behind the scenes: Working on the B revision voltage standard

[JohnnyBerg]

I expect to finish the design of my revision B voltage standard in about 2 weeks. I want to share some toughs of things I am considering, revoking or other interesting stuff.

Currently I have tested a voltage doubler for the 10.00000 V version. I need a voltage doubler, because I want to use one 9V battery. One of the IC's I'll be able to use is the LT1021. The board will have a layout, so that other, pin compatible, IC's will fit.

I already tested a nifty layout with an 8 pin dip, and a so8 package in the same place. 

Anyhow, I was reading in the datasheet of the LT1021 and found this graph:



I marked a spot yellow. At this point a 14 bit converter can have a 3 ppm/°C TC, when the delta operating temperature is 10°C

In normal human language: 
When the operating temperature of a 16384 count (= 14 bit) ADC is between 18°C and 28°C (delta=10°C), at least a 3ppm/°C reference is needed.

Interesting ..

[JohnnyBerg]

Here is that nifty construction with a choice between DIP8 and SO8:



The 3D view only shows the DIP8, but there is a choice



I am in doubt over the mechanical stress relief. Should it be as shown, or are 2 slides on the edge of the pcb sufficient?
Hmm .. I will have a good night sleep over that.

From the datasheet:
Effects of Air Movement on Low Frequency Noise
The LT1021 has very low noise because of the buried zener used in its design. In the 0.1Hz to 10Hz band, peak-to-peak
noise is about 0.5ppm of the DC output. To achieve this low noise, however, care must be taken to shield the
reference from ambient air turbulence. Air movement can create noise because of thermoelectric differences
between IC package leads (especially kovar lead TO-5) and printed circuit board materials and/or sockets. Power
dissipation in the reference, even though it rarely exceeds 20mW, is enough to cause small temperature gradients in
the package leads. Variations in thermal resistance, caused by uneven air flow, create differential lead temperatures,
thereby causing thermoelectric voltage noise at the output of the reference.


So, is it necessary to get some kind of shield over the reference IC? And where do I get something like that? Has to be custom made, I presume.

[Vgkid]

I do not believe that the mechanical reliefs are necessary. And for the shield, is it (the board) going to be mounted in a case?

[JohnnyBerg]

I do not believe that the mechanical reliefs are necessary.

There is a recommendation about that .. I thinks it was from Lineair. They did a video about that.

And for the shield, is it (the board) going to be mounted in a case?

Nope .. on a battery holder ..

[nuno]

Sometimes the reference chip is on a PCB corner, one I did once is in the middle but having 2 cuts (U1 is the reference chip, the black line is the board outline so there's a cut on both left&right):



Still the chip is on the corner on purpose.

[blackdog]

Hi JohnnyBerg,

Why do you not make a circuit board, that fits into one of these nice aluminum boxes.
http://www.eoo-bv.nl/index.php?_a=viewProd&productId=15907

"Dutch Mode on"
Weet je nog, een referentie hoort in een metalen behuizing! 
"Dutch Mode off"

Kind regarts,
Blackdog

[JohnnyBerg]

Why do you not make a circuit board, that fits into one of these nice aluminum boxes.
http://www.eoo-bv.nl/index.php?_a=viewProd&productId=15907

If I make a very stable one (< 2ppm), I consider that.
Problem with that enclosure is that it needs a lot of work (machining and making it attractive), before it is a product that can be sold.

The never ending battle between technics and sales 

"Dutch Mode on"
Weet je nog, een referentie hoort in een metalen behuizing! 
"Dutch Mode off"

Ja .. ik weet het. Maar zo'n metalen bakje smoelt gewoon niet lekker op een foto 

[JohnnyBerg]

I finished placement of all the components on the board. Sort off

For mechanical stress relief I looked again in some datasheets, and I decided to do it like this.

I added 3 slot holes around the reference IC. In there I can put some kind of metal shielding. It will protect against noise radiated in, and drought over the board. As the backside of the board will be a ground plane, the whole reference IC is well shielded.



