Ultra Precision Reference LTZ1000

[Andreas]

Hello,

I always use a dummy LTZ when trying a new layout or a new OP-Amp to see if there are oscillations or other issues.
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg829328/#msg829328

with best regards

Andreas

[CraigD73]

Just starting a LTZ1000A project. After many hours of reading some great posts on the forum I have a fair grasp of what is needed.

I could still use a couple of recommendations for the +12V power supply regulator and OP amp to use for the ordinal 7-10V converter.   I assume one would want a regulator with low noise and low TC.  For the OP amp I was thinking about the zero drift devices with low input offset.  But I am not up to date on who has the best parts in these areas.

What about a switching regulator to allow 5V or Li battery power?  They have switching noise which I believe could be mitigated.  Don't think I would put the switcher on the same board and it could be followed by a linear.  Battery power does have some appeal.

I am not really a voltnut but I would like to to the best I can as a learning experience.  Most likely not going to use the extra pricey .005% resistors just the Vishay low TC .01% foils.

I have had an LTZ1000A running on a test board for about two weeks now.  Built the board to burn-in the LTZ1000 and for something to play with.  The little cube next to the board is a pseudo-LTZ1000 built with a zener, heater resistor and a couple of transistors potted in some foam.  It worked but I forgot about the Vbe and used a 7.5V so it makes and 8.1V.  It was also some what unstable because of thermal lag.  The resistors are just junk box .1% MF's and while I did pay attention the Kelvin paths I didn't optimize the layout at all.  I was actually going to hand wire or island mill some copper clad but at the last minute I decided to send it to JLC PCB in China along with a 3GHz prescaler board for my HP34401A counter.  The quality and delivery were really great, received 10 pcs in 5 days all for $2.00 US.

The references reads 7.0707V on my HP 6.5 digit  HP34401A.  This is some what lower than what others are are reporting but still within the specification.   

[try]

Hello craig73,

I am not really a voltnut but I would like to to the best I can as a learning experience.  Most likely not going to use the extra pricey .005% resistors just the Vishay low TC .01% foils.

why do you intend to use 0.01% or even 0.005% ones?
They should not provide any advantage for the output voltage of the reference in terms of temperature or long-term stability over the simple ones in 0.1% precision.

I see two trimmers on your board. What do you use them for?
The picture is a nice one but a wiring diagramm might show a reason for the relatively low voltage on the output.
It could be interesting where you diverge from the original wiring in the data sheet.

Regards
try

[CraigD73]

I am new to all this and not in any way an expert.  What I have gleaned reading lots of posts the two critical areas of the circuit are the divider used to set the heater temp and the resistor used to set the Zener current.  For the divider, the ratio rather than the resistor values is the important parameter,  it needs to have the best TC and long term drift characteristics. There is some debate as to exactly what ratio to use, seems to depend on your view of long term drift of the device vs max environmental usage temp.  The 120 ohm Zener current sensing resistor also needs a good TC and long term stability but it value depends on what Zener current you wish to use, this seems to be a noise vs long term stability issue.  Not knowing enough I am just going to stay with the recommended values.  For the three critical resistors your choices seem to be WW or metal foil.   Not looked too much at the WW resistors but have looked at the Vishay Z201 and VSMP smt metal foils.   Their sweet spot seems to be in .01%, they don't offer much at lower precision.   A normal .1% film resistor is going to have way worse TC and drift than the Vishay foil resistors.  Going to .005% or .001% adds a lot to the cost.  Still need to look at WW resistors.   A custom Vishay .01% VSMP runs about $13 so it's not a killer for the three critical resistors.   Vishay can make almost any value by programming shunts included on the raw stock material.

One of the trimmers is used for the 7-10V converter -- didn't want to select resistors.  The other trimmer is on the heater divider to play with the temps but will be fixed in the final design.  The board uses normal 1% resistors and not built to be an accurate reference, it's just for exploration and burn-in.
Craig

[Kleinstein]

The board picture shows a LT1179  quad OP, that is rather slow. This can interfere with the LTZ1000 voltage loop and even make it oscillate. If you have a single supply the LT1013 is usually the best choice for the OP: accurate enough and single supply and not a chopper that can produce etwa trouble.  So for the LTZ1000 ref. + 7 to 10 V amplifier it usually takes an LT1013 and a single chopper OP or high precision BJT based one (e.g. OP177).

For the temperature setting divider, a very high quality thin film resistor array may be an option - though it is hard to find suitable ones in small quantity.

