Ultra Precision Reference LTZ1000

[Kleinstein]

The switched capacitor divider is not effected as much by the speed of the transitions of the control signal. This is different from PWM. The tricky part here are parasitic capacitance from the caps and lines to ground and so on. So humidity may change this. The other problem is charge injection.

There is also the version to go from 7 V to a nominal 10.5 V via switched capacitor circuit, and than down to 10 V with the resistiv divider. Here only the 0.5 V difference is really effected by the drift of the resistors, so its less sensitive than the direct 7 V / 10 V step by about a factor of 5. It gets even better it you start at 6.8 V or so.

[HighVoltage]

There never was any consensus on the best way to get 10V from the 7V[ish] reference output:

What a nice summary.
Thank you !

[Andreas]

There never was any consensus on the best way to get 10V from the 7V[ish] reference output:

What a nice summary.
Thank you !

Thank you Ken.

So I have only to give some hints:

My "temperature chamber" is shown in the following thread:
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg462298/#msg462298
A peltier car box with a added heater. (it is better to heat against the cooler than switching on/off the cooler).
A 30 deg C temperature range around room temperature is possible.

Up to now I have done no 7 to 10V transfer.
The latest idea for a fixed 10V reference is to use a cirquit similar to the Fluke 7000 system.
A (halfway) stable resistor divider (low noise) which gets a correction by a processor controlled DAC (12 Bit?).
I want to correct absolute voltage (drift) and the T.C. of the resistor divider with a resolution around 0.1ppm.
So the resistor divider has to be adjusted to the reference voltage within 100ppm.
I do not want to build an oven around the resistor divider since I want a battery supply.

For a variable output calibrator the idea is to use a 7V reference + LTC1043 capacitive divider * 3 / 4.
This gives around 5.2-5.4V Reference voltage for a correction loop ADC.
The main output is done by either 2 16 Bit DACs or a 24-30 Bit PWM DAC + a resistor divider for the 10V output.
The correction loop is responsible for T.C. correction and linearity.

With best regards

Andreas

[alanambrose]

Pah, I wrote a longish post yesterday that has been lost in the ether...

Thanks Andreas for the pointer to the temp chamber thread.

This is where (~centre / right - marked as 'HEATER' on CLIP pdf page attached) the 3458a LTZ board heater connection interferes with the 3458a input amplifier. Anyone make sense of that?

Regards, Alan

[Kleinstein]

The resolution of the PDF file  makes a little difficult to read all the values. The heat signal seems to have the option to be coupled back to amplifier in the higher amplification (likely 100 times) mode after quite some attenuation.
So this would add some offset and intentional drift (depending on heater signal) to the output. This is somewhat strange, as I can't see an adjustment to choose the optimal amout of correction and in addition the heater signal is not linear in temperature. So this will not compensate a linear TC, but a therm that approximately depends on the square root of 90 C - case temperature.
It also does not make sense to use this as a way to measure the heater current (e.g. during startup) for diagnostic purpose - there should be much better points to choose the input.

Llikely most of the time the relay will turn off this extra component anyway. To me this does not make much sense, as the relay is quite some efford and it will likely make temperature drift worse, not better. If I were to correct thermal dirf, I would add a simple sensor and do the corrections in the µC. The sensor for the case temperature is likely allready there.

So my best guess is this part will not be populated in many units. Possibly just a kind of joke to make the chineese think about magic. 

[lukier]

So, I *think* I found a link between the LT1088 and the LTZ1000 [probably the special die mount technique of the LTZ1000A, and the special die arrangement of a circular heater with a diode in the center].  Jim Williams may have left us a clue.  Look at the bottom of Ap-Note-22, and reference #3:  "Renez, Repus ..."

http://cds.linear.com/docs/en/application-note/an22.pdf

"Renez Repus" is "Super Zener" spelled backwards!

Great find! One has to read Jim Williams very carefully My personal favorite are footnotes saying "See note above" 

[plesa]

wow, nice finding. It is a pitty that LT1088 is obsolete.

[branadic]

So, I *think* I found a link between the LT1088 and the LTZ1000 [probably the special die mount technique of the LTZ1000A, and the special die arrangement of a circular heater with a diode in the center].  Jim Williams may have left us a clue.  Look at the bottom of Ap-Note-22, and reference #3:  "Renez, Repus ..."

http://cds.linear.com/docs/en/application-note/an22.pdf

"Renez Repus" is "Super Zener" spelled backwards!

-Ken

If you google Renez Repus you find more about that:

http://analogfootsteps.blogspot.de/2014/03/pranking-prankster-jim-williams-and.html

[Bud]

Look at the bottom of Ap-Note-22, and reference #3:  "Renez, Repus ..."

http://cds.linear.com/docs/en/application-note/an22.pdf

"Renez Repus" is "Super Zener" spelled backwards!

