Ultra Precision Reference LTZ1000

[muvideo]

The comparator is referenced to VREF = 2.5 V. Nothing goes negative relative to the potential at the GND level.

No!
I refer to figure 1 of the article.
The complete OP amp's supply are referenced to ground. The supply of the reference also.
The voltage on R1C can only be positive above ground. But a positive voltage feed
into RN2A (buffered by IC2A) would  (with a negative supply voltage for the OP) result in a negative output voltage of IC2B.
But without negative supply, the output voltage will always stay near zero volts or an erroneous phase-reversal occurs.

From a rapid eye to the circuit, I'm with you on this.
If I understand correctly when the die cools down the
voltage at IC1 pin 1 goes up, in order to begin a new
heating cycle IC2C pin 10 must go above pin 9.
So IC2A/B must buffer and amplify the signal in
non-inverting configuration, so I dont understand IC2A/B
in this circuit. Also IC2D could not work as intended if
opamps input offset voltage is on the "wrong" side 

[quantumvolt]

Well ... I guess the schematic needs some adjustments.

The last line in the article reads: "Ensure that the ?5-V supply is on whenever the 12-V supply is on. With the ?5-V supply off and the 12-V supply on, the controller loop applies the heat current continuously."

I am sure the circuit (which is nothing but a meta-thermostat) will work after debugging...

[CaptnYellowShirt]

I just finished with a telephone conversation with Bob Dobkin, and he was short on time, so I was not able to get all of the answers to every question I [and we] had, but here are the ones I was able to get:

Did Bob give you any indication as to the spread of drifts of the LTZ (or 399) after burn-in? Or the percentage that are rejected even before they are sold to someone like HP?

 

[CaptnYellowShirt]

No one wanted their name on the thank you card so I just ended up writing a simple note on some plain 'Thank You' stationary:



Dear Bob,

Thank you for taking time out of your schedule to answer our questions on the LTZ1000. That level of support from you and Linear means a lot to us!

Sincerely,
The [Volt-Nuts]


Went out in today's mail.

[Andreas]

Did Bob give you any indication as to the spread of drifts of the LTZ (or 399) after burn-in? Or the percentage that are rejected even before they are sold to someone like HP?

The more interesting thing would be how the criteria for selections are.
For a instrument with 35ppm/year the drift of the reference should be below 18ppm (the other ppm for resistor + tempco).
To guarantee (3 sigma) that the measured drift of the reference alone should be even below 6 ppm/year or 2ppm/kHr (assuming sqrt (9khrs) = a factor of 3 difference between 1kHr and 1year).

From my measurements about 50% of LM399 are below 1-2ppm/kHr after 2000 hours ageing under power (15 hrs on/7 hrs off per day) without previous burn in.
Now after 4000 hrs further devices seem to stabilize below this limit.

But probably you will have further tests (0.1Hz - 10Hz) noise or even lower frequency noise to filter out the bad references.

I do not know if it is a really good idea to do a burn in at 125 degrees C. Since the heater is off above 90 degrees the aging mechanism will probably change. And changes to the heater will create relative large changes to the output voltage. E.g. changing the heater voltage supply will create several ppm/V output change especially around low (10V) heater voltage.
Although the temperature and so the heater power are regulated on chip.
Of cause it could be that all aging is mainly not related to the chip itself but related to the die attach (epoxy) between chip and package. In this case a burn in at higher temperatures would  make sense.

Good aged LM399 based instruments have about 1-2 ppm drift per year.

with best regards

Andreas

[rf-design]

Did Bob give you any indication as to the spread of drifts of the LTZ (or 399) after burn-in? Or the percentage that are rejected even before they are sold to someone like HP?

The more interesting thing would be how the criteria for selections are.
For a instrument with 35ppm/year the drift of the reference should be below 18ppm (the other ppm for resistor + tempco).
To guarantee (3 sigma) that the measured drift of the reference alone should be even below 6 ppm/year or 2ppm/kHr (assuming sqrt (9khrs) = a factor of 3 difference between 1kHr and 1year).

