Buffer LTZ1000 DR. Frank

[Dr. Frank]

By the way Dr. Frank, where do you source your "tuner boxes"?  Are those TEKO enclosures?  Perhaps 393.16?  http://www.tekoenclosures.com/en/products/family/RF/series/37-39

https://www.eevblog.com/forum/metrology/mx-reference/msg1297126/#msg1297126

Yep, it is TEKO 393, bought @ Reichelt

[MisterDiodes]

Latter documents and the underlying experiments are simply an evidence, so no one has just to 'believe' something.

The T.I. paper you're citing, also does not support your claims, first because the LTZ zener, but also the whole circuit is relatively low-impedance, as you yourself already mentioned, and 2nd, as the 100nF output capacitor in the circuit by Andreas will dominantly absorb these chopper spikes in first place.

Your proposed measurement with and w/o RC filter, on several different LTZ1000, over 3-5years, is very unpractical, at best.
That's more an unrealistic approach, due to the diversity of any of two LTZ1000s, and due to the obviously very small expected magnitude of the effect.
Being noticeable only on a 5 or even 20 years timescale, as you already stated in another thread, seems to me to be completely irrelevant for our application here.


This eevblog forum contains a lot of open source material, like freely available designs, measurements and real facts from many other engineers and scientists.
Therefore I have a general credibility problem with all other contributors, who are not willing, or are not able to openly publish profound substance, and are just arguing in a hand-waving manner.
This way, you also withdraw yourself from critical peer reviews, which is a striking aspect of any scientific / engineering publication of this kind.


Frank

Yes, as I said in the first line of my post I can't release client data - it isn't open source.  Sorry.  Someday I can.  I'm trying to leave at least a few bread crumbs for you guys from the real world beyond EEVblog - I'm trying to give you at least a clue as to what you want to look at to make your LTZ age well.  Use the information or not - up to you.

Believe me, if I thought it wasn't important, I sure wouldn't waste time posting.

I 100% agree that some of these effects are not observable on a hobbyist scale, that's why you -must- listen to the manufacturer's own advice.  LT/ADI builds chopper amps. but they do NOT recommend you connect one directly to the LTZ die - they do not want the chopper noise on the LTZ die.  Isolate chopper inputs with an RC filter if you're connecting the chopper inputs to any other substrate or PN junction (applies in almost every case).  On your circuit even a modest RC between the LTZ and your chopper amp input will cause no harm and no error - so why not give your LTZ the best chance?  If you're worried about Vref error across the resistor, bootstrap the cap to prevent cap leakage error.  Done all the time as good engineering practice. (Stack 2 caps, top cap has your Vref signal fed by low value R, bottom cap prevents leakage on top cap and charged from same signal by high value R - R's are not critical at all...see Art of Electronics for example).

I 100% agree the scientific approach is best, and use every opportunity to question everything for yourself.  DON'T LISTEN TO ME!  I sure as hell wouldn't, we agree on that point!   That's why you build a TIA and look at the input current spikes on your chopper amp (This info is usually not published, so find out for yourself like we did), and then explain why it -doesn't- affect the LTZ die.  The LTZ Zener would be considered low impedance yes - but it's a current mode device and with the crystal lattice is being modulated by the chopper spikes.  Guaranteed to happen without an RC between the chopper and LTZ.  HINT:  It's not always the amplitude of the spikes, but the rise time.  Also notice the spikes are very asymmetrical - that should be another big clue if you're a student of slower crystalline structure aging.

What I'm trying to get across:  I don't think you've measured the '2057 (or any chopper amp) input current noise directly?.  Until then, please don't discount mine or Kleinstein's points either.

Again:  If you don't have the resources to test these effects long term on many LTZ's, I recommend using even a modest RC filter on your chopper.   LT/ADI recommends the same practice, noted in several conversations with their apps engineering team, and they see a lot more feedback from bigger customers.   I also recommend you fix the thermal issues on your LTZ circuit before adding extra caps that wouldn't normally be required.  Especially look at drawing heat from Pin 6 of your LTZ - sometimes that thermal problem looks like an electrical issue.  Not saying that's exactly what you've encountered, but that is a real "trap for young players".

Dr. Frank:  Our circuits are always in a primary steel housing, with a minimum 2mm wall, usually 3mm or more.  Box is a part of the inguard system.  By the time it's mounted in the field it will have a double steel enclosure, like any piece of quality test gear.  Sometimes we'll use a drawn MuMetal box depending on application.   Ambient temp range can be spec'd 10 ~ 40C, but customer will want testing done usually -10~50C or more and looking for noise spec 10mHz to 10Hz or 1mHz to 10Hz.   GENERALLY we have to spec drift rate for 24 hr, 30 day, 90 day and 1 yr , and specify a maximum cal test interval as part of documentation deliverables.  That's why we're burning in LTZ's 18 ~ 24 months before they go into service.  Not hobbyist level specs, but something to learn from...because that's how we have to do it.  Your needs (and circuit topology) might be different.

