Ultra Precision Reference LTZ1000

[chickenHeadKnob]

  Ken, thank you for doing this. I don't think I can add to  the list of questions generated by you and Dr. Frank, save for one area. What does LT recommend with respect to circuit board mounting methods and device characterization before and after. What I am getting at is that both LT and subsequent "major customers" are testing and aging devices in sockets (presumably) and then usually solder mounting with the exception of the recent Agilent 34461-lm399. What do they know that they are not telling us? Agilent might want to keep trade secrets but LT should be able to share their statistical knowledge. If a device receives a thermal shock from soldering what does it do to the long-term aging?

  I am contemplating using gas tight gold-plated sockets of my own design instead of solder simply because I don't know  what the resultant effects of  shock/reset will be, and I don't have the resources to buy large sample lots and perform  the necessary tests.

[CaptnYellowShirt]

Most of the manufacturers that are using the LTZ1000A are building it onto a module, and then "burning-in" the whole module-- not just the LTZ. 

Do you know if the modules are powered during this burn-in time?

[fmaimon]

According to the "rumors"-- yes.  And their outputs are monitored as well.  The module for the 3458A has a cheap resistor on it that doesn't appear to be used for anything [R419, 2K67]-- not even on the main board that you plug the reference board into.  The resistor is connected from -15V to ground-- and nothing on the board uses -15V.  It's my theory that they are "watching" the drift of this cheap resistor [in the "burn-in" fixture] to know how far they have come in the burn-in process.  I could be wrong, but it's the only thing that makes sense.

Maybe this resistor is "nulling" the gnd pin current from the main board to the module. 15V @ 2K67 is about 5.6 mA, about the right value considering 5 mA for the  zener curent, considering that the heater goes to another power pin...

[branadic]

Have you ever seen a board with solder stop mask that was driven to  temperatures up to 125°C or even 150°C over and over again for hours, weeks or months? Its color changes, the color of the solder joints, the board itself and several parts on the board too.
I've done a high temperature storage at 120°C on my LM399 reference board for more than 300h and the color has changed clearly. I  can't see such an evidence on the pictures of the reference boards or even the pcb material on the 3458A reference boards.
Whatever burn-in means for several manufactors, power the circuit up or power the circuit up driving a temperatur profil, most things discussed help keeping a myth alive.
I appreciate an offical publication that do away with all that fairy tales.

[Dr. Frank]

I've done a high temperature storage at 120°C on my LM399 reference board for more than 300h and the color has changed clearly. I  can't see such an evidence on the pictures of the reference boards or even the pcb material on the 3458A reference boards.
Whatever burn-in means for several manufactors, power the circuit up or power the circuit up driving a temperatur profil, most things discussed help keeping a myth alive.
I appreciate an offical publication that do away with all that fairy tales.

leaded solder gets soft from 105°C onwards, leadfree a little bit higher.
That softening (euthetic phase) will harm the solder junction.

Therefore, any storage or cycling of soldered PCBs above 110°C will definitely destroy the solder junctions, or at least deteriorate the reliability greatly.
It's possible only to make such  high temperature storage of non assembled components. Here, the max. die temperature of about 150°C is a limit.
Getting close to that temperature will also harm the silicon structure already, due to Arrhenius law.

Burn-In usually is done on components, which have a big drift rate intrinsically, and the purpose of the burn-in is to accelerate those drifts to a state, which otherwise would be reached only years later.

Another technique is to burn-in a completely assembled  PCB (at 90°C max.) to detect early failures and so to get more reliable PCBs, especially under rough conditions as spacecraft, military and automotive applications.
But latter goal is not ultimately required for volt references.
 
So I stick to my opinion, that a real burn-in on ultra precision components, i.e. on components which have low drift rates by design already, will do more harm than really improve the stability further.
 Frank

[Dr. Frank]

The highest performance DMM's on the market today are the HP/Agilent/Keysight 3458A and the Fluke 8508A.  They both use the LTZ1000A, and they use "reference modules" that are burned-in separately from the rest of the instrument.  They are *not* doing this "because it seemed to be the right idea"-- they are doing it because they have spent a ton of money and manpower on this problem, and they both reached the same conclusion.

Hi Ken,

what do you know about the interior of the Fluke 8508A?

I also assumed, this instrument is very similar to the Datron 1281 (very similar manual), an dwould have the same LTZ1000 reference module(s) as in the Datron 4910, but quarks has sent me a photo indicating, that the 8508A is very differently built.
The reference has a resistor ceramic hybrid on it, which resembles more the Fluke 732B assembly inside the oven part.

I'm still collecting all reference types (LTZ/A, SZA, LTFLU) of the instruments, and their different stabilization temperatures.

Fluke 8508A and the Keithley 2002 are still lacking.

PS: The engineers who designed the 3458A reference board have ignored many rules for getting a stable reference, as too high stabilization temperature, using the A version, and exposing the solder junctions to airflow. That's the only reason, why they are always struggling with the stability, and why they have to make such a big effort in selecting the references for 8 / 4 / 2 ppm/yr.

Thanks Frank

[branadic]

I'm still collecting all reference types (LTZ/A, SZA, LTFLU) of the instruments, and their different stabilization temperatures.

By the way, LTFLU is currently available on ebay eBay auction: #111063925120.

Fluke 8508A and the Keithley 2002 are still lacking.

The pictures by TiN didn't help?

[Dr. Frank]

I'm still collecting all reference types (LTZ/A, SZA, LTFLU) of the instruments, and their different stabilization temperatures.

The pictures by TiN didn't help?

