New LTZ1000 10V Reference build (WAS: Agilent Keysight 03458-66509...)

[Dr. Frank]

Regulator ICs like the LTZ1000 make possible very stable voltage standards. But after only a short time futzing with 2 of them, I realized a precision oven is required, because the external components add a lot of temp drift. I completed one, fully enclosed, very stable at 10VDC. but temp drift of 1.0ppm/C. Even though the regulator is temp regulated, and even if the heater control is well implemented, the other components have to be inside a precision oven to obtain the results experimenters want.

An outer oven is not required at all!
Maybe you have used low quality resistors only, or you made another design error.
Please provide your Bill of Material, and your schematic.

Please do not make such misleading statements in this forum, that for the LTZ1000, a precision oven is required!

Frank

[dietert1]

I think a temperature gradient of 1 ppm/K is a lot and easy to have if the 10 V to 7 V divider is made from 2 ppm/K resistors or any other low TC resistor. Dr. Frank did not understand that statement.
Of course the statement that an outer oven is required is a bit strong. Usually people get away with compensating the residual TC, which is fairly easy to do and has been decribed in many places.
As the residual TC contains nonlinear terms, the best solution is tuning the residual TC to zero at a certain temperature and run the circuit inside an oven of that temperature. Others attempted to compensate the nonlinear residual TC, too.

Regards, Dieter

[Dr. Frank]

I think a temperature gradient of 1 ppm/K is a lot and easy to have if the 10 V to 7 V divider is made from 2 ppm/K resistors or any other low TC resistor. Dr. Frank did not understand that statement.
Of course the statement that an outer oven is required is a bit strong. Usually people get away with compensating the residual TC, which is fairly easy to do and has been decribed in many places.
As the residual TC contains nonlinear terms, the best solution is tuning the residual TC to zero at a certain temperature and run the circuit inside an oven of that temperature. Others attempted to compensate the nonlinear residual TC, tool.

Regards, Dieter

Hello Dieter,
no, I understood his various statements very well.
Please read his first two sentences.

There he stated, that the LTZ1000 circuit (the "regulator IC", as he calls it  ) requires an outer oven, because the external components add a lot of T.C.

This statement is ultimately wrong, as all the volt-nuts, who really contributed with a T.C. characterization of their LTZ circuit have demonstrated, that the overall T.C. is well below 0.1ppm/K when using appropriate external resistors, and especially when T.C. compensation is used to bring that T.C. down to near zero .. INSTEAD of using an outer oven, of course.
Especially due to the attenuation factors of the resistors instability parameters, those external component in contrary do NOT contribute a lot to the overall T.C.
 
I also remember that we already had this big discussion, that an outer oven would degrade the performance of the LTZ circuit, and is counter-productive.

Then he obviously mixes the following gain stage, to 10V, into his consideration.
But that's a completely new aspect.
Sorry Dieter, I fully understood that as well!

Sure, if he uses not T.C. - matched resistors, or ones which have too high a T.C., this 10V output will have too a high T.C. as well.
You can of course "heal" that design fault by using an oven... that's called over-engineering,I'd guess.

In his last sentence, he concludes, that only due to the mediocre T.C. of his 10V gain stage, that the WHOLE circuit must be inside an oven.

And that simply shows, that he has not understood the basic LTZ circuit at all.

Dieter, anyhow, I hope you and your family are doing fine now in Brazil!

Frank

[dietert1]

When someone designs a 10 V reference, there will always be the problem to make a 10 to 7 V divider with low TC. One solution is a precision PWM. When using a resistive divider, even the best available resistors will likely produce a residual TC of 1 ppm or more. Plus amplifier temperature drift, thermal EMF and the like. I think that is what alligatorblues meant.
In general it is easy to tune the linear TC to zero or near zero at a certain temperature. Due to nonlinear terms, the usable temperature range will still be limiited, e.g. to about +/- 4 K around the zero TC temperature for less than 1 ppm error. Using an oven is a valid method to get rid of that last ppm. Of course if you have a 23 +/- 1 °C lab everything becomes easier.
A peltier oven is fairly easy to make and does not require a higher than ambient temperature.
Another consideration is whether it makes sense to TC adjust the 7 V reference and the 10 V gain stage separately. I`d guess most of us would try to save some work and put everything onto one PCB that fits inside the oven, with one TC adjustment for the whole circuit. With some luck one can even save the TC adjustment and instead adjust the oven temperature to the zero TC temperature.
As far as i understand the ADR1001 implements these concepts to a certain level. Similar to other AD products it compensates TC to some level and then relies on oven temperature stability.