As I plan to deliver a calibration report with each device, there has to be a serial number on the board. I do not write values any more on the board, as I did in the A rev.

The potmeter is drawn straight, but it will bend on the pcb. I replaced the smd potmeter with a 10 turn type. The bending is necessary for transport. After calibration I seal the potmeter position.



I have some nice gold plated testpoints. They will replace the hand made testpoint from rev. A. I won't be using this switch anymore. It is likely to damage during shipment as it sticks out on the top. Actually the same reason as with the potmeter.

[blackdog]

Hi JohnnyBerg,

If you want, i can buy/test one of your first sample's, to see how stable it is and test the noise level?
Maybe its a good point, to remove the pins on the DIP that are not in use, cut them! (pin 1, 3, 7 and

Kind regarts,
Blackdog

[JohnnyBerg]

If you want, i can buy/test one of your first sample's, to see how stable it is and test the noise level?

Yes, please! I send you one for free!
Edit: speaking with Louis van Gaal: "I hope for your unsalted opinion" 

Maybe its a good point, to remove the pins on the DIP that are not in use, cut them! (pin 1, 3, 7 and

I think that sales has a problem with that 

For the 10V version I've got a LT1021DCN8-10 in order. I'll think I will order a REF5010AIDR for the smd version, as the board provides for both packages.
Nothing to fancy yet, all above 2.5PPM/°K (typ.)

I'll have a good night sleep over the shielding ..

Yeah, deadline is still 2 days away .. 1 ready, 2 to go 

[metacollin]

10.00000 V version

This is a terrific project, the board looks very well done, and I don't mean to disparage it at all.  But, assuming an LT1021 as the reference...it seems like there is 1 too many significant digits there.     Well, or maybe not, I'm just asking if that number is correct or a typo.

10.00000 translates into 1,000,000 counts - meaning you need stability as good as 1ppm.  I've seen LT1021s with temp co's better than 0.2ppm at room temperature in the wild, and indeed, they're not even that uncommon, and that would give you a small but not impractical temperature window for 1ppm stability.  But you'd still have to do some pretty serious cherry picking of a not particularly inexpensive part to achieve that.  The bigger issue though is the long term stability.  The epoxy DIP packaging does absorb moisture over time and is definitely not hermetically sealed, so I doubt any parts will do much better than the 10ppm/kHr drift stated in the data sheet.  Indeed, I found a post by Andreas on the volt-nut mailing list stating that these parts see 0.5ppm drift per % change in relative humidity of the environment.  As a desert-dweller with an evaporative cooler... the relative humidity inside can go from 10% to 70%+ no problem, which is more or less a hell tailor-made for the LT1021.  Or really any reference in a non-hermetically sealed package.

So, you would need to hermetically seal those puppies somehow... I am not really sure what is involved in that, as being completely encased in epoxy seems pretty sealed to me, but obviously it is not.  And I can't find any data on if that helps with the 10ppm /kHr drift.  Maybe rock out with your caulk out and go nuts on the DIP part? I dunno heh.

Basically, 10.0000, or 100,000 count (10ppm drift budget) is very reasonable.  10.00000 (1ppm drift budget) seems doable, but you'd need some real special sauce going on that you haven't mentioned in this thread.  But that'd be pretty sweet . 

Just wanted some clarification.  Regardless, its a great little standard regardless of ±1  one significant digit.  Again, I don't mean to disparage the project at all, but just wanted to double check what was going on.  Anyway, looking forward to seeing how this project evolves!

[JohnnyBerg]

Just wanted some clarification.  Regardless, its a great little standard regardless of ±1  one significant digit. 

You are absolutely right!

I was a bit over enthusiastic, and I wrote 10.00000V. That is what my 34401A shows. I know that 1 of my 2 34401A's is only 1 or 2 count out, but I would have to say that I'd be able to trim to 10.00000V within 35ppm accuracy. (being the guaranteed accuracy of the 34401A). And I would have to specify a guaranteed time period.

