Ultra Precision Reference LTZ1000

[Dr. Frank]

Swapping pin1 and pin2 of the heater was my first fail about 15 years ago, on my prototype.

That way I discovered the internal diode across these pins, and in turn caused the first revision A of the LT datasheet, where you should now see this diode.

Simply don't swap pin1 and pin2 !


Frank

[TiN]

Received PCBs. Sex on a stick!

Assembling first board (by hand as usual):

[Kjelt]

Hi TiN,
Looks awesome (and expensive  ).
What is your goal? A 9,5 digit accuracy so a stable last digit ?

[syau]

3458B 

[TiN]

And first board is now online! Need to trim and measure each cell performance, before I start tests on combined output.

No smoke, no sparks, everything worked from first try so far. Board eats ~115 mA from +18V rail and some mA from -15V rail.



Also next photo gives some idea how many expensive resistors did I sort to find a match for 1 quad board.

[Kleinstein]

I don't see a real need to match the resistors for a board with multiple references in parallel. It can help to match the LTZ1000 chips, to have a reasonable equal voltage so that the averaging resistors get even less important.
There could be some adjustment needed for current compensation if used.

[MiDi]

Obfuscated parts, hidden collector resistors, a link leading to nirvana...
A Thriller

[TiN]

My 3rd 3458A is not golden, so need to goldify it myself. 
Main goal to get <0.2 uVpk-pk noise, so I can avoid need of external 10V reference for calibrations.

[exe]

TiN what's the cost of the board? Looks like >$500 to me...

[TiN]

Total cost in parts for one board is about $3k.

[Kosmic]

Really nice board Illya, thank you for posting.

Starting to look like a Valhalla 2720gs 

[antintedo]

If I squint really hard I can see OPA2140A and ADA4522-2. How is my eyesight?
Edit: that does not look like $3k worth of resistors and ICs, is the PCB $1.5k per piece?

[Kjelt]

Total cost in parts for one board is about $3k.

You were meaning it literally, not kidding 

My 3rd 3458A is not golden, so need to goldify it myself. 

[Theboel]

@ Tin
Just wondering to lowering the overall cost maybe You sell for the unmatched resistor. 
not a disc price but cut the waiting time

[2N3055]

If I squint really hard I can see OPA2140A and ADA4522-2. How is my eyesight?
Edit: that does not look like $3k worth of resistors and ICs, is the PCB $1.5k per piece?

That is probably total cost of all resistors and parts from which he selected ones that he put on the boards.

[3roomlab]

My 3rd 3458A is not golden, so need to goldify it myself. 
Main goal to get <0.2 uVpk-pk noise, so I can avoid need of external 10V reference for calibrations.

what if you lower the temperature of the entire 3458a? then maybe have a LTZ heater at 30C?
https://www.eevblog.com/forum/metrology/dmm-adc-noise-comparison-testing-project/450/

dont know what is the set point of jorns LTZ, maybe 65C ?
that post, jorns stdev = 0.37uV @ 18C, is this better than most DMM/LTZ?  and its a 34470a
maybe a cooler LTZ with smaller R1 ... 10mA ? but looking at some other posts, it seems to indicate that the noise of LTZ is not smallest at higher current. i cannot remember the post, after 3mA, noise increases?
did you try your select LTZ at different R1 and see if each has a low noise at a different current? I am wondering if each LTZ noise is exactly the same at exact same current, which i think is not true. that could mean sort by noise instead of Vz? but i dont know ... never build one of these before ...

i always have a thinking that stability can be mitigated by removing temperature problems (that is after reading all the wonderful tempco experiments of cos)
but noise? can noise be mitigated by tweaking temperature as well?

kill the mad cow tempco, murder the humidity, now .... what do we do with the noise ...
i did see some articles about cooling to reduce 1/f noise ... but i dont know. im sure its not the LN type of cooling, i think electronics are not specially made for LN. esp when there is humidity around

[dietert1]

My answer would be a very serious study of LTFLU, which also has four zeners inside. Appearently it can be used at 4x 3 mA current. But that belongs to the LTFLU thread. I will also try a low impedance capacitive 7V -> 10V stage to improve noise. Yet another thread. I will also try to use a TEC "oven" at 23°C (with desiccant inside) in order to keep noise and aging down. Made a small, high efficiency TEC controller to run from battery. There are lots of interesting details.

I think TiNs approach is more about engineering. Rapid results! To whom it matters: Recently i found that OPA189/OPA2189 has about 5 times less current noise than ADA4522 (in the specs). Maybe it's a new generation of chips, similar to AD4528. If i have the calc right at 50 KOhm input impedance total noise will be roughly the same as a OPA140.

Regards, Dieter

[Kleinstein]

For the LTZ1000 circuit itself the OPs noise and drift is not that critical. It is attenuated by about a factor of 200. So the LT1013 is still well good enough and no need to worry about AZ spikes.  The interesting point of this 4 reference circuit would be how to combine the outputs: the low side (GND) could probably use individual current compensation. For the high side with averaging resistors it would likely need a buffer amplifier, and this would likely be an AZ type, like ADA4528, AD8628, LTC2057 or OPA189.

With so much extra power, one may have to check the thermals inside the 3458. Worst case by getting a -5 V rail for the current compensation and possibly powering the LTZ1000 heater from some 8 V rail.

Normally using 4 reference should about half the noise. So the expected noise level should be more like 0.6 µV_pp (.1-10 Hz), maybe a little less with selected LTZ1000. A little of the 10 Hz end may be removed by filtering, but not that much. However the relevant bandwidth would be more like 0.01 - 1 Hz anyway. With 13K/1K dividers and LTZ1000A the temperature is already relatively low - this also helps keeping the power for the heaters low.

[mimmus78]

So this is noise of my dual LTZ A9 board (700nV to 800nV p2p) ... time to order a couple of 50 ohm resistor and check how it performs at high current.

[exe]

So this is noise of my dual LTZ A9 board (700nV to 800nV p2p)

What do you use as an amplifier?

[MiDi]

I guess it is the one from the DIY-LNA-Thread with 0.1-10Hz and x10k amplification.

[mimmus78]

Yep this one

I guess it is the one from the DIY-LNA-Thread with 0.1-10Hz and x10k amplification.

[mimmus78]

And this is the board "slap together" ... not nice as Tin ones but it seems to works. Since one week it tracks sub ppm the other 3 LTZ I built. I'm considering to put everything in one pcb instead of having it stacked.

[TiN]

Wow those leads are long

[mimmus78]

Wow those leads are long

They will be cut when on the final pcb board ;-)