KX Reference

[CalMachine]

Well, after all, your KX'es belong to me
Decided to give a helping hand to CalMachine, with his two attempts.



Obviously, my design constrains are bit unfriendly to newcomers, especially tricky SMD film capacitors.
Once I fix these two modules, I'll log them to establish stable voltage levels and send back to CM for his calibration vs their 732B standard.

That will help to also help to keep my lab calibration uncertainty as "reverse-calibration", as my last Volt calibration was done in February 2017, and is overdue.

I Appreciate the help and it is a great opportunity to get some international volt transfers going!

The main problem I was observing with the 1 good board was, temp co's of ~40-50 ppm/ºC.   I swapped out many components, even including the LTZ1000, and still not a working module.  My lack of building experience has got the best of me on this project.

[Andreas]

The main problem I was observing with the 1 good board was, temp co's of ~40-50 ppm/ºC.   

That should have given the hint that the heater cirquit is not working at all.

with best regards

Andreas

[TiN]

Both are fixed and logging now. I'll test tempco on weekend to see what is what. Expecting 0.2-0.3ppm/K as these are unmodified B01 PCB version. Simple trace rework should take care of that.

[TiN]

China seller has no shame at all  , I think world should know the "hero" with listing at Taobao. Can somebody who has account report this? 

Meanwhile I wrote some code to reduce experimentation setup efforts for the multiple thermal tests.
Not sure if it may be any interest for members here, but I might make in open-source project.
Currently supports K2510,K2001/K2002 and HP3458A.

Temperature curve setting:

Code: [Select]

               Peak_temp
               ______
              /      \
             | Step   \
            /          \
   Slope_pos            \ Slope_neg
          /              \
time_start Speed_pos / neg\  time_end
  ------/                  \--------- TEC temp
 Sv_start                    Sv_end
   2h     9h     2h    9h       2h


[Zucca]

Bastards, hope they will take it down.

[ManateeMafia]

TiN,

Sorry. You know my calibration costs are high. I have to make money somehow. 

[TiN]

Maybe so, but without thermal coefficient evaluation service you have no business   

First run data on CM's KX1.

Worst case tempco calculation: -0.084 ppm/K (12 °C Δ based on PCB MAX6610 sensor).
Best case tempco calculation: -0.039 ppm/K (19.4 °C Δ based on TEC box sensor).

[d-smes]

TiN-  It might be enlightening to dwell longer at the 20C and 40C temperatures to make sure everything fully stabilizes.  For example, between Midnight and 1 AM, voltage jumped upward while temperature was at constant 40C.

Other interesting events in the data are jumps in reference voltage with abrupt changes in ambient T and RH at 6-7:30 AM and 6-8 PM.  Reference voltage jumps down for both cases: a jump up in ambient T and RH (morning) and a jump down in T and RH (evening).

Lots to ponder.  Keep up the great work!

[TiN]

It's hard to have better stability, as essentially the meter that used to perform the measurement have same stability than the DUT, so there is no margin.
Running second time now with twice the speed and peak temperature 50C instead of 40C.

[kj7e]

It's hard to have better stability, as essentially the meter that used to perform the measurement have same stability than the DUT, so there is no margin.
Running second time now with twice the speed and peak temperature 50C instead of 40C.

What temp set resistor values are being used (13K/1K) on the DUT?  It will be interesting to see it runs out of temp regulation over 40C.

[TiN]

It's 12.5K/1K, which is about 55-57C, so there is still few degree of margin.
Current live data.

There is no much to see when LTZ's oven out of regulation, your tempco become 35-45ppm/K as result 

[BFX]

Another KX Reference is heating up to get its stability

[TiN]

BFX
Looking good.

I got the tempco data for both of CalMachine's KX references. One of which is particularly nice, showing <0.005ppm/K TC in range +34 to +43C (by box method).
Measurement setup I had covered before number of times, it's same die cast box with 40W TEC, controlled by K2510. All measurements done over 3458A in stable room temp (+/-1C), so own DMM tempco nulled out.
Except MX1 ref data, which was tested in similar box, but controlled by ILX 5910B controller instead. Meter is 3458A as well.

Data results



Legend and RAW data

1. NVS/KX1-1097 reference - this is one of my DC bank references, using KX A01 prototype PCB. 13K/1K + 120R Z202 resistors and LT1097 opamps, using LTZ1000A 2014 week 24. Board Photo. RAW data.
2. MX1-REF - this is MX1 prototype, with LTZ1000CH 2016 week 49. LT1013 + Riedon 13K/1K + 120R BMF. RAW data.
3. CalMachine's KX B01 reference unit 1 - KX B01 board, 12.5K VHP100T + 1.0K VHP103 + 120R VHP319895 + 2x80K Z201. LT2057's, LTZ1000CH 2016 week 41. RAW data.
4. CalMachine's KX B01 reference unit 2 - KX B01 board, 12.5K VHP100T + 1.0K VHP103 + 120R VHP + 2x80K Z201. LT2057's, LTZ1000CH 2016 week 41. RAW data.
5. KX B03 reference I made for member - covered in Tale 1. 13K/1K VHD200 + 120R VHP101 + 2x75K VHPs, LTZ1000CH 1990 week 15. RAW data.

MX1 reference definately needs more work, so stay tuned 

[BFX]

Thx TIN but I'm curious why my LTZ is relatively cold (31.5 Celsius) (I had to put some black isolation tape to LTZ of course)
I'm using 75KOhm resistors instead of 70k (for now).
Do you think its ok?
I will start some measure next year now its heating.

