Ultra Precision Reference LTZ1000

[Dr. Frank]

But what would be the best operation point for LTZ1000A then?
Asking in practical reason, as have A-based reference in design as well.

Well, 15-20°C above max. ambient temperature, therefore 50-60°C stabilization temperature (12k5 - 13k over 1k).
If you can guarantee metrology grade environmental conditions, i.e. 18-28°C max. all the time,  45°C,  equivalent to 12k over 1k, is tightly sufficcient. But pay attention to sample variation of the LTZ, as +/-5°C can happen easily.

For my references (non A version), running on a nominal  45°C +/-5°C, at 23°C max. RT, the regulation headroom is >20°C, which is safe under all conditions.


Perhaps it is necessary to ship the reference for comparison at a fellow volt-nut, then the expected temperature there has to be taken into account.

Btw.: I cannot understand why people still buy the A-version, even directly from LT.
It is much more expensive (12$) and less suited for high stability.

Frank

[TiN]

Ok, thanks.
I plan to have 40°C as max ambient, so this gives 55°C as reference temperature.

I bought A, because of experience lack, and A-version opamps from LT usually better specs (per datasheets), so thought same regarding zener, lol.
It's still first zener I got, so $12 not a big deal for knowledge.

And actually i still did not decided if I'll have it running battery-backed either. Need check proper battery size/weight to keep reference hot during shipping (I'd take 2 weeks for it, as average).

[iTist]

Hi,

has someone experience with the technique described in this Patent:

http://www.google.de/patents/US5369245

Greetz

Oliver

[Dr. Frank]

Ok, thanks.
I plan to have 40°C as max ambient, so this gives 55°C as reference temperature.

....

And actually i still did not decided if I'll have it running battery-backed either. Need check proper battery size/weight to keep reference hot during shipping (I'd take 2 weeks for it, as average).

For 40°C ambient, better use 13k/1k, i.e. 65°C.
I pimped my HP3458As reference to that value, because a dirty fan filter will rise the interior from 34°C to 40°C, max. (at 23°C rooom temperature.
This still would give 1ppm/yr. as default.

...

Batteries will give less output noise / disturbance compared to AC mains supply.
The LTZ itself does not need permanent operation; stability is <1ppm with intermittent operation.

Possibly, resistance based dividers/amplifiers (to 10.0000V) may need constant powering for achieving <  1ppm stability.

Frank

[Dr. Frank]

Hi,

has someone experience with the technique described in this Patent:

http://www.google.de/patents/US5369245

Greetz

Oliver

That's the well known Pickering patent used in the Wavetronik/Fluke 7000 reference, for conditioning the LTZ1000 after power down.

I have collected some experience: I did several tests concerning temperature cycling (refrigerators / heater), but not using that schematic.

This patent at first indicates, that temperature excursions of the LTZ induce hysteresis, and if this hysteresis is not removed, by symmetric cycling, it will  in turn induce creeping drift.

From that it directly follows, that any "burn-In" process is bad for the LTZs stability!

I have seen hysteresis of several ppm on both of my LTZ references, but only if the temperature deviation from the nominal value is > +/-25°C. (I accidentially heated one of the LTZs to >90°C.)

If the temp. excursion is less, there is no noteworthy hysteresis, perhaps a few tenths of ppm only.

That means, at 45°C stabilization temperature, as in the Fluke 7000, the conditioning does not improve uncertainty that much.. The reference at this temperature is already very stable without further measures, also in intermittent operation.

I measure that on the HP3458A, each time I power it on again, and two times already in the last 4 years, when I removed my constantly powered LTZ references from the mains for several weeks. Each of those 3 LTZs return to their default value to < 1ppm deviation, every time. Exactly the same goes for the Fluke 5442A calibrator, which uses 2 SZA263.

Only if very high (>70°C) or very low (<15°C) ambient temperature excursions  are encountered, this feature might theoretically have some effect...

But then, as the lower conditioning temperature is limited to room temperature, i.e. 20°C, the complete or symmetric conditioning circle, as described, can not be passed through. 0°C would be necessary, to reset a trip to 90°C, as shown in the  patent in fig. 2.

Therefore, this patent is nice for theory, but did obviously NOT work in the Fluke 7000.
Perhaps this was one of several reasons, why Fluke terminated that product.

Frank

[SimonSatCom]

Hi all

It is a nice little project but it has not been easy. Information about both the LTZ1000 and the LTZ1000A is pretty
hard to come by. I do think that Linear Tech could do better on they datasheet.