And it was published in "Private communication, Linear Technology"
 

[slashdev]

The test structure in the die center is not used. it is designed to offset effects described by Counts Theorem (see References).

5. Counts Theorem, propounded by L. W. Counts at the 1986 ISSCC, states that 9 != 10.

What a great Easter Egg!

[TiN]

There are few LT1088's on ebay for $40, in case someone feeling extreme... I skip for now, enough toys here.

[retrolefty]

There are few LT1088's on ebay for $40, in case someone feeling extreme... I skip for now, enough toys here.

 I saw that when I checked, several sellers all around the same price. Do you think these are pulls or NOS?

[TiN]

Seem to be NOS, pins not soldered and date code is 2001. But I think it's somewhat offtopic here.

To stay on topic, anyone have photos or know what is different on Fluke's 3458A/HFL ref compared to usual HPAK 3458A? I'm trying to make up my mind around various options of 3458A. Better resistors around LTZ?

[HighVoltage]

To stay on topic, anyone have photos or know what is different on Fluke's 3458A/HFL ref compared to usual HPAK 3458A? I'm trying to make up my mind around various options of 3458A. Better resistors around LTZ?

I have been reading about the 3458A only and I do not have one - may be in the future.
But What I have read is that the 3458A HFL has as the only difference, the lower reference temperature.
May be there is more.

[richiem]

I *think* the Fluke HFL 3458A version not only has a much tighter spec for the reference, but also has better/more robust/less thermally sensitive dividers for the 100V and 1kV ranges.

[plesa]

Several months ago I make a research related to differences between 3458A, 3458A opt. 002 and 3458A HFL.
And 3458A HFL should have better 40k Ohm resistor for Ohm AutoCal. 3458 Opt002 is using LTZ1000A and not LTZ1000.
The rest seems to be identical, maybe selected from bigger batch.

I have question about burn in of LTZ1000. Are you burn-in reference with heater on or off? For me logic was that heater should be off, otherwise zener temperature will be out of operation range window ( 125°C + 50°C in my case for LTZ1000 and for LM399 125°C+90°C)

[TiN]

So far all 3458A references I saw online are using LTZ1000A, not the usual one.

Then temperature, LM399 regulation is internal, so if you run in +125 ambient, heater would be off anyway.

As of LTZ, you have both control using heater and ambient. Modules I had were left running +90c (heater setpoint only) for 2 month, then to normal temp. Last time I measured 4 of them they were within 5 ppm of last reading (likely closer, but I had used 2001 with bit higher ambient temp).
I'll have to do new testrun after rusty 3458A back in business (i hope )

[Andreas]

I have question about burn in of LTZ1000. Are you burn-in reference with heater on or off? For me logic was that heater should be off, otherwise zener temperature will be out of operation range window ( 125°C + 50°C in my case for LTZ1000 and for LM399 125°C+90°C)

Hello,

I do not burn in my LTZ1000A references up to now. (only a normal run in phase of 6-12 month before first calibration).
After a burn in it is very likely that you get a more or less large hysteresis.
(as I experienced by accidently shorting the output and setting the heater setpoint to infinity).

By thermal cycling you can partly remove this hysteresis.
But the last ppm will be stored for at least some months.

With best regards

Andreas

[plesa]

Thanks both for input.
So I will keep my references in oven running at 125°C with heater on for next few weeks and we will see.
At the end I'm planning to make cyclyng according to Pickering patent or similar to cycling described in one document from CERN (they are burning LTZ by internal heater, I have also LM399 so it is not possible for me):
The burn-in procedure is the following:
- Put all the measurement probes on the testpoints corresponding to VB1
- Set the heater OFF by using the appropriate jumper switch
- Set the oven to 80ºC
- After the temperature of the oven is stabilised (some hours), measure and note
down VB1
- Set oven to 120ºC
- After the temperature of the oven is stabilised (some hours), measure and note
down VB1
- Set oven to 130ºC
- After the temperature of the oven is stabilised (some hours), measure and note
down VB1
- Put the heater ON and change the measurement probes to measure the value of the
current – I . This current is measured through a 10Ohm resistor.
- Set the oven to 80ºC and let the temperature of the oven stabilise (some hours).
- Adjust the current “I” through the use of the potentiometer box, to a value of
19mA.
- Change the measurement probes to measure the value of VB1, wait the necessary
time for the temperature to stabilise (if you were forced to open the oven door)
and then measure and note down VB1. This value should be within the interval
VB1@120ºC and VB1@130ºC
Adjust the heater current using the potentiometer box to have on VB1 the average
value between VB1@120ºC and VB1@130ºC.
- Measure the current on I testpoint and note it down. It should be close to 19mA
- Keep the setup at 80ºC during 3..4 weeks but pulsing the supply voltage to have
the heater + zener powered during 45minutes and unpowered during 15minutes
per hour. Each 2 days, adjust the current to the value corresponding to a VB1
which is the average value between VB1@120ºC and VB1@130ºC.

