Ultra Precision Reference LTZ1000

[branadic]

How are the Edwin's resistors manufactured? Any details on that? Is the resistive wire enamel coated? Bifilar winding (low L)? For such a resistance like 70k/12k you would need a pretty long and thin wire to wind on that small core, imho..

Have a read here.

-branadic-

[Andreas]

did anyone have load life specs for his resistors? xx ppm yy C @ zzzz hrs 90min on 30min off ?

the datasheet is here:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg463479/#msg463479

with best regards

Andreas

[hwj-d]

...
Handled them all with maximum care and bent them with a special tool but one of them is really dead.

Hm, those nice views. 

[MegaVolt]

Inspired by the boards of branadic and Dr. Franks schematic i made my own board, adding a double RC-filter at the input of the LTC1052
and move the power supply to a separate board (not built yet).

Can you share the circuit on which the printed circuit board is made. Very interesting wiring.

[F64098]

I have no own schematic, just used Dr. Frank's schematic and added the two R/C-filters.
The two OPs got also their own R/C-filters from 10Ohms/10µF/100nF for their power supplies.

Best regards

Frank

[Qmavam]

A couple days ago I got up at 2:45am to make sure I was awake to bid at 3:17am on an LTZ1000ACH.
When I saw it the price was $1.25, When I got up it had increased to $3.25.
With 15 seconds left I bid $4.00, I was over bid at $4.50, But I had time to put in a $5.00 bid.
It sold for $5.50.
 Boy, do I feel stupid.

 It was coming from China, don't know if it was any good.
But I would have paid $6.00 to find out! :-)
  Mikek

PS. I don't even know what I would with it!

[Magnificent Bastard]

A couple days ago I got up at 2:45am to make sure I was awake to bid at 3:17am on an LTZ1000ACH.
When I saw it the price was $1.25, When I got up it had increased to $3.25.
With 15 seconds left I bid $4.00, I was over bid at $4.50, But I had time to put in a $5.00 bid.
It sold for $5.50.
 Boy, do I feel stupid.

 It was coming from China, don't know if it was any good.
But I would have paid $6.00 to find out! :-)
  Mikek

PS. I don't even know what I would with it!

I think someone did you a favor.  REAL {and in-spec} LTZs go for a lot more than $6.  This was probably a FAKE device.  If you want a REAL LTZ, get some samples from analog devices.

[maginnovision]

Yea, if you're getting one for $6 you could be VERY lucky. I'd suspect you're getting a total fake, or a damaged one instead.

[feedback.loop]

Let's get the logic straight. It was an auction. The fact that it ended with $5.50 does not mean that the part was fake. It may be a fake, but using the end price as a confirmation is not logical.

[GigaJoe]

100% fake ...
auction's  image:
https://pasteboard.co/IuyosXD.jpg
pin orientation,  fonts, printing ...

[ckocagil]

Does anyone know what happened to the cheap LTZ1000s from China? I searched all over taobao but can't find any cheaper than $45-50 which is nearly what you'd pay for a brand new one. Did this well dry up?

[3roomlab]

a 2005 report about investigation of vref noise in metrology zeners and JVS (using allan variance)
zeners tested Motorolas, Linears LTZ, waveteks
and motorolas have lower noise floor?

[chuckb]

Just my view of this article -

Fifteen years ago I was talking to a participant of the North American Intercomparison of the 10V standards. Every few years Fluke F732B standards were shipped all over the country, each lab determines what they think the voltage is. Those results are sent to the pivot lab for analysis. The four GOLDEN Fluke 732Bs used in the intercomparison were probably the same F732Bs with Motorola chip that were used in this test. The Acknowledgments section says as much.

The comment by a participant was "those four F732Bs are the best that Fluke ever made". While I'm sure that the results of Dr Witt and Dr Tang evaluation are accurate (I have met both of the gentlemen and toured their labs) they may not be representative of real world production parts.

BR

[dietert1]

Can you explain what you mean by "not real world production units"? Are they cheating like using the four references as one? When they write "Motorola", it isn't Motorola? Or somebody outside Fluke reworked the devices?

Anyway it appears like even ultra precision references are not the real winners. I mean 3 ppb white noise at 1000 s sampling 15 years ago leaves a lot of space. Would like to know what the best ones a capable of nowadays.

Regards, Dieter

[Kleinstein]

If I understand the article right the data in the histogram are for the low frequency limit due to 1/f noise, so the very low frequency noise  (e.g.  < 0.01 Hz), or the minimum / asymptotic value for the Allan variance. This is not just comparing the pure Zener references, but the full 10 V references: the Motorola ref. (SZA263) are older Fluke 732B, the Linear ref. should be newer Fluke 732B with LTFLU and the wavetec refes should be LTZ1000 based.
Especially the lowest frequency noise can very well also depend on the quality of the temperature control and 7 to 10 V circuit. There is also a good chance that the refs used are not just of the shelf ones, but possibly selected ones. I would more consider it a comparison the the 10 V refs., not just the zeners.

The higher frequency noise (e.g. 0.1 -10 Hz) could be different - in the article the sample LTFLU had lower white noise but more 1/f noise than the SZA263 sample.

