KX Reference

[feedback.loop]

This is my build of the KX reference. Overall the project went very well. The reference worked on the first attempt. The output voltage was monitored for about 46 hours so far using HP 34401A and Datron 1071. The graphs look great overall, but with some small unexplained glitches.

https://youtu.be/mgHhGDRKZjo

[branadic]

Would you mind plotting voltage (y-axis) over temperature (x-axis)? This way you can better see dependency.

[Dr. Frank]

Both DMMs are much more noisy and less stable (about 10 times) than the LTZ1000 circuit.

The 34401As internal LM399 may produce these 1..2ppm steps.

The bigger glitches may be produced by the LTZ1000 circuit, caused by RF disturbances, from switch mode power supplies in your lab. Watch out for LED lighting..
xdevs circuit lacks additional capacitors, which suppress most of these disturbances. I added 2 capacitors to the LT circuit, and things improved a lot, already.
Such glitches vanished completely, for my circuits.

So please consult the Super LTZ1000 thread for details on Andreas' circuit.

Also, you might trim the circuit for smallest T.C.
For that, you would need a 2nd LTZ circuit at constant temperature.
Frank

[SvanGool]

To save time on scanning through 80 pages of the "Super LTZ1000" thread, I think Dr. Frank is referring to Andreas' explanation on removing distortions and his overall explanation of his circuit.

[feedback.loop]

Would you mind plotting voltage (y-axis) over temperature (x-axis)? This way you can better see dependency.

Here is the plot.

[CalMachine]

Both of my KX references are up and cooking!  I currently only have 1 meter at home, so in about a week ill start logging some data on one of the references to share!   

[Ordinaryman1971]

Is there somebody making more of those boards? I would be interested in building one myself and even more in making few of the low cost LM399 ones. TiN, do you have some spares please

[kj7e]

Both of my KX references are up and cooking!  I currently only have 1 meter at home, so in about a week ill start logging some data on one of the references to share!   

Got all of my parts ordered today.

[TiN]

As I received fresh LTZ chips from LTC, time to get them installed and start logging. I've used my two last KX boards as test venicles here.
Both PCBs unlike rest of the references have AUGAT socket, and LTZ chips installed without soldering, with trimmed legs. (1 x CH and 1 x ACH).

PCBs bit different too, one is from Sjef, latest version B03, and one of my own older ones B01 with few patches. Detailed config:

* SvanGool's purple KX B03 PCB, 2xLTC2057, Edwin PWW 12.5K+1K bodged & crooked, VHP202Z 120?, 75K? <1ppm/K PTF56 x 2. LTZ1000A 2017 chip in AUGAT socket.

* xDevs.com B01 PCB, 2xLTC2057, few extra caps per Andreas schematics, AE 15K+100K, S102 174+312R, AE 50K+30K, AE 50K+30K, AE 1K foil resistors, LTZ1000CH 2016 49week chip in AUGAT socket, 390K compensation resistor present.



Few more photos of board.




Initial datalog started, ppm levels are reset for new references:

Code: [Select]

Channel 1: VM1=7.072588550, VM2=7.072588602, AVG = 7.072588576, -0.2825 ppm
Channel 2: VM1=7.129833446, VM2=7.129833183, AVG = 7.129833314, 0.3918 ppm
Channel 3: VM1=7.136680531, VM2=7.136680303, AVG = 7.136680417, 0.1275 ppm
Channel 4: VM1=7.184637357, VM2=7.184637357, AVG = 7.184637357, 0.8055 ppm
Channel 5: VM1=7.129231640, VM2=7.129231482, AVG = 7.129231561, 0.3414 ppm
Channel 6: VM1=7.136664333, VM2=7.136664053, AVG = 7.136664193, -0.2630 ppm
Channel 7: VM1=7.128692837, VM2=7.128692784, AVG = 7.128692811, -0.5456 ppm
Channel 8: VM1=7.134149909, VM2=7.134149594, AVG = 7.134149751, -0.3853 ppm

Channel 1: VM1=7.072587902, VM2=7.072587989, AVG = 7.072587946, -0.3716 ppm
Channel 2: VM1=7.129833831, VM2=7.129833901, AVG = 7.129833866, 0.4692 ppm
Channel 3: VM1=7.136680426, VM2=7.136680478, AVG = 7.136680452, 0.1324 ppm
Channel 4: VM1=7.184636866, VM2=7.184637094, AVG = 7.184636980, 0.7530 ppm
Channel 5: VM1=7.129231534, VM2=7.129231377, AVG = 7.129231455, 0.3266 ppm
Channel 6: VM1=7.136664018, VM2=7.136663948, AVG = 7.136663983, -0.2924 ppm
Channel 7: VM1=7.128692346, VM2=7.128692136, AVG = 7.128692241, -0.6255 ppm
Channel 8: VM1=7.134150031, VM2=7.134150137, AVG = 7.134150084, -0.3387 ppm

Channel 8 is VK5RC's ref externally from the chassis. All powered from linear +15V LDO supply, scanned thru Keithley 7168 + RPI and sampled by 3458A NPLC100.

Now I don't have any more LTZ references left, time to move on to other projects related  .
Will be interesting to compare drift of my older references (which are 24/7 for few months already) to this two brand new fresh ones. 

