Ultra Precision Reference LTZ1000

[TiN]

janaf
I can do measurements for you as well, what is this intended for tho?

Also decided to make a calculator for LTZ, since there is LOADs of data already on web, but no place where everything valuable consolidated in one page.
Picking up bits here and there across all forums is kinda troublesome for voltnut beginners.
Here's what I got so far: http://xdevs.com/ltz1000/
If anyone have ideas - feel free to suggest.

[MK]

My take on the calculator is that one does not need R2 and R3 to be equal, the transistors are slightly different sizes anyway, and for R3, lower gives less noise, so the temperature stability can be as good as possible, whilst for R2, the zener is noisy anyway, so you might as well run at a lower current there for a higher vbe tc, that will reduce the sensitivity of the zener transistor combination a little bit.

[janaf]

I can do measurements for you as well, what is this intended for tho?

The intention is to see how sensitive the circuit is to resistor stability which simply tells where to put the money when selecting resistors. My measurements indicate that some of the data in the datasheet are not valid. As an extension, it may be possible to select resistors characteristics to compensate each other.

[TiN]

But how you can judge from different references we have, as well as measurement gear and resistor themselves?

[janaf]

Long story, please see my previous posts. A summary of my measurements is:

• The relative changes in output voltages versus changes in resistor values seem quite repeatable/predictable.
• The changes in output voltages versus changes in resistor values are not dramatic but rather smooth or linear over wide ranges.
• This means measurements can be done in fairly large steps and are therefore quite easy to measure.
• The data sheet values and my measurements match for some resistors, but are very different for other.

So what I'm asking is if someone else can make the SIMPLE & FAST test I suggested, to verify (or reject) the results I got.

[max666]

But how you can judge from different references we have, as well as measurement gear and resistor themselves?

He has to compare his results to other references, that's an important point. If he finds a particular resistor isn't as critical as mentioned in the datasheet, but other members can't replicate the result on other references, then it's not a general result. It might just be a fluke with his reference.
Different measurement gear should not be that critical, because he's only asking how big the relative change is. Besides, all the voltnuts here probably first thing in the morning they switch on a 6 1/2 digit (at least) multimeter, before they even have coffee
And regarding different resistors, I think there isn't a model for the reference, if I'm not mistake. So having these measurements at different resistor values would be necessary for building a model.
I'm guessing this being the case, from reading along the thread. I'm not good enough to be able to call myself a voltnut yet 

[TiN]

Alright, you got it going. But to make it proper, it will take two days.
I had my 2002 with open hood last days, got it assembled back and took one of LTZ's to do janaf's measurement request.

KX module #3 (Red), +7.1367 VDC, LTC2057, VHP203: 70K, Z202's: 13K+1K tempset, 120R. Powered from +12V moto battery.
KI2002 input - LTZ output.
KI2001 = Battery voltage monitor.

Tonight it will capture data with 120R, tomorrow with 10K000 VHP202Z.

[janaf]

Hi,

I wrote a summary of some of my measurements on R1 and made an effort to make it understandable, the why, how and results.

Attached.

[janaf]

Alright, you got it going. But to make it proper, it will take two days.

Great!

I also just posted a summary page on some of the measurements I did on R1.

[TiN]

UPDATED: With both data collected.

Initial reference data with 120R only:



With 10K 0.01% VHP203Z in parallel.



RAW datalog 120R: TXT file, 1.5MB
RAW datalog 120R+10K : TXT file, 1.5MB
Format (columns left to right) is battery voltage (input for reference), LTZ1000A output, hour, minute, second, day, month.

Measurement setup:





After 10K added to 120R:



Output voltage raise ~ +134.9 uV, or +18.90 ppm.
Measurement little bit crippled, as battery voltage not exactly same, but should not be big enough impact.

Thank you for PDF, I like your way of reporting things using fixed PDF, which easy to save/store for future use, rather than forum post

[janaf]

Thanks! Calculating exactly like for my values, I get a change of -0.00159 ppmV/ppmR, slightly higher than my value, -0.00128. Lymex posted a value of 0.0014 some time ago and so did I (-0.0014) from another series / another board and LTZ1000 but all four are quite similar and all are far away from the 0.01 in the datasheet!

Maybe the data-sheet data where taken on the LTZ1000 and different, not adjusted for the LTZ1000ACH??

Anyone has a plain LTZ1000 (non-ACH) to test?

[TiN]

I have LTZ1000 version as well, with different setpoint tho (12K0 + 1K0).
Can repeat test on it as well. Also your PDF is somehow corrupted formatting-vise.

[Dr. Frank]

TiN,

you obviously get the same behaviour, i.e. an attenuation of the instability of R1 of about 630.

(dR/R = -11857ppm, dU/U = +18.7ppm)

To first order, R1 together with the transistor constitute a constant current source for the zener diode.

The change of R1 changes this zener current, and so you can calculate the zener diode differential resistance, in this case, that is 2.97 Ohm (*)

That explains the very small influence of R1.

The datasheet specifies a change of typical 80mV, maximum 240mV by changing the current from 1mA to 5mA.. From that very big change, one would calculate 20..60Ohm for the differential zener resistance.. maybe LT has taken this value for the stability calculation.

As the LT references seem to be manufactured very similar, I assume also, that this is a generally valid result.

And yes, please, repeat that measurement with LTZ1000, as I also have 5 EA waiting.

Frank

(*) 0.45V Ube assumed

[janaf]

Reply to this older post. This may make the R1 even less sensitive, can't quite wrap my brain around it. I think I'll do some measurements...

