LTZ1000 vs. LTFLU aka FLUKE vs. DATRON

[zhtoor]

-

[branadic]

(i also hear that there is/was a heater on board LTFLU which was never used)

You can find pictures of LTFLU-1 and LTFLU-1AHC here:

https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg957684/#msg957684

https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg1313201/#msg1313201

-branadic-

[Kleinstein]

It takes a lot of stress to make silicon move. The usual silicon is essentially free from dislocations (this is different from some other semiconductors like GaAs, GaN). I once bend a silicon crystal, but that was at really high temperature (600 C) and with really high stress (AFAIR some 200 MPa).

The internal temperature gradients are tiny and thus the stress levels due to gradients are also very small. The bigger problem is different thermal expansion of the die (rather low) and the metal case and glue /solder.  So there is some thermally produced mechanical stress. But this is mainly due to thermal expansion mismatch and not due temperature gradients. The main part to settle is the glue or solder used to attache the die.
Another effect happening is humidity entering / leaving the oxide layer on top of the chip.

In chip temperature gradients may limit the precision of the on die temperature regulation. The gradients depend on heater power and thus on external temperature. Still I see no good reason of not using the temperature regulation on die. Instead of replacing it with external regulation one could still consider a 2 nd layer of temperature regulation around it (and maybe more of the circuit). It this case the on die heater would run at a rather constant low power level and dampen small errors from the outer oven.

The LTZ zener itself is a bit smaller than the one in the LTFLU/SZA263 and could thus be slightly lower noise. Without a temperature regulation the TC is smaller and can even be tuned to 0 with the external resistors. So the system approach is different:
LTZ uses precision temperature regulation to get low TC, despite of non zero TC of the ref. itself.

The LTFLU system uses adjustment of the TC close to zero through external resistors. For demanding applications an additional, less critical temperature regulation is used. So this ref. can be used with or or without heater (e.g. DMM7510, Fluke 8046?) and with a heater (e.g. 732x, calibrators). The downside is it also needs the individual adjusted resistors and an external oven usually is larger and needs more power.

[TiN]

KI DMM7510 does use onboard heater for LTFLU on the reference suspended area.
All given, getting even and uniform thermal gradients in tiny LTZ die area is much easier than getting same uniform gradient over much bigger oven box/module size. Just because heat leakage issues are much smaller and better controlled.

[Kleinstein]

I see no real sense in using the LTZ internal circuit like that of the LTFLU. It stresses Q1 to its limit or above (negative VBE) and it makes temperature sensing difficult. The LTZ1000 circuit might look a little odd at first sight, but it is a good choice to make a low noise and stable 7 V. The LTFLU might have a slight advantage when it comes to directly making a stable 10  V or similar reference (with no buffered 7 V). However suitable OPs for the 7 to 10 V step are not that difficult to get anymore. This might have been an issue in the old days when the SZA263 was new -  so back than the different architecture might have been important.
Today I usually would prefer to have both the stable 7 V raw reference and scaled 10 V or whatever is needed. So drift of the scaling resistors can be checked independently with a good DMM.

I consider the LTZ1000 internal circuit to be rather good as it is. The main point I would like to be different would be a lower intrinsic TC instead of the 50 ppm/K. This way some of the resistors could be less critical. Transistor and zener current gives some room for fine adjusting the TC (e.g. some 10-20 ppm/K). However to bring the LTZ intrinsic TC close to zero this way would be at a current that is too low to be practical (< 1 µA for the transistor).

Thermal conductivity of silicon is rather high. So I don't see a big deal if the heaters are stronger or weaker at the outer ring. The main part important would be thermal gradients in the inside part so that transistor and zener temperature would be different.

[branadic]

I consider the LTZ1000 internal circuit to be rather good as it is. The main point I would like to be different would be a lower intrinsic TC instead of the 50 ppm/K. This way some of the resistors could be less critical. Transistor and zener current gives some room for fine adjusting the TC (e.g. some 10-20 ppm/K). However to bring the LTZ intrinsic TC close to zero this way would be at a current that is too low to be practical (< 1 µA for the transistor).

Thermal conductivity of silicon is rather high. So I don't see a big deal if the heaters are stronger or weaker at the outer ring. The main part important would be thermal gradients in the inside part so that transistor and zener temperature would be different.

The thing I would like is a post process of the silicon after the bipolar process. Wouldn't it be great to have the burried zener diode on an etched island with only 4 bridges running to the island? Maybe the outer wafer itself bonded to glass and some piece of glass lid on top of the silicon so the zener including to transistors are already hermetic on chipscale? What a wonderful idea.

