Ultra Precision Reference LTZ1000

[martinr33]

Try Texas Components. They are a Vishay licensee and are pretty quick. They don't carry the full range.

You will probably have to call them.

[syau]

After several email exchange with an “authorized” distributor in Shenzhen, I gave up and get a quote from ELcoPC LLC in the State for few set customized VHP/VHD series resistors.

It will take 5 months to get it deliveries even if under Express Line.

[Dr. Frank]

Hello,

maybe you remember my LTZ 1000 tuner box assemblies.




Inside the tuner box, there's space for a simple linear regulator, like an LM317 based one.

Those boxes were all powered by transformer wall warts, which I salvaged from different devices. I use a 2.8VA, 15V, 180mA transformer as a replacement for the original ones, plus bridge rectifier and electrolytic capacitor for an unregulated DC output. Here's an old  transformer case from Nokia.



It was always my intention to have a proper mechanical assembly for these tuner boxes, inside an outer case.
At first, here's the LTZ #5, solution copyright by Andreas, battery powered, where I added CuTe Pomona jacks for output, Guard (connected to isolated tuner box) and case ground.



Those Pomona jacks are long enough to mechanically connect outer case and the tuner box, adding an isolating plastic plate, and plastic spacers, so that outer box and tuner boxes are electrically isolated from each other, and also avoiding long connection cables.



On the rear side, the tuner boxes are screwed on to another plastic plate, using an electrical connection to the tuner boxes, serving as a common Guard.



Then I added the Guard and Case Ground jacks, a simple 12V LM317 stabilizer, and connected all 4 modules in parallel.



I would have of course preferred 4 isolated supplies.
This way, the 4 reference grounds differ by 50.. 250µV, because of differences on the supply cables, due to different oven supply currents. It's also not possible to connect any of these 4 reference grounds together (to make differential measurements), as this directly shifts the references outputs.

That's ok for me at the moment, as I regularly compare all of my 12 voltage references against each other, by absolute measurements of each by means of my 3458A.
Maybe somebody has an idea, to commonly supply several LTZs and properly interface the reference grounds. I currently have no idea, how this is done inside the Datron 4910.

Here's my Tower of References, where I maintain Volt and Ohm.



Frank

[Kleinstein]

The setup looks nice.  If there is an EMI problem, one may consider adding some common mode choke / ferrite ring just inside the reference boxes.

For the ground problem, one could consider seprate transformers. Some mains transformers come with 2 fully separate winding and thus could power 2 ref modules.

If a single supply is wanted, one could build a circuit to shift the low sides of 3 modules a litte (e.g. +-1 mV range), so that the reference low side is at the same level as the one unit taken as a master and common ground for the output. So the low side would be all from 1 box and the other low sides on the extra extra amplier / compensation part. Shifting the low side would take quite some current, though only at a low voltage.

An alternative would be to directly connect the reference gounds and make sure the negative supply leads get the same current as the positive side. This would add some current regulation / mirroring and maybe the voltage regulation. So more circuit,  but likely less extra noise and drift. The refrence modules should not provide a significant current, so I don't think one would have to treat that current separate.

[dietert1]

With a bipolar supply one can use ground current compensation similar to the HP 3458A reference module.
With our LTFLU references, in order to use a simple +15V supply without negative supply, i am using a silicon diode in the Gnd connection to simulate a negative supply of -0.7 V. Then one can use chopper amplifiers to equilibrate the Gnds using the Gnd sense lines from each reference.
Or use a transformer with multiple secondaries to get separate supplies for each reference.

Regards, Dieter

[Kleinstein]

Ground current compensation is with the 3458 is a good idea: the actual reference current is constant and thus possible to compensate with a constant current, and the heater and OPs can use a separate ground.  It is still not clear if this works well with the existing boards.

[BU508A]

I would have of course preferred 4 isolated supplies.
This way, the 4 reference grounds differ by 50.. 250µV, because of differences on the supply cables, due to different oven supply currents. It's also not possible to connect any of these 4 reference grounds together (to make differential measurements), as this directly shifts the references outputs.

That's ok for me at the moment, as I regularly compare all of my 12 voltage references against each other, by absolute measurements of each by means of my 3458A.
Maybe somebody has an idea, to commonly supply several LTZs and properly interface the reference grounds. I currently have no idea, how this is done inside the Datron 4910.

