KX Reference

[kj7e]

Sorry for all the photos Conrad 

Had some closed cell hi-temp foam laying around so made top hat, reduced the heater current draw by almost 1mA while out in the open with moving air in the shop.









Edit, I found air moving under the LTZ1000A had just as much effect as directly on the metal can.  So slight modification to the foam hat.  Enlarged the hole for the can so there is no direct contact and made it taller to come all the way to the board.  This very effectively stabilizes the effects of air turbulence.  Originally I wanted to keep the foam off the board and traces.  Think I may use some silicon to tack it to the board. 

Made a short video showing the effects before and after;
https://youtu.be/sYhQBudCLiU

[Andreas]

Hello,

do not forget to (air-)shield the pins from the other side of the PCB.

with best regards

Andreas

[rigrunner]

@TiN, thanks a lot, but
please give me a tip, from where i can get these pcb's one for lm399 and one KX?

Prompted by Svangool's LTZ project upload, and with help from the guys at OSHPark for the drill file conversion, i've uploaded a shared project for TiNs LM399 KX-ref PCB Version S02 here: https://oshpark.com/shared_projects/oaEZl86b (with TiNs permission).

As per Svangool's project the minimum order at OSH Park is three boards, these cost in total $49.80 ($16.60 per board), including standard international shipping, excluding any import taxes or other costs.

These are as yet untested.

RR

[Muxr]

Has anyone had luck running the KX reference using the 100K values for the original? 70K   (R6,R7). I thought I saw somewhere that this is a viable substitution. I got all the other resistors but 70K is hard to find.

I would like to substitute them with this resistor: http://www.mouser.com/Search/ProductDetail.aspx?R=Y0062100K000T9Lvirtualkey59450000virtualkey71-S102KT100K00TB

[hwj-d]

I will try 80k, 2x2 of them:
http://www.ebay.com/itm/110983067006

They and other parts+pcb's are on the way. So, assembly will be next month so far.

[0.01C]

I will try 80k, 2x2 of them:
http://www.ebay.com/itm/110983067006

They and other parts+pcb's are on the way. So, assembly will be next month so far.

try
https://world.taobao.com/item/521116800640.htm?spm=a312a.7700714.0.0.2LAU7Z#detail

[Alex Nikitin]

Hmm, I'm not sure if it is OK from the forum rules POW  , I have a number of Vishay S102C 87.15K 0.01% resistors, about 18 years old, still very accurate. That is what I plan to use for my own LTZ1000 boards in place of 70K resistors, however I can sell some of these for LTZ reference use and much cheaper than prices on Digikey/Mouser/etc. for S102 types. PM me if it is of interest.

Cheers

Alex

[kj7e]

Been working with a Die Cast aluminum Bud Box where the KX reference and buffer/follower boards will reside.  Decided to ovenize it, set at a reasonable 35 deg C (one of the reasons I decided on using higher temp 13K for the heater).  The oven control circuit is working very well, and seems to hold the internal air temp to well within 0.1 deg C.  Still playing with the feedback gain and placement of the thermister for best performance, its a very slow process.  Meanwhile the KX board cooks away with 1 cent metal film resistors and has not budged, Ill have a 1000 hours on the can by the time the Vishay parts arrive.  My plan is to build everything in a small portable enclosure with 10AH of battery power.  Photos and test results to come.

Used some closet shelf angle brackets to make a nice mount for the KX board;


Initial testing to see how much power would be needed to heat the box;


Decided on 2x 25 Ohm 10 watt resistors in series for a 50 Ohm heater load;


This will heat the box from 20 deg C to 35 deg C in about 20 min with 20 volts at 0.4 amps (8 watts)


Playing with the feedback gain;


Still very much in proto stage, but its coming along nicely;


Edit, some temp ramp and stability graphs;

From 20 deg C its within half a deg of 35 in about 30 min.


Zoomed in, 0.5 deg FS, it holds within 0.1 dec C.  I'm at the limits of my temp measuring ability, not sure I can tune anymore ripple out, but I'm happy with this.

[kj7e]

Building the portable battery powered enclosure;









After final placement of the 10K NTC sensor in the oven and a second layer of foam insulation temp stability is far below +/- 0.1C.  even better than +/- 0.05C best I can tell.  I'm now using a second precision 10K NTC sensor and my 34465a 4W to measure the air temp inside the oven, this seems to be far mare accurate than the K type thermocouple .  The sensor at 35 Dec C is 6,531.0 Ohm and 270.5 Ohm/Dec C at this point.  So 27.05 Ohm/0.1 Dec C, with the trend chart 27 Ohm FS, the line is nearly flat.



