quite better, but on what price?

[Andreas]

Hello branadic,

kovar has a TCE of around 5.9 if I found it right.

with best regards

Andreas

[MisterDiodes]

 I've been following along here on the LTZ1000 reference discussion for a few months - this is a really great forum and the best discussion I've seen on LTZ1000a!

I've built several hundred of these over the years for specialize production test equipment on laser diode/detector fab lines - and so far we've never had an issue with using just wirewound resistors (3ppm units typical, some have 1~2ppm parts for the 120 ohm resistor), no slots or gimmicks, lots of copper on a small board and put the ref board in a can, and use a linear pre-regulator.  I promise you that the units in the field for 20 yrs have gotten very, very stable, and drift rates of less than 1ppm / yr are really pretty typical. LTZ1000a devices, and we did some LM199 / 399 also (those are very good also) .  No voodoo. 12.5k over 1k WW for heater circuit, and everything basically as in app notes.   The only boards we had trouble with were when we bought the expensive hermetic magical Vishay resistors - Dr Franks discussion on the hysterisis effect has been our experience also, if the temp gets too warm especially - they will take quite a while to recover.    The Vishay's worked OK but were never worth the extra expense (at least for us). The regular wirewounds - and heck the 5/ 10 /20 ppm surface mounts work fine also if you have reasonable temperature control.  Unless you're running the ref outside, you should have good results with even moderately good resistors.  We've tested the expensive resistor-parts units, and after a 5 & 10 & 15 year tests we're reminded how optimistic Vishay's datasheets are.  LT1013 for the op-amp, we've never had trouble with - plastic package or metal TO in the early days. Humidity has never been an issue for us since our boards are running in a humidity-controlled environment anyway in the clean rooms.  We've never really had any of these refs fail outright, and the older they get the better they are, and its really amazing how well they hold up over time.  Kind of like me (I wish) 

And then tonight I stumbled across this -   We've been doing it  all wrong all these years...

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=261407717248

What could go wrong?  I asked the seller what the drift rate per year is, and he says you don't need to keep it on - its always good and it never ages... But you don't want to keep it on a lot...  He says he had an Electrical Engineer with 40 yrs experience look at it and was astounded at the results!  I see he's sold 4, as of tonight. 

Thanks again Dave & Co. for the great forum and discussion!

[CaptnYellowShirt]

He says he had an Electrical Engineer with 40 yrs experience look at it and was astounded at the results!

Hmm... only 40 years of experience? For my money, his imaginary friend would need at least 41 years.

[MisterDiodes]

I would be astounded too, if it works.  I was almost going to buy one for fun, then I realized I had better use for the time & money.  I also asked what the load regulation was, and no answer from seller yet.

I think I'll stick to stuff I know works, and works well for a long, long time.  Like Buried Zener refs.

[MisterDiodes]

This might not exactly belong in this thread with LTZ1000a's (except to do a direct comparison about vrefs for general knowledge), but this is the seller's response when I asked about drift rates on his  ebay device.  I'm not even sure I can understand it.  It might be a pot attached to an LM317 for all I know. I think he's convinced himself its better than a 732a:

We do a burn-in to minimize drift. These units do not need to be run continuously like a Fluke 732A/B. So, the drift over time is not coherent. These units have a specified drift per first 1,000 hours of use (~10ppm, and much less for the second and subsequent 1,000 hours). They require roughly five minutes to settle to maximum accuracy. As long as the unit is not powered on for extremely long periods (250 hours or more), it should hold the standard indefinitely.

The stability is greater than the maximum error of the 732A. In the null setup photo, the unit drifted from +3uV to -4uV. -4uV was the maximum error in relation to the 732A. The 732A has a maximum error of 0.6ppm, which @ 10VDC is 6uV. So, the maximum error of the unit using the 732A as a reference is 10uV. We spec the unit @ +-75uV, because reference standards require integrity and reliability. We may change that spec when we have more experience with the units.

We have an electrical engineer with forty years experience, conducting R&D on special mock-ups of this device, and the results have been astounding. But for now I'm sticking with the original spec of +-75uV. But the temperature at calibration is marked on the each unit. If this temperature is observed, the accuracy is in excess of 10x the official spec. Our testing has shown that the units are unaffected by repeated power ups. The drift is in relation to cumulative hours of power-up

[CaptnYellowShirt]

There's nothing to understand. Its just double speak and audiophoolery.

[CaptnYellowShirt]

Maybe we should offer an independent test? Intercompare it with our standards?

[jd]

3) What is the voltage reference inside the box?  Is he not saying because it would be embarrassing?  Is it a voltage reference with a non-hermetic package?  What is the temperature coefficient [TempCo]?  Is the TempCo compensated?  What is the hysteresis of the part if it experiences radical temperature excursions during shipping?

Hi,
The listing "item specifics" says that it is a LT1021-10. So one of those in a little plastic box I guess.

JD

[MisterDiodes]

...or keep the plastic Vref package at constant temp, and give it a jacket of Tfe coating, then its impervious to the local weather.  That helps for just about any Vref.  Most of the stuff on eBay just isn't built that carefully.

The plastic package absolute Vref can sometimes be used for precision work to -some- degree -  IF its really the same die inside (as metal version), and you have good thermal management & humidity control- which normally we do anyway.  We do that on all Vrefs to give them the best chance, and when we build a box with multiples for averaging calibrators - say 35 vrefs in one temperature-controlled box - each Vref circuit or Vref group is replaceable as its own module.  You can spot the drifters that way.  But you have to keep an eye on them - if you have a wonky drifter in the group that's not playing well with the other Vrefs to contribute to a low-noise average, he has to go.

The reason we have to use plastic package Vrefs sometimes is the availability on the Metal cans means kind of a long wait for production, and sometimes we can't do that.   Sometimes a customer can't wait, and we're forced to use plastic - and take all the precautions.

Regardless of plastic or metal package, the big trouble with any absolute ref on a die is they tend to drift -much- more over time than a well-built, stable LTZ1000a / LM399.  The LTC6655 does pretty well (for an absolute ref) , although we don't have any to test for 10yr drift - yet.

Sorry for the thread interruption - back to the good LTZ1000a's...




 

[CaptnYellowShirt]

Dr. Frank has discussed the effects of heater set-point temperature on the LTZ1000's drift rate. A lower temperature results in a lower drift rate, however I am wondering about the long-term total drift between a 'hot' and 'cold' LTZ. Will they end up with the similar amounts of total drift over the course of many years -- temperature simply affecting the rate at which they asymptotically approach that point? Or will a higher temperature tend to diverge in total drift when compared to a lower temp reference?

[acbern]

I was wondering the same, and have nowhere seen any comments/tests. In general, higher temp means faster aging, a simple law of physics (arrhenius law). aging means drift. that is in line with the temp setting observations of the 3458a ltz1000. of course, always, there are additional effects. the ltz, according to the data sheet, needs some time to settle and decrease in drift rate. and of course, there are more and less stable devices. it is beyond me why a 3458a opt. 2 may or may not be powered and still meet its accuracy while all precision references must be on. also, this is a little bit like saying if you have a precision test equipment, keeping it on is better. in my experience, the opposite is the case. of, course, not the exact same thing, and not as simple as that, but anyway.
it woud be worth while doing a test with a set of ltz1000, keeping them on initially until stabilized per data sheet, determining their drift rates, and then switching some off. i bet there is not much difference, maybe the ones in off-state are even more stable, given their initial drift is already over.