[RandallMcRee]

Can you explain what you mean by "not real world production units"? Are they cheating like using the four references as one? When they write "Motorola", it isn't Motorola? Or somebody outside Fluke reworked the devices?

. . .
Regards, Dieter

I'm not chuckb but what he says seems pretty clear: "not production units" is referring to the fact that those four went through a special selection process. Hence, as we all know from statistics class *not* randomly selected and thus not representative of a sample. E.g. to spell it out I can't buy one from digikey and be guaranteed to get that result. But maybe I could buy 100, and put them through a selection process, etc.

[chuckb]

Can you explain what you mean by "not real world production units"? Are they cheating like using the four references as one? When they write "Motorola", it isn't Motorola? Or somebody outside Fluke reworked the devices?

Anyway it appears like even ultra precision references are not the real winners. I mean 3 ppb white noise at 1000 s sampling 15 years ago leaves a lot of space. Would like to know what the best ones a capable of nowadays.

Regards, Dieter

The impression I had was that Fluke selected the very best units from the normal production for their own internal use and for special tests. So these are probably the best 0.1% of units produced. Some parameter like flicker noise might be 2x better than normal for some units. Of course some parameters (long term drift) can not be determined accurately at the factory during the few months they have them. There are probably some production units that are just as good as the units in the article. It's just statistics.

The Allan Deviations of my five Fluke 732A units are around 40nV for 100-1000 sec (Morotola SZA263). 

[dietert1]

Zlymex showed a table of voltage reference specs, where the 732B had 0.06 ppm noise as compared to the 732A with 0.1 ppm noise.
https://www.eevblog.com/forum/metrology/teardown-voltage-standards/msg902849/#msg902849
When you start from the 732A Allan deviations and calculate 40 nV * 0.6 then you roughly get the numbers of the paper.

In Zlymex's table the 732B is listed as "LTFLU-1" while in the paper they say "five 732B instruments fitted with Motorola reference/amplifiers". Does this mean the noise difference is not attributable to the references M or L? Somehow i don't get it.

Regards, Dieter

[SilverSolder]

If a lot of the reference noise is white noise...   would it not be reduced by using a filter of some kind on the reference?

[Kleinstein]

Zlymex showed a table of voltage reference specs, where the 732B had 0.06 ppm noise as compared to the 732A with 0.1 ppm noise.
https://www.eevblog.com/forum/metrology/teardown-voltage-standards/msg902849/#msg902849
When you start from the 732A Allan deviations and calculate 40 nV * 0.6 then you roughly get the numbers of the paper.

In Zlymex's table the 732B is listed as "LTFLU-1" while in the paper they say "five 732B instruments fitted with Motorola reference/amplifiers". Does this mean the noise difference is not attributable to the references M or L? Somehow i don't get it.

Regards, Dieter

AFAIK there are older 732B with Motorola Ref. and newer 732B with LTFLU. Its also possible there was a missprinp and the Motorola units were actually 732A units.
The SZA263 is long obsolete - likely the reason Fluke got the LTFLU as a replacement.
Noise is not the primary concern, it is more about the long term drift.

Noise filtering  is tricky, as at the very low level capacitor leakage, DA and temperature effects can cause more trouble than good at the very low frequencies. In many uses a DMM or similar would be used that does low pass filtering anyway. In many cases one would case about noise at some .01-.1 Hz  Though even if only white noise, filtering would be tricky, with huge caps (100s of µF of polypropylene (better PTFE) caps).

[truser]

Bright thoughts to all!

Have the technology stopped improving over the past 20+ years or did something more accurate than the great-grandfather LTZ1000 (which have already been discontinued (for unknown reasons) from production) or so?

And...

Maybe someone will advise modern amplifiers (instead of LT1013) for a reference voltage source?

[Kleinstein]

So far there is no newer replacement for the LTZ1000.

The LT1013 is still be best bet for the reference circuit. The LT1013 is well good enough and not too expensive  - no need for experiments there.

[razvan784]

Where are you getting these rumors that the LTZ1000 is discontinued?
It clearly states "In Production" on the Analog Devices (which bought Linear Technology) website: https://www.analog.com/en/products/ltz1000.html
It may be old, but there is no real need for a replacement -- it's a very good design, with no obvious flaws, and it's a Zener after all, you can't really improve much on basic physical phenomena.
There are currently only two ultra-stable, ultra-low-noise references in integrated circuit form: the LTZ1000 and the LTFLU1, the latter being a custom design for Fluke (now also Keithley, they are under the same parent corporation) and not available for general sale. And as I said, there is really no need for more, there is not much one can improve about them.
You can do the math on how much the LT1013 offset voltage, bias current, open-loop gain and noise affect the reference voltage and you will probably find it's also adequate for the job.

[iMo]

Is there any comparison on the performance of the LTZ1000 made in say 80ties/90ties and the latest silicon (especially after the merge LT+ADI)?

[branadic]

Is there any comparison on the performance of the LTZ1000 made in say 80ties/90ties and the latest silicon (especially after the merge LT+ADI)?

Well, more or less, but still ongoing: https://www.eevblog.com/forum/metrology/vintage-ltz1000-from-1986-nib-!/

-branadic-