Warm up plot from 25 Deg C to 35 Deg C;


Full scale is 0.1 Deg C;


Edit, final oven control circuit attached.

[cellularmitosis]

Very cool!  Er, I mean warm!  Duh, nice work!

[rigrunner]

I was planning on ovenizing my LTZ boards.  Just waiting on parts to finish off the second build.
Why did you choose to mount the resistors outboard of the enclosure kj7e?

The OSH lm399 boards have arrived.

[RandallMcRee]

Question for the LTZ1000 experts.

I am putting together a new KX reference board.

I have some LTC5400 resistor packs which I am planning on using for R4/R5. On hand are the "LT5400-8 1k 9k    1:9" and    "LT5400-3 10k 100k 1:10". Looking for a nominal 1:12 ratio.

The -8 with suitable parallel series combos gives a ratio of 900 ohms/10000 ohms.
The -3 gives a better fit to 1:12 with 9090/110000 ohms.

Is it a concern that the -3 absolute values are 10 times higher than the data sheet spec?

[kj7e]

The heater resistors?  Two main reasons, easier to wire, less erroneous energy/possibility of noise within the shielded housing.

Made some progress on the overall enclosure;






Still need to mount the pair of 5s 5000mAh Li-Po's and build the input power switching circuit (auto switch to external power if over 18v in).  Once the Vishay resistors come in Ill finish the KX board and mount it in the oven.  After a month of running it reads just a few uv below where it did when I first powered it on, and that's with cheap 1/4 watt resistors.

[Kleinstein]

A much higher absolute value for the divider is a problem. There is some base current flowing through the divider and this current can change with aging. Often also 100 K resistors are less accurate, as this is in a range where leakage starts to get important.

The ratio of 1:12 is still a rather low temperature.

[RandallMcRee]

Is there a formula that determines the oven temperature given R4, R5?

I've read extensively but not seen it...

Thanks!

[cellularmitosis]

Is there a formula that determines the oven temperature given R4, R5?

I've read extensively but not seen it...

Thanks!

I haven't read through every page of the LTZ thread yet, but I did find this: https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg380638/#msg380638

From that, I determine that T = 310 - 7*500*R5/(R4+R5)

Here's a table from my notes:

Note: I do recall reading somewhere that the formula isn't exact: there could be up to a 5 degree offset from the formula due to manufacturing variations.  Don't quote me on that though.

[Muxr]

The right resistors finally arrived. I put my KX in one of those mini Pelican cases, nothing special:



The heat shrink tubing is hiding a protection diode and some ferrite beads. The waiting (burn in), begins.

[Muxr]

I am pretty impressed with what I am seeing so far in terms of TC and noise. This is compared to my best reference, Doug Malone's 10V LT1021 (battery powered). LTZ1000 KX is powered of a lab power supply. Built from scrounged up foil resistors from Ebay/China, and no doubt still has a lot of hysteresis to burn through. It's only been on for a few days, since the soldering iron touched it.





I have enough resistors to build a second one. Any advantage to going with a LTZ1000ACH version for the 2nd one?

edit:
Battery powered and semi shielded by a tin box:

[TiN]

Try without aircon, as those oscillations aren't helpful . If I'm reading your graph correctly, you see 1.6ppm pk-pk? That's bit much, I usually get around 0.5-0.6ppm pk-pk for week long captures, or 0.3-0.4ppm pk-pk daily. Example here and here, with K7168 switch scanner in place.

Before fellow voltnuts here start blaming chopper opamps, CH2 reference on my graphs is "proper" HP 3458A A9 for comparison reasons .

Also if you measure ref like it's shown on photo - add foam cap on top and bottom of the zener, even when it's inside the box. That tame drafts even in still air a bit.

As for ACH version, if you want to try - sure, why not, but unless you running on battery, you wouldn't be able to separate data result from either CH or ACH.

[MisterDiodes]

Muxr: Use the "A" version.  The non A requires that extra TC compensation resistor and in 5-10 yrs time that will need a re-trim for best performance - otherwise it can add more trouble than it solves in the future.  That's the whole point of the "A" version (it doesn't need the extra resistor), and that's why it was developed for 3458a.   

Randall:  LTC5400 diffused resistor packs don't have a lot of use around LTZ circuits.  Suggest you just get some PWW made from Pettis or GR or Rhopoint, etc.  Those will give you better performance for a cheaper price.  You don't want a 12:1 ratio either.  If this is your first unit stick with 13k/1k as datasheet tells you or if you know it'll never get warm maybe 12.5k over 1k.  Plenty of decades 3458a's don't drift 1ppm /yr with their standard 15k / 1k, and learn from that. 

