The LTFLU (aka SZA263) reference zener diode circuit

[Dr. Frank]

Hi,

A little more research.

I have attached a pdf with the Fluke 431B schematic. It uses a reference similar or the same as the LTFLU.

Here is the reference section from the 335D voltage standard:

...

The manual says that R13 is selected to set the tempco. R9 trims the output voltage. This gives a clue as to how the LTFLU chips should be trimmed.

This is a heated reference.
The heater is on the die, similar to a LM399. The heater is fed with constant voltage and the heater is a non-linear PTC that controls the temperature. A1 is a 723 voltage regulator.

Jay_Diddy_B

That is not correct.

The 332/335 reference amplifier from T.I. had no internal heater.

The 4 pole package was externally temperature stabilized by a KLIXON oven.

Assembly can be found in this thread:

https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg393627/#msg393627

There you can also find a T.C. estimation of the reference amplifier w/o oven stabilization, i.e. about 2ppm/K.

Frank

[Dr. Frank]

This is the 8840A from 1987 He never been calibrated after 1987.

So it will definitively go wrong: 1990 the definition of Volt has changed by 9.264 ppm.
http://www.nist.gov/calibrations/upload/tn1263.pdf

With best regards

Andreas

It depends, where this instrument was bought or manufactured.

Afaik, the change of the S.I. Volt in 1990 in Germany / Europe was much smaller, than in U.S.


Frank

[Mickle T.]

Fluke 732B

[janaf]

Fluke 732B

You got a box full?!?

[Mickle T.]

All of the photos above was taken from bbs.38hot.net

[Jay_Diddy_B]

Hi group,

What happen in 1970 that caused a shift in the Volt?




This graph shows the uncertainty between various government labs. It would be kind of interesting to plot the typical uncertainty found in home labs over the same period.

In my own case it would be some thing like this:

Early 1970 around 2% analog meter. 20,000 ppm

1979 bought a Fluke 8050A, from Fluke, 0.03% 300 ppm

1990 HP3455A / 3456A  25ppm

2004 Fluke 732A  6ppm

2008 Datron 1281


Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

I found my 8842A parts unit. This unit had the Motorola SZA263 reference. Here is a photo of Z701 resistor network:



I removed the resistor network. Between pins 1 & 2 I measured 13.4569 k Ohms and between pins 2 & 3 I measured 0.99855 k Ohms.

Regards,

Jay_Diddy_B

[Dr. Frank]

Hi group,

What happen in 1970 that caused a shift in the Volt?




Regards,

Jay_Diddy_B

That's no volt shift, that's a redefinition of the S.I. volt, and a different mise en pratique.

Until 1972, the national volts were defined and realized by Weston cells, each having a nominal value of about 1.018638 Volt.
Afaik, there also was no precise definition of the Ohm at that time, and the Ampère definition also was not so good to be realized, so the S.I. volt was very unprecise in terms of the base units.

As a consequence, each of the different metrology  institutes had their own volt, (because the chemistry of every Weston cell always is a bit different, but the assigned value was always assumed to be the nominal one), and all differed and varied , as it is obvious in the curve.

In 1972, the single junction Josephson volt delivered an independent definition and realization, only based on the definition of h, e and the second.
Latter constants of nature were already relatively precisely known, and so was the definition, then. Anyhow, in 1972 the responsible BIPM conference probably had chosen the wrong value for the Josephson constant, off by about -7ppm, or so.

But as a single junction delivered a few mV only, the realization (mise en pratique) was also not very precise, on the order of about  10nV / 2mV ~ 5ppm, I would bet.

You can easily see, that consecutive measurements in one metrology institute already varied by several ppm.

Until 1990, the JJ array (10V) had been developed, which greatly increased the uncertainty of the realization of the volt  (factor of about 5000 better), and also e and h meanwhile had been determined more precisely.

So in 1990, the redefinition gave a much more precise value for the Josephson constant, a much more precise mise en pratique, and just recently, then, the von Klitzing Hall effect delivered the quantum Ohm. So the realization of Volt and Ohm was uncertain to better than 1e-9, but within the S.I. both were still unprecise to around 0.2 ..0.3 ppm, and that's still the situation of today.

The next step will probably happen in 2018. Then, Volt and Ohm may get exact values in the S.I., i.e. zero uncertainty for the Josephson and Klitzing constants, as then e and h will be 'exact', as is already the second.

Frank

 

[Jay_Diddy_B]

Dr. Frank,

Thank you for your explanation. It is very clear. How much of shift do you think we will see in 2018?

Regards,

Jay_Diddy_B

[Dr. Frank]

Dr. Frank,

Thank you for your explanation. It is very clear. How much of shift do you think we will see in 2018?