[ArthurDent]

Here is a quick update on where my Dr. Frank board assembly is right now.

I got my circuit board from OSH Park and have assembled as much as I can but still need a resistor or two for the amp/buffer although I have enough parts so I could start testing to see how the circuit works.

I had found a good deal on the WIMA .1mF film caps with really fast shipping but the ones I got are rated 250V and are physically thicker than the 63V ones so there was a minor placement problem in a couple of locations. All the precision resistors are Vishay RNC90Z 2ppm that I bought on Ebay from China or Poland and they all look very good according to my HP3457A. The divider uses 988R/12K931 and two 68K000 and 120R for the others around the LTZ1000ACH. Everything is a compromise and this is what I decided would be adequate for my needs especially where it is my first voltage reference board. That is also why my first board is the Dr. Frank board with the 10V output where the 10V output is important for my use.

I used a schottky 1A diode in series with the + supply lead to prevent my accidently frying the chips while testing and put a 10mF tantalum cap on the board to keep the supply rail low impedance. Before applying power I checked everything several times to make sure I didn’t have any part backwards and end up damaging the LTZ1000ACH. No magic smoke, whew! Voltage seems to be about 7.12774 and the resistors for the amplifier are 12K000 and 4K750 plus about 74R and the trimpot. I still need one resistor and do some fine tuning for the 10V.

I also modified the output by adding an NPN 2N3904 emitter follower within the feedback loop to help protect the rest of the circuitry from output problems and increase the load capacity of the amp. Note there are two cuts from pin 6 of the ICL7650 to keep the feedback cap (C10) on the outside of the feedback loop. The 150R collector resistor is not a critical value and is used as short circuit protection of the output by limiting the current through the 2N3904. The voltage drop across the resistor will be essentially zero with no or high impedance loads. Adding the two resistors and the transistor required cutting a few runs and drilling a few small holes for the leads. The finished modification looks pretty good. The thermistor pads/runs were used for the transistor but I wasn’t going to use the thermistor anyway so it wasn’t a concern for me.  Most of the modifications I’ve made didn’t involve any deep thought so I still have to check out the finished product to see if they work the way I hope they will. After I finish I’ll clean all the flux off the board and put it in a metal box.

[eurofox]

@MisterDiode,

Be so kind to provide the kind of filter you implement between the LTZ1000 and the opamp with schematic and theory behind it.

eurofox

[floobydust]

If I understand this right, hard capacitive loads presented to the LTZ zener from, for example, a CAZ amp or op-amp feedback cap (C7 on MX Reference) are to be avoided.

But what amplitude/rise time of LTZ noise current are we taking about, that causes long-term aging through lattice defects? We have no scale or measurements, on either end (the LTZ or the 2057).

LTC2057 datasheet shows 525R internal input resistors and 100kHz chopper current pulses "0.7nA RMS" from charge injection.
It sort of seems difficult to get current pulses enough to disturb the lattice, but it leads to questioning what value of resistor is reasonable?
Adding it is as Dr. Frank mentions, a new thermal EMF but it could be cancelled with dummy component/junctions.

[zhtoor]

Yes, as I said in the first line of my post I can't release client data - it isn't open source.  Sorry.  Someday I can.  I'm trying to leave at least a few bread crumbs for you guys from the real world beyond EEVblog - I'm trying to give you at least a clue as to what you want to look at to make your LTZ age well.  Use the information or not - up to you.

Believe me, if I thought it wasn't important, I sure wouldn't waste time posting.

I 100% agree that some of these effects are not observable on a hobbyist scale, that's why you -must- listen to the manufacturer's own advice.  LT/ADI builds chopper amps. but they do NOT recommend you connect one directly to the LTZ die - they do not want the chopper noise on the LTZ die.  Isolate chopper inputs with an RC filter if you're connecting the chopper inputs to any other substrate or PN junction (applies in almost every case).  On your circuit even a modest RC between the LTZ and your chopper amp input will cause no harm and no error - so why not give your LTZ the best chance?  If you're worried about Vref error across the resistor, bootstrap the cap to prevent cap leakage error.  Done all the time as good engineering practice. (Stack 2 caps, top cap has your Vref signal fed by low value R, bottom cap prevents leakage on top cap and charged from same signal by high value R - R's are not critical at all...see Art of Electronics for example).