First about the temperature set point. Again...I have not been able to find any formula so that I could calculate the
exact set point. A collegue of mine figured out that my setpoint is about 46.5C. I did not have a 13k000 precision
resistor...but I did have a 12k100 resistor. In the article "The Ultra-Zener...is it a portable replacement for the
Weston cell?" I was able to see what happened if the value of R4 is changed. I understand now that this is the LTZ1000
(not the A-version) but since no information was available I decided to try it out. In the document it is claimed that
the aging (ppm/year) will be lower with a lower temperature setting (which makes sense). This leads me to the next
question about the temperature in my lab.

My lab is located in the basement of my building. The temperature in the basement varies from 15C to 20C and the plan
with this reference project is to ovenized it the entire thing. The idea is to set the temperature to 35C since I can
not see the temperature rise to more then that (its always cold in Denmark). In the LTZ1000 and LTZ1000A datasheet I
did not find that much help only that the A-version is set 10C higher then the "not A-version".

And Dr. Frank thinks that I made a mistake by using the A-version...could you specify why that is? I got the A-version
as a sample from Linear Tech and thought the A-version was an improvement over the LTZ1000. But you don't seem to
think so...why is that?!

Do any of you have the formulas used to calculate your LTZ1000 projects? It seems there is a lot of discussions about
the correct settings for all the precision values. How did you (Dr. Frank) reach the value for your project? And do
you have a lot of experience with the LTZ1000?

Best regards

Simon

[Galaxyrise]

Do any of you have the formulas used to calculate your LTZ1000 projects?

From page 5 of the datasheet:
"With the values given in the applications, temperature is normally 60°C. This provides 15°C of margin above a maximum ambient of 45°C, for example. Production variations in emitter-base voltage will typically cause about ±10°C variation. Since
the emitter-base voltage changes about 2mV/°C and is very predictable, other temperatures are easily set."

The values given in the datasheet for R4:R5 are 13k:1k across the ~7 output of the reference.  You then have to know that Vbe has a negative temperature coefficient.  From that, I determine that T = 310 - 7*500*R5/(R4+R5)

And Dr. Frank thinks that I made a mistake by using the A-version...could you specify why that is? I got the A-version
as a sample from Linear Tech and thought the A-version was an improvement over the LTZ1000. But you don't seem to
think so...why is that?!

From the electrical characteristics section of the datasheet, we see the only difference between the two is the thermal resistance.  80C/W for the LTZ1000 and 400C/W for the LTZ1000A. And on page 5,
"Because higher temperatures accelerate aging and decrease long-term stability, the lowest temperature consistent with the operating environment should be used. The LTZ1000A should be set about 10°C higher than the LTZ1000. This is because normal operating power dissipation in the LTZ1000A causes a temperature rise of about 10°C."

So the LTZ1000A runs hotter in most applications making it less stable.  So why have the LTZ1000A at all? The datasheet says "To simplify thermal insulation".  In other words, so you don't have to put it in a temperature controlled box, and it will stay warm with very little power.  If you put a LTZ1000 in an oven which is 10C above ambient, then it will run at the same temperature as an LTZ1000A sitting outside the oven, with potentially similar performance.  If you're going to go through the trouble of ovenizing, then the non-A is preferable.

[SimonSatCom]

Ahh...okay...now I get it. So I should consider getting the non-A version.

The choice of 12k100 works fine at the moment...but would it make sense to increase it to 14k000 (which I have in the resistor box)...?

[Dr. Frank]

Ahh...okay...now I get it. So I should consider getting the non-A version.

The choice of 12k100 works fine at the moment...but would it make sense to increase it to 14k000 (which I have in the resistor box)...?

Definitely not 14k! That would give 75-80°C, and higher annual drift
12,1k is ok, depending on your envirnonmental condition. Simply calculate.
Perhaps you got another 400Ohm resistor to add for 50°C.
Afaik, you got low TC resistors, wirewound, n'est pas?
And perhaps, you might heat the whole basement to 20-21°C during winter (as I do),

As you have received the LTZ1000A for free, it's definitely no fault, but you have to live with it.

It gives a fine reference also, totally sufficient for ones needs.

Frank

PS: I think I have some experience, as I published the theoretical background and made solid measurements on nearly all of them.