Regards,

Jay_Diddy_B

Well, less than these ominous 0.2 or 0.3 ppm.

According to a recent publication by our German PTB, which obviously includes the latest intermediate measurement results from the Watt balance and the Si sphere, they expect a jump of about 0.02 ppm for the Ohm, and about 0.1 ppm for the volt.

Frank

[Jay_Diddy_B]

Hi,

Dr. Frank,

Thank you for your explanation. It is very clear. How much of shift do you think we will see in 2018?

Regards,

Jay_Diddy_B

Well, less than these ominous 0.2 or 0.3 ppm.

According to a recent publication by our German PTB, which obviously includes the latest intermediate measurement results from the Watt balance and the Si sphere, they expect a jump of about 0.02 ppm for the Ohm, and about 0.1 ppm for the volt.

Frank

I think I will put my reference building project on hold until we know what the right value is. 

Jay

[Dr. Frank]

I think I will put my reference building project on hold until we know what the right value is. 

Jay

No, the correct consequence would be, to improve all references of the different projects here, in the next 4 years to a level of 1e-9 stability.

Currently, 0.3ppm is good enough, but then, that will be the new uncertainty limit.

Frank

[janaf]

While I like the attitude I can not see how the current reference projects could be improved by orders of magnitude.

It think it would take radical new approaches, new components, new paths. And we still have the low frequency noise floor to cope with.

No, the correct consequence would be, to improve all references of the different projects here, in the next 4 years to a level of 1e-9 stability.

Currently, 0.3ppm is good enough, but then, that will be the new uncertainty limit.

Frank

[janaf]

I received some used LTFLU-1TH in the mail today.

So now, I'll find out if I burned money or not.

- Six of them, all 4-pin.
- Date codes? 89 and 90
- One looks externally bad,
- Four others are used but appear decent,
- One looks new, never soldered, leads un-cut.

Does anyone have the pin-out? 

[Dr. Frank]

I received some used LTFLU-1TH in the mail today.

So now, I'll find out if I burned money or not.

- Six of them, all 4-pin.
- Date codes? 89 and 90
- One looks externally bad,
- Four others are used but appear decent,
- One looks new, never soldered, leads un-cut.

Does anyone have the pin-out?

As in the 732B both the SZA263 and this LTFLU have been assembled in the same place, it must be the same pinout as the SZA263.

Therefore, grab a 732A, or 731A, or a 5200A service manual and search there for the pinout.

Frank

Here's a snapshot from the Fluke 5720A calibrator, which also uses two LTFLU as reference, U6, U7.

The pinout should look like this, as identified from a photo of the 5442A.

[janaf]

I looked in all manuals I can find but none give the pin numbers, only generic symbols.

I think I'll measure, using 1uA current limiting.....

[janaf]

Dr Frank; Ah, yes, the Fluke 5720A schematic snapshot you posted has the pin numbers! Thanks

[Dr. Frank]

Mickle T. has one photo, showing the pinning.

My CAD drawing is correct, therefore.

Frank

[janaf]

I think what Frank meant, correct me if I'm wrong, was that there is a need to develop non-JJ, every-day-usable standards, transports, calibrators, that are significantly better than now.

What I meant was that break-trough performance leaps will take much more than tweaking, even the most brilliantly built reference from single users will not be enough. There has to be some break-through scientific development.

In the mean time, I bought some used LTFLU from ebay 

[janaf]

Back on subject.

I did some measurements on the LTFLU ICs. I've so far measured the zener only, at room temperature only.

There is one big difference compared to the LTZ1000 zener: In the normal current range of a few mA for the LTZ1000, I measure a resistance of approximately 33ohm. For the LTFLU it's around 7ohm at current of 3-10mA!

The zener voltage of the LTFLU is a bit lower than the LTZ1000's, around 6.3V vs 6.5V for the zener only.

I attach a file. Comments welcome!

 

[Dr. Frank]

I think what Frank meant, correct me if I'm wrong, was that there is a need to develop non-JJ, every-day-usable standards, transports, calibrators, that are significantly better than now.

What I meant was that break-trough performance leaps will take much more than tweaking, even the most brilliantly built reference from single users will not be enough. There has to be some break-through scientific development.

In the mean time, I bought some used LTFLU from ebay 

Yes, that's exactly, what I wanted to focus on..
The uncertainty difference between the upcoming exact definition by JJ array and the usable standards is much too big..

I doubt that zener references might be improved by one or two orders of magnitude..

I have read about MEMS voltage references, which have a totally different working principle..
and might be an alternative..See here, for example:
http://lib.tkk.fi/Diss/2006/isbn9513868605/article4.pdf

Well, and JJ arrays on liq. Nitrogen temperature also have been demonstrated.. about 1e-8 uncertainty..See here:
http://juwel.fz-juelich.de:8080/dspace/bitstream/2128/2069/1/19406.pdf

I think, a compact JJ array, at much lower price would be possible for everyone..