From my measurements about 50% of LM399 are below 1-2ppm/kHr after 2000 hours ageing under power (15 hrs on/7 hrs off per day) without previous burn in.
Now after 4000 hrs further devices seem to stabilize below this limit.

But probably you will have further tests (0.1Hz - 10Hz) noise or even lower frequency noise to filter out the bad references.

I do not know if it is a really good idea to do a burn in at 125 degrees C. Since the heater is off above 90 degrees the aging mechanism will probably change. And changes to the heater will create relative large changes to the output voltage. E.g. changing the heater voltage supply will create several ppm/V output change especially around low (10V) heater voltage.
Although the temperature and so the heater power are regulated on chip.
Of cause it could be that all aging is mainly not related to the chip itself but related to the die attach (epoxy) between chip and package. In this case a burn in at higher temperatures would  make sense.

Good aged LM399 based instruments have about 1-2 ppm drift per year.

with best regards

Andreas

The LM399 shows two possible entrys for a temperature change related mechanical stress effect. The first and most reasonable is that the regulation temperature change with mechanical stress. The second would be that the buried zener voltage change because of the stress. The second argument is not reported up to now but for the first argument there is much material.



from:

Package Shift in Plastic-
Packaged Bandgap References
Vishal Gupta
Prof. Gabriel A. Rincón-Mora

So the hysteresis effect is reported by LT as be a square law dependence on the temperature amplitude.



To my understanding the aging as well as the temperature hysteresis depend on the mechanical stress change of the die attach epoxy over temperature and time. If the temperature change for instance at warmup or at the start of the burn in mechanical stress will built up but will realease after some time to lower level. This effect will repeat but with memory from previous cycles.



Zymet Zero-Stress Adhesive

There are alternates for bonding. Interesting is the reduced stress at higher temperatures for silver-filled expoxy. But the zero-stress point is above 150°C.

So the best burn in will be at target operation. A burn in at higher temperature will possible leave a greater stress at lower operating temperature. The effect will either change the regulation temperature or direct the bandgap voltage. So it will impact the obsolete LTZ1000 and LM399 buried zener as well as newer bandgap based designs.

Does anyone know about a IC level temperature regulated bandgap?

[Andreas]

Does anyone know about a IC level temperature regulated bandgap?

The LT1019 has a 400 ohms heater on board. (Pin7 to GND).
The info is deleted from newer data sheets. But the heater is still there. (see AN42 p 15 figure 66)
The disadvantage is that the heater current has no separate GND return path, so it affects the output voltage.

And unfortunately the metal can case is obsolete.

With best regards

Andreas

[rf-design]

Does anyone know about a IC level temperature regulated bandgap?

The LT1019 has a 400 ohms heater on board. (Pin7 to GND).
The info is deleted from newer data sheets. But the heater is still there. (see AN42 p 15 figure 66)
The disadvantage is that the heater current has no separate GND return path, so it affects the output voltage.

And unfortunately the metal can case is obsolete.

With best regards

Andreas

I guess that the heater is for curvature trimming on wafer w/o complete wafer chuck heating. It is possible be faster or simpler. I think it is not for normal operation because otherwise there should be some temperature control. There are 3 DNC pins which are missing in the schematic. Do you know if the third DNC is connected somewhere?

BR

Reiner

"The LT®1019 is a third generation bandgap voltage reference
utilizing thin film technology and a greatly improved
curvature correction technique. Wafer level trimming of
both reference and output voltage combines to produce
very low TC and tight initial output voltage tolerance."

[branadic]

Somewhat off topic, but does anyone know what kind of oil is used in Vishay resistors VHP100, VHP101, VHP102 & VHP103 series (Ultra High Precision Hermetically Sealed Resistor with Almost Zero TCR) or similar products with hermetical sealing?

I know that in some pressure sensor application silicone oil is filled in, but silicone is unwelcome.

[Andreas]

The discussion on Volt-Nuts speaks of "white mineral oil" (purified oil) similar to "baby oil" (without flavor)

with best regards

Andreas

[branadic]

The discussion on Volt-Nuts speaks of "white mineral oil" (purified oil) similar to "baby oil" (without flavor)

Thanks, I found that pure mineral oil is available in drugstores, I will give that a try.