[SimonSatCom]

Hi

Okay, I will see what I can do. But I think that I will try to get the LTZ1000 version instead of the A-version. I will of course stay away from the 14k000. One of the reasons for ovenizing the whole circuit is because the precision resistors I use are "only" 2ppm resistors. So I figured that keeping the entire circuit (LTZ reference and 7V-to-10V amplifier circuit) at a constant temperature would be a good idea.

If I switch to the non-A-version I guess it would not be a problem with the 12k100 when the complete circuit would be inside a 35C environment. Or what do you think?

Thanks for clearing things up. I had a hard time finding the information I really needed.

Btw...do you have a homepage with more LTZ info?

[Dr. Frank]

Hi

.. One of the reasons for ovenizing the whole circuit is because the precision resistors I use are "only" 2ppm resistors. So I figured that keeping the entire circuit (LTZ reference and 7V-to-10V amplifier circuit) at a constant temperature would be a good idea.

If I switch to the non-A-version I guess it would not be a problem with the 12k100 when the complete circuit would be inside a 35C environment. Or what do you think?

Thanks for clearing things up. I had a hard time finding the information I really needed.

Btw...do you have a homepage with more LTZ info?

Well, you just should start READING (RTFM).

It's all inside the datasheet, or within in this blog. No, I don't have a personal HP.

12k or 12k1 would be perfect for an LTZ1000.

If you avoid that external oven @35°C, and keep the reference below 30°C, 12k1 is also fine for the LTZ1000A.

What the heck do you mean with resistors, 2ppm???

2ppm of tolerance or 2ppm/yr. or - if assume right - you mean 2ppm/K wirewound resistors?

That is just perfect! That's top notch, not beaten even by those miraculous Vishay metal foil resistors.
I use 3ppm/K econistors, and the circuit is rock stable.

Why?

Once again, please simply read that f****d specification of LT!!!

All instabilities or variations are attenuated by a factor of 100 at least!!

A 2ppm/K drift on any resistor will translate into 0.02ppm/K drift of the reference, only.
A 20ppm/year (which is typical for such wirewounds) will translate into 0.2ppm/year for the reference.

Both are phantastic values!!

Therefore, an additional oven is not necessary, if you feel ok with that level of stability.


But somehow, I just feel if I repeat what I already have stated earlier in this thread...

Please, begin to READ this thread thoroughly.

Frank

[Galaxyrise]

For my references (non A version), running on a nominal  45°C +/-5°C, at 23°C max. RT, the regulation headroom is >20°C, which is safe under all conditions.

While we're on the subject, how'd you get the +/-10°C mentioned in the datasheet down to +/-5°C?  I still haven't decided how I'm ultimately going to thermally stabilize the reference, but most of my ideas involved measuring Vbe in situ and computing a resistance from that rather than choosing a resistance to target a temperature.  That +/-10°C was just too big a spread.

[babysitter]

Maybe we should split this thread into a section with fairy dust and one without

best regards,
the double babysitter
(cured from ltz-esotherics) 

[Andreas]

the double babysitter

[off topic]
congratulations! and my best wishes for your growing family.
[/off topic]

best regards

Andreas

[babysitter]

Thank you Andreas ! 

[Dr. Frank]

While we're on the subject, how'd you get the +/-10°C mentioned in the datasheet down to +/-5°C?  I still haven't decided how I'm ultimately going to thermally stabilize the reference, but most of my ideas involved measuring Vbe in situ and computing a resistance from that rather than choosing a resistance to target a temperature.  That +/-10°C was just too big a spread.

Well yes, I didn't remember the max. variation in the datasheet correctly..

I also measured the VBE at room temperature carefully, i.e. with small current and without heating, and then estimated the stabilization temperature. It was very close to those 45°C with 12k/1k.

Frank

[CaptnYellowShirt]

Dr. Frank: Could you detect the AfE tower demolition in the output of your LTZ?   

[babysitter]

@Dilligent Minds: The fluffy alternative to foaming enclosures!  O0 but I would never be scared to start fires with such filling - I wouldnt even expect melting temperatures for polyolefines in such low power circuits, without Tantal Capacitors there is just a low risk of flames forming and (ceramics, ok, different story) and in a sleeping bag filling I would expect some flame retardants.

Also, I am the "tuner tin can guy", dont forget !  this will contain the fire if it breaks out I guess, without having analyzed the risks yet.

Hollow fiber insulation doesn't belong to the fairy dust class, as it is effective: On the northern hemisphere you can try this with a corresponding sleeping bag.

