Group-buy for 10V LTZ1000A reference? [Completed, no more boards].

[TiN]

Considering Decided to make 10 completely assembled and tested modules to provide +10.00000 +/-4 ppm buffered output.
One of the challenges is cost and time involved to test and calibrate these things.

Device is based around xDevs.com FX module PCB, which have next design specifications:

Reference source: LTZ1000ACH with datasheet schematics, using LT1013ACN.
LTZ Temperature SP: 13K/1K which provide about +55C oven temp
Resistors: VPG BMF HZ for 120ohm, temperature SP. PTF56 5ppm/K for 70-80K bias resistors
Output 7V->10V stage: Chopper opamp ADA4522-2 in SO8, Hermetic resistor network or BMF resistors. Each unit adjusted by trim resistor for tight output tolerance.
Onboard supply: Linear LT3042 LDO, configured for +11V output to LTZ circuit.
PCB: 4-layer FR4 with ENIG plating
Input power requirements: +12V and -12V.
Power consumption : ~0.6W when operated with +24C ambient
Additional features : isolated section with I2C EEPROM with calibration data/config/history (RPI python app to be provided to maintain data), MAX6610 PCB temperature sensor for tracking
Calibration traceability: <4 ppm to 732B, uncertain with 0.6ppm.
Expected performance: <4 ppm/year long-term stability, <0.1ppm/K tempco over 18-28C ambient temperature.
Assembly type: Hammon metal box with Pomona 3770 low-thermal binding posts.

I'd like to check if there is any interest in expensive hobby-level reference among the members here.
If enough people would be willing to buy one, it will be easier for me to pull the trigger and tie in 4-5K$ into this project.
Estimated delivery date would be March 2018 October 2018, with final good cost $500 USD.

Specs above are on safe side, so the module likely to be 2-5 times more stable, however I wouldn't guarantee it just yet. There is no replacement to actually running reference for 3+ years to determine actual long-term annual stability.

Feel free to comment

Member list, who interested in module

montemcguire - 1 unit [CONFIRMED] [DELIVERED]
dr.diesel - 1 unit [CONFIRMED]  [DELIVERED]
chriswebb - 1 unit [CONFIRMED] [DELIVERED]
PeLuLe - 2 unit [CONFIRMED]
MM - 1 unit [CONFIRMED] [DELIVERED]
bck - 2 unit [CONFIRMED] [DELIVERED]
USA cal club - 1 unit [CONFIRMED] [DELIVERED]
CM - 1 unit [CONFIRMED] [RESERVED]
zucca - 1 unit [CONFIRMED]  [DELIVERED]
No more units available., unless somebody change their mind.

Prototype board overview, test data samples



Initial output after first assembly.

[Vgkid]

Sounds interesting, I would do it if I ever get the money. That would be around 6mo + from now.

[montemcguire]

I'd buy one - count me in!

[lukier]

Does it really need -12V? It would be much simpler to have just one battery (12V lead-acid or 4x18650 LiPo).

[dr.diesel]

Count me in, prepay is no problem to help offset costs.

[maxwell3e10]

Sounds interesting. What would be the specification for 0.1 to 10Hz rms noise? That is primarily what I am interested in.

[mimmus78]

I'm interested too but as I'm building my own version I'm not interested in buying one.

I have a DHL contact with good fee for international shipping ... I can offer few trips for free is some of this board wants to became a traveling standard among EU guys or for just a stop and go.

[nidlaX]

I'm interested, but it's only a run of 10 units?

[chriswebb]

I'd also be interested in picking up a unit if available.

[Kleinstein]

Does it really need -12V? It would be much simpler to have just one battery (12V lead-acid or 4x18650 LiPo).

The circuit should run with less than -12 V. It should be possible to make it run on just a single supply, maybe 14 V. For battery operation something like 12 V min would be nice, but this would not leave much room for driving both the positive and negative side. So getting it down to 4 Li Cells could be a little tricky.

[TiN]

montemcguire, dr.diesel, chriswebb - OK, I'll keep you guys in mind.

Does it really need -12V?

-3V can be fine too, I checked today, output change was less than 0.4ppm.
But I will be validating only +12/-12 +/-5%. Negative voltage is used on the output stage, not the LTZ-ref itself.

Sounds interesting. What would be the specification for 0.1 to 10Hz rms noise?

I expect ~3 uV perhaps. Will know better once I test first 3 prototypes I have now.

I'm interested, but it's only a run of 10 units?

