[FT] xDevs.com KX LTZ1000A fairy-tale, or the story of little jumper.

[plesa]

Sooner or later you will needs decent enclosure. I used this one
http://www.ikea.com/se/sv/catalog/products/40288354/
It is better than polystyrene box used in past.
Grounding of enclosure is made by cu tape placed over the gasket.

[mmagin]

Well, I'm finally giving into serious volt-nuttery.  It seems like once I manage to obtain resistors I'll probably be building a couple of these.  I'll post updates, and I will probably be interested in the unofficial calibration service.  At the moment, this is all better than the best meter I have, an HP 3457A

Is it reasonable to consider the LTZ1000A and precision resistors water washable?  My inclination is toward using entirely water soluble flux, initially washing with warm water and detergent, then multiple rinses of distilled water + isopropanol.

[plesa]

Well, I'm finally giving into serious volt-nuttery.  It seems like once I manage to obtain resistors I'll probably be building a couple of these.  I'll post updates, and I will probably be interested in the unofficial calibration service.  At the moment, this is all better than the best meter I have, an HP 3457A

Is it reasonable to consider the LTZ1000A and precision resistors water washable?  My inclination is toward using entirely water soluble flux, initially washing with warm water and detergent, then multiple rinses of distilled water + isopropanol.

3457A is quite good DMM. If you purchase in future 3458A you will know how to operate within few seconds
I can hook it to LTZ setup to provide comparison between 3458A nad 3457A. For example K2001 is noisy but stable, 34401A are silent but fluctuating with temperature/humidity.

I will not recommends any washing nor immersing in water. Use IPA only. Clean by brush and small amount of IPA. Some of resistors are sensitive to humidity. LTZ board(s) do not need to be superclean.

[mimmus78]

And here you go. Initial tests with brutal TC estimation tell me all assembly is 0.1 ppm/°K.

[mimmus78]

plesa

I used 12K/1K as temperature set poit for the LTZ1000 ... what is temperature delta vs ambient temperature inside this IKEA thermos?
Wondering if it will be too much high temperature during summertime (here gets as hot as 30 degree during summer time).

[mimmus78]

Anyway during this days I discovered my K2001 is just 5ppm apart from the 3458a on the 10V range (the other ranges are much apart).
Not bad considering last calibration (sticker) is 15 years ago.

[plesa]

plesa

I used 12K/1K as temperature set poit for the LTZ1000 ... what is temperature delta vs ambient temperature inside this IKEA thermos?
Wondering if it will be too much high temperature during summertime (here gets as hot as 30 degree during summer time).

I'm using same ration sot temperature setpoint (~40°C), but I'm not measuring temperature inside IKEA thermo. That can be interesting.
This vacuum thermo flask is good because it is also humidity tight. With some 3D printed holder and cap over LTZ1000 it is ready for transport.
I do not think the heat generated by LTZ mainly will be problem.

[Andreas]

Hello,

you should detect the upper temperature limit when doing final T.C. check (with final housing)
of the LTZ cirquit over the temperature range of interrest.

With best regards

Andreas

[VK5RC]

Re temperature of LTZ in insulated-enclosed environment.
I have been using an old reference board from HP (ex 3458), it is in a metal box with insulation inside the box and some over the LTZ. Tied to the board close to the LTZ (about 6-8mm away) is a Fluke temp probe (brown cable) also inside the box is a linear PSU which  drops from 18 to 15V at about 30mA - so total power inside the metal box is about 0.5W. The metal box is inside a plastic tub (further insulation) see photos below. If ambient is 17-25C (my thermal limits!) it takes about 1-2hrs to stabilise in temp but is  between 30-31C .
Grey cable DC in
Brown Fluke temp probe
Green DC out
Edit; Box is now shut and cables routed through small hole

[Kleinstein]

If the whole ref circuit is inside a relatively well insulated box, the LTZ internal heater (and the driver transistor) will also work to stabilize the temperature of the whole circuit. It depends on thermal insulation and will not be perfect, but one could expect something like having a circuit temperature at something like half way between the fixed internal temperature and the room temperature. So temperature variation would be reduced by a factor of 2 in this example. To use more of the effect, a slightly higher temperature set point and better insulation could help. Also the amount of insulation directly around the LTZ1000 as an influence - more insulation is not always better.