I 100% agree the scientific approach is best, and use every opportunity to question everything for yourself.  DON'T LISTEN TO ME!  I sure as hell wouldn't, we agree on that point!   That's why you build a TIA and look at the input current spikes on your chopper amp (This info is usually not published, so find out for yourself like we did), and then explain why it -doesn't- affect the LTZ die.  The LTZ Zener would be considered low impedance yes - but it's a current mode device and with the crystal lattice is being modulated by the chopper spikes.  Guaranteed to happen without an RC between the chopper and LTZ.  HINT:  It's not always the amplitude of the spikes, but the rise time.  Also notice the spikes are very asymmetrical - that should be another big clue if you're a student of slower crystalline structure aging.

What I'm trying to get across:  I don't think you've measured the '2057 (or any chopper amp) input current noise directly?.  Until then, please don't discount mine or Kleinstein's points either.

Again:  If you don't have the resources to test these effects long term on many LTZ's, I recommend using even a modest RC filter on your chopper.   LT/ADI recommends the same practice, noted in several conversations with their apps engineering team, and they see a lot more feedback from bigger customers.   I also recommend you fix the thermal issues on your LTZ circuit before adding extra caps that wouldn't normally be required.  Especially look at drawing heat from Pin 6 of your LTZ - sometimes that thermal problem looks like an electrical issue.  Not saying that's exactly what you've encountered, but that is a real "trap for young players".

Dr. Frank:  Our circuits are always in a primary steel housing, with a minimum 2mm wall, usually 3mm or more.  Box is a part of the inguard system.  By the time it's mounted in the field it will have a double steel enclosure, like any piece of quality test gear.  Sometimes we'll use a drawn MuMetal box depending on application.   Ambient temp range can be spec'd 10 ~ 40C, but customer will want testing done usually -10~50C or more and looking for noise spec 10mHz to 10Hz or 1mHz to 10Hz.   GENERALLY we have to spec drift rate for 24 hr, 30 day, 90 day and 1 yr , and specify a maximum cal test interval as part of documentation deliverables.  That's why we're burning in LTZ's 18 ~ 24 months before they go into service.  Not hobbyist level specs, but something to learn from...because that's how we have to do it.  Your needs (and circuit topology) might be different.

@MisterDiodes

here is what i understand from your comments:-

1. thermal equilibrium of the die is extremely important for short settling (aging) times.
2. since pin #6 is the base of the temp. sensing transistor, the amount / rate of heat removal from it determines the sensor transistor response time.
3. something special about this pin is also evident in the die, it has 3 bond wires attached (heat transport capacity?)
4. would it help to make an externally tunable heat-reservoir (copper pad, area - tunable) on pin #6?
5. any fast rate of change impulses fed into the die would result in some kind of piezoelectric interference in the crystal lattice settling times in response to thermal gradients.
6. if all of the above are (semi) true, then i can understand why fluke chose not to put (use?) the heater on-board LTFLU-1(A)CH for their standards (and retiring Datron 7001)
7. if thermal management is such a critical issue, why use the on-board heater?
8. is there a way to use the heater as a strain-gauge to estimate the relaxation state of the LTZ1000 die?

please do look at:-
https://www.eevblog.com/forum/metrology/off-label-ltz1000/

regarding some discussion about the thermal issues.

best regards and keep educating us however much you can.

-zia

[Andreas]

3. something special about this pin is also evident in the die, it has 3 bond wires attached (heat transport capacity?)

it is not Pin6 but Pin1 (heater) with the 3 bond wires. (caution pinning in data sheet is from "bottom side")

[zhtoor]

3. something special about this pin is also evident in the die, it has 3 bond wires attached (heat transport capacity?)

it is not Pin6 but Pin1 (heater) with the 3 bond wires. (caution pinning in data sheet is from "bottom side")

thanks for correcting my mistake.

courtesy of TiN there is another very good image of LTZ1000 die at:-

https://doc.xdevs.com/doc/xDevs.com/KX/Die_LTZ1000_small.jpg

1. there are 7 concentric rings of which 6 look like they are part of the heating element.
2. there is a third ring (from outside) which does not appear to be part of the heating element, what is it?
3. are the 3 groups of 2 rings each connected in parallel ?

best regards.

-zia

[Dr. Frank]

Yes, as I said in the first line of my post I can't release client data - it isn't open source.  Sorry.  Someday I can.  I'm trying to leave at least a few bread crumbs for you guys from the real world beyond EEVblog - I'm trying to give you at least a clue as to what you want to look at to make your LTZ age well.  Use the information or not - up to you.