[cellularmitosis]

I am new to all this and not in any way an expert.  What I have gleaned reading lots of posts the two critical areas of the circuit are the divider used to set the heater temp and the resistor used to set the Zener current.  For the divider, the ratio rather than the resistor values is the important parameter,  it needs to have the best TC and long term drift characteristics. There is some debate as to exactly what ratio to use, seems to depend on your view of long term drift of the device vs max environmental usage temp.  The 120 ohm Zener current sensing resistor also needs a good TC and long term stability but it value depends on what Zener current you wish to use, this seems to be a noise vs long term stability issue.  Not knowing enough I am just going to stay with the recommended values.  For the three critical resistors your choices seem to be WW or metal foil.   Not looked too much at the WW resistors but have looked at the Vishay Z201 and VSMP smt metal foils.   Their sweet spot seems to be in .01%, they don't offer much at lower precision.   A normal .1% film resistor is going to have way worse TC and drift than the Vishay foil resistors.  Going to .005% or .001% adds a lot to the cost.  Still need to look at WW resistors.   A custom Vishay .01% VSMP runs about $13 so it's not a killer for the three critical resistors.   Vishay can make almost any value by programming shunts included on the raw stock material.

One of the trimmers is used for the 7-10V converter -- didn't want to select resistors.  The other trimmer is on the heater divider to play with the temps but will be fixed in the final design.  The board uses normal 1% resistors and not built to be an accurate reference, it's just for exploration and burn-in.
Craig

Hi Craig,

It turns out that the datasheet's ranking of which resistor is more important seems to be a bit off (one of the 70k resistors is actually more important than the 120R resistor).  See https://www.eevblog.com/forum/metrology/resistor-set-for-ltz1000-positive-standard-7v-circuit/msg1386014/#msg1386014

[CraigD73]

Yup --the LT1179 is all I had in my parts stock -- will have to order some LT1013's
Craig

[try]

I am new to all this and not in any way an expert.  What I have gleaned reading lots of posts the two critical areas of the circuit are the divider used to set the heater temp and the resistor used to set the Zener current.  For the divider, the ratio rather than the resistor values is the important parameter,  it needs to have the best TC and long term drift characteristics. There is some debate as to exactly what ratio to use, seems to depend on your view of long term drift of the device vs max environmental usage temp.  The

Forget the useless debate. 
It is the maximum environmental usage temp that puts a floor under the operating temperature. If you want the reference to be stable with varying ambient temperature the heat regulation has to work properly. Give it some headroom by fixing the regulator temperature adequately above. Even with the heater disconnected, you have a temperature increase. In my case of the "non-A"-version it was like 3-7 degrees celsius if I remember correctly.
There is a really helpful description how to measure that temperature posted by doc Frank somewhere.
That description discloses the real operating temperature of your reference.
The precision of that measurement depends on two things:
1. Your LTZ, being off, having adapted ambient temperature.
2. Your multimeter being fast when measuring the initial Ube-voltage relating to ambient temperature.

Once you found that temperature figure you can calculate how to change the operating temperature taking into account the individual Ube of your LTZ in question.
...

Not looked too much at the WW resistors but have looked at the Vishay Z201 and VSMP smt metal foils.   Their sweet spot seems to be in .01%, they don't offer much at lower precision.

In case of the VSMP the table in the datasheet looks like the absolute resistor value determines the possible precision and TC.
If you need 120 Ohm you have no choice and the resistor does not get better if you order it in 0.005%.
I have the impression that everybody is using through-the-hole resistors for the divider instead of SMD ones.

Your possible options are:
1. Using simple metal film resistors (p.e. Yageo MF).
I used them in my prototype and was rewarded with a shaky output  of a couple of ppms up and down upon temp variations of +- 5 degrees celsius in my room. 
2. Using something slightly better, like PTF56. At least the spec sheet shows better figures than the standard Yageo metall film resistors. I replaced the 120 Ohm and the divider and my prototype looked stable then.
3. Use metal foil or WW resistors. Examples: Vishay S102, GR's "econistor" sold by Rhopoint or the often quoted "Ultrohm Plus" manufactured by fellow member Edwin Pettis.
4. Use something special with good tracking ratios:
      a) poor man's tracking solution: try to couple your resistors thermally with copper foil (see pictures from Andreas, p.e.)
      b) LT-5400 can be used for a certain divider ratio
      c) Vishay 300144 + 300145 and their Z-variant offer you a divider product manufactured according to your specs
5. Expensive solution with high impress factor: Order your resistors or your divider in an hermetically closed housing. (see pictures from tin). Doc Frank did the calulation, conclusion: This luxury does not improve the performance. But if it makes you sleep better, why not?