Regards, Dieter

[Rax]

[...] the usable temperature range will still be limited, e.g. to about +/- 4 K around the zero TC temperature for less than 1 ppm error. Using an oven is a valid method to get rid of that last ppm. Of course if you have a 23 +/- 1 °C lab everything becomes easier.
[...]
Regards, Dieter

I have not participated in all the original pursuits and research - obviously, hence this thread - but I am getting close to putting together at least a couple of kits (probably at least one LTZ1000 and one LTZ1000A), though could build several with different bits I put together. My environment doesn't fit this description, so I'll likely have some data to report along these lines.

[Dr. Frank]

When someone designs a 10 V reference, there will always be the problem to make a 10 to 7 V divider with low TC.
...
 When using a resistive divider, even the best available resistors will likely produce a residual TC of 1 ppm or more. Plus amplifier temperature drift, thermal EMF and the like.....

In general it is easy to tune the linear TC to zero or near zero at a certain temperature. Due to nonlinear terms, the usable temperature range will still be limiited, e.g. to about +/- 4 K around the zero TC temperature for less than 1 ppm error. Using an oven is a valid method to get rid of that last ppm. ...

Regards, Dieter

Hello Dieter,
I still do not agree at all to your assumptions / proposals.

Here's my LTZ #3 which I use as a traveling standard. Recently ab precision compared both reference outputs against his 732B with about 0.3ppm uncertainty, and on return, it agreed within 0.3ppm with my reference group as well.

I used 2 FLCY from AE (those two blue components, I think 10k and 4k) for the 10V gain stage.
Per datasheet, they have a T.C. of 2.5ppm/K, but I already knew from another characterization, that they are good for about 1ppm/K.
I have not measured these in advance, only in situ, i.e. the overall T.C.
The T.C. of the direct LTZ output is about 0.02ppm/K.

The diagram shows a box - T.C. of the 10V output of about 0.7ppm/11k = 0.06ppm/K.
The chopper amplifier, either an LT1050, or an ADA4522, have of course no influence, and emf's are very low by design, i.e. in the 0.02ppm ballpark.

For lab temperatures, any non linearities can be neglected.

I also see no big timely drift of this 10V reference, I will review my monitored data and will report back.

Frank

[Kleinstein]

The 7 V reference part does not need an extra oven, though a stable temperature would not hurt, unless it gets rather hot. With the trim resistor (R10 in Franks circuit) one can trim the linear TC, though this can add some 2nd order term. It is still good enough for a normal indoor range like +-10 K.

For the 7 to 10 V step a oven for a stabilized temperature is a possible way to go, especially if the temperature variations are large.
However even than one should still have rather good resistors as the resistors also effect the long term drift. With a small temperature range one may than get away without an oven.  If really needed one can trim the TC by adding a little copper wire / trace to contribute to one of the resistors. This may still be easier than building on oven.

[Andreas]

When someone designs a 10 V reference, there will always be the problem to make a 10 to 7 V divider with low TC.

Mhm,

with a statistical divider I got 0.026 ppm/K difference between 6.6V and 10V output on my ADR1000A#1
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3931196/#msg3931196

The only problem is the ageing drift of this divider (but now stabilizing after ~2 years)

with best regards

Andreas

[Kleinstein]

Getting 0.03 ppm/K is likely some luck, but a resistor array (e.g. 8 equal resistors) is a real option. With my ADC I got some 0.5-1 ppm/K for a circuit with 2 skale factors  (ref. amplification 7 to 28 V) and the ADC itself (more like an -1 gain amplifier).
The 2:1 ratio for the ADR1001 is a little easier, but one can still get a 10:4 = 5:2 ratio (for a 7.1 V ref) from 7 equal resistors.

[iMo]

I think one cannot generalize on the 7->10V stage based on a single (or couple) of successful builds with such small TCs as above, imho. Those are almost outliers

What Dieter indicates is that the version with the ovenized 7->10V stage is a safe and least expensive solution than an elaboration with statistical arrays (randomness) or manual fine-tuning of TC (time, labor costs). It means ie. with an 0.1C oven and 5ppm/C resistors you will always get the total TC<1ppm/C with no effort or risk.