Well, I'm not expecting, that you would ever really reveal any profound data.
That sentence is really disturbing me: '..to leave at least a few bread crumbs for you guys..', because that's always the feeling I had, after reading through your posts.

Believe me, if I thought it wasn't important, I sure wouldn't waste time posting.

...

I 100% agree the scientific approach is best, and use every opportunity to question everything for yourself.  DON'T LISTEN TO ME!  I sure as hell wouldn't, we agree on that point!  

Mr. Diodes, I really would like to listen to you, if you would prove your statements by evidence, but this way, again, it's quite useless for me, even if it's on a hobbyist level forum only

That's why you build a TIA and look at the input current spikes on your chopper amp (This info is usually not published, so find out for yourself like we did), and then explain why it -doesn't- affect the LTZ die.  The LTZ Zener would be considered low impedance yes - but it's a current mode device and with the crystal lattice is being modulated by the chopper spikes.  Guaranteed to happen without an RC between the chopper and LTZ.  HINT:  It's not always the amplitude of the spikes, but the rise time.  Also notice the spikes are very asymmetrical - that should be another big clue if you're a student of slower crystalline structure aging.

What I'm trying to get across:  I don't think you've measured the '2057 (or any chopper amp) input current noise directly?.  Until then, please don't discount mine or Kleinstein's points either.

I did not doubt the existence and the weight of these spikes, that's already well known and documented.. so I don't have to measure that again, on my own.
Up to date, I never have seen any documentation about that said destructive effect on Si lattices, so maybe you can explain or cite instead, why this energy should be high enough to get into the necessary activation energies for measurable changes of the reference voltage.

Again:  If you don't have the resources to test these effects long term on many LTZ's, I recommend using even a modest RC filter on your chopper.   LT/ADI recommends the same practice, noted in several conversations with their apps engineering team, and they see a lot more feedback from bigger customers. 

That's also a argument, sorry.
As you don't even at this point present any numbers for that said effect, nobody can possibly judge, if this effect is measurable with reasonable amateur equipment, or the other way round, which level of equipment or stability would be required to provide evidence for that.
As the information from LT seems not to  fall under NDAs to your customers, please provide details, numbers, facts.

In fact, my both prototype references run since about 2004, or so, the first with an 7650 practically directly coupled to the LTZ1000, and the other w/o any chopper behind.
Both drift meanwhile in parallel over several years, about -0.5ppm/year drift.
Conclusion from that? None in your direction.. Only, that it's possible to realize a very low drift of < 1ppm/year, as expected from the mentioned publications.

I also recommend you fix the thermal issues on your LTZ circuit before adding extra caps that wouldn't normally be required.  Especially look at drawing heat from Pin 6 of your LTZ - sometimes that thermal problem looks like an electrical issue.  Not saying that's exactly what you've encountered, but that is a real "trap for young players".

Exactly. I don't see any thermal issues, in contrary, the DIY references perform extremely well over temperature, with a 'simple' T.C. trimming, but w/o any mysterious gimmicks.
I still don't understand, how additional caps would influence the T.C.? Thermal Problem??
These additional capacitors clearly suppress dips and spikes of the output voltage.
In the Datron 4910, they've used exactly the same extra circuit, for some reason.
I'm really curious about the reverse engineering of the 7001 reference, by TiN, maybe we'll meet again such caps..

Dr. Frank:  Our circuits are always in a primary steel housing, with a minimum 2mm wall, usually 3mm or more.  Box is a part of the inguard system.  By the time it's mounted in the field it will have a double steel enclosure, like any piece of quality test gear.  Sometimes we'll use a drawn MuMetal box depending on application.   Ambient temp range can be spec'd 10 ~ 40C, but customer will want testing done usually -10~50C or more and looking for noise spec 10mHz to 10Hz or 1mHz to 10Hz.   GENERALLY we have to spec drift rate for 24 hr, 30 day, 90 day and 1 yr , and specify a maximum cal test interval as part of documentation deliverables.  That's why we're burning in LTZ's 18 ~ 24 months before they go into service.  Not hobbyist level specs, but something to learn from...because that's how we have to do it.  Your needs (and circuit topology) might be different.

In the end, you've finally lifted the curtain a bit..
Therefore, your applications are like the 3458A, inside a doubly shielded box, probably w/o any direct influence from the outside onto the LTZ1000s output.
The needs for a voltage calibrator are definitely different, apart from the temperature range, that's similar.. and the drift rate is also on an annual timescale only, not these 5 .. 20 years, you've mentioned.

Again, you also missed out the information, if the drift specification of your LTZ is really much less , than the usual specs for 732 / 4910 / 7001 references, what a pity.