[CaptnYellowShirt]

Since we've been talking about burn-in methods, I thought some of you might enjoy this paper I'm reading now:

http://link.springer.com/article/10.1007%2FBF02504039

It talks about Datron's burn-in method for its 4000 calibrator and a proposed 'accelerated' method to get the job done more quickly using "soft thermal shocks".


edit: I'm realizing now that many of you might not have access. PM me if you want the paper.

[CaptnYellowShirt]

Does anyone have experience with this kind of 'oil': http://en.wikipedia.org/wiki/Krytox

In some of my reading I've come across the use of this stuff in precision references as a thermal transfer fluid. Its Fluorocarbon nature made me wonder if it was oxygen or water vapor resistant?

[babysitter]

As I had trouble with the spoiling of plant oil and mineral oil (you smell it..) i went at work to a silicon oil sold in small quantities as modelcar differential oil. It is in a makeshift thermal bath (not cal lab grade). Have a good feeling about that stuff. Vacuum and heat extraction sounds resonable.

[SArepairman]

Somewhat off topic, but does anyone know what kind of oil is used in Vishay resistors VHP100, VHP101, VHP102 & VHP103 series (Ultra High Precision Hermetically Sealed Resistor with Almost Zero TCR) or similar products with hermetical sealing?

I know that in some pressure sensor application silicone oil is filled in, but silicone is unwelcome.

Vishay Precision Group uses silicone oil when filling their hermetic packages with oil.  How do I know?  I asked them.

Mineral oil breaks down over time and becomes acidic.  I would not use mineral oil for a permanently sealed package.  OTHO-- silicone oil will dissolve silicone rubber compounds, so it may not be compatible, and you may be forced to use mineral oil-- just make sure you leave a way to exchange it.

All of these oils absorb both water and gases.  If you don't want these in the oil, you have to extract them first.  Silicone oil absorbs less than mineral oil, but this may still need addressing, depending on your application.

Vishay Precision Group uses a proprietary method of extracting water and gases from the silicone oil before sealing it in a hermetic package.  [Probably a combination of heat and vacuum].

they may also have more elaborate gimics for gas separation, I imagine that making it "rain" oil in a vacuum environment would help degassing, essentially you want to increase the surface area of what you are degassing. Or maybe just a flat dish..

[Blackart]

Hi Guys Ive been away for a while, just caught up with the thread.

Thanks Dillegent minds for organising the chat with Bob Dobkin... great stuff, to hear from a legend in the field. I contacted Bob Pease a few years ago as he mentioned he still used HP 3400As for noise measurement. I was having trouble with mine as the chopper neons were almost dead. He replied immediately and said he'd discuss it with Jim Williams. A few weeks later I got several hand written notes from Jim ( on graph paper ) outlining mods etc to the 3400A. Its terrific that even blokes at there level can take the time to disseminate their knowledge down the chain to us.

I was interested in the thermal fluid discussion as my refs have arrived ( pron pics attached ) so Im onto the design of the burn in rig. I got some nude vishay foils which I thought I might glue to each other for thermal tracking so maybe a oil bath would also improve matters. I downloaded the MSDS for the Krytox and was a little concerned about the "flu like symptoms" which result from burning it - What happens if the ref developed a shorts etc and fills the room with some horrible toxic smoke film etc. Other alternatives Silicone and white mineral oil seem much better in this respect with Silicon having a hight flash point than the mineral oil. I found the MDDS for the mineral oil on the Vishay site incidentally labelled MM ( the ones that make strain gauges etc ). So I might play with some designs using the oils.

Interesting to hear Bobs idea on the LM399s as I got 10 of them and was planning something like the Pease app note. It will be interesting to compare with the LTZ1000 design. Im way behind a lot of you blokes but plan to get the refs all burning soon. So I can get on with the design of the final units.

Lucas

attachments : Pics of Refs as packed from Linear tech - good and easy to deal with, took about 4 weeks to arrive in Oz.
 