You select your heater resistor ratio for your enclosure and application.  You don't want the die too hot, but it has to have a setpoint warm enough that the heater circuit can control the temperature.  You know if it's too cold when the ambient temp rises and you see the LTZ lose control of the die temperature, and that is pretty obvious.  You do this by testing.  There are charts out there but not very accurate because each die is different.  Watch your Del-Vbe of the heater sense transistor to see what the die temp is on -your- setup at your ambient temp.

The problem with the assumption that warmer temps will cause the die to drift faster was made by people not really understanding the underlying mechanism that causes drift in the substrate crystal lattice in the first place.  If you have a -very- stable die to begin with, it's going to run warmer in a very stable condition also.  If you happen to get a die that has a higher internal strain, yes the higher temp will cause a somewhat faster drift rate initially...BUT a lot of times what happens is those higher strain die begins to assume a more relaxed state - and in say 10 ~ 12 years they will settle down and become very good performers as the crystal strain-relieves itself, atom by atom.  This is not an experiment for impatient people, but you do learn what the reality is after a while.

Again, the time frame you're looking at for initial break in is on the order of months, and then you look at the first 5~8 years for initial performance, and then start looking at the 10~25~30 year performance and you see a bit different story as the die really settles into it's final state.

You don't have to believe me, ask any 30yr old 3458a that goes in for cal and hasn't moved except maybe 1ppm or less per year.  And has been doing that for the last 10 years or so.  Not all do that of course but a lot do...and they all run 15k over 1k for the heater resistor ratio, because it's warm inside that box.

[Muxr]

Try without aircon, as those oscillations aren't helpful . If I'm reading your graph correctly, you see 1.6ppm pk-pk? That's bit much, I usually get around 0.5-0.6ppm pk-pk for week long captures, or 0.3-0.4ppm pk-pk daily. Example here and here, with K7168 switch scanner in place.

Before fellow voltnuts here start blaming chopper opamps, CH2 reference on my graphs is "proper" HP 3458A A9 for comparison reasons .

No way can I live without aircon (Summer in Florida) but fret not Tin.. I am building a micro environmental chamber, so I should be able to get readings from a much more controlled environment. (although can't put my hp3458a into it) In winter I can do measurements without aircon.

I am still very impressed with the results, usually my results are in 5ppm p-p range or more under my usual lab conditions, lots of airflow and aircon activity as you noticed.

[Muxr]

@MisterDiodes

Thanks for the info.. will try the LTZ1000ACH next. The heater setpoint I am using is 1K - 12K7 closest I have in Vishay foil variety. I need to checkout some of these other sources of low TC resistors you outlined.

[RandallMcRee]

Randall:  LTC5400 diffused resistor packs don't have a lot of use around LTZ circuits.  Suggest you just get some PWW made from Pettis or GR or Rhopoint, etc.  Those will give you better performance for a cheaper price.  You don't want a 12:1 ratio either.  If this is your first unit stick with 13k/1k as datasheet tells you or if you know it'll never get warm maybe 12.5k over 1k.  Plenty of decades 3458a's don't drift 1ppm /yr with their standard 15k / 1k, and learn from that. 

I took your advice, above, and tore out my ltc5400 and replaced it with two TE Conn 15ppm resistors with a 1:12.5 ratio (these were on hand). Previously, it was exhibiting strange drifting. Gone. Very stable at least for the the first five hours.

Thanks again.

[MisterDiodes]

Randall:  LTC5400 diffused resistor packs don't have a lot of use around LTZ circuits.  Suggest you just get some PWW made from Pettis or GR or Rhopoint, etc.  Those will give you better performance for a cheaper price.  You don't want a 12:1 ratio either.  If this is your first unit stick with 13k/1k as datasheet tells you or if you know it'll never get warm maybe 12.5k over 1k.  Plenty of decades 3458a's don't drift 1ppm /yr with their standard 15k / 1k, and learn from that. 

I took your advice, above, and tore out my ltc5400 and replaced it with two TE Conn 15ppm resistors with a 1:12.5 ratio (these were on hand). Previously, it was exhibiting strange drifting. Gone. Very stable at least for the the first five hours.

Thanks again.

You might have been running just a little too cold, but glad it's working now!!

[Muxr]

@TiN

I've updated the post https://www.eevblog.com/forum/metrology/kx-reference/msg1250845/#msg1250845

With a short run powering the reference from a 12V battery. I am getting 0.6ppm peak to peak with that setup. I suspect I can further improve the performance by improving my cable / probe situation (I am just using the tiny micro test clips with a twisted pair which are not the best).

In either case, that standard deviation is an order of magnitude better than I've seen from any other reference I have.