Like in the Rb atomic clocks, the microwave part of the circuitry might be cheaply be realized, also a simple current controller, and so on..
Only the superconducting array would be the really expensive part.. either on Nb base (with He4 cooling) or on 77K with High Tc superconductors (which would be not as stable as a Nb system).

Frank

[janaf]

Thanks for the feedback and the info is really interesting!

What your little bird tells about the zener size matches my resistance measurements, about 1/4 of the LTZ1000. I guess the the lower zener voltage of the LTFLU also indicates that. It of course means the LTFLU is less sensitive to the current in that respect, or proportionally more stable given the same current stability.  The less fun part with the LTFLU is that there are three different currents added to the zener....

Measurements; they are done in pulse bursts at 1kHz, ie one sweep of 100 points takes 0.1s. I could reverse the sequence (highest current first) or use a much longer delay to see if it makes a difference. I'll dig into the docs to see what the actual pulse length is.

If time allows I'll do measurements on the transistors in the weekend. Then I'll need to make arrangements to measure at different temperatures.

[Dr. Frank]

Back on subject.

I did some measurements on the LTFLU ICs. I've so far measured the zener only, at room temperature only.

There is one big difference compared to the LTZ1000 zener: In the normal current range of a few mA for the LTZ1000, I measure a resistance of approximately 33ohm. For the LTFLU it's around 7ohm at current of 3-10mA!

The zener voltage of the LTFLU is a bit lower than the LTZ1000's, around 6.3V vs 6.5V for the zener only.

I attach a file. Comments welcome!

It looks like you are using continuous current to measure the LTZ-A.  Because it [intentionally] has low junction to case thermal conductivity, you need to measure it with short pulses, with a rest time in between pulses to let the device cool.  The LTFLU will not be affected as much, because it is direct-mounted to it's package.  If you do this the upward slope that the LTZ curve shows will probably vanish.

A "little bird" told me that the LTFLU has 4 times the area for the buried Zener, and so you can run it at 4 times the current [resulting in half of the DC-10Hz noise that the LTZ has].  The interconnects on the LTFLU were done better, and it has much lower interconnect resistance.  The transistor in the LTFLU will actually carry the same current as the Zener, unlike the LTZ where the transistor would be destroyed if you tried to do this-- it is a small signal amp only.

There really needs to be a new super-Zener reference chip designed, but who will pay for the development?  I'm guessing a minimum of US$50K, but more money is likely needed.  We know a lot more these days than we did back when the LTFLU and LTZ were designed, and there are some new and possibly better processes that can be used; for example, copper [and/or low-TCR] interconnects, MEMS processing to eliminate die separation stress, enlarging the Zener area, etc.  These might lead to much better performance over time.

Ok, I can't decide from that info which reference is more stable, potentially..

But didn't your little bird, i.e. Mr. Dobkin, tell you recently, that the LTFLU and the LTZ1000 were very similar, especially concerning the buried zener?
I assumed, that LT used the same bipolar design cell, and only arranged the zener and the transistor in a different way? And both components drift the same direction, statistically, i.e. about -1ppm / year?

In the end, both chips really look totally different, anyhow.

Frank

[janaf]

Ok, I can't decide from that info which reference is more stable, potentially..

Frank; was that referring to what I wrote "proportionally more stable given the same current stability"? I was referring to the lower local resistance only. My mind was into "how important are each of those those expensive high stability resistors?" I have recently been measuring voltage from the LTZ1000 circuit, making small incremental changes to each resistor value. A lower zener resistance should make the external current setting resistor (R1 / 120R on the LTZ1000 datasheet shematic) less critical. But the LTFLU implementation has to be different anyway and I wasn't really going to get into that here. For the LTZ1000 circuit, my measurements pretty much show the same thing as the LTZ1000 datasheet except that R1 is less critical than given in the datasheet while R2 is more critical than the datasheet indicates. I will get back with the LTZ1000 measurement results when I have double-checked them.

[janaf]

You are right. I re-did the measurements, ramping down instead of up, doing fewer points and then the results show lower resistance for the LTZ1000AHC but for the LTFLU, the difference is marginal. Better insulation of the LTZ1000AHC is one reason, the other, the apparent difference in die size.

I attach a new file with some updated plots. 

It looks like you are using continuous current to measure the LTZ-A.  Because it [intentionally] has low junction to case thermal conductivity, you need to measure it with short pulses, with a rest time in between pulses to let the device cool.  The LTFLU will not be affected as much, because it is direct-mounted to it's package.  If you do this the upward slope that the LTZ curve shows will probably vanish.