Ultra Precision Reference LTZ1000

[UPI]

Reminds me of this relatively quick/inexpensive idea:

http://www.romanblack.com/xoven.htm

[SeanB]

Not a vacuum, but in a sealed enclosure with a large sachet of silica gel inside to adsorb water. Heat it up in an oven to dry it out well and then place in the dry enclosure hot and then seal it up while it is still hot. Then it will make a very low humidity interior. you will need to have IP68 rated connectors in the box, and preferably a smaller inner box to hold the reference in an insulated place. You can have heaters in the outer box to keep it at a stable temperature, and this will keep the inner stable. All you need on the outer is to keep it to within 2C, as the inner insulation will attenuate the variances. I have seen this ( without the dessicant) in an appnote to get a inner box to be temperature stable to within millidegrees.

[chrome]

Can anyone tell my why 6.95V?

It seems like such an arbitrary value.

I'm sort of assuming it's inherit with the design maybe?

[SeanB]

Reminds me of this relatively quick/inexpensive idea:

http://www.romanblack.com/xoven.htm

I did that years ago, and they work well.

6.95V as it is a stable point where the voltage is not quite avalance breakdown and not quite zener action. Lower and higher voltages have drift with temperature, this is a sweet spot in the characteristic. Often you get a 6V2 zener diode internally in series with a silicon diode to make a 6V8 unit, as this is also very stable as the tempco's of the zener and the diode are nearly the same but of opposite direction so they cancel out nearly totally.

[(In)Sanity]

Thanks for the replies about the vacuum,  Potting I assume would be subject to moisture absorption and too slow to respond to temperature changes or corrections ?  So if it got cold for example it would take far to long to stabilize and perhaps have too much hysteresis?

Also which would be better for a 10.0000 volt reference,   a 6.95 and an op-amp or two 6.95's and a voltage divider ?   Which would be more stable ?

I'm full of questions at the moment.

Thanks,

Jeff

[SeanB]

Potting would also stress the references. the biggesst issue is the resistors you use, not the opamp. You can get an opamp that is chopper stabilised, which removes it's own drift, but the resistor drift determines the stability. Better to use a single reference and multiply it up to 10V, one less source of error.

[(In)Sanity]

Potting would also stress the references. the biggesst issue is the resistors you use, not the opamp. You can get an opamp that is chopper stabilised, which removes it's own drift, but the resistor drift determines the stability. Better to use a single reference and multiply it up to 10V, one less source of error.

Thanks Sean,   that's what I was wondering.   I was looking to see which would be the lesser of the two evils.  To build one of these references correctly appears to be a very expensive endeavor with $90 references, $28 resistors and an equally expensive trim pot which should be avoided if possible,  ohh and let's not forget the chopper amp.   I may still pick up an LM399 just for experimentation sake and not breaking the bank.  I'm fortunate that I have a full finished basement and the temperature down here stays pretty constant relative to the rest of the house. So my 3457A also stays pretty stable.  I'm pretty good with long term projects,  so this could be fun.

Jeff

[SeanB]

Often the resistors are selected to be close, then a variable is used to get the final adjustment and is replaced by a fixed value. Leads to using Vishay foil on glass resistors in series and with final trim being by soldering a few low TC metal film resistors and then cutting resistor leads to trim up to the right value.

[PA4TIM]

http://www.pa4tim.nl/?p=2531 my LM399 ( the last one at the bottom, all my reference projecs are on this page)

Quarks: i have the ESIs but I did no stability tests with them. If i use them to calibrate I measure all off thecterminals and then take the avarage. Then measure all in serie and that must be the same value as the 12 separate. I have the bar that sets them parallel and I'm told this is enhances the precision by averaging out the differences.

Mine all measure very constant. No significant differences between terminals of one unit.
Problem of monitoring drift, tempco ect ( Vrefs, resistors, capacitors ect) is you never know what is drifting, the DUT or the instrument or both 

Is it because australia is very far away or is it just me. This forum is very slow, it takes ages for pages to open, before I can type comments and many notifications never seem to arrive. ( do I made wrong settings 's ? )
It would be great is i could set notifications on permanent z

[SeanB]

Server is in the USA. They must be awake and watching movies............

[(In)Sanity]

Server is in the USA. They must be awake and watching movies............

Haaa,  the only movies I watch are electronics related for the most part.  Sadly I can't stream anything in HD anymore,  even though I have a 20 Meg connection.  Everything is going to crap lately..but that's another topic.