[TiN]

If you used LTZ1000A, it's case does not get very hot. 75KOhm's are just fine, any resistor in range 65-90K are working just fine.

Is the right side zener current opamp? It should be so hot, usually it's just barely above the ambient, and heater NPN is few C warmer.

[ManateeMafia]

I have a second MX board that will be populated similar to the first board except I now have some of Edwin's resistors. This will help determine how much of an effect the resistors had versus the not so great layout.

The MX thread will be updated once it is up and running.

I am hoping to get some LTZ1000A from LT in the next few weeks. They are backordered but DigiKey shows a mid Nov delivery for them.

[BFX]

If you used LTZ1000A, it's case does not get very hot. 75KOhm's are just fine, any resistor in range 65-90K are working just fine.

Is the right side zener current opamp? It should be so hot, usually it's just barely above the ambient, and heater NPN is few C warmer.

This opamp is little bit hotter but not so much (around 33.2°C ).
Do you think its too much?

[TiN]

I've decided to put up little index on community modules for upcoming article. 



If you want to be included and have working KX module, feel free to send me photo, voltage output and used instrument, I'll add.
It's just crazy, never expected to see so much interest.

[TiN]

Little sucker eated all the instrument time for a week already.
Perfectionist in me could not rest with +0.06 ppm/K tempco and parabolic-type curve on temperature extreme.



Running for a cure now.. 

Also looks like leadtime on Digikey for LTZ1000A and LTZ1000 is now mid-december 2017/january 2018. 

[Andreas]

Hello Illya,

are you shure that you are not running out of temperature regulation above 40 deg C environment?
(which temperature setpoint?)

with best regards

Andreas

[TiN]

Yes, I'm sure Setpoint is bit over 60C. You can easily see when it is running out of oven margin by very visible temperature curve. I have my old module with same setpoint, which does not show this curvy stuff and behave in nice linear way.

[Kleinstein]

The TC compensation with R9 (400 K nominal) gives a curvy contribution (proportional to square root of heater power), with a stronger influence when the heater power is low. So a rather low temperature set-point or a lot of thermal insulation can contribute. With a higher temperature the contribution of R9 is expected to be linear.
The residual TC of the LTZ1000s can also vary. If too much compensation via R9 is needed the curved shape can be a problem.

Also the heater resistance has an influence - so the 400 K nominal value for R9 is an order of magnitude values only for best performance it should be individually adjusted.

[TiN]

Kleinstein, 
You are right on, amazing  .

I got same guts feeling, so first thing I trimmed TC compensation resistor. Well, that ended up to be a rabbit hole. Judge youself:
Not to bloat the thread with images, I'll just add links to each run log:

* Test run 1 this is original setup, board rev.B01 , no modifications, 400K resistor. Low peak: -2.2ppm at 43C, High peak: -1.0 ppm at +55C. Worst TC box: -0.122 ppm/K.
* Test run 2 same, just faster run speed, 400K resistor. Low peak: -2.1ppm at 44C, High peak: -0.9 ppm at +55C.
* Test run 3 board rev.B01 , 226K resistor. Low peak: none, baseline, High peak: +6.0 ppm at +55C. Worst TC box: +0.194 ppm/K.
* Test run 4 board rev.B01 , 340K resistor. Low peak: -0.8 ppm at 39C, High peak: +1.75 ppm at +55C. Worst TC box: +0.055 ppm/K.
* Test run 5 board rev.B01 , 305K resistor. Low peak: -0.7 ppm at 38C, High peak: +2.2 ppm at +55C. Worst TC box: +0.064 ppm/K.
* Test run 6 board rev.B01 , trace mod to B03, same 305K resistor. Low peak: -0.4 ppm at 28C, High peak: +3.7 ppm at +55C. Worst TC box: +0.108 ppm/K.
* Test run 7 board rev.B01 > B03, 680K resistor. Low peak: -2.1 ppm at 48C, High peak: -1.4 ppm at +55C. Worst TC box: -0.066 ppm/K.
* Test run 8 board rev.B01 > B03, 470K resistor. Low peak: none, baseline, High peak: +4.0 ppm at +55C. Worst TC box: +0.117 ppm/K.
* Test run 9 board rev.B01 > B03, 563K resistor. Low peak: -0.65 ppm at 38C, High peak: +0.7 ppm at +55C. Worst TC box: -0.05 ppm/K.

FYI, each run was about 14 hours.
I think I'll settle with Run 9 results, this already sucked 120 hours of 3458A runtime.

Here's my LT1097/LTZ1000A that acted as TEC setup sanity check, no curved business.

[Andreas]

Hello Illya,

do you have really environment temperatures of 55 deg C in your lab?
I would cease work above around 30-32 deg C.

So with some safety margin I never do calibrations above ~40 deg C.

with best regards

Andreas

[TiN]

Lab temperature visible on the graph, dashed cyan line, Ambient C. It's around 27-29 this week, as I didn't do much stuff, also needed to keep it constant so I can compare run1 to run9 (3458A runs at same ACAL +/-1c span since last weekend, to avoid shifts in reading from own 3458A TC).

I never know what CalMachine's intention to use these two modules, but since he installed 12.5K resistors to get 60C setpoint, that's what I test to, minus the 5C margin room. Maybe he desire to fit one module into K2001 instead of original LM399, and "ambient" temp inside that box can reach well around +50C. Obviously since test include 20C-55C points, it's easy to process data to smaller span without any effort.

I'm done with these anyhow, and hooked HP 3458A A9 module (STD) for a spin. Let's see what she is made of.