[Dr. Frank]

I think one cannot generalize on the 7->10V stage based on a single (or couple) of successful builds with such small TCs as above, imho. Those are almost outliers

What Dieter indicates is that the version with the ovenized 7->10V stage is a safe and least expensive solution than an elaboration with statistical arrays (randomness) or manual fine-tuning of TC (time, labor costs). It means ie. with an 0.1C oven and 5ppm/C resistors you will always get the total TC<1ppm/C with no effort or risk.

Well, this technique of T.C. matching of resistors and T.C. characterization / trimming of the complete circuit has been done by FLUKE for all of their old reference instruments, like 730, 731, 332/335, and this statistical array plus T.C. trimming has been done by John R. Pickering in his Datron / Fluke 7000 references, very successfully.
I don't regard an ovenized 7=> 10V stage neither as a least expensive, nor as a simple solution, especially in regards of volt-nuts, who mostly strive for most simple and cost effective solutions, which also work practically. That's always my approach, to invest some good ideas and a bit of measurements for simple solutions, instead of over-engineering things. Btw., I work in R&D of the Automotive Electronics industry, where you have to save every tenth of a penny.

I can't remember that anybody has published his fully ovenized, practically working and fully characterized (10V) LTZ reference here. So, please think again about "no effort, no risk".

It's clear that my low T.C. 10V reference is achieved by chance, but I never made any statement, how probable or improbable this is. I also did not state any generalization of my result.
Therefore, I don't value both your statements, either that the likelihood for a pure resistive divider is rather high to end up with > 1ppm/K, or this is almost (always) an outlier result. None of us has a statistical idea, or can present a test series, how the probability for a good T.C. matching is, for the different suppliers of such PWW or BMF resistors.

Anyhow, John Pickering used 2 EA of those TiN (?) resistor arrays, in an entangled schematic, to create a stable statistical divider. It's obviously very successful in series production, as they could guarantee a very good timely stability .. the T.C. is trimmed internally, but I assume, that the overall T.C. is already quite low as determined by the arrays themselves.

branadic made similar experiments with a single array, and trimmed the T.C. by means of a short piece of copper wire. The timely variation was quite high, afaik.
If I understood Andreas correctly, he achieved a very low T.C. with a single (?) array as well.
Therefore, this method is not that elaborated, but already approved for volt-nuts.

In the end, why I insist on simple, already validated solutions is the following:

I know Andreas lab on-site, and RAX also described his man's cave in detail. Both labs suffer from big temperature variations (+15°C / - 5°C, or so), therefore, to anyhow do ppm/sub ppm metrology, one needs ultra low T.C.s for the references. I mentioned my exchange with AB Precision @ 0.3ppm transfer level; both our basement labs might have different room temperatures, (in fact, exactly the same), therefore this requirement is as well mandatory.

Even with such very constant temperature conditions (+/-0.2°C) in my lab, the residual T.C. of any precision instrument can be quite annoying .. like the T.C. of a transfer instrument like the HP3458A. I promise to soon write an article about my recent T.C. measurements on the 03458-66509 reference board, and the overall T.C., and how this might disturb sub ppm transfers.

Frank   

[Kleinstein]

A small oven for a constant temperatre is not that difficult to build. It is quite common for very stable quartz oscillators.
However the oven can only solve the TC part and not the long term drift, that is often also important and maybe even the more important part.

Though no guaranteed by the specs, using a combination of resistors from a single resistor array can give quite good TC and also long term stability.
Combining resistors in series / parallel is a way to get the wanted ratio from of the shelf parts (e.g. 8 equal resistors as TDP, or NOMC). Statistical averaging to get better matching is a wellcome bonus and if needed a 2nd array is still affordable.

For the voltnuts some trim (e.g. with the extra copper wire) and try, test and if needed rework is not so bad as it is for commercial production.

[Rax]

RAX also described his man's cave in detail. Both labs suffer from big temperature variations (+15°C / - 5°C, or so)

Frank   

Frank - that must be Andreas', but in my case, there's nowhere near such wild variations. As I said before, during the spring through fall, my bench variations have been at worst +-2-3C (around a midpoint of around 25C). In a typical/moderate daily cycle (so no heatwaves, etc.), there's less than 1.5-2C delta. Southern California (at least where I'm at) doesn't typically get frost at all the entire winter. Our normal yearly variations are very small, comparatively, and therefore, we almost never need to use our HVAC system, year round.

Interestingly, the garage (my bench) is always the warmest room of the house (meaning, it never gets cold in there, even if the larger rooms in the house may be a little chill). This space always stays very even and comfortable. The one exception is late in the afternoon when the sun hits the garage door directly AND it's during a heatwave, then I have to turn on the heat pump I have in the garage. Not ideal (a portable unit), but it can shave off the additional heat when otherwise the garage would go beyond 27-28C.