So I'm really keen on learning about the 7001 topology.. the supply filtering looks extremely interesting, and I think that there really lies the design focus, besides appropriate shielding, and other means for EMI suppression.

[MisterDiodes]

Dr. Frank -

I can tell you what I can, because that's the same info LT / ADI will tell you when you work with their apps engineers.  Otherwise all our customer test data is owned by our customers.

So I guess that makes me full of crap - but it is expected.  I totally get it.  As I pointed out - I'd run my own tests and verification too.  That's a good way to do it.

What we see for drift rates:
In general, drift rates on the LTZ's we've built are fine, rarely no more than 4~5ppm/yr, usually less.   After 20 ~ 30 years too.  Some outliers down in the 1ppm/yr and over 6ppm yr, probably at least 20% are 2~3ppm.  Depends on each individual and application.  Basic datasheet circuit - sometimes with resistor mods as requested.   We don't run the LTZ's particularly cool, but not hotter than they need to be.  A lot depends on the thermal conditions  they see in actual use too.  Typical heater ratios are 13.25k to 15k over 1k.  Maybe even 16k over 1k if that is called for.  Depends on what cooling / heat sinking is available or not.

We don't ever bother adding the 400k resistor to adjust TC.  If you dial that in to an ideal setting today that might not be the ideal setting for later on - and that's only useful for a narrow temp range anyway.  What's more important (for us) is a stable and predictable TC - then you can add a temperature compensation factor into the system downstream (Which RHB should really like!!).

Never have we had to add extra caps to meet noise spec. - but then again we don''t run naked boards at customer site either - that's not the best idea for precision circuits.  If adding caps is the only way they work for you then that's how your setup works.  I won't argue with success  That being said I'm looking at an LTZ right now that's sitting in its open box, 13.25k over 1k, LTZ covered,  running on a linear power supply at 15V, PWW's and it's noise is <<1.2uV p-p.  Which is spec.  Just the datasheet circuit, nothing at all fancy, no slots or cutouts, no expensive resistors, etc.   

Zhtoor - Yes, it's all about heat flow.  Be careful about probing the circuit while it's in operation, because that probe in itself changes the thermal characteristic.  All pins really want to be same temp, but not all pins have the same thermal resistance to the die!

IN GENERAL - A chopper amp with a lower Ron on the input switches and higher charge injection combined with faster switch time can be more influential on the lattice aging.  I'm talking about chopper amp's effect on lattice aging, not outright destruction of the LTZ (although edge cracks are a possible failure mode of the LTZ die - but you usually see that on extreme thermal cycles or short circuits,  etc.).

Chopper amps USUALLY don't have the input current spikes spec'd AT CHOPPER FREQ.  BUT one suggestion is when you measure an amp for yourself you can then build an LTSpice model to play around with how the front end works.  The chopper amp simulation in LTSpice DOESN'T include the input current noise spikes in the chopper model...but there is another way.

Here is a CONCEPT ONLY example that models the switch currents of a 2378-20 SAR ADC; this is interesting since the front end switches are a cousin to some of the newer AZ choppers built by LT/ADI...This example is for an ADC not a chopper amp I know - it is showing what happens when the input switches are switching against the differential input caps.  But you can use a very similar technique to model the input of an AZ chopper amp.  You'll dial down and re-arrange the caps to simulate FET charge infection injection, but the process shown here should give you an idea how to build the chopper amp front end model in LTSPice to show you charge injection spikes, and maybe show how even a small RC filter will help:

http://www.analog.com/en/technical-articles/ltspice-simulating-sar-adc-analog-inputs.html

There is a video link at the end of the article -  especially interesting is the 50mA current spikes shown around 7 minutes into the video.  Which should pique your interest since you started out with just a 40pF cap on the chip and 1nF on your filter.  Low Ron of the switcher FETs and faster rise/ fall times will get you some serious current flows for sure!   

Again:  That's NOT the input to a chopper amp, I picked that as an online example as to how you might model chopper FET switching effects in LTSpice.  Be aware that some of the ADI choppers use a 4 phase or higher clock system - so you'll have to play around with the concept.

[floobydust]

Thanks to all for discussing something extremely subtle, playing pool with atoms getting bumped around. It led me to another question, I can move post to LTZ mega thread if people want that.

Why does this aging mechanism not exist for LTZ Q1 (temp. compensating) transistor on the other end, with datasheet 0.022uF capacitor at pin 5?
I would expect large (sink) noise current pulses there, and it operates at low collector current 90uA and zenering the E-B junction causes permanent damage, so it seems like a fragile transistor.

I notice the Datron 4708 reference using 2k7/1nF instead of a hard capacitive load (on Q1), and at the zener, a large R (39k?) to isolate the LT1013 22nF NFB cap. They were 'in the know'.