-branadic-

[Kleinstein]

...
so LTFLU topology is superior in regards to raw tempco tuning?
....
-zia

The topology (transistor with base to the zener or transistor with emitter to zener) has not much influence to tuning the TC (the LTFLU might allow a little more in a more complicated circuit). The main tuning option is from zener and especially transistor current. So not much difference here.

The different unheated TC is likely mainly due to a slightly different zener / transistor voltage. The LTZ1000 ref voltage is a bit higher (some 200 mV AFAIR) than with the LTFLU. The LTZ is a bit on the high side to reach zero TC.  For some reason it looks like LT did not have a lot of LTZ with slightly too low a voltage (e.g. 6.9 V) and TC  - otherwise we would have likely seen some LTZ1000B.

[dr.diesel]

make our own LTZ1000EEV 

TiN has already got the patent on this, and a few prototypes hiding under his pillow.

[Kleinstein]

The adjustment is there adjust the linear TC. So after adjustment one would have a residual temperature curve that is about a parabola. The peak (likely a maximum in voltage - this is the normal case with compensated zener diodes) temperature would depend on the fine adjustment - that is what happen when a linear part is mixed with the parabola.

[Kleinstein]

In the LTZ1000 circuit adding R4 to adjust the the unheated TC will require higher quality resistors. R4 and R1 need to be very stable.  So one needs to find the right value for R4 (can be slightly different on different units) and than get a very stable resistor or that odd value. So for most cases it is not worth to compensate the TC to make it less temperature sensitive. Things would be different if the initial TC would be in the +-10 ppm/K range (and thus R4 < R1 / 20), as the R4/r1 ratio would than get less sensitive than the temperature set divider.

There still it the problem with the residual square law part and this is best coped with by a stable temperature, though no very stable temperature needed anymore.

The LTZ1000 circuit is made to run with temperature regulation and the temperature regulation can work very good. One big advantage is that this does not need an individual adjustment - the individual adjustment needs quite some time, and custom super stable resistor are also slow to get.

Having a slightly higher temperature also has some advantages: The fast relaxation processes can settle in reasonable time and thus only the slower processes would contribute to long term drift. So a lower temperature is not always better here, unless you can afford to age the circuit for a few decades before use. A year at a slightly higher temperature is more realistic.

[kj7e]

Zia,  please don't take this the wrong way, I'm just curious.  You can also tell me its none of my business and to bug off 

You have many threads on LTZ1000's, LM399's and other various precision voltage references and alternatives (Poor-Mans..., Off-label...), I'm curious what equipment you have that can make use of or measure the drift or TC of such devices?  Theory and ideas are great to explore, but do you have any LM399 or LTZ1000 based references now to play with now?  If not, I would encourage you to build one and play with it.  IMHO, building one is far more fun than talking about it.  I know you have a set of standard cells, but I think you will find even a simple LM399 will out perform them.  The LTZ can cost you more, but an LM399 can be built on the cheap.  If finances are a concern, I'm sure there are a few guys here who would contribute some parts to help you out.

I don't recall if you mentioned what gear you have to measure these devices, maybe you have but I don't recall seeing it.  This could help others in advice offered to you.

[cellularmitosis]

advantest R6581T - 8 1/2 digit.

[eurofox]

advantest R6581T - 8 1/2 digit.

[Kleinstein]

Using higher current with the LTZ1000 will slightly increase the sensitivity to resistor changes/aging. The main advantage would be slightly lower noise, but not that much. In addition there might be slightly more aging at higher current - though not for sure.

One has to find a balance between long time stability and noise. Higher current offers slightly lower noise, lower current gives slightly better long time stability.

Quite often long time stability is more important than noise. Not many application can make full use of a noise level lower than the LTZ1000 running at 4-5 mA. Not many meters have a lower noise reference.

A higher current is even more a problem for the LTZ1000A as there is more self heating, so if at all a higher current would prefer the non A Version. For just lower noise the Chinese 2DW232 (with some selection ?) would be the even better choice - it is also much cheaper, but would need external temperature stabilization.

[Kleinstein]

I don't like the quadrature correction very much. It may be possible in a digital way, but analog it tends to be drifty. Already linear TC compensation does not work all that well with the LTZ. The internal temperature regulation usually works much better and is easier.

So if you want to go unheated, with TC compensation a reference like 1N829 or 2DW232 would be a better point to start.   Especially the 2DW232 might be interesting for this, as there is access to the diode part of the compensated reference too - so one has a build in temperature sensor. In addition with many samples it is reasonable to adjust the current to get zero linear TC. However the residual square law part is difficult if done analog.