Even to make 10 boards, lot of time and money will be tied in, so I'd rather spend some resource for other projects, using this reference, instead of making just the reference. I originally did not plan to make any extra's at all, and still need to run the numbers to pull the trigger.

Also each output resistor network must match the used LTZ chip value, so it's NOT simple assemble/hook to fancy meter/write value/sell process.

[pelule]

Count me for 2 (if such many left available), min. 1.
Prepayment possible.
BR
/PeLuLe

[wn1fju]

Could I buy one volt for $50USD

[VintageNut]

Hi TiN

I am curious about the 0.1 ppm/deg C temp dependency. If the temp changes +/- 3 deg C, then the output will change +/- 3uV?

The temp in my home/office can change +/- 5 deg, which is 10 deg p-p. That would translate to 10uV p-p or 1 ppm p-p.

My 731B which cost $250 is pretty much that good for temp dependency.

I think that if your LTC ref were in an oven like the KE DMM7510, THAT would be very attractive. \

One reason that I like my DMM7510 so much is that the ovenized reference makes the DMM rock solid stable in my home/office where the temperature is not stable.

[TiN]

VintageNut
Will have to see if can do better. Current data looking like this:



Ignore the steep jump on the end, that shows reference ran out of oven thermal margin, with resulting +150ppm raise of zener voltage and output voltage.

[texaspyro]

So 17 uV over 10 degrees C...

My ancient scrap heap salvaged EDC MV100 seems to drift around 20 uV over 3 degree C... not too shabby for an un-modified, non-ovenized 1N829 zener diode referenced clunker with a 6 decade adjustable divider... I've been looking at seeing what some LTZ1000 / LM199 / better chopper love could do for it.

[bson]

I'd be interested in a few (say 3) known to be genuine LTZ1000 references, just the parts without a board - if you can get a few extra!

[TiN]

So 17 uV over 10 degrees C...

Actually more like 22 uV over 22.6C, here's live data.
So that's about 1uV/C, or -0.1ppm/K.

LTZ reference itself was tested last week to ~-0.03ppm/K over 20-50C span. TC is not exactly linear due to compensation.
Sweet spot with nearly zero TC is to run with ambient temp 30 to 40C.

I'd suggest to move discussion about the design to thread.

There will be no parts/kits, only fully assembled and tested module is in question.

[texaspyro]

I already have a couple of 732A's and a 732B,  but if you need a another taker to meet your minimum build I'm game.  Might even be interested, otherwise.

The main difficulty I see with the unit is the +/- 12V power.   I'd like to do a small unit with LiPo (18650) backup batteries.  There are some nice, isolated, low noise DC-DC's out there that should be able to take care of that.. hopefully without slathering the system with switching noise.

[Echo88]

Maybe a capacitor charge pump suffices for the negative voltage while ensuring low noise? Or a big LED + solarcell. 

[Zucca]

TiN let's the magic happen, it's Xmas time!

I am in. Yes Yes Yes!

PS: anyone with connections for custom resistor arrays/dividers (Fluke 2464707)

https://www.eevblog.com/forum/metrology/tin-livestreaming-layout-of-new-ltz1000a-10v-reference-module/msg1345273/#msg1345273

[TiN]

zucca
OK. What did you do with previous ref, btw?

ALL
No more boards, so far I got all allocated, unless somebody from the list wanna change the mind.
Ordering parts (LTZs from LT direct, resistors from VPG direct, rest are from Digikey) next week.

[Zucca]

zucca
OK. What did you do with previous ref, btw?

I needed first a HP 3245A, now I am going to upgrade it...
Following your path.. very slowly my friend.

[TiN]

LTZs, opamps and LDO's ordered from LTC  .

[TiN]

Linear parts and Fluke network resistors arrived.

[TiN]

Got quote from VPG resistors (13/1K network hermetic, 120R hermetic, 70K epoxy).
However lead-time is looking around 11-16 weeks, which makes it around May'18.

If anyone want to drop off waiting that long, let me know.

[Pipelie]

Got quote from VPG resistors (13/1K network hermetic, 120R hermetic, 70K epoxy).
However lead-time is looking around 11-16 weeks, which makes it around May'18.

If anyone want to drop off waiting that long, let me know.

the lead-time of VPG resistors easily goes around 16-18 weeks , we ordered several times from Vishay in the past 3 years.

[ramon]

Hi Tin! What is the price for the 13/1K hermetic resistor network? I am thinking about building two LTZ1000 boards .... someday. So I can wait.

[TiN]

It's around $70ish so far.