Thanks for the information about not using a trim pot.  I kind of knew that in the back of my head,  but somehow was in denial about it actually being done that way.   I wonder if anyone has made any good references by balancing out the positive and negative temperature coefficients of the devices in the circuit?   I know most DMM's and such have a thermistor or ten to obtain somewhat of this behavior.  I guess it would be no better then just using an oven no matter which way you rolled it.   Perhaps a combination of the methods.   I'm such a newb at temperature stability issues.   

Also can't everything be corrected in software

Jeff

[quarks]

http://www.pa4tim.nl/?p=2531 my LM399 ( the last one at the bottom, all my reference projecs are on this page)

Quarks: i have the ESIs but I did no stability tests with them. If i use them to calibrate I measure all off thecterminals and then take the avarage. Then measure all in serie and that must be the same value as the 12 separate. I have the bar that sets them parallel and I'm told this is enhances the precision by averaging out the differences.

Mine all measure very constant. No significant differences between terminals of one unit.
Problem of monitoring drift, tempco ect ( Vrefs, resistors, capacitors ect) is you never know what is drifting, the DUT or the instrument or both 

Thx again,
I do also have  the optional parallel, serial-parallel compensation networks and the shorting bars and have measured allmost all possible  combinations.
What I am still interested in, is how resistance transfer works. If you could share some knowledge that would be great. But this is OT here  and I can, if there is any interest, update the other post with more details.
Tomorrow I will also pick up some more related ESI gear from custom.

[Conrad Hoffman]

FWIW, my Analogic 8200 6 1/2 digit voltage/current reference uses the LM399 and it's remarkably good. If you can find the circuit on-line, it uses analog switches to change the output from a master divider chain and is definitely worth studying.

[quarks]

Hello Conrad,
analog switching sounds interesting and with LM199/299/399 I remember a Bop Pease article where he used multiple of them together. I will definitely try to have a look at it.

A first quick search for a Analogic 8200 schematic unfortunately showed no result for manual and service manual.
But the search lead to
https://www.eevblog.com/forum/projects/kelvin-varley-divider-(and-precision-voltage-source)/msg71081/#msg71081
so I tried also Data Precision also with no luck. Do you maybe have a link or can post a photo/scan?

thx
quarks

[muvideo]

quarks:
http://www.ko4bb.com/Manuals/09%29_Misc_Test_Equipment/DataPrecision_8200_6.5_Digit_Calibrator_Service_Manual.pdf

I'm reading it now, it's a wonderful manual, almost an application note

[robrenz]

I believe this is it

Edit: Sorry muvideo I was too slow

[quarks]

great, thx a lot

[quarks]

in the Wavetek 1271 I just saw, they use a LM399 and a LTZ1000 Board

[free_electron]

I believe bob pease had a circuit using a few lm399s to make a nanovolt reference.

[PA4TIM]

The Czar of bandgap references himself, that must be a decent design.
Do you know where it was publiced ?

[muvideo]

I remember only one of his "What’s All This ___ Stuff, Anyhow?" series,
this one: http://electronicdesign.com/archive/what-s-all-long-term-stability-stuff-anyhow
But I'd like to know if there is some more in-depth article...

Fabio.

[free_electron]

i think it's in one of his Bob Pease Show video's... hang on ...



or it may be in his book analog troubleshooting. he uses like 10 of em ...

[jnd]

i think it's in one of his Bob Pease Show video's... hang on ...



or it may be in his book analog troubleshooting. he uses like 10 of em ...

I saw this one just week or two ago, there is one LM399. BTW TI's channel on Youtube have them reuploaded in better quality (I tried to search there for "anyhow") which is interesting. The new version is here, it's still pretty relevant to the precision talk here:

[nukie]

These are 3 of my LTZ1000 references and geller svr 5V. Along with two Motorola SZA263 used in the Fluke 8840a and Fluke 731B voltage standard and Fluke 732A reference standard. Still waiting for free time to attack those SZA263 with solder iron. Please note, without high precision supporting passive components these parts are as good as nothing.




You can read more about it here
http://home.51.com/jj3055/diary/item/10053954.html

Schematics here
http://bbs.38hot.net/read-htm-tid-36472.html

And finally here how to convert 7V to 10V
http://www.crystalradio.cn/thread-229749-1-1.html

[quarks]

@ Mechatrommer, thx for your reply. I know it will be hard to get all the ideal parts but I at least want to try and find out.

@ nukie, that looks very promising. What you already have is exactly what I thinking to do. thx a lot for sharing.
I will read through your links and come back to you.