[cellularmitosis]

Oh, interesting, is that where the "R60" resistor comes from (~20R above the LTZ zener)?  https://doc.xdevs.com/doc/Datron/4708/4708_reference.png

[Edwin G. Pettis]

Floobydust,

There are no spikes generated internally in the LTZ1000/A, those 'spikes' everybody is talking about is coming from the use of an AZ amp instead of the intended LT1013.  Use the components specified in the LT data sheet and there are no spikes except from an external source such as the power supply or EMI or the incorrect use of an AZ amp.

I don't see why these is such preoccupation with Datron designs, if anything, they were flawed, all those extra components were nothing short of band aids, there was absolutely no reason for a 2.7K/1nF snubber in place of the specified capacitor.  There is no reason to 'isolate' the LT1013 22nF capacitor either, more bungling....no they weren't 'in the know', the standard LTZ circuitry put out by LT was the 'in the know', it has been tested for over 30 years and proven correct and reliable, the only modifications were done for unusual requirements.

[Dr. Frank]

Floobydust,

There are no spikes generated internally in the LTZ1000/A, those 'spikes' everybody is talking about is coming from the use of an AZ amp instead of the intended LT1013.  Use the components specified in the LT data sheet and there are no spikes except from an external source such as the power supply or EMI or the incorrect use of an AZ amp.

I don't see why these is such preoccupation with Datron designs, if anything, they were flawed, all those extra components were nothing short of band aids, there was absolutely no reason for a 2.7K/1nF snubber in place of the specified capacitor.  There is no reason to 'isolate' the LT1013 22nF capacitor either, more bungling....no they weren't 'in the know', the standard LTZ circuitry put out by LT was the 'in the know', it has been tested for over 30 years and proven correct and reliable, the only modifications were done for unusual requirements.

Mr. Pettis,
serious question to you, have you ever built and operated any LTZ1000 circuit yourself?
Operated in this voltage reference manner?

What you state here, is utter bullshit, sorry to say it so harsh! 

These spikes appear in these regular datasheet designs, with proper LT1013. That has absolutely nothing to do with chopper amplifiers.

These spikes definitely are coming from outside, i.e. EMC disturbances which mostly affect the oven regulator circuit, which is very sensitive in the regular design.

The reason why other people like Mr. Diodes (or you) don't see these, and therefore deny their existence, is that his  / your application of the circuit is inside a double shielded environment with no direct connection to the outside world, like in the 3458A.

And that's totally different to the Datron voltage references 491x and 7001, which are also open to the environment originating from the output jacks, maybe also from the PSU.
I bet, that TiN will find the very same EMC blocking capacitors inside the 7001 reference.

And these devices had been the only one known commercial products using an LTZ1000 as a voltage reference.
All other applications so far known are shielded, inside a closed box.

And then there are these dozens of LTZ1000(A) reference designs within this forum, which suffer from spikes, more or less, depending on the EMC environment in their houses.

I would really appreciate, that you won't confuse these people, as you very obviously don't know the facts correctly.

Frank

[zhtoor]

And these devices had been the only one known commercial products using an LTZ1000 as a voltage reference.

and both retired. i wonder why?

best regards.

-zia

[Andreas]

Hello,

Fluke bought Datron so there was no place for 2 concurring designs.
Commercial decisions are not always logical on mergers.
But perhaps 2 resistors of a ref-amp are cheaper than 5 for a LTZ.

with best regards

Andreas

[MisterDiodes]

Mr. Pettis,
serious question to you, have you ever built and operated any LTZ1000 circuit yourself?
Operated in this voltage reference manner?

What you state here, is utter bullshit, sorry to say it so harsh! 

People are sure getting downright nasty, disrespectful and rude here.  I'm done and outta here.

Dr. Frank:  As a parting gift, before you accuse others of being full of crap, I'll suggest you get your own facts straight, and do some research on physics of crystalline materials.  You'll learn something.    The people that designed for Datron didn't get it all right either.

The very first thing you do on a sensitive precision analog circuit is keep it well shielded.  LT/ADI will tell you that.  That is not hard nor is it expensive for a hobbyist - even a standard drawn steel electrical box can work well. Yes the spikes you see are coming from OUTSIDE the circuit but you're NOT doing your LTZ any favors using a direct connected chopper amp either.  Those effects all work together on how well your LTZ will work, and Mr. Pettis is 100% correct.  Don't listen to me or Edwin, listen to LT/ADI, or Keysight (or Fluke or Keithley).  That's who I work with, and they get lots of feedback too.