I'll be able to extend my observations on my bench temperature into this winter, which is a season I haven't had such close observations as I did during the past spring through fall.

[Rax]

I am able to start populating a Frank board, sourced from OSH Park (Jason designed).

What would be some of the recommendations for building it? I am aware of the warnings on using too long a time on soldering leads of the LTZ1000 and the Z201s. I've been soldering from childhood, and also having (hand)made and built PCBs for about as long, am obsessively attentive to keep the soldering time as short as possible. I never do over 4-5 seconds at a time, and that's plenty enough for perfect soldering joints/points. I'm also considering using dissipating aluminum clips for each leg removing some heat during soldering to protect the part.

I'm also going to be careful with touching the board (I actually always solder with vinyl gloves). I'll probably wash the board to be populated with IPA before starting. Post soldering I always wash the board off rosin and all crude with IPA.

There's also a 3D print cover for the LTZ1000 on xDevs - have people here found that useful? I assume it'd remove any residual sensitivity to air drafts due to convection and whatnot.

[Jacques]

There's also a 3D print cover for the LTZ1000 on xDevs - have people here found that useful? I assume it'd remove any residual sensitivity to air drafts due to convection and whatnot.

In my experience adding a cover already made a measurable difference for the LM399, so I'd certainly add one with the LTZ1000 as well. I normally use 3D printed ones (e.g. nylon); maybe others here have also experimented with metal covers?

Edit: note that this was within a larger instrument, however.

[Dr. Frank]

I am able to start populating a Frank board, sourced from OSH Park (Jason designed).

What would be some of the recommendations for building it? I am aware of the warnings on using too long a time on soldering leads of the LTZ1000 and the Z201s. I've been soldering from childhood, and also having (hand)made and built PCBs for about as long, am obsessively attentive to keep the soldering time as short as possible. I never do over 4-5 seconds at a time, and that's plenty enough for perfect soldering joints/points. I'm also considering using dissipating aluminum clips for each leg removing some heat during soldering to protect the part.

I'm also going to be careful with touching the board (I actually always solder with vinyl gloves). I'll probably wash the board to be populated with IPA before starting. Post soldering I always wash the board off rosin and all crude with IPA.

There's also a 3D print cover for the LTZ1000 on xDevs - have people here found that useful? I assume it'd remove any residual sensitivity to air drafts due to convection and whatnot.

RAX, do you have self-clamping tweezers with copper plating?
It's most important to not heat the LTZ1000 via its legs, so I clamp such a tweezers on the top side, while I do a very short soldering on the bottom side, probably 1..2 sec. only. I also use these tweezers for the precision resistors.
Both the LTZ, as well bulk metal foil resistors show permanent hysteresis of several ppm, when getting heated or cooled down too much.

I still use leaded tin, btw.

I also do not shorten the legs of the LTZ, that helps to keep the chip cool, and the length of the legs goes into the T.C. budget.
My assembly is w/o a cap on the top side, as displayed in the pictures. I assume, this also goes into the T.C. budget, but I never experimented, what the impact of an additional cap would have on the overall T.C. W/o a cap, the whole thermally insulated assembly / circuit gets warm, and therefore .. the whole circuit is sort of thermalized.

There's also no cap on the soldering side, as my PCB design and mechanical assembly is intended to create an iso-thermal plane on the bottom side, as recommended in the datasheet.
That means, if you orientate the PCB horizontally, add a thin thermal insulation around the whole PCB, like in my pictures, then all solder junctions should remain on the very same temperature.
The assembly / PCB should never be tilted when in operation, as this would create temperature differences on the junctions, as well the LTZ chip itself is prone to tilting, which might create shifts of several ppm. This effect was found by Andreas. I even tilted my HP3458A, and have seen this effect, which is not mentioned at all in the specification, in contrary to specifications for HP OCXOs.

Frank

[antintedo]

I found that BU-34C copper alligators work well, very convenient to use on small legs. When designing a board from scratch it's worth elongating the pads, heavily pretinning them and adding flux before assembly, soldering time can be reduced to just half a second

[Rax]

I started putting this together.

Wanted to collect some input on using sockets for the op amps ("IC1" and "IC3"). I'd highly prefer them (honestly, I always use sockets... for ease of repair or upgrading), but maybe there's the risk of getting some parasitic EMF created between dissimilar alloys on the leads vs. socket etc. But maybe that just gets added to the overall voltages budgets, internal to the circuit.