A normal .1% film resistor is going to have way worse TC and drift than the Vishay foil resistors.  Going to .005% or .001% adds a lot to the cost.  Still need to look at WW resistors.   A custom Vishay .01% VSMP runs about $13 so it's not a killer for the three critical resistors.   Vishay can make almost any value by programming shunts included on the raw stock material.

Yes, but the precision (.1%) is not the determining factor. You get Vishay S102 in 1% and all standard foil properties apply regardless of precision. The only difference with 1% is that the laser cutting the resistor templates stops cutting earlier.

One of the trimmers is used for the 7-10V converter -- didn't want to select resistors.  The other trimmer is on the heater divider to play with the temps but will be fixed in the final design.  The board uses normal 1% resistors and not built to be an accurate reference, it's just for exploration and burn-in.
Craig

Out of the data sheet for a simple trimmer in 10K:
General offer ±250 ppm/℃ (±100ppm/℃ is available if    => 250ppm through the wiper! 
Pettis recommends Bourns 3250 and 3290 (expensive, but. mil-grade)
Fellow member branadic used one of those Vishay Models: 533, 534 or 535 (still expensive)

Regards
try

[cellularmitosis]

This luxury does not improve the performance.

Are you sure about that?  I was under the impression that non-hermetic vishays had something like 1ppm per % relative humidity.  That’s at least a 20ppm swing from season to season, which is quite a bit larger error than what you will see from temperature fluctuations in your lab.

Granted, with the error attenuation of the LTZ circuit, this should be a relatively minor effect on the output, but I wouldn’t say that it doesn’t improve performance.

[RandallMcRee]

  b) LT-5400 can be used for a certain divider ratio 

THe LT5400 is not recommended in this application (ref Mister Diodes who claimed that this came from LT app engineers). I did try it and got bad results. I moved to a Vishay divider set.

Presumably in the LTZ circuit, there is too much current to eliminate self-heating.

[try]

Hi cellularmitosis,

This luxury does not improve the performance.

Are you sure about that?  I was under the impression that non-hermetic vishays had something like 1ppm per % relative humidity.  That’s at least a 20ppm swing from season to season, which is quite a bit larger error than what you will see from temperature fluctuations in your lab.

Roughly the same. Say 0,05ppm/K * ( (28-15) K ) =0,65 ppm (I have no exact figure of the TC!)
20ppm resistance variation translates into 0,52 ppm (if I have set up my spreadsheet on voltage sensitivity correctly)

At least this is cyclical and does not cause an unidirectional trend.

Granted, with the error attenuation of the LTZ circuit, this should be a relatively minor effect on the output, but I wouldn’t say that it doesn’t improve performance.

No, I am not sure because I only looked at the long-term drift effect of the hermetic enclosure. The drift reduction does not play a role because the zener drift dominates if I got doc Frank's statements correctly.

I am not aware of your impression as I ignored the humidity aspect of it!
My environment varies roughly from 40% to 60% only.
I assume Lars Walenius or branadic could contribute here.
Do you have any findings?

Regards
try

[try]

Hi RandallMcRee,

  b) LT-5400 can be used for a certain divider ratio 

THe LT5400 is not recommended in this application (ref Mister Diodes who claimed that this came from LT app engineers). I did try it and got bad results. I moved to a Vishay divider set.

Presumably in the LTZ circuit, there is too much current to eliminate self-heating.

Could you quantify your bad results somehow? What might be considered a bad result for Andreas could be a good one for me, p.e.
Where do you use the LT5400? To set the temperature or for the voltage raise from 7V to 10V?

[RandallMcRee]

Hi RandallMcRee,

  b) LT-5400 can be used for a certain divider ratio 

THe LT5400 is not recommended in this application (ref Mister Diodes who claimed that this came from LT app engineers). I did try it and got bad results. I moved to a Vishay divider set.

Presumably in the LTZ circuit, there is too much current to eliminate self-heating.

Could you quantify your bad results somehow? What might be considered a bad result for Andreas could be a good one for me, p.e.
Where do you use the LT5400? To set the temperature or for the voltage raise from 7V to 10V?

Was specifically referring to R4/R5 ratio set.
7v-10v (opamp gain setting) is fine. OPamp gain set is what the LT5400 was designed for.