A crude temperature regulation (just an oven around the whole circuit) can be easier than square law correction and is very effective, when it comes to the small residual TC after adjusting for much of the linear part. So this would be similar to LTFLU: low (e.g. adjusted via resistor) TC and oven outside.

[Andreas]

Hello,

I also would not do that with my own samples. 

Usual specification on transistors is >= 5V.
So above 5V you cannot tell if there is unwanted leakage.
(From device to device different and maybe also temperature dependant).

with best regards

Andreas

[Andreas]

No not a single uA:

"But when you are all done with this transistor, you should (virtually) throw it away, as the Vbe and beta have been degraded by the zenering" Bob Pease.

I am also a very poor man missing a JJA (and a room for it) and a couple of 3458A.

With best regards

Andreas

[Andreas]

Hello,

Bob Pease was at National Semiconductors. (Not LT)
The citate I found when looking for "transistor degradation reverse operation"
so it is a general problem of bipolar transistors. (did not find the article).

I would like it much more if you leave off the "r" in pray   

best regards

Andreas

[eurofox]

hello friends,

question:

is there an inherent difference between LTFLU topology and LTZ topology wrt. generation of popcorn noise (telegraph noise)?
an aside question is: is the zener in an LTFLU a "burried zener"?

also see:
https://www.eevblog.com/forum/metrology/vlfn-characterization-of-reference-devices-topologies/

best regards.

-zia

I just wonder why you want to use components in a different way than the way they are design for?

Some have been developed by the best engineers on the planet, tested in real applications and commercial product for some more than one decade.

You don't think if there was a possibility to use them on a better way nobody already discover this?

In many of your post you it is the same story

[RandallMcRee]

hello friends,

seems to me that a working ref. implementation (even maybe lab-references like 732A)
*should* be implemented as an averaging parallel combination of at least 2 primary elements
like ltz/ltflu/lm399 etc. reason being that the probability of all of them going into a "popcorn" fit at the
same time may be a *lot* lower, the more the merrier. i think the same strategy should be
followed in measuring instruments for the same reason.

best regards.

-zia

The above is true, I think. But there is a more nuanced view. Rather than building a circuit which averages a bunch of references we should build a circuit capable of measuring each individually. In this digital age it is a simple matter to then compare each with one another (the essence of metrology, right?) Outliers can then be discarded if necessary. Drift calculations performed and so on and so forth. It is much more flexible.

[pigrew]

I like the idea of measuring them separately.

I think you can use a cheaper ADC by doing relative measurements between references instead of measuring the reference values themselves.

(If I have time, this world be a fun project to work on).

The other half of the coin is how to generate a low noise output. We could use a low noise reference with bad long term stability, and then do a fine correction of that with a low noise DAC (and tempco lookup tables for the various references)?

[Kosmic]

The Valhalla 2720GS (Vdc calibrator) is doing something like that. There is up to 4 boards with 2 vref per boards. I'm not sure if the vref on 1 board are averaged or if they are tracked individually.

some info on the 2720GS: https://doc.xdevs.com/doc/Valhalla/2720GS/

[pigrew]

The Valhalla 2720GS (Vdc calibrator) is doing something like that. There is up to 4 boards with 2 vref per boards. I'm not sure if the vref on 1 board are averaged or if they are tracked individually.

some info on the 2720GS: https://doc.xdevs.com/doc/Valhalla/2720GS/

Ah, prior art! The datasheet says that they individually digitize each output using a 30-bit ADC. I didn't find a text description of the circuits (only found schematics), so I don't yet know if it is doing single-ended or differential measurements between the reference and the DAC.

I hadn't considered batteries, but that does make sense as a voltage source. Personally, I was thinking of summing the LTC6655 output with a DAC output to get 10V. The LTC6655 would have the advantage over the battery of a much lower TC. I don't know how the noise level would compare between the LTC6655 (or other references) and a battery. However, with more transistors/die area in signal path of the LTC6655, perhaps the chance of RTN (popcorn noise) is increased.

What is useful to me would be to build a low voltage calibrator (10V, 1V, 0.1V) to help me synchronize my DMMs. I don't personally NEED a ultra-low-noise 10V reference, but it'd be fun to make.

There is still the trouble of going from 6.2V to 10V. It needs a highly linear ADC. Doing the differential measurements as I proposed only helps with outputting a very low noise value close to the "raw" reference voltage.

(I find myself writing too many words; I hope that they are at least a little bit interesting. I will start my own thread if I continue with the above idea).