[Kleinstein]

For a 10 V reference the 13K/1 K divider at the LTZ1000 is not that important. The divider for the 7 V to 10 V step is way more important (e.g. about by a factor of 50).  It would only be for an possible 7 V output that the stability of the temperature set point can matter. There is a kind of natural limit on how much to spend on the resistors - at some point it would be more effective to have a second ref. chip or a way to measure the voltage at the 13K/1 K divider - possibly to make it an extra 0.4xx V reference output.

[cellularmitosis]

It's around $70ish so far.

The VHD200 12.5/1K divider I ordered from Texas Components was $46.

Edit: lead time is 3 weeks.  should be here by 1/29.

[chriswebb]

Is there anything you need from me to get my order confirmed? Payment sent or something? I think I missed when you were getting the orders confirmed.

[TiN]

Naw, you all good. Just sanity check for myself.

[1anX]

A little off topic but how well does a AD584KH compare used in temp controlled (aircoditioned) lab?
LiPo battery powered versions are readily available https://www.ebay.com/itm/High-Precision-4-Channel-2-5-7-5-5-10V-Voltage-Reference-Module-AD584KH-NEW-coi/202035834876?hash=item2f0a462ffc:g:4D0AAOSw0ThZoOBg
Would this be a waste of money to use to check calibration of lab equipment?

Update: I did a forum search and found my answer! http://www.markhennessy.co.uk/ad584_references/

[Jxw6637]

I am interested in getting one, is there an extra?

[TiN]

Sorry, no more PCBs on this project.

[TiN]

Yay, finally something arrived.

 

70K resistors will be used in above-mentioned FX 10pcs run. 1K/13K VHP203T's are just for temporary testing, final boards will  have 1/13K VHD divider instead.
Rest of resistors (120R, divider) expected to arrive next month.

Resistors were ordered February 7, 2018, so this gives us ~17 week delivery time for these and ~20weeks or so for remaining.

Started tempco test already.

[ManateeMafia]

you need more 3458A   

[johartley2001]

are these still available to buy ?

[TiN]

All FX units are allocated for members, and I'm still waiting for final part delivery - resistors. It is typical to have 17-24 weeks lead time order for precision resistors.

Perhaps somebody on EEVBlog have another LTZ reference to sell.

[eurofox]

are these still available to buy ?

Build one yourself, there is enough information in this thread to do it.

Mine in aging right now and start to be very stable, it include a 10V buffer, it is mounted in a teco box and I will add a oven control on it to bring the temperature to 35°C +/- 0;1° C.

In fact I build different versions most without buffer to 10V.

Only problem is final calibration 

[SKPang]

Hi TiN,

Are you going to do another batch ? If so put me down for one.

[TiN]

Nope, due to high cost only 14pcs of these modules will exist in the world, sorry.

KX one is available on OSH Park for PCBs : https://oshpark.com/shared_projects/HfKcqjV3
(LTZ chips you can buy from ADI and resistors from Edwin Pettis here or VPG).

[TiN]

After 181 days of waiting, complete VPG order is finally here.



Now I can go ahead and build stuff!

[Vgkid]

Damn , all of those resistors cost more than everything I have bought in my lab...    

[TiN]

Okay, Digikey box from USA arrived nice and safe.

Wait is over, we are finally ready to start assembly. Enjoy, that's 6K+ in parts... 

[Zucca]

[TiN]

Got some parts for low-noise regulators assembled.



Close-up.

[TiN]

Had little chance to work on assembly tonight.

LDO regulators on 5 boards are good and quiet, now putting in LTZ section parts (except BMF resistors and LTZ itself, obviously) and output stage parts.



BMF resistors are still going thru sorting/binning so it will take some time till i get overall result table to pre-sort those before installing into boards permanently.

[branadic]

The ceramic caps are quite a bit big for the footprint, aren't they?

-branadic-

[Andreas]

perhaps they are only soldered on the wrong edge 

with best regards

Andreas

[TiN]

I got that one messed up in order, Caps are 1210 on 1206 pads. 

[TiN]

Almost done on matching resistors. 

[TiN]

Sets ready for tempco trim of 7V stage.

Ref 3 is are my "old" prototypes, unlabelled one is USA CalClub version.

[Dr. Frank]

And what about this 7001 reference?
Is it defect, or could you determined its  oven setpoint?
Frank

[Philipp23]

Hello

I hope it is OK to write down this here. Is any one of the lucky owners of the 10V LTZ1000A reference willing to trade it against other precision metrology equipment? For example a National Instruments PXI-4071 7½-Digit FlexDMM or a very nice digital Multimeter that was way ahead of its time the Prema 8017 Multifunction Meter?

