Resistor Set for LTZ1000 positive standard 7V circuit

[pitagoras]

Hello,
LTZ1000A + LT1013 are on their way. Several of them. 
Meanwhile I read and read, and re-read again the datasheet as a kid waiting for Santa.
My main problem now is the resistor set.
I asked a quote to Mr Pettis but first I want to understand exactly (well...) what I'm doing. Because budget is not infinite... and I could privilege building "many" references over "ideal" ones.

I understand that R1 (120) is very critical as it sets the current through the zener. I guess it has been chosen very near the sweet spot. I understand that changes in this resistor (either due to temp or long term drift) would obviously change it's voltage drop and thus affect the output voltage... So for this resistor, the best tempco, humidity coefficient, long term drift, etc. that one can get will pay in the end.  So far so good (but see below).

Then there are R4 and R5. They set the operating temperature. For this the datasheet indicates to use the lowest temperature consistent with the environment, adding 10°C for LTZ1000A. "Consistent" here means (if I understand correctly) take a margin for the ±10°C variation due to production. So all this implies making it work at
Ta_max + 10°C [production variability worst case] + 5°C [extra safety] +  10°C [if LTZ1000A]

From this information and the note below the circuit, one might conclude that the exact ratio is not *that* important, provided the ratio is as stable as possible. So if I had 1% resistors with exactly the same tempco (and other drifts) [Of course I understand that probably no such thing exists] , is better than 0.01% resistors with very different tempcos.

Then there are the two 70k resistors (R2, R3). Their drifts impact 300 (500 for R3) times less. So drift for those does not look like a super critical parameter. What I don't know is how critical is their nominal value (as 70k is a rather hard to find value).

So to resume, if I had not an infinite wallet to bet:

1) Put as much as possible in R1
2) Make sure R4 and R5 are the same kind, their values must respect the ratio giving the desired temperature, and selected for same "drift due to the elements". . This can be tricky, so better put some money in those too.
3) if money is left, bet on R2 and R3. Else just put some cost reasonable metal film,  "about 70k" each.


So, is this a question? I hope so, and not a flame war. To summarize, and avoid unnecessary casualties, I think what I'm asking is
1) The rationale of resistor selection when in a budget is correct? (in a limited budget)
2) What is all this 13k stuff anyway? I'm aware that there are discussions about the 13k value, for example, discussions whose origin I don't know, but apparently it's not the datasheet?
3) is 70k important. Would 50k do? and 72k?
4) bonus track: why HP used 111 for R1 and 15k/1k for R4/R5  My guess is they changed the operating temperature with R4/R5 and then selected a new sweet spot with R1, but most people here knows better for sure...

[Andreas]

Hello,

This has been discussed in the LTZ1000 thread many times. So the summary is:

R1 is not the most critical resistor.

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg833226/#msg833226

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg656529/#msg656529

So put your money into R4/R5 first.

You will at least need 0.1% resistors. Otherwise you have to put a further safety margin for the temperature setpoint.
(-2mV per deg C against the resistor variations).

the 13K / 1K gives about 60 deg C temperature setpoint.
12K5 / 1K around 53 deg C so this can be used with a LTZ1000A for a operating temperature up to around 35 deg C.

The 70K is not that critical. It is mainly to meet the 100uA collector current for which the device is specced in the data sheet.
But since the unheated tempco of the LTZ increases about 10% when reducing the value to 50K I would rather use a 68K-75K resistor.

with R1 you want to set the 5mA setpoint for low noise.
the 120R with 12K5/1K give around 4 mA only.

So if increasing the temperature in HP with 15K/1K
you would have to compensate for the -2mV/K.
(havent calculated it yet)

with best regards

Andreas

[Dr. Frank]

I want to confirm completely, what Andreas stated already, especially that the stability of R1 is not critical, also not its absolute value.

Use precision wire wound resistors,  either from Edwin Pettis, which may take a while, or from G.R., the econistors, which are all available from stock, 0.1%, T.C. 3ppm/K, and use a set of 12k5/1k, 120 Ohm, and 2x 70k. This will also give the least cost.

Argentina, especially Buenos Aires, I assume to be located in a moderate climate, so 35°C ambient temperature max. for the circuit @ about 50°C oven temperature should be fine. I also recommend constant laboratory temperatures, for example in your basement, for all precision measurements.