Nope. Also not the BOM of the circuitry, which I've found on the Keithley site..
I've expected precision divider resistors in the range between 1k:12k and 1k:15k, but there are no resistor relations like this, neither multiple of those.

So I assume, Keithley uses a different scheme to set the temperature. perhaps some PWM or DAC, or so.
It would be useful, if somebody would determine the temperature by measuring the UBE voltage directly.

Frank

[TiN]

I'll do all that later, when recovering schematics.
I just don't want to disassemble meter every time, for this or that, it's still likely in cal.

[Andreas]

Somewhat off topic but since the LTZ needs precision resistors too:

perhaps this fits to Franks results for the TC of the resistors.

With best regards

Andreas

[casinada]

Is anybody familiar with this unit?
VALHALLA 2720GS
http://www.adret-electronique.fr/valhalla_2720/adret_valhalla_2720.html

It has 0.01 ppm and suppose to be able to calibrate 8 1/2 digit DMMs
10.0000000V

[quarks]

I am a big fan of Valhalla gear and saw this unit before.
But so far I have never seen a real one and only have seen very litte details. 
see att. Prices and HSR specs

[CaptnYellowShirt]

"2 Year Spare Parts" ?

Does it require regular oil changes?

[Mickle T.]

Valhalla 2720GS = Solartron 7081 (PWM ADC) + Datron 4000A (8x 1N829 zeners and PWM DAC) 

2720GS has a very old and noisy PWM type ADC for a such measurements. 0.01 ppm resolution is absolutely useless.

[TiN]

There is dead one on ebay for sale for a while for 1K.

[babysitter]

Freshly calibrated 34401A@work arrived from cal today, will measure my LTZ source soon

[Andreas]

Freshly calibrated 34401A@work arrived from cal today, will measure my LTZ source soon

Too late: you are already outside the 24hrs calibration window.
For the 90 days tolerances the accuracy is +/- 190uV for a 7V source.

And did you measure the 34401a before shipping?
Or how can you tell that the device did not drift during transport back from calibration?

With best regards

Andreas

[babysitter]

Hi Andreas,

its too late to use the 24h-window, yes.
But its not too late to do exactly what I am able to do, although it will be one more day I am afraid.

I have a second 34401A sitting here which was compared to the cal'd one right before shipping, should be able to give a hint from it how the difference between departure and (late) arrival is.

However, Agilent screwed up a bit this time, the meter was out for about one month, they requested more and more order confirmations totalling at 3 (upfront at InfoLine, on paper with signature in the box and after their 3rd request as email again), cal was on 6th and arrived here on 11th.

It only gives a bigger window of where the actual voltage might be. But I dont expect the LTZ to have wandered too far.

[branadic]

Thanks for sharing all this information.

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.

9) For the LM399 and both of the LTZ parts, he said a good burn-in routine would be to operate them at 125C [in an oven] for 2 weeks.  After that, you can cycle the power on and off 10 to 15 times at normal operating temperature, and this will get them to settle down.  This process should remove most of the initial drift that these devices exhibit.  [So, my initial guess for a burn-in cycle was pretty good-- there is nothing wrong with a burn-in 10 times longer than this.]

That confirms the way I realized my first LM399 reference, a slotted pcb design with a thermal encapsulation on both sides and a thermal burn-in of the board for more than 300h, but sure without powering the circuit, as I had no adequate cables at hand. The reference seems to be very stable and that makes me believe that the humidity depended drift I can observe is the drift of the 34401As reference itself, because all the mentioned actions are not realized in this DMM.

Once again, thanks for sharing.

[Vgkid]

Thnks for sharing Dilligentminds

[quantumvolt]

Thank you very much to Bob Dobkin and to you DiligentMinds.com.

I have been working a lot with the SZA263-type AmpRef/RefAmp circuit, and have found that HP used the circuit/component already in 1963, while Fluke seems to use it first time around 1971. The circuits are so similar that it is imo unreasonable to argue for any originality in the Fluke 731A.

So may be it could be an idea for US based volt-nuts to contact Fluke and ask them to "free" the LTFLU-1.

[CaptnYellowShirt]

Did you happen to get an address that we could send a thank you card to?

[quarks]

  great information
Thanks a lot for sharing

[robrenz]

Thanks for the great information

That gives me confidence to continue my practice of only turning on my 8846A when using it.

[quantumvolt]

Thank you very much to Bob Dobkin and to you DiligentMinds.com.

I have been working a lot with the SZA263-type AmpRef/RefAmp circuit, and have found that HP used the circuit/component already in 1963, while Fluke seems to use it first time around 1971. The circuits are so similar that it is imo unreasonable to argue for any originality in the Fluke 731A.

So may be it could be an idea for US based volt-nuts to contact Fluke and ask them to "free" the LTFLU-1.

Good luck with that!  Fluke was purchased by the evil Danaher corporation (along with Keithley and Tektronix)-- they pretty much fire everyone over the age of 30 [thus losing the people that actually know *why* things were done in the past], then they cancel anything that isn't making huge profits, and then milk their "captive customers" for everything that they are worth.  I seriously doubt if they are going to budge even one inch in giving us access to *anything*.  But you can *try*... Let us know how that goes!

I sure will    Nothing to lose...

I do however not expect much. After seeing the 6-in-1 launch timer/clock running negative overtime far to long after count-down, I sent an email to Tek's Quality Dept. The timer was taken away a few hours later. But they didn't thank me - they are now repeatedly filling my email inbox with US located seats for some post-launch sessions.

But I have seen a video on YT with Fluke's Chief Corporate Metrologist Mr. Jeff C. Gust. If it is not all circus for the masses, I suppose he will at least answer my mail.

I'll be back in due time