[deepfryed]

You can get gold plated dual wipe DIP sockets, they're a bit more pricier but I've seen them used in some test jigs. Alternative is gold plated ZIF sockets, but I reckon everything additional would add some TE offset / noise.

[deepfryed]

heavily pretinning them and adding flux before assembly, soldering time can be reduced to just half a second

I've seen mentions of lower temperature cadmium tin solder in literature, anyone used it before ?

[Andreas]

Hello,

I've seen mentions of lower temperature cadmium tin solder in literature, anyone used it before ?

a) its highly toxic. (even the damps during soldering).
b) you have only advantage when soldering copper to copper. No advantage when soldering Kovar (LTZ1000) leads to copper (PCB).

with best regards

Andreas

[aronake]

heavily pretinning them and adding flux before assembly, soldering time can be reduced to just half a second

I've seen mentions of lower temperature cadmium tin solder in literature, anyone used it before ?

I got hold of a roll of 60% cadmium and 40% tin solder. Also found and bought a small bag each of laboratory grade pure cadmium and tin. Still not used it, and well aware it is toxic. Plan to do some experimentation with this on TEMF.

[MK]

heavily pretinning them and adding flux before assembly, soldering time can be reduced to just half a second

I've seen mentions of lower temperature cadmium tin solder in literature, anyone used it before ?

I got hold of a roll of 60% cadmium and 40% tin solder. Also found and bought a small bag each of laboratory grade pure cadmium and tin. Still not used it, and well aware it is toxic. Plan to do some experimentation with this on TEMF.

The fumes when you try and melt the cadmium into the tin are meant to be quite severe. Needs a LOT of care to do safely from what I have read.

[Rax]

Interesting progress here, and I think I figured out some useful steps through the process (thank you Frank!). I'll likely update the original post to include a simplified, clear step by step guide to put this together. Again, my goal here is to help others working on this after me so they'd have a streamlined guideline available (and stop bugging the veterans!). The below assumes the builder is located in the US.

Once one has the PCB at hand:

• Populate the board, excluding IC3 and associated parts
• Dial in R10 to provide temperature compensation
• Measure the exact output off the particular LTZ part used
• Calculate the gain ratio necessary for that particular value
• Order appropriate resistors from TCC. You can be extremely close to 10V due to the capacity to customize very high quality resistors.
• Kick back and relax for the few days it takes them to fabricate your custom resistors. Single malt recommended for the evening completing the steps above. Make sure your LTZ kit is running to start aging
• Populate the rest of the kit and slightly adjust the gain ratio ladder, if necessary, to hit the exact target value. I wager that, if R12/R13 are judiciously chosen per LTZ output, R18/R15 can do all the rest of the work to sub-ppm levels.

Example: My LTZ outputs 7.159V. I started with a 10k/4k pair for R12/R13, but this landed me too high (10.0226V). If I instead get an 10.1k Z201, that'll land me at Vout = 9.99425V without any additional resistance on top of that "ladder." With a 20 ohm pot, I get a top of adjustable range of 10.02263V. An 10 ohm R18 makes that adjustable down to 9.99897.

So, with the values above, one should be able to adjust between 9.99897V and 10.02263V. A more narrow selection of R12 and R13 should allow for an extremely narrow trim range (low ppm or sub-ppm).

Now I have to figure out dialing in R10 - welcome recommendations - but I assume it involves running days-long logs of the LTZ output with R10 at different values between 200k and 1M and the one that returns the flattest curve on the largest delta T wins the big prize.

[Rax]

I'd be very interested in hearing about some methodologies employed by others to tweak the TC resistor (R10, 400k) when using the LTZ1000 version of the reference.

I've started by letting the kit run - so I'd also age it as much as I can - and logging 8.5 digit measurements (20s integration) over 24hrs (this far) on my Prema 6048. I guess 6.5 digits would probably be enough - and maybe a different type of reference would provide more reassurance of dissimilar tempco for relevant measurements - but why not have the additional granularity. I'm planning to do the same with different values for R10, and see which value seems to exhibit the lowest Vout variation per 1C over random deltas during the days each individual value is examined.

I've searched the relevant older threads to nausea but didn't come up with a clear and simple methodology. I've also looked at Frank's recent "3458A-related" series of articles (which obviously have an "A-version" at their center) in the main LTZ thread, but I'm looking for a simpler approach and practical points to set it up. If anyone has a link to such, please share. Many thanks in advance.