If you'd like to go with isolated supplies, I'd recommend two Talema toroidal transformers, e.g. the 70033K. It has 2x 15VAC output and 2x 3.5VA. It is available at Bürklin for ca. 17.50 EUR/unit plus shipping.
https://www.buerklin.com/de/Ringkern-Flachtransformator-7-VA-15-V-15-V-2x115-V-Talema-70033K/p/38C766



Sizes (LxWxH): 50mm x 50mm x 24mm (ca.)

Attached the datasheet

[Echo88]

An input filter, two of those https://www.mpaudio.net/product-page/audio-grade-transformer-60va and a nice PCB with CMCs and some LT3042s, preferably with good layout that actually achieves the stated PSRR -> 4 independent isolated supplies.
Yes, those transformers have way too much power, but they are the only ones i could find with shielding.

[BU508A]

If you really want to, you can have a custom toroidal transformer made, including a shield winding and four separate secondary windings.

Multi Circuit Boards is offering this service but it is not cheap.
https://www.multi-circuit-boards.eu/en/products/toroidal-transformers.html

[Andreas]

I would have of course preferred 4 isolated supplies.

Hello,

that is why I prefer to use batteries during measurement.
Since the linear regulated DC wall plug adapters are now unobtanium due to regulations the 2nd best option is to use 4 AC wall plug adapters.
If you dont have enough old outdoor Christmas Light transformers still some AC adapters are available.
I would take 4 of these:
https://de.rs-online.com/web/p/steckernetzteile/1391764

with best regards
 
Andreas

[dietert1]

Bürklin also have these shielded 12 VA transformers with two secondaries:
https://www.buerklin.com/de/Universal-Transformator-12%C2%A0V%C2%B7A-10%C2%A0V-12%C2%A0V/p/35C100
The Schmidbauer part number is UNT01223012,
These transformers have a residual coupling of about 3 pF around the shield.

Regards, Dieter

[RoadDog]

Been looking at ideas and other peoples builds to figure out building some references the last few days.

The KX board looks pretty cool from the write up on XDevs.

How do I get the Vishay resistors? Digikey doesn’t stock them and I’m guessing VPG wants you to order them in bulk? Are there others I can use with the same performance?

I like the ability to just order three boards from OSHPARK and building three references on a proven design. Searching here I’ve found a lot of possible options and seen that folks have pooled money to purchase lots of resistors in the past. I’ve searched a lot here. Hard to sort through it all and the shortages only compound things





 

[TiN]

You can order resistors in 1 pcs from VPG, if you ok with waiting times 6-8 month (been always like that even before any modern shortages madness).

KX design is old and not the best (overpriced unnecessary 70kOhm VHPs, unbuffered unprotected output, chopper opamps).
I'd suggest looking at FX design, which is more traditional and much more friendly and robust for general use.

But you gotta have LTZ or ADR chip first, without it not much can be built.

[RoadDog]

But you gotta have LTZ or ADR chip first, without it not much can be built.

Does it have to be the ACH or can I use the LTZ1000CH?  I can find those. ACH is out until the end of August unless I want to take my chances with eBay or “quotes” from unauthorized dealers

I tried your cart trick for that ADR on the website. They’re backordered and looks like you have to buy 100 at a time. No more small quantities I think.

Cool build. Love how you just hack random parts from different Fluke products into something.

If I wanted to go that direction are there gerber files somewhere I can download to use at OSHPARK?

[Ole]

If I am not mistaken the only difference is the thermal resistance.
The LTZ1000 is at 80°C/W and the 1000A is at 400°C/W. (Source: https://www.analog.com/media/en/technical-documentation/data-sheets/LTZ1000.pdf )
I am still quite new to the field of electronic measurement but, if you are using an insulating material it should matter relatively little. You might need a bit more insulator though.

The LTZ1000CH#PBF does seem to be in stock at Analog Devices Website, and you can, once its availible again, purchase the 1000A aswell in 20-packs

[julian1]

Is there a single-supply op, like the lt1013, but with a 'normal' soic dual pinout?

I ordinarily use opa2277 with +-rails, but get start-up problems when the rail supplies are not sequenced - so I want to switch to single supply.

Most of the modern low noise ops are quite fast/jfet  >10MHz and probably more susceptible to emi. But with an Andreas/Dr Frank compensation capacitor (R/C=10k/10n), i suspect it may not matter.

I found op213, which looks interesting - probably overkill for ltz1000.

lt1013      0.55uVp2p 0.1-10Hz.   drift 2uV/C.   GBW ~= 800kHz-1MHz

op213.   120nV p2p 0.1-10Hz.  offset drift 0.8uV/C best grade. low noise, low drift, single supply op.  4-36V.  GBW 3.4MHz.