At work I just got a sample of a Lackwerke Peters Polyurethane potting material which is quite nice to work with. Good if you are confident that you never need to repair the embedded circuit, but I like the thermal mass and thermal conduction properties. No chance for even forced airflows below unpleasant levels

[Rerouter]

you could always do a low voltage spot weld to the copper if you wanted to get fancy (where you have the weld tips in contact before applying current from a super-cap or single turn output transformer) that reduces it to what ever fun the copper and the lead material cook up,

it just comes down to if this is a better or worse thermocouple

[Rerouter]

I would think as you want minimal heat loss from the reference and power planes anywhere near it would soak up way too much, where by tiny little copper traces on essentially an insulator are almost invisible to the heat-source (now mounting that thing in a shielded box might be an option if you where that concerned with noise coupling, but you would still risk it coming in on your leads,

equally star grounding and ground planes can be very bad, it really comes down to what your doing, the best way to lay out references like this is to treat every trace with resistance and every connection as a thermocouple, and pick where you connect your traces wisely, for instance the 2 input pins of your chopper amp, if there is a small temperature gradient across the board from the references heat source then you could have a few nV-uV more generated in the input pins solder joint closest to the reference and offset your reading,

[Dr. Frank]

Hello folks,
                 attached the top and bottomview of the voltage reference circuit of the ultimate HP/Agilent 3458A - 8 1/2 digit DMM. What struck me observing the approach of the PCBs is the lackof a solid ground plane, short - wide power rails, wide power tracks, star grounding - any thoughts about improving the overall quality of these PCBs - would a ML-4 or ML-6 layer approach contribute to this ?

Thanks !

This reference is one of the weak components of the 3458A.

HP ignored several design rules to make it more stable.
One of those rules (see LT datasheet for the LTZ1000), is the thermal shielding of all solder junctions, another one - having supreme negative impact - is keeping the stabilization temperature at  65°C or lower.

Some simple improvments on this reference, for metrological use:

- Reduce Tstab. from 95°C to < 65°C. Simply replace the 15k metal foil by 13k (or parallel it with 100k)
- Replace LTZ1000A by LTZ1000 and set Tstab to 45-50°C (12k-12.5k)
- Replace both 70k resistors by metal foil or wirewound types.
- Use instrument at Tamb < 30°C, i.e. at 23°C +/- 5°C only and keep fan always clean => max. internal temperature of < 38°C

=> will give <= 1ppm/yr. DCV stability.

- Put complete reference PCB into a closed box to avoid any air draught from the rear fan.

=> willl reduce medium term random noise

Any further measures are not necessary!
The original LT circuitry, as it is also realized on this PCB, is stable enough to achieve an ultra stable reference.

The reference is prone to external AC noise, which might also cause random unstabilities.
But obviously this is no problem inside the HP3458A, as the PSU and shielding measures around the DC conditioning PCB is extremely good. Therefore, any additional layers or star points, or whatsoever is also not necessary.

Frank

[Mickle T.]

Two of the high-end digital multimeters - Prema 6047 and 6048 was designed in the 1988. Both was build on the Prema's ASIC ADC chip (but not a well-known PRI5610). Model 6047 have a LM399 voltage reference, 6048 - LTZ1000, but in DIP8! Is it possible? 

[JBeale]

Well, it is possible to bend the 8 leads of a TO-5 or TO-99 can to fit into the through-hole locations of a DIP-8 footprint...

[Andreas]

Any further measures are not necessary!

Is there no calibration after changes necessary?
My LTZ1000A have around 50ppm/K drift when changing the temperature setpoint.

- LTZ1000, but in DIP8! Is it possible?

I do not know any datasheet which shows a DIP8 package.
At that time many manufacturers used plastic spacers which spreaded the TO-99 pins into a DIP8-pattern on the PCB side.

With best regards

Andreas

[Galaxyrise]

Last year, Linear Tech introduced the LTC2057(HV), which is a zero-drift op-amp that also has very low noise-- even lower than the LT1013.  There is simply no excuse anyone can make for using anything other than two LTC2057's in this circuit if it were redesigned today.

It's nigh impossible to find new parts in hermetic packages any more.  I can find some CERDIP 1013s on ebay, though.  Are these opamps humidity sensitive at all?  "For humid environments, surface coating may be necessary to provide a moisture barrier." is the only datasheet reference I can find to it (in the 2057 datasheet.)