Kind Regards

Philipp

[Echo88]

@Kleinstein: Could you elaborate on the possibility to omit the Negative Voltage Supply for the bipolar output buffer stage which is used in the FX-reference?
Im at the moment testing this output-design and thinking about doing the same, but cant think of a solution which wouldnt add drift and noise when shifting the ground.
Im using an LT1054-based charge-pump for now, to generate the needed negative voltage, like its done in the 732B.

[Kleinstein]

The circuit to provide separate negative side drive and sene pins only needs a relatively low voltage. Depending on the OPs use some 0.5 or 1 V could be already enough, if one limits is to compensating normal small lead resistance.  One can create such a small negative supply with just something like 2 forward diodes to get a small voltage drop. So the external supply negative side would be the negative supply, the reference circuit ground would be behind some 2 diodes, some 1.2 V higher.  This would need the external supply to be higher by the drop of the diodes. So 12 V external supply could be too little or at least quite close to the edge. A 14 V supply could be sufficient.

Here it helps that modern AZ OPs often are Rail to Rail. So just -1.2 V or even -0.6 V could be enough to compensate a few 10 mV of drop on the wires.

[Echo88]

Indeed it works, thanks.
Im wondering about the noise contribution of the diodes and would have expected that the output-noise increases, but it stays the same no matter if i use a dedicated -6V Supply or the diodes (Noise/Noise2.asc).
Simulations are attached. Maybe i made an error? Didnt simulate noise in a long time.

[Kleinstein]

Creating the negative supply from the string of diodes makes not big difference over a real regulated -6 V. The OPs PSRR  should be quite good and the voltage over the diodes is usually quite low noise (most regulators are higher noise), just a little temperature dependent.

The noise simulation is not really correct, as Vout is relative to the ground symbol. So the negative side sense part is not correctly included. To get it better, move the GND point to the negative output.

The OPs seem to not include noise (at least not in the models used on my PC).  The noise model of the reference may also be not that accurate.

[Echo88]

Weird, if i connect the gnd-symbol to the negative supply (Q3 Collector) as you suggested, then ill actually get less measured noise than when ground-symbol is connected to the lower side of R3 across Vout.  This happens in both posted versions, with diodes and Voltage Source as negative supply.
Did you simulate your ground-symbol-suggestion? I use the normal Solver-Method and cant recall that i changed any other tolerance/etc. that might influence the simulation.
I should use other OP-Amp-Models if it doesnt include the OP-Amp-Noise and i was suspicious wether LTSpice would simulate diode-noise correctly.
I think the diode-solution might produce less noise than the cap-switcher which produces EMI, but ill have to read more about noise-simulation to get a better fitting simulation of the diode-based-solution.

[Kleinstein]

I have not done the simulation with the moved GND.  Getting a lower noise with the correct GND point is not such a surprise, as the OPs are modeled without noise.  Though the difference should be rather small as it's mainly the diodes and a few resistors.

One can kind of add some of the OP voltage noise to the simulation with a high resistor in series to one input. Some, likely more modern models do include noise for the OPs too.

AFIAK the noise of forward biased diodes is modeled reasonably good. So the diode part is definitely low noise, just a more drift for the negative supply voltage.

The charge pump chips tend to produce quite some ripple / spikes, depending on the layout - not really something to see in a simulation. Chances are a switcher with inductor is lower EMI.

[TiN]

Guess it's a big (sorta) day today.

First preliminary FX was shipped few hours ago.







Not too impressed with Pomona 3770 posts. While job getting done, they are still kinda flimsy to use. Certainly do not ooze quality as "proper" 25$ 5-way posts. 

I'm still undecided if should use LEMO multi-way connector to provide input +/-12V power for module, isolated power for EEPROM/PCB temperature sensor and I2C data, or just stick with simpler dual BNC connectors for +/-12V only. LEMO option is nice but it's expensive and over the planned budget already. If somebody have better idea, feel free to tell.

[Zucca]

Thanks for the update, personally I would love the LEMO. I wanted a Ferrari, you are getting pretty close to perfection.

"Abbiamo fatto trenta, facciamo trentuno!"

It is a phrase that is historically attributed to Pope Leo X who in 1517 had to name some new cardinals. With great work, he came to draw up the final list, reaching as many as 30 names, already a substantial number. In the end, however, he realized that he had not added a prelate of great merit and fame, not being able to exclude any cardinal, he added the name to the list, pronouncing the famous words: "We did 30, we do 31!"