HP originally designed the reference for up to 55°C ambient temperature, which requires an oven temperature of about 90°C.
So they violated the LTZ1000A, and operated it practically like the LM399, so that the LTZ1000A also would have a similar drift rate like the LM399. Only by run-in and selection, they could guarantee these - very mediocre - 8ppm/yr.
As the base-emitter voltage of the reference transistor decreases by about 60mV at 90°C, compared to the regular 60°C oven temperature, the zener current would also decrease by about 10%, so they decreased the 120 Ohm to compensate for that effect.


In a stand alone reference, and in real metrological environment, you might even decrease the oven to 45°C for a LTZ1000, non-A version, typically then having < 1ppm/yr. from scratch.

Frank

[MK]

Dont forget that the 1K/13K or 1K/15K ratio is just that, a ratio, that resistor chain is running at 700uA, to stabilise a transistor running at 100uA, so it could easily go to 2K/26K or 2K/30K, also 5K/65K would do, it would help if you want to run it on batteries. going much higher would make it more sensitive to rfi.

when I made a spice model, using 5087? 5089? as the small signal transistors the whole circuit was very sensitive to rfi, even prone to oscillating.

I have three ltz1000a's but in amongst other projects and work currently not enough time to lay out the PCB. I will definitely be using Edwin's resistors for R2, R4, R5 but suspect I may just use RC55's for the others.

One thing I thoght about is to limit the heater power so that the D(temp)/D(time) is controlled to reduce the turn on stress to the die as I suspect that is possibly the cause of slight changes after a power cycle that has been noticed by some.

[try]

Hi Frank,

I want to confirm completely, what Andreas stated already, especially that the stability of R1 is not critical, also not its absolute value.

Is that true?
Why bother buying high end resistors then?

Regards
try

[MK]

Hi Frank,

I want to confirm completely, what Andreas stated already, especially that the stability of R1 is not critical, also not its absolute value.

Is that true?
Why bother buying high end resistors then?


Regards
try

Because as hobbyists it is actually cheaper to overspecify a bit than to fit a resistor then later have to replace it and spend again because the unit is found to be not stable enough.

And also a few people just copy what has been done before and have not done all the error budget calculations for themselves.

[Kleinstein]

The importance of R1 seems to be usually lower than specified in the LTZ1000 data-sheet. If taking the values from the data-sheet R1 would be about as important as the R4/R5 divider. However experience from build circuits showed lower sensitivity to drift in R1. This could be due to the LTZ1000 getting better or a rather conservative specification. So the important resistors are R4/R5 and less important by about a factor of 3-5 the other resistors (120 Ohms, 70 K).

The 70 K value is kind of a compromise in low TC of unheated reference (would prefer larger resistor) and low noise (would prefer smaller resistor).  Both aspects are only marginal for the overall performance. So no need to exactly get 70 K: something like 50 K to 100 K should be acceptable. The low noise circuit shown in the DS even uses 30 K.

[Dr. Frank]

Hi Frank,

I want to confirm completely, what Andreas stated already, especially that the stability of R1 is not critical, also not its absolute value.

Is that true?
Why bother buying high end resistors then?

Regards
try

Yes, it is not as critical, as given in the LTZ1000 datasheet.
This has been proven by several volt-nuts consistently.
The real, measured attenuation factors for drifts are about +/-90 for R4/R5, -600..-800 for R1, -250 for R2, and about -1000 for R3.
The factors mitigate the resistors T.C.s and their timely drift, but all 5 drifts sum up, of course.

As you want to achieve < 0.05ppm/K for the whole reference, it's obvious, that you also want the less critical resistors to have low T.C.s.
So, with a matched set of PWWs @ <5ppm/K , you'll achieve these 0.05ppm/K quite comfortably. Below that, it's getting a tough job.

Same goes for the timely drift, the sum of all the resistors shelf life drift (attenuated by their individual factors) should be far less, than the typical -0.8ppm/year @ 45°C for the LTZ1000 itself.
This parameter is specified very rarely.

Therefore, only good quality resistors with about < 10ppm/year drift should be used.
Make a simple calculation and variation on your own, to get a feeling, how less well specified resistors will influence the performance.

On the other hand, choosing only highest quality and price resistors (e.g. Vishay hermetically sealed, low T.C. MBFs)  is an engineering and monetary overkill, as you'll anyhow end up with a residual T.C. which you may further compensate by several other techniques.

Frank

[Andreas]

Why bother buying high end resistors then?

because we can!