[CraigD73]

Ok, I just did a quick test on the sensitivity of R2 the 70K for the Zener.   Added a ~70meg in parallel with the 70K which by my calculations reduced the value by about .1% (1000ppm), I see no change in output voltage on my 6-1/2 digit HP DVM.  That is not to say it didn't change but I can't see it.   Should be fairly easy to get std cost .1% or even 1% with 50ppm TC's.

Guess I will need to start the hunt for an 8-1/2 digit meter.

If I get time I will chedk to see what it takes to see a change in the output.    Got to be a little careful I don't want to smoke something.

Craig

[try]

Hello CraigD73,

Ok, I just did a quick test on the sensitivity of R2 the 70K for the Zener.   Added a ~70meg in parallel with the 70K which by my calculations reduced the value by about .1% (1000ppm), I see no change in output voltage on my 6-1/2 digit HP DVM.  That is not to say it didn't change but I can't see it.   Should be fairly easy to get std cost .1% or even 1% with 50ppm TC's.

Guess I will need to start the hunt for an 8-1/2 digit meter.

This is not necessary.
Your 1000ppm change of R3 should yield a 3ppm change in output voltage.
According to measurements from lymex rather 4ppm. There is a table around in this thread that compiled the sensitivies measured by different forum members.
I don't have access to my own measurements right now unfortunately.

Maybe the contact failed?
Try 10M Ohm instead.
The effect can be seen on a 6,5-digit multimeter.

Regards
try

[CraigD73]

Did another test on R2, by paralleling my 71.5K resistor with a 10 Meg.  I see about 200uV rise  in the Vz which seems to calculate out to .39 PPM for a 100 PPM (.01%) change in resistance.   A little more than what the data sheet indicates but not way out of wack.   I am somewhat limited by the lack of voltage measurement capability with a 6.5 digit meter but its not orders magnitude off.   Here are my numbers -- someone can check my math.
R2  = 71.5V   (1%)
R2 ||  10Meg  (5%)  =  70992.4
Vz @ 71.5K = 7.07056  Vz @ 70.993K  = 7.07076   Delta V = +200uV  (good enough for first order approximation)

Think I will stick with the a 69.8K .1% Panasonic or TE  a +/-10PPM TC MF at about a buck.

[kj7e]

You want the lowest TC parts here, the exact value is not very critical.  So if you find a 3ppm/C TC 1% resistor go with it every time rather than a 0.005% 20ppm/C TC part.

[try]

Hi CraigD73,

 someone can check my math.
R2  = 71.5kV   (1%)

Your math is OK.
By the way, I measured a sensitivity of 0,4ppm as well in my prototype board.

[MegaVolt]

Guess I will need to start the hunt for an 8-1/2 digit meter.
This is not necessary.
...
3ppm change in output voltage.
...
The effect can be seen on a 6,5-digit multimeter.

34465 6,5-digit multimeter. 24 hour error = 13ppm

How can they measure changes in 3 ppm?

[try]

Hello Megavolt,

that measurement does not take 24 hours.
You measure the voltage, that might take some seconds.
The multimeter does not jump 13ppm when you are measuring with the parallel resistor a few seconds later.

[MegaVolt]

measuring with the parallel resistor a few seconds later.

I was not attentive It was about jumping relative measurements in a short time.

Thank you corrected.

[keen101]

So, i know this is many years too late,  and maybe there is zero Interest in this anymore, but there were many people who had wanted an LTZ1000 board way back when i posted my HP/Agilent Clone on my blog several years back (wordpress.com/2013/02/12/hpagilent-03458-66509-7v-dc-reference-clone/).

Since so many had asked me for one,  i finally made a few boards and put them on ebay. It seems no one wants one anymore,  so maybe it was a waste of time, and if so, oh well, still a neat little board I'm proud of.

Here is the ebay link if anyone is interested.  Thanks. (https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F113011267007).

[TheSteve]

I think the problem may be that your price is a little bit too high. You can buy original 3458A A9 ref boards for less quite often on ebay and they use the LTZ1000ACH.

[Dr. Frank]

Nice board, but the price is way too high.
Also, it's lacking some information, like oven temperature set point.. is it still 95°C (15k/1k divider), or did you adapt it to the non-A version (at least 10°C less), and for metrology application, i.e. smallest possible expected drift, like 13k/1k and 65°C?

Frank

PS: And the module should have been trimmed for lowest T.C.

[Echo88]

About 130€ BOM and you want 350$. Price is way to high and would only be okay if it would already have a long-term-burn-in and calibration data so one can use it directly in a used 3458A for example.