[robrenz]

... I contacted Bob Pease a few years ago as he mentioned he still used HP 3400As for noise measurement. I was having trouble with mine as the chopper neons were almost dead. He replied immediately and said he'd discuss it with Jim Williams. A few weeks later I got several hand written notes from Jim ( on graph paper ) outlining mods etc to the 3400A. Its terrific that even blokes at there level can take the time to disseminate their knowledge down the chain to us.

Hello blackart, any chance of sharing that information with the forum?

[Blackart]

Robrenz

Sure, I was originally planning on putting it all on a webpage anyway along with other 3400A stuff I've learnt while keeping mine going. However, I had to move house and the mentioned letters from Jim and Bob are still safely stored for upload. Unfortunately both of them died before I could ask them re permission but I would assume its ok now. Ill post to the eevblog forums under another more suitable thread. ( but it may take a while ). PM me if you want my recollections of them sooner. ( I've got 9 out of which 3 are in spec, the rest we sacrificed for the 3 )

Lucas

[robrenz]

Thanks Blackart,  I am in no hurry I don't even have a 3400 yet. I do have a 3410 that I am going to restore.

[CaptnYellowShirt]

Unfortunately both of them died before I could ask them re permission but I would assume its ok now.

I think you'd be doing the world a service by posting it.

Which reminds me... does anyone know if they donated their "professional papers" to a library or other organization? (e.g. http://spec.lib.vt.edu/scitech/ )

[quantumvolt]

As I said I would do in an earlier post, I contacted Fluke with a letter to Mr. Jeff Gust, Chief Corporate Metrologist, Fluke Calibration.

I got a very kind and helpful answer, and it seems that the Fluke management probably is not willing to actively do anything to facilitate the LTFLU-1 being a product for unrestricted sale. I am not willing to reproduce the email here due to privacy reasons. If anyone wants to read the full email, I will have to seek permission first. But the general information given below is really all there is to know.

Mr. Gust gave the advice to work with LTZ1000 and to pick up used LTFLU-1 devices. He wrote "... there is very little difference between the LTZ1000 and the LTFLU-1.", which is more or less the same as Mr. Dobkin from LT said. Furthermore - and this is all he said:"What makes the LTFLU-1 special is what we do to the product after we purchase it from LT."

Everyone's guess ... I guess. Whatever it is ... - a big "Thank You" to Mr. Gust for a swift and kind answer.

[bdivi]

... there is very little difference between the LTZ1000 and the LTFLU-1.

I cannot understand how LTFLU-1 is similar to LTZ1000.

LTFLU-1 is a 4 legged design which is thought to be a remake of SZA263 from the 70s. These are often specified as Reference Amplifiers and are not ovenized - just a zenner and compensation emitter junction.

[quantumvolt]

I do not want to bother Mr. Gust anymore. He was just about to leave for a trip abroad.

I just sent an email to the "Contact us" address on Fluke's website marked "Attention Mr. Jeff Gust, Chief Corporate Metrologist". Anyone can send their questions this way. But as everyone knows - Fluke is not LT ...

To the post over there is nothing to say but: The LTZ1000 is just a RefAmp (even though the configuration of the 6.2 V diode and the tempcompensating transistor is turned "upside down"). But instead of putting the RefAmp in an oven with a thermistor, LT put a tempsensor transistor and a heater element in the same can as the RefAmp.

If the LTZ1000 was in any way far superior - no Fluke box would ever be used to check/calibrate any HP3458A ...

[CaptnYellowShirt]

If the LTZ1000 was in any way far superior - no Fluke box would ever be used to check/calibrate any HP3458A ...

All beers are superior to Bud Lite. But still people drink it. 

[Rigby]

All beers are superior to Bud Lite. But still people drink it.

I've seen people get their asses whipped and hospitalized because they WEREN'T drinking Bud Light.  The very one-sided fight was preceded immediately by "Check out that fag and his fag beer!"

Bud Light people could fall off the Earth and I think the world would be at least twice as awesome as it is now.

edit: speeling