IF you find extra UN-necessary caps work in yours and Andreas' circuit that should be a BIG clue for you to look closer to se e what is really going on.  Put a sensitive current probe on that new cap (or build one) that works down into the few 10's uA range or lower and if you see those same spikes on your current flow  - you're affecting LTZ lattice aging.  To what extent will depend on an individual LTZ die.  Just like operational temperature heater setting affects each LTZ in a unique way for each LTZ die. 

Until you understand if your EMI is electrostatic or magnetic in nature, you DON'T just add caps.  "Oh I saw it on a schematic on a product that was a failure, that should work...what could go wrong?"

But then you say: "MisterDiodes - I added the cap and the noise went away".  Yes, as measured as a voltage spike you saw the spike reduced - but the spike is STILL THERE!  You're just not looking for it the right way.  Look at the situation with a CURRENT probe on your new cap: if your EMI was an H-field problem, adding a cap just made things worse!!.  Which is usually what happens when you add caps to the wrong places in precision analog systems.  Now you've got a low impedance pool of stored energy ready to quickly flow across a PN junction on a sensitive precision analog circuit.  That is never a desired effect for keeping  stress in the lattice low.

Again: Small currents, but what your asking the LTZ needs to do in terms of long term stability is even on a smaller scale  still.

IF you found out that maybe your EMI was more of an E-field problem, then adding a cap might be a good solution for YOUR  particular application.  BUT your shielded enclosure should fix that in the first place.

IN both cases:  If you did homework on proper shielding, and if Datron figured out what was really wrong with their LTZ Vref before adding caps, your higher-than-normal spike problem wouldn't be there as a VOLTAGE or CURRENT spike.

All of this information is available online or right under your nose on the test bench or at the manufacturer.  And then verify for yourself what works best for the application at hand. If you're going to make ppm measures, do it right.  That's how we do it, and we're measuring with fully maintained 732 calibrated references & 3458a meters, time & frequency standards etc....no stuff like "A friend of a friend had the old eBay meter checked at Maker Fair 5 years ago and it was OK, and these relative measures will be perfectly fine to ppB".   Sure.  If your passionate about PPM, then spend some money and get some equipment in cal, and that will help you keep your measurement uncertainty within reason.  It doesn't have to be expensive - a good, working 3456a is cheap to get cal'd, and is just as accurate or more so than a 34470a.

If you complain that you don't have the resources to test this yourself, then I suggest you listen to the people that do.

Adios guys, these LTZ threads are yours.  I've got better things to spend time on.  Have fun!

[dr.diesel]

Adios guys, these LTZ threads are yours.  I've got better things to spend time on.  Have fun!

I value your opinion, as well at Dr. Frank's, we all don't always agree, please keep any possible constructive criticism going!   

[zhtoor]

Mr. Pettis,
serious question to you, have you ever built and operated any LTZ1000 circuit yourself?
Operated in this voltage reference manner?

What you state here, is utter bullshit, sorry to say it so harsh! 

i second Mr. Pettis' and MisterDiodes' excellent opinions and appreciate the time
that they take to guide us with their *extremely* valuable opinion.

but i guess, as they say:-

it takes a jeweler to value a diamond.

@MisterDiodes, @Pettis:-

my hats off to you.

respect from pakistan.

-zia

[lowimpedance]

Adios guys, these LTZ threads are yours.  I've got better things to spend time on.  Have fun!

Would be a real shame to loose the input from real industry experience, all your posts contain valuable information that add to the wider pool of information for all the volt nut experimenters out there whether they are regulars here or not.   

 ********************************************************************************************************************************************************************
 So there may be other ways to 'skin the cat' , good go for it that's what experimenting is all about. And always never loose site of the fact its meant to be a fun hobby and accept all who contribute as we all benefit from it. Vitriol has no place here.
 Surely it would be a very quiet place if everyone couldn't get along and went back to doing there own thing.

[zhtoor]

Dr. Frank:  As a parting gift, before you accuse others of being full of crap, I'll suggest you get your own facts straight, and do some research on physics of crystalline materials.  You'll learn something.    The people that designed for Datron didn't get it all right either.

 

that is exactly the conclusion that i have come to regarding their "die-annealing" scheme in Datron 7001.
and i have *never* seen a physical LTZ1000(A)CH as of now (maybe in future), let alone do *anything* with it.

best regards.

-zia

[floobydust]

MisterDiodes, I greatly appreciate your voice here and invite you to let it roll like water off a duck's back.
The forums are limited for their signal/noise ratio and who has time to write a book explaining something a person knows is real, gained from wisdom and intimate knowledge, and stay within corporate limits.

The discussion has at least four possible sources of noise (current) pulses which could cause excessive aging in the LTZ crystal lattice. So any thread on this is going to derail.