[plesa]

video related to LM399 and LTZ1000 references, it is quite long and in French:-(
https://jipihorn.wordpress.com/2015/07/15/reference-de-tension-utilime-ltz1000-comment-ca-marche/

[Gyro]

Sorry if this has already covered, I've been following as many of the 90 pages as I could over the past months.

Has anyone had significant success with the "Adjusting Temperature Coefficient in Unstabilized Applications" typical application circuit on the LTZ1000 datasheet? The one where you cycle the heater on and off at one minute intervals while tweaking the bias current for minimum TC.

I was wondering how low a TC it is possible to achieve using this method. Given that the datasheet specifically states that drift increases with high temperature, then for an ultimate goal of minimum long term drift, I'm wondering if this first order temperature compensation, maybe coupled with a thermometer (standard cell style of working) and/or basic external oven is a viable option. Minimizing Kovar lead thermocouple issues too.

It looks from the "LTZ1000 - Heater resistance - Possible fakes" thread that HP are generally thought to have done the wrong thing in using the LTZ1000A variant at higher temperature rather than the non-A at lower.

As I say, sorry if this has already been covered. 

[Dr. Frank]

Has anyone had significant success with the "Adjusting Temperature Coefficient in Unstabilized Applications" typical application circuit on the LTZ1000 datasheet? The one where you cycle the heater on and off at one minute intervals while tweaking the bias current for minimum TC.

I was wondering how low a TC it is possible to achieve using this method. Given that the datasheet specifically states that drift increases with high temperature, then for an ultimate goal of minimum long term drift, I'm wondering if this first order temperature compensation, maybe coupled with a thermometer (standard cell style of working) and/or basic external oven is a viable option. Minimizing Kovar lead thermocouple issues too.

As I say, sorry if this has already been covered. 

Hello,

no problem, I think nobody ever has given a final conclusion about that feature.
And I think, I have read every entry here!

First of all, you are mixing T.C. and long term drift.
Both characteristics have nothing in common, and must be discussed separately.

Long term drift is determined by the oven temperature, the lower, the better.
45°C gives about -1ppm/year.
High quality external resistors (5 EA) add about +/- 0.3ppm/year or less, as their impact on all drifts is attenuated by factors of 1/100.. 1/1000.


The T.C. for a stabilized solution (using the internal oven) is determined only by the external resistors ( to first order), and therefore being < 0.1ppm/°C in most cases (using < 3ppm/°C resistors)

The reference element itself has a T.C. (un-stabilized, no oven) of about 50ppm/°C.
Only if you do NOT use the oven, a trimming of this characteristics is needed /makes sense.

In the stabilized case, the oven regulation is so good, that it reduces these 50ppm/°C to  unmeasurable values, i.e. being much lower than the impact of the external resistors.

This fact and the probable complication of the trimming resistor  might explain, why nobody has put much effort in this feature.

Frank

[Alex Nikitin]

The reference element itself has a T.C. (un-stabilized, no oven) of about 50ppm/°C.

50ppm/°C at room temperature or at elevated temperature?

Cheers

Alex

[Gyro]

Thanks Frank,

Glad I haven't missed any too obvious discussions then. 

Yes, agreed, Drift and TC need to be considered separately but I don't think moving from active heating to passive TC compensation should adversely affect drift. With well aged external resistors and low overall power dissipation (eg, no op-amp and transistor drive for the heater) hopefully at room temperature drift would be very low indeed.

For the TC, I'm not sure if your 50ppm/°C includes the Vbe of the compensating transistor or not, I'm assuming that it does (I can't find a figure in the datasheet). That particular application circuit seems to be tweaking the zener current to optimise the tracking of the zener and compensation transistor TCs (hopefully by shunting a 200R long term stable resistor with a high value less critical shunt one). It's a question of how close that tracking / compensation can get over a limited (room) temperature range, which it sounds as if nobody's really looked into.

It also occurs to me that if the heater is not in active use then there is also the sense transistor available which could maybe be employed in a second order compensation circuit?

At least having the on-chip heater would make the thermal cycling nice and easy for trimming... and maybe for 'relaxing' any hysteresis too.

Of course I don't actually have an LTZ1000 but the possibility is making me edge towards about a Digikey order!

EDIT: Though I suppose if adverse effect on drift is already so low when heated, then there may not be too much milage in this approach. I just like the elegance of a room temperature solution that doesn't have as many issues with thermocouple effects with a heated package.

[Mickle T.]

The old good Datron reference board