And of course because the AN86 specifies on page 46 the VHP-100 resistors with 0.1%.
(That information is "cut out" in the data sheet).

with best regards

Andreas

[Andreas]

Hello,

by the way.
has anyone tried a thin film resistor array for this?
A 4 resistor array with 2*1K+2*10K could give a 12.1 : 1 ratio for a LTZ1000 (non A) version.

https://www.digikey.de/product-detail/de/vishay-thin-film/ORNTA10-1T1/ORNT-10K-1KCT-ND/1771027

with best regards

Andreas

[feedback.loop]

Hello,

by the way.
has anyone tried a thin film resistor array for this?
A 4 resistor array with 2*1K+2*10K could give a 12.1 : 1 ratio for a LTZ1000 (non A) version.

https://www.digikey.de/product-detail/de/vishay-thin-film/ORNTA10-1T1/ORNT-10K-1KCT-ND/1771027

with best regards

Andreas

They seem to have 25ppm/C tempco

[Andreas]

But not for the ratio

with best regards

Andreas

[Kleinstein]

The TCR matching is specified at +-5 ppm/K. So this might be acceptable for a lower cost version (not super stable). However the long time drift specifications are no not that good: 150 ppm for 2000 h at 70 C. So if useful at all it would need an extensive burn in to get rid of much of the initial aging at least. Also the SMT case might be somewhat sensitive to board stress.

[MK]

OT

[MK]

OT

[MK]

OT

[pitagoras]

I like very much your discussion about LM399, but it's not somewhat off topic here?
I opened this thread because there are 80+ pages of LTZ1000 thread, and this one focuses on resistors only. If it starts deviating then it would not be useful anymore.
Your discussion is very interesting, would you mind opening a new topic please? I will subscribe to it 

[lars]

If we go back to the resistors for the LTZ1000.

If I use simple thin film resistors for the 120 and 70kohm and even the 1k and 13k resistors do I risk a lot except long-term drift and temperature coefficient? For example LF noise, jumps? As far as I understand HP used simple thin film resistors for the 3458A.

As I think I have said before my tests of RC55 in through hole and RN73 in SMD0805, ERA6 and PCF0805 in the range 1-100kohm have been less than 100pm drift the first year and after that less than 20ppm/year. Temperature coefficients has been within maximum specs from manufacturer (sometimes quite close). Humidity sensitivity has been in the same range as wire wounds I have tested, that is < 2ppm/%RH typical 0.5-1ppm/%RH for 10kohm. This has been solder to FR4 proto boards and without power in my normal lab room (16-32C).

Should be interesting if others have long term data from real tests. Anything I have missed on the internet?

Lars

[lars]

The TCR matching is specified at +-5 ppm/K. So this might be acceptable for a lower cost version (not super stable). However the long time drift specifications are no not that good: 150 ppm for 2000 h at 70 C. So if useful at all it would need an extensive burn in to get rid of much of the initial aging at least. Also the SMT case might be somewhat sensitive to board stress.

Two discrete resistors with max +-3ppm/C will give +-6ppm/C ratio so worse than +-5ppm/C. If the about max 0.05ppm/C that will give to the LTZ design is not good enough I probably would go for a better resistor set or trim the TC of the complete LTZ board before measuring and selecting individual resistors.

For me it seems that data from the resistor data sheets are very difficult to apply to the conditions of an LTZ1000 based std in a normal hobbyist lab say 15-35C. So probably 150 ppm for 2000 h at 70 C says nothing?

For R4-R5 it is probably a good idea to have something in the same package like the ORNTA or NOMCA resistor networks. Another possibility is to have a series/parallel combination of the same value. I have at home 10 pcs 3kohm PCF0805 5ppm/C from Digikey that I intended to test in a combination to give 1k:13k (7pcs will give 1k:12k) (let’s see if I ever do the test). My tests of other resistors with the same values seems to indicate that both drift and humidity sensitivity will be in the same direction (but not TC) so a “statistical array” would be of use I guess.

Curious if someone has any data on bending on boards with eg SMT resistors? How likely is it with bending in a DIY home built LTZ design eg similar to  the KX? If this is a problem in the LTZ design with the small sensitivity from the resistors it should be a large problem in other applications?

Lars

[lars]

I have several times read that you should not have the resistors radial in the LTZ design but how much will it actually affect the performance? For me it seems you will have a gradient but it will not change a lot if the board is in a case and the box is not tilted. If the box is tilted is it better if they are mounted axial??

Lars

[Dr. Frank]

If we go back to the resistors for the LTZ1000.

If I use simple thin film resistors for the 120 and 70kohm and even the 1k and 13k resistors do I risk a lot except long-term drift and temperature coefficient? For example LF noise, jumps? As far as I understand HP used simple thin film resistors for the 3458A.

HP indeed used thin film resistors for both 70k, 10ppm/K.
The others were Vishay BMF resistors, 1.3k/ppm.

