[volt-nuts] Traveling Standards

[Bob Smither]

WB6BNQ wrote:
> Hi Bob,
> 
> I have been monitoring your thread since it got started.  Personally, you would
> be better off spending your money on a used Fluke 731B off of eBay.  You would be
> way ahead of the game compared to where you are at with your current concept.

Hi Bill,

Thanks for taking the time to provide your very thoughtful feedback on 
this project.

The last 731B I saw on EBay was ~$800.00.  Unless I had it independently 
calibrated (another expense) I would end up with another instrument in 
my lab that I could not trust.  Even if calibrated, after 30 days it 
would have 10 ppm uncertainly and after a year 30 ppm.

> First, the biggest issue is not having your circuit set to a "cardinal" voltage
> point; i.e., 10 volts.  By having some other level the measuring party would have
> to ratio against his standard to report a value.  That adds an unnecessary level
> of complexity and increases the error limits in reporting the uncertainty.

My objective with this project is to obtain a stability of 10 ppm, 
verified by having several independent readings made by other volt-nuts 
members.  If this objective can be realized, another advantage is the 
unit can be passed around to help other list members.

Establishing a cardinal value such as 10 Volts would require an 
adjustment.  I am hoping to avoid that uncertainty.  On a 10 volt scale 
the ~7 Volt output of the LM199A provides 1.4 ppm resolution on a 6 
digit VM - adequate for my objective of 10 ppm.

> Second, you need to provide separate power supply regulators for any heating
> system and the reference voltage circuitry.

Good point!  I did overlook this.  The heater of the LM199AH draws about 
18 mA after it stabilizes and is fairly constant.  It will certainly 
change less than 5 mA over the expected temperature range of a lab.  The 
load regulation of the LM78L15 is (worst case) 1.5 mV / mA resulting in 
an error at the LM199AH of 0.13 ppm.  Although small, this can be 
reduced further by powering the heater of the LM199AH directly from the 
24 Volt supply and I will do that.

> Third, your choice of a power supply regulator is very poor.  You should consider
> something more exotic like the Linear Technology LT-3080 or the LT-3083.  In
> particular, pay attention to the noise reduction considerations in the
> specification sheet.  Preferably a two stage affair using a LT-3083 (handles 3
> amps) as a pre regulator set to 18 volts and then the separate LT-3080 (handles 1
> amp) devices to their respective circuitry.

Although not shown directly on the posted schematic, the design does 
provide two stage regulation.  As detailed in my notes on the design 
page ( http://www.c-c-i.com/ts ), there is a separate external 24 volt 
supply.  The specs on that supply are !% line regulation and .02%/C TC. 
  At the LM199AH, after secondary regulation from the LM78L15, these 
errors translate to .06 ppm and .0011 ppm/C respectively.

The LT3080 is a great part.  I want to keep the package as small as 
possible, but I don't do surface mount.  The through hole version of the 
LT3080 is large compared to the LM78L15 (TO-92) that I am using.

The LM78L15 drifts at about 1 mV/C resulting in .018 ppm/C at the 
LM199AH.  If we assume that labs will be between 20 and 30 C this adds 
.2 ppm.

I have an idea to substantially reduce this error source and will try it 
when I get time.

> Fourth, you need to pay particular attention to ground loops and your ground
> plane layout.  Think along the lines of a star pattern, i.e., single point of
> return.

My instructions will be to measure only one output at a time.  Each 
output (from the LM199AH, the MAX6350E, and the LM34) has a separate 
return terminal and the return current is brought back to the return pin 
of each device, effectively removing the effect of other loads in the 
ground wiring.

> Fifth, to do it right you should provide four (4) separate 10 volt references
> with their own outputs in the package.  This would provide for independent
> analysis of the "bank mean" and allows for determining if any individual
> reference shifts outside of a nominal "bank mean."

> Sixth, in designing the 10 volt reference, the output loading should provide the
> highest possible isolation as to affecting the actual reference cell.  The output
> circuit should provide a minimum of 10 milliamps without a any change in the
> output level.

The intent with this unit is that it will be measured using a voltmeter 
or in a nulling arrangement.  I am assuming that the input impedance of 
the measuring instrument will be high.  A 10Mohm load (.7 uA) on the 
LM199AH introduces .1 ppm.

> Seventh, if any adjustable elements are used in the design, they should be of the
> smallest possible size as to resistance value.  This will help to minimize
> temperature affects.  A "cermet" potentiometer type should be used with values
> under 50 Ohms.  For example a 89pr10 (or 20 etc.).
> 
> Eighth, if it is going to be powered from the mains voltage, then a faraday
> shielded transformer should be used.  Proper EMI & RFI shielding is also needed.

The external 24 Volt supply I picked has 240 mV p-p of noise.  After the 
LM78L15 this is reduced to 3 mV p-p or .018 ppm at the LM199AH.

> Just food for thought and, by the way, having done all of the above, you just
> constructed a 731 if you just do one reference.

To recap - my objective is a shared standard that can be inexpensively 
shipped to other volt-nuts.  After several of us have measured the unit 
it will hopefully provide a reference that can be trusted to calibrate 
other instruments, if only to the 10 ppm level.

Many thanks for your thoughts!

Best regards,
-- 
Bob Smither, PhD                                 Circuit Concepts, Inc.
=======================================================================
    "We are over in Iraq trying to devise a new constitution for them.
     Why don't we just give them ours?  We're not using it any more."
     --Jay Leno
=======================================================================
Smither at C-C-I.Com           http://www.C-C-I.Com           281-331-2744
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smither.vcf
Type: text/x-vcard
Size: 140 bytes
Desc: not available
URL: <http://www.febo.com/pipermail/volt-nuts/attachments/20110830/ed54e463/attachment.vcf>