For the 10V version I've got a LT1021DCN8-10 in order. I'll think I will order a REF5010AIDR for the smd version, as the board provides for both packages.
Nothing to fancy yet, all above 2.5PPM/°K (typ.)

There you have it. That is what I expect to get.

This project will have 2 siblings, a low cost version with a LT1004 and another REF5050-isch. These siblings will have a precision current source and 2 precision resitors on them. They are intended for multimeter testing and calibration.

Behind the scenes I am playing with some other stuff, but i'll save that for later 

[blackdog]

Hi JohnnyBerg,

This is the only language i speak => "I hope for your unsalted opinion"
I am a Dutch and you too, you will few problems, concerns saying it like it is. 

Yo can find my e-mail adress on C.O.

Kind regarts,
Blackdog

[JohnnyBerg]

After some iterations with the autorouter, and some manual work I'm done. And satisfied 



Backside almost complete copper. Good shielding



Unless I get some brainwave the next few days, this is how its send to manufacture.

[blackdog]

Hi JohnnyBerg,

If you let me see the schematic, i can give you a better review.

Think about the mounting hole next to: "REV. B"
If you use shielding, there is no place for the screw.

Kind regarts,
Blackdog

[JohnnyBerg]

Think about the mounting hole next to: "REV. B"
If you use shielding, there is no place for the screw.

Good point! Thanks. I was getting tired and wanted to finish. 

I can't move the mounting hole, as it has to match the battery clip holder. Hmm .. have to move the reference IC, and the slots .. and the rest. 

Tomorrow is a new day 

[blackdog]

Hi JohnnyBerg,

Just a lesson of a old man ;-)
Do not push yourself,  your creativity disappears with fatigue and you start experiencing tunnel vision.
The efficiency can become nagative!!!

Kind regarts,
Blackdog

[JohnnyBerg]

Just a lesson of a old man ;-)

Haha .. you're not the only old man 

Do not push yourself,  your creativity disappears with fatigue and you start experiencing tunnel vision.
The efficiency can become nagative!!!

Yeahh .. you're right.

On the other hand, I try to send out a order for some panelized, multi design pcb at the end of each month.
I don't wan't to design forever

[Andreas]

So, is it necessary to get some kind of shield over the reference IC? And where do I get something like that? Has to be custom made, I presume.

Hello,

something like this?
(a coin cell holder and a metal cage for a optic fiber sealing ring)

can be etched like a pcb or made by every PCB manufacturer (similar to SMD-Stencil)

With best regards

Andreas

[JohnnyBerg]

Pff .. another iteration. Unless someone comes up with something realy good, I'm done 



Backside, almost complete copper




I will try to finish the low cost version today, hopefully

[JohnnyBerg]

something like this?

that is exactly what I meant. 

[nuno]

Looks excellent  !

[blackdog]

Hi JohnnyBerg :-)


Is there a reason why u use a vertical model for the trimpot?

Kind regarts,
Blackdog

[JohnnyBerg]

Is there a reason why u use a vertical model for the trimpot?

It's a 3296W model, from Bourns. I'd rather had the 3296P model, but they are not as widely available as the 3296W model.

Back in the 80's my mentor on my first job after I graduated brainwashed me: "is there a second source available for the part you are using?" he kept asking ..pff

So, I had to make a choise between bending the legs or use a not so common potmeter. I choose bending the legs. The 3296W has support for horizontal mounting, by the way.

Edit: horzontal mounting is only needed when sending 1 piece out, by normal mail. Bigger orders go in a karton box, and then the horizontal mounting is not nessasary.

[blackdog]

Hi JohnnyBerg,

OK that's clear :-)

But....

Mmmm, LTZ1000A, LTC2057, ADA4077, ect. i wil remove them from my stock parts.
And my Agilent 3458A, i will throw him out of the window. 

Kind regarts,
Blackdog