Why do you bother about noise and jumps?
The LTZ1000s purpose is superior stability.

As I think I have said before my tests of RC55 in through hole and RN73 in SMD0805, ERA6 and PCF0805 in the range 1-100kohm have been less than 100pm drift the first year and after that less than 20ppm/year. Temperature coefficients has been within maximum specs from manufacturer (sometimes quite close). Humidity sensitivity has been in the same range as wire wounds I have tested, that is < 2ppm/%RH typical 0.5-1ppm/%RH for 10kohm. This has been solder to FR4 proto boards and without power in my normal lab room (16-32C).

Should be interesting if others have long term data from real tests. Anything I have missed on the internet?

Lars

PWW and molded BMF resistors are specified to have shelf life stability of < 20ppm/yr, so the complete circuit is more stable from the beginning, than yours, after one year.
I can't imagine, why you're using such unstable resistors with the LTZ1000, that's simple a waste.

Frank

[lars]

Why do you bother about noise and jumps?
The LTZ1000s purpose is superior stability.

That is the reason I bother about noise and jumps from the resistors. Of course I know you can have outliers of both resistors and the LTZ for all kind of parameters. My objective is to solve if normal thin film is not useful at all.

PWW and molded BMF resistors are specified to have shelf life stability of < 20ppm/yr, so the complete circuit is more stable from the beginning, than yours, after one year.
I can't imagine, why you're using such unstable resistors with the LTZ1000, that's simple a waste.

Frank

That is the specification. From real life with about the same conditions soldered on FR4 board I have had up to 30ppm drift the first year for S102. Some S102 tested at the same time had very low drift so the variation was very large. Also remember that both BMF and WW in epoxy have humidity sensitivity that is much worse than the drift. With 1ppm/%RH and a 40 %RH variation you will have a seasonal variation of 40ppm not getting better with time. So even if I have seen a lot of WW and BMF probably having 1-2ppm/year, I say probably as with the large humidity dependence I have had difficulty to get the long term drift.

A side note: My intention two years ago was to make 3-4 LTZ1000 based references but I only have one prototype with thin film resistors and BMF's so far. The reason is I bought two sets WW (except the least sensitive 68k) but found they had up to 2ppm/%RH. So it stopped the project.

With a price of up to 10USD for WW compared to about 1USD for thin film 5ppm/C or 10ppm/C resistor from Digikey, Mouser and others I think it is relevant to see how much each type worsens the complete design. Of course hermetic BMF's for about 50USD each will make the ultimate design?? Sorry I couldn´t resist I know from all your previous post you have a serious attitude.

Lars

[Dr. Frank]

..
With a price of up to 10USD for WW compared to about 1USD for thin film 5ppm/C or 10ppm/C resistor from Digikey, Mouser and others I think it is relevant to see how much each type worsens the complete design.

Of course hermetic BMF's for about 50USD each will make the ultimate design?? Sorry I couldn´t resist I know from all your previous post you have a serious attitude.

Lars

By using T.F. resistors, you very probably have worsened the design upfront, if not by humidity sensitivity, then by their other mediocre stability parameters.

And your further writings just puzzles me, maybe you're not so serious about me?
I have used PWW resistors only in my LTZ1000 circuits, as I've always stated, that BMF resistors have no real advantage, maybe the hermetical ones, but these are engineering and monetary overkill for that design goal.

Frank

[Kleinstein]

As the drift sensitivity for the resistors is in the range of 1/100 and less, noise of the resistors should not be a real problem.

With TF resistors there is also a large variation between different versions. The more difficult properties are long term drift and humidity - the TC is just the parameter that is relatively easy to measure. The humidity effect can be less critical if used in an always warm circuit.

The high quality hermetical sealed TF resistors can be better in stability than many PWWs. Of cause the price is also high and availability is difficult with these.

It somewhat depends on the application whether long term drift is more important than noise.

[lars]

Hello Dr.Frank,

I am absolutely serious about you! Sorry if I missed that you only used WW. I really appreciate what I have read from you.

But I really want to know about T.F not just that they just are worse.

The LTZ in itself are not perfect even for stability and noise and the design of the LTZ is such that it allows worse resistors to some extend as both resistance shift and noise is suppressed. For the resistance shift we have found the attenuation coefficients but not for noise as far as I can see. Also the data sheets for resistors reveals just part of the truth I think.

Having resistors that cost 5-10 times more and add upp to some 25-50USD needs to be considered. Having all wire wound seems not as the final solution if you are cost conscious.

Lars