....Bob Dobkin's ap-note on page 6 of the data sheet shows a way to make the LTZ1000 more stable at room temperature-- but this same idea can be extended to a heated device.  The circuit shows a potentiometer-- and this is probably not a good idea-- this resistor will be smaller than the 120-ohm resistor [because you need less than one Vbe for proper compensation].  As a wild guess, it will be somewhere between 15-ohms and 25-ohms. ...

[janaf]

Somehow I managed to export the whole raw ugly data...  File updated now.

I have LTZ1000 version as well, with different setpoint tho (12K0 + 1K0).
Can repeat test on it as well. Also your PDF is somehow corrupted formatting-vise.

[janaf]

I got the question on the other resistors. In my measurements

- R2 is slightly more important than the datasheet indicates, -0.4ppmV/100ppR compared to 0.3ppmV/100ppR in the datasheet. Not a very significant difference (apart from the sign!)
- R3 was less sensitive. I got -0.07ppmV/100ppR compared to 0.2ppmV/100ppR. A quite significant difference but low values anyway.
- R4/R5 I got slightly higher value than the datasheet and same sign, positive. Not very significant difference.

- For R2 the sensitivity decreased a bit with increased resistor value (range 40K-100K), but not very significant
- For R4/R5 the sensitivity decreased a bit with higher ratio/higher temperature setting, but not very significant
- For R3, the sensitivity was unchanged over the range (40K-100K)
 

[janaf]

For a more complete model, one should do the full Jacobian matrix, i.e. vary each resistor versus all the other. That would take at least 15 measurement series, each with, for example, 10 data-points. Quite an effort.... I'm only trying to map which ones are significant and which are not.

On the whole, the result seems to be a confirmation that the R4/R5 divider are the most critical part without comparison, while R1 seems less important than R2. R3 is relatively unimportant, it seems no exotic hi-end $$$ resistor needed there.


While at it: if the R4/R5 are important, a divider to reach 10.00000V is much more difficult than R4/R5........

[janaf]

Thanks, I should have read better.... I guess I should do some measurements on that....

[TiN]

Updated post with measurement 120R vs 120R+10K. Sorry for little drift in second case, as ambient temperature reduced from 21.5°C to 19.5°C, raining days here 
Time to test LTZ1000CH.

[janaf]

Thanks,

If I take data from 20% to 40% of the 120+10K file, would that be approximately the same temperature as the second half of for the 120 only?

I'm not looking for ultimate accuracy as there may be individual differences anyway, I just want confirmation just if the attenuation is in the order of 1:100 or 1:700......

PS: This circuit is very immune to power supply voltage variations as it only powers the OP-amps and the heater transistor. I have verified this by measurements by varying the supply from 15V to 9V and it did not show any output change, down to the resolution of my 7.5 digit DMM. (It might depend a little a bit on what op-amp you use). But I would not worry about supply variations.

[TiN]

I do agree regarding input voltage immunity. I use LTC2057 on all except one references.
Reason why have battery power is to isolate reference from input AC power noise, as it's not on primary spot which had dedicated mains line during this test.

LTZ1000CH test:

Unit Rev B01 PCB, LTZ1000CH, LTC2057, 12K/1K heater , base 7.1304749 VDC @ +23.6°C ambient (power on from cold state).



1. 120R



2. 120R+10K



About +134uV as well, so same with LTZ1000CH

[Andreas]

Reply to this older post. This may make the R1 even less sensitive, can't quite wrap my brain around it. I think I'll do some measurements...

....Bob Dobkin's ap-note on page 6 of the data sheet shows a way to make the LTZ1000 more stable at room temperature-- but this same idea can be extended to a heated device.  The circuit shows a potentiometer-- and this is probably not a good idea-- this resistor will be smaller than the 120-ohm resistor [because you need less than one Vbe for proper compensation].  As a wild guess, it will be somewhere between 15-ohms and 25-ohms. ...

No, it won't make R1 less sensitive, but it will make the R4/R5 ratio less sensitive [about 1000:1 vs. 100:1].

But what does it "cost"?

Even if its only around 20 Ohms with 5 mA it will increase the output of the reference by 100mV.
These 100 mV have the full tempco of the 20 Ohms resistor.
So the 20 Ohms resistor contributes 1/70 to the 7V output voltage.
So now this gets the most sensitive resistor in the cirquit.
The datasheet value is 200 Ohms -> only 1/7 reduction on resistor drift.

According to the datasheet of a Z201 all resistors below 100 Ohms have a higher tempco spec than those above 100 Ohms.

Maybe I am wrong?

So Ken: what are your measurement results to this suggestion?

With best regards

Andreas

[janaf]

So basically, we'd de-sensitise R4R5 but get a new pair R1-"R6" that is slightly more critical?

[max666]

If you have them wirewound by hand anyway, why not wind both resistors on the same bobbin?



Like example 2 or 4, but feed all 4 ends out separately.

[lars]

If the 20ohm resistor "R6" according to Andreas has 1/70 sensitivity and 120ohm R1 according to Janaf, Lymex and TiN has about 1/700 sensitivity it seems strange that they should track?

If it was only the same 5mA current through the resistors R1 as "R6" it should be more like 6 times higher sensitivity on R1?

Has I missed something?

Also if Mr.Pettis can make matched wire pairs of 20 and 120ohms he might be able to do 1k and 12 or 12.5k instead of an extra precision resistor?

Lars Walenius