[Andreas]

If somebody have better idea, feel free to tell.

For power supplies I usually use DIN (Diode) connectors.
e.g.
https://de.rs-online.com/web/p/din-steckverbinder/1564224/
https://de.rs-online.com/web/p/din-steckverbinder/0776088/
https://de.rs-online.com/web/p/din-steckverbinder/0776167/

But you could also use D-Sub connectors.

with best regards

Andreas

[guenthert]

Congrats!

Pity that they're not for sale, but I hope that, once I'm reunited with my gear, I can get on the US CalClub list and meet one of those units then ...

[TiN]

Playing with matplotlib and python. Here I can present tempco plot, captured from eight xDevs.com FX LTZ1000A references (before trim, as is):



Purpose of the test is to establish baseline for TC before readjustment, to make sure we have no problems with temperature stability. My final goal is to have each ref 10V output deliver better than 0.03 ppm/K.

In test - each reference scanned twice, one value for raw LTZ1000A output (7.x-ish volt) and second value is 10V output (before trim adjustment). Except one output, channel 12 is connected to my xDevs/Fluke 792X FX ref +9.9999779(22) VDC, outside of chamber to verify stability of scanner setup. Worst ref = channel 14, 10V output -0.059 ppm/K. 
Scanner = Two Keithley 7168 nanovolt JFET cards. Detector : Two Keysight 3458. 

Temperature swept by Keithley 2510 in DIY airchamber box, with 40W watercooled TEC and Xeon copper fansink inside as heat radiator.
Temperature in the box measured by Fluke 1529 thermometer and Honeywell RTD PT100 probe.

Used RAW data file, DSV format.
Python 2.7 app to plot graph. Depends on scipy, numpy and matplotlib.
 

[TiN]

Alright, after 95 weeks, time for long awaited update has come.
Few units are completed and delivered while ago as test drive, and feel happy so far:

MM - 1 unit [DONE]
USA cal club - 1 unit [DONE]

Now I need these people to send me their shipping information in PM, so I can proceed.

montemcguire - 1 unit [PENDING]
dr.diesel - 1 unit [TBD]
chriswebb - 1 unit [TBD]
PeLuLe - 2 unit [PENDING]
bck - 1 unit [TBD]
CM - 1 unit [TBD]
zucca - 1 unit [PENDING]

I understand it was a loooong wait, so If somebody changed mind and don't want a ref anymore, that is OK, just let me know so I update the list accordingly.

[niner_007]

If anyone changed their mind or you got extras, I’m interested, please let me know

[syau]

If anyone changed their mind or you got extras, I’m also interested, please let me know

[wolfy007]

Same here, if theres an extra, Ill take it.

[Zucca]

This is like good italian wine, the more you wait the better it gets.

PM sent.

[nexus]

Any chance that the design will become open source? I would love to build one myself and send it in for calibration!
Only components I see as being difficult to source are:

-fluke custom film resistor network, P/N 2464707
-matched precision resistors as covered in your article: https://xdevs.com/article/tcr_test/#vpg2018

If you don't plan on releasing the design for end-user assembly, I'd still be interested in buying a complete module.

Thanks!

[dr.diesel]

Merry X-Mas to me!

Many thanks to @TiN for his efforts and support of information here on the forum.  I've not yet powered this up and don't intend to until I figure out how I want to 24-7 power the unit, I don't want to just hang it off one of my linear supplies when needed.

 

[notfaded1]

Really nice presentation of this... even a carrying case aye!  The labels are super neat too.

Bill

[Zucca]

Just to say thanks to the immense TiN.



His strive for perfection and the results he delivers are truly an inspiring example for me.

[pelule]

still patiant waiting and interested. PM sent./Pelule

[TiN]

still patiant waiting and interested. PM sent./Pelule

Unit picked up for you today. My apologies for delays.

Here's report for your unit:

[franklin]

Hi TiN.
Any chance that this project will be restarted?

[TiN]

Unless somebody finds magical resistors, I'd say chances are 0

[Blue]

https://vpgsensors.com/ sells them right?

[niner_007]

https://vpgsensors.com/ sells them right?

Not the Fluke one.

My design choice would have been to use custom orderable parts from Vishay and no unobtanium resistors. After aging you could order custom ration / value VHD200 resistors to trim to 10V. This would take a long time aging though, and before any units are made available.

[SigurdR]

I am interested!

How long do you think it would need to be aged? A year? 3 years?

[binary01]

Just a quick +1 from me... I'd also love something like this if the unobtanium component issue could be overcome.