It is a good idea to have an extra temperature sensor somewhere at the circuit to see if the temperature for any reason is too high. With such an insulated bottle, much of the heat loss is at the top and through the cables.

[dr.diesel]

but nobody still built it in public.

I just got word from Edwin that my resistors have shipped!  I bought enough to populate 3 KX boards, leaving 2 more for future thoughts/lessons learned.  I have all the other necessary components already.

[mimmus78]

Almost finished experimenting with mine ... I'll keep it running for some days and than publish what came out.

In all this Odyssey I learned a lot and got some ideas for the next one.

1. Always buffer the reference. I risked to short it many many times in the most not obvious ways. In my case I think this buffer also kinda lowered some noise (but I still have to check measure and confirm this data).

2. I think to get better long term stability and noise is mandatory to add some sort of oven too. This double oven (one in the LTZ and this other) will improve TC and keep resistor at a stable temperature. From my little experience with TC of resistors, keeping resistor at the same temperature will reduce long term drift. But than you have to keep it on always or otherwise the bigger temperature excursions can create more drift/hysteresis to those resistors than the normal t.amb variance.

3. LTC2057 was causing more than 1ppm jumps depending on what power source I was using or if I used shielded leads or not. Replaced opams with ADA4522 and now it's 1 to 2uV delta swapping power supplies or using different leads. The most of the offset was caused by the opamp driving the zener. Changing the opamp for the heater didn't change anything really measurable. Using the same LTC2057 as buffer seems to have a very little influence. Anyway I'm still not convinced about this drift was caused by  LTC2057 ... another LM399 based reference is using the same chip and seems to be not susceptible. Unfortunatelly noise of LM399 is much higher so making such evaluation needs much more care than I really wanted to put in the measurement.

[Andreas]

3. LTC2057 was causing more than 1ppm jumps depending on what power source I was using or if I used shielded leads or not. Replaced opams with ADA4522 and now it's 1 to 2uV delta swapping power supplies or using different leads. The most of the offset was caused by the opamp driving the zener.

Good to know that the ADA4522 behaves better.
I have solved the EMC-issue on the LTC2057 with a 100nF cap.
(but also use a 22R series resistor and a capacitor in the feedback loop).
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg846835/#msg846835

With best regards

Andreas

[mimmus78]

Well yes slightly better ... I solved susceptibility with Siglent power supply (drift passed from 10uV to 2uv) and apparently reduced problems with leads ... but now I had Idea to test with very short twisted copper ethernet cable (15cm) and I'm getting 10uV down drift. Deftly need to try your caps and resistor for EMC mitigation after buffer. See you in a few hours.

[mimmus78]

It's not EMC the buffer just follow what it find on input ... still investigating 

[dr.diesel]

First board up and running! 

(12k/1k ratio, Edwin resistors)

[splin]

That's a remarkable meter you used to measure the reference - apparently with maximum and minimum voltages displayed at 7 1/2 digits, uV resolution, but manages to calculate the peak to peak to a further 7 digits, 6.8025031uV p-p. 100 femto volt resolution on a 10V range is quite incredible.

Too incredible obviously. The difference between the displayed max and min values is 7uV. There may be another digit undisplayed so 6.8uV p-p may be plausible even if the .8uV digit is mostly noise. But why would the developers choose to display a further 6 digits of completely bogus resolution? How could the value have been derived?

[mimmus78]

What's the meter?

Inviato dal mio Nexus 6P utilizzando Tapatalk

[mmagin]

That's a remarkable meter you used to measure the reference - apparently with maximum and minimum voltages displayed at 7 1/2 digits, uV resolution, but manages to calculate the peak to peak to a further 7 digits, 6.8025031uV p-p. 100 femto volt resolution on a 10V range is quite incredible.