[Conrad Hoffman]

Is there somebody making more of those boards? I would be interested in building one myself and even more in making few of the low cost LM399 ones. TiN, do you have some spares please

Me too!

[hwj-d]

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

[Muxr]

There is a project on OshPark that lets you order (3 is the smallest run). https://oshpark.com/shared_projects/HfKcqjV3

Dunno if anyone has any spares.

[kj7e]

I have three LTZ1000 purple KX B03 PCB boards on order, I should only need one provided I don't mess it up.  So I should have two available if someone wants.  Send me a PM.

[hwj-d]

@kj7e
send PM

[RandallMcRee]

Three more on order from me, as well! I think will keep two so if anyone wants one, ping me.

Did not see this thread before ordering mine from OSH!

Should have a few more LTZ1000s in the US Cal Club. Exciting. I have to confess I am going to try and go cheaply (TE Connectivity 15 ppm) on the resistors and put them in an oven, which I already constructed. See how that compares to the much more expensive Vishay/Pettis builds.

Randall.

[ManateeMafia]

LT has been good to me too. I received an order last week from them. My board will be more of the traditional, boring, thru-hole design.

[RandallMcRee]

Why 5? Seems like an interesting number...tell us more.

[ManateeMafia]

My first rev will be one or two populated boards. Once I feel comfortable with the component choices and fix my initial errors, I will get the second improved batch made.

The first rev pcb has a header for a piggy-back board that will allow me to try different buffer amps and 10v boost designs without building several LTZ boards. Once I settle on something, the next rev will have that design added to it and I can build 3 more for long term testing. I have a Data Proof scanner I plan on using for long term data collection.

[TiN]

ManateeMafia
Sounds like somebody was working Looking forward for the data.
What resistors and temperature setpoints are you going to use?

Meanwhile, first 12 hours of data :



Note remind: Whole setup was cold, powered off for 2 days. Now added two fresh LTZ chips (Channel 5 is LTZ1000A and Channel 7 is LTZ1000).
First two hours it took to warm up and stabilize with ~1ppm change on all of references.

Channel 1 is good old jumper LTZ1000A from eBay . It's stable, just jumpy. Good for ADC resolution testing, I bet, as jumps are fairly constant level ~0.7ppm.
Channel 2 is non-modified HP 3458A A9 reference module, rest are my KX boards.

P.S. ambient is about 25C, reading has correction +2.0c due to proximity of BME280 to FPGA board.

[MK]

Perhaps you could try running channel1 jumper LTZ at a higher current to see if that makes the jumps go away? Something for a rainy day when you feel you have enough history and they all seem to have stabilised.

[d-smes]

@TiN-  What's the slight step down at 6:30?  Did the sun rise and shine on your test leads?    Even your eBay jumper follows along...

[SvanGool]

@TiN

Thanks for sharing !

Can you detail a bit more which reference is on which channel?

My gamble:
Ch1: "Good old jumper"
Ch2: HP 3458A A9 reference module
Ch3: VK5RC KX Reference
CH4: B03 reference
CH5: B01 reference with Andreas' capacitors
......

Andreas added following components with reference to your VREF_B02 schematic, to improve the noise and stability of the standard design:

• 100 nF between U1-6 and GND (lower noise)
• 100 nF between U1-4 and GND (lower noise)
• 100 nF between U6-2 and U6-3 (lower RF noise)
• 100 nF between U5-2 and U5-3 (lower RF noise)
• 100 nF between U5-6 and GND + 10K between U5-5 and U1-4  (improve loop stability)
• 100 nF between VREF+ and VREF- at the output connector (block RF from output)

Andreas could only do item 6 when item 5 was implemented, with the opamp he used (LT1013), which, when needed, would probably require some PCB-trace/guard changes on the KX-board.

You mentioned you implemented some of Andreas' suggestions, what did you add on the B01 board of above items ? 

[kj7e]

Maybe a B04 version with some of the filtering in the future?

Working on my B03 board now.



Going to use 13K/1K on the heater with the LTZ1000A.  Also going to buffer the output with the LTC1150/LT1010 combo and place in side an heated (35 deg C) die cast box.  Plan is to build it all in a nice battery powered chassis.  But one step at a time.

[Kleinstein]

Filtering likely only makes sense with an additional buffer. It would reduce higher frequency noise only, which is not a big deal in many application (e.g. reading with an DMM) anyway. But of cause if a buffer is build, one could include at least some filtering, if this is not adding too much DC errors and additional low frequency noise.

In this case a just an AZ OP (also with an additional power stage) might not be the best choice as a buffer, since this could easily add again some higher frequency noise (though usually different frequency band). To really make full use of a filter, it would need some a low noise and background one (e.g. compound with the AZ Op only for the LF part).

[kj7e]

It will be a few weeks (maybe months) before the Vishay parts are in, but I wanted to get it fired up and let it burn in until then.  Who needs precision resistors anyway?

Instantly, like with in 2 seconds it was reading 7.1330703 and its not moved.   








I also decided to space the LTZ1000A above the board slightly.  Since I will be placing the reference and buffer boards in a thermally stable box, this only made sense to me;


Edit, LTC1150/LT1010 buffer working on the bread board.;