1.  I see the CAZ with LTZ issue debated here for over a year; https://www.eevblog.com/forum/metrology/(ft)-ltz1000a-fairy-tale-or-the-story-of-little-jumper/msg1114903/#msg1114903
I notice Dr. Frank's implementation has large 100nF capacitance, enough to absorb commutating charge injection and possibly why he does not see a problem.

2.  EMI - this is probably ESD as well

3.  The 22nF cap C1 on Q1 collector (I asked if it's similar to the zener issue in aging effect).

4.  Adding bypass cap C2; I hear the repeated cautions against adding bypass caps. https://www.eevblog.com/forum/metrology/mx-reference/msg1297826/#msg1297826


I did a rough LTspice sim adding some voltage noise (1.2uVpp) to the zener, to see what resulting noise currents would look like into bypass capacitance on the LTZ. This is how I understand the bypass cap risk. I don't have flicker and LF noise there, not an exact noise model. Believe it or not, gas VR tubes also have issues with bypass capacitance for aging and noise even though plasma is of course different than zener conduction. This is why I'm a believer, lol

But the current pulses look very tiny, hard to get even 30nApp. The MX took C2 to 100nF with extra LT1013 RC (10k/100nF) added to maintain stability, and not much increase.
The ~20R zener impedance and 120R resistor seem to moderate it. Or people can tell me what is missing in my understanding.

[zhtoor]

hello floobydust,

here is a a (gu)estimated noise model for LTZ1000 @ 4ma zener current estimated from the datasheet graph.

best regards.

-zia

[floobydust]

I did see that model, but it's in the frequency domain not time domain and I could not see how to implement that in LTSpice

[ArthurDent]

Instead of getting into the previous discussion about how many angels can stand on the head of a pin, I have what I consider to be some practical observations about the Dr Frank LTZ1000A board I’m assembling and testing that I would like to share-for what it’s worth. To recap, I have most of the necessary parts to complete the board soldered in their proper locations and I have enough of the parts so I can do some testing. One resistor I need is still on order so I got the value I needed by putting two lower spec-ed resistors in series with the leads just poked into the holes on the pc board which is far from ideal but allows me to continue testing.

One problem I had was the 20R trimpot used as the adjustment for the 10V voltage divider was driving me nuts. The ‘granularity’ was such that the smallest adjustment I could get would cause the output to jump in steps of about .00005 volts. This was caused by the limited number of turns of resistance wire on what is essentially a one-turn pot with mechanical gearing. As the wiper would move from one wire turn to the next there was a much larger resistance change than desired. Also with the fixed 20R resistor across the 20R trimpot, that combination gave from 10R-20R which resulted in far too much adjustment range for my liking.

I calculated I needed 4K822 for the upper half of the voltage divider so I ordered a 4K750 RNC90Z 2ppm resistor like the other resistor I’ve used and found I had a nice salvaged 72R14 WW resistor to put in series with the 4K750 to give me the 4K822. I also had a small 5K thirty turn pot I put in series with a 5K resistor to put across the 72R14 resistor that gave me a smaller adjustment range (.5R) and excellent adjustability. Once I get the 4K750 resistor and check its exact value I’ll remove the 20R and 20R trimpot and jumper/put different resistors in that location.

Once I got the output adjusted to 10V I decided to check how loading affected the output with the 2N3904 emitter follower I had put inside the amplifier loop to get better drive and protection for the rest of the circuit. I soldered a couple of 9” #24 wires to the output pads and had a decade resistor I could use to adjust the load. I was quite disappointed in the variations caused by the load and after a lot of checking and head scratching concluded that by measuring the output voltage at the end of the two wires I was actually measuring the output through two low value resistors which had a small voltage drop. It doesn’t pay to worry about the voltage drop on a couple of mm of runs on the pc board and then put about 500 mm of wire in series with the output.

What I did to correct this was connect the negative lead from the power supply to what would be the negative binding post and connect the end of the negative wire from the board to the negative binding post so no load current was going through the wire going to the board. On the positive lead I connected it so only the emitter of the 2N3904 was going to what would be the positive binding post and ran a ‘sense’ wire back to the top of the 10V divider resistors. This kept the output current carrying wires only going to what would be the binding posts and the sensing points going directly to the binding posts and not seeing any load current. I did measure the drop across just the current carrying wires and at a 1K load the drop totaled about .0004 volts so the changes I made took that possible drop out of the circuit. Once I made these corrections I could drop the resistance load from 10M down to 500R and the output looked stable. Granted, you probably will never have such a low load on the voltage reference but it is worth making such a simple correction to make sure it isn’t a problem.