Too incredible obviously. The difference between the displayed max and min values is 7uV. There may be another digit undisplayed so 6.8uV p-p may be plausible even if the .8uV digit is mostly noise. But why would the developers choose to display a further 6 digits of completely bogus resolution? How could the value have been derived?

Somewhere that programmer's high school science teacher is cringing.

[dr.diesel]

untruncated calculation precision is not uncommon!     

[splin]

untruncated calculation precision is not uncommon!     

Very true. But in this case what calculation is involved in working out the peak to peak value other than subtracting the minimum from the maximum values? That doesn't create any extra digits, spurious or not, unlike mean and standard deviation calculations.

I'm struggling to think of possibilities but I guess it's conceivable that they could be allowing for the meters own noise by adjusting the maximum and minimum readings by subtracting/adding the meter's known peak noise (adjusted for range, integration time and filtering levels etc.). That could make the peak to peak readings more accurate - on average - but could under/over read true maxima and minima. This seems very improbable though given that the benefits/detriments would likely be very dependant on the statistical properties of the signal being measured.

So where could those extra digits have come from? I'm sure there's a simple explanation but I'm not seeing it at the moment.

[TiN]

dr.diesel
Beauty. Now add some foam protection around both sides of LTZ, drop everything in the metal can, and let it sit for a month or two powered (about enough time to build a DIY thermal chamber for future tempco tests ). Then we can see some data.

I can't say I particularly like these standing way to mount those resistors though. I know why you did it, but not sure if that cause issues due to uneven thermal gradient over resistor body (PCB side leg is much shorter, with much larger thermal mass on it). Usually when using axial resistors, I bend legs in "radial-way" and mount them with body parallel to board.
Maybe if you plan building more boards, you can try that so we can compare?

splin
I'd expect programmers just implemented standard FP math without actually thinking much how real are the numbers, be they down in the white noise, or have real meaning. Doesn't bother me much though, as you already know what is real and what is not.

[Andreas]

First board up and running! 

Mhmm,

7uVpp for a source that should have 1.2uVpp.

Which NPLC?
Is it the DMM or the zener?
Does the heater run properly?

with best regards

Andreas

[mimmus78]

So time to publish some result.

It seems I almost sorted out all my problems with EMI and leads and power sources ... now no matter what I do, I do not have anymore super large drift of the output (almost 2ppm before) the max drift by changing swapping all this conditions is now usually 0.1 or 0.2ppm with the larger one of 0.3 ppm when I turn on a programmable power supply.

I will explain all I did in a further post, I think I can live with it for now.

Anyway here is last chart from yesterday when I applied heater caps ... (RMS noise went down of almost 100nV).

The image is very large ... so it's better you open in another window.
The red line at the bottom is RMS noise. 0nV at base of the chart and 250nV at every division, RMS is calculated on last 100 samples.
The tin green line that is my room temperature.
The green dots are the single readings (100NPLC with autozero sync)
The blue line is the median value of the previous 100 readings.
The orange dashed lines are the temperature of last ACAL and the internal temperature of the 3458A (temperature are chart with -10°K to make the chart more compact)
The last black line is internal reference temperature.
The chart is one 10uV large ... so 1.4 ppm.


PS: first drop in the chart is me changing leads.

[dr.diesel]

7uVpp for a source that should have 1.2uVpp.

Yeah, don't think anything of this, it was only a test power up, just to make sure no shorts, it works, I didn't screw up etc.

TiN

Two more boards will be finished today, one I will build like you describe (parallel to the board), same 12k/1k ratio, and the 3rd with 15k/1k ratio, both Edwin resistors.

That will leave two spare boards, I was thinking of building one with off the shelf DK resistors, to demonstrate what's possible without special ordering.  Unsure and taking suggestions for the 5th.

My 3458A mods/SRAMs/re-cap is complete and it's headed to CalMachine's next week for calibration.

I'm also happy to help/provide testing to others.