[Kleinstein]

The OPs drift and noise are attenuated quite a bit. So the OP does not absolutely be low dift and low noise.
The TLE2022 is possible, but not with the best performance.
There are quite a few FET based OPs that could work (e.g. TLC272 / TLC277 / OPA2196 ) - though not highes performance. Not all are super fast.

The ADR1000 DS suggsts the ADA4084 , though this one could be tricky with the input bias current changing just at around 0.5 V where it is used. So it would not be my choice.

The OP213 would also not be that good, as it has quite some current noise and the source impedance would be around 50 Kohms. So it would be more noisy than an LT1013 and maybe even LM358.

[Andreas]

Hello,

I´d probably go with the ADA4522-2. (because of higher Large-Signal Voltage Gain).
And of course it is not forbidden to put a weak pull up on the zener for start-up.

with best regards

Andreas

[julian1]

hmmm. that's a bad miscalculation on my part. input bias current (leading to current noise) of op213 is high 600nA max, versus 15nA typ lt1013, versus 2-4nA typ opa2277.

It is interesting that bipolar ada4084 is recommended for ADR1000 despite 16MHz GBW.

Is there a reason not to consider jfet input? Bjt has better gain and lower flicker noise. But the datasheet specs of modern jfet are also very good - presumably due to extra complexity around the input stage.

For example, opa2140 (Input voltage range includes V– supply) - and using a compensation/bypass capacitor to slow things down/address emi.

I am using mc33172 right now as a test - basically equivalent to lm358. 

[julian1]

And of course it is not forbidden to put a weak pull up on the zener for start-up.

Ohhh. very interesting, this might be simplest solution. I was also considering the discrete jfet thing in the 34401a used with the lm399 to bias the startup, but its tending to too much complexity.

[Kleinstein]

An Az OP can be tricky for 2 reasons: one is the start up (there may be some delay, though usually not too bad) and the other is the the input current, that often includes short current spikes that can upset the reference.  So an AZ OP may look good from the drift and low frequency noise, but may still cause extra noise. An AZ OP may want some extra filtering and good supply decoupling. So it may not work well in an unmodified circuit. The change for the slightly unusual pinout of the LT1013 is likely easier.

The ADA4522 is very low voltage noise, but this comes with relatively high current noise and also likely with relatively large current spikes.
So the AZ OP of choice would be more a type more suitable for a higher impedance like LTC2058, OPA2187, NCS21912 or AD8639. Less heat from a lower supply current also helps. The choice of AZ OPs for higher voltage is somewhat limited.

I see no absolute reason not to use a JFET type. The OPA2145 or OPA2196 may be an option.
Even the MC33172 may still be not that bad - the OPs noise and dirft are attenuated by about a factor of 200. So the OPs noise would still be a small contribution.

[Cerebus]

And of course it is not forbidden to put a weak pull up on the zener for start-up.

Ohhh. very interesting, this might be simplest solution. I was also considering the discrete jfet thing in the 34401a used with the lm399 to bias the startup, but its tending to too much complexity.

On the schematics for the 34401A that I have the startup circuit for the LM399 is a pull-up via a diode (1/2 a BAV99) and then 1k resistor to the +5V rail. Where did you you see a JFET?

[julian1]

On the schematics for the 34401A that I have the startup circuit for the LM399 is a pull-up via a diode (1/2 a BAV99) and then 1k resistor to the +5V rail. Where did you you see a JFET?

You're right, 34401a is 1k bias to 5V.  I was thinking of this self-start for lm399, https://www.eevblog.com/forum/metrology/influence-of-resistors-in-lm399-reference-circuit/msg1212712/#msg1212712

[julian1]

Ok, went back to the dual supply opa2277. Then applied the 344401a zener startup method.   

3.3V rail -> mmbd4148 -> 1k as pullup.

It appears to work well. The 3.3V digital supply is not ideal, given potential capacitive coupling, but is sufficient for the moment.

The question now is - can the in-loop diode at the op output be eliminated? I don't think serves any other purpose except as a bias - and is a potential source of emi rectification?
 
Edit. seems to work fine, with in loop diode removed. edit2. now seems not to be 100%.. ok, lowering the resistance, to increase pullup, and better control over the op loading, and it seems reliable.

[dietert1]

That diode you eliminated prevents the opamp from blocking startup. Startup will then depend somewhat on the opamp input and output characteristics. One can bypass the diode with a 10 nF cap to prevent RF rectification.

Regards, Dieter