Richard:

I have one 731B that I have had for 15 or more years.  I have only been

plotting drift for the last 10 years.  This was done against a bank of four
732A references.  The maximum drift that I have observed is about -1.5 ppm
over that 10 year period.  This is as best as I can determine without having
a Josephson Junction Array.  During the first 5 years of that period it
seemed to drift around +0.4 ppm total, but then in December 2010 I moved to
a new home and at that time it seems to have a -1.5ppm total drift over the
next 5 years.  Not too sure why.  So as far as yearly drift I guess you
could say that it averages around -0.15 ppm per year over that 10 year
period.  As far as temperature dependence, all of this is over a range of
20C to 24C depending upon the temperature of the room.  I cannot separate
the temp variations from the DC noise but just a casual look at the chart
for the past 6 years it seems to indicate that this 731B output goes down
with a higher room temp and vice versa but less than 0.4 ppm total over that
range of temps.

Does this one beat specs, you bet!  Others may not be this good however.

Bill

----- Original Message -----
From: "Richard Moore" richiem5683@gmail.com
To: volt-nuts@febo.com
Sent: Saturday, March 21, 2015 1:02 PM
Subject: [volt-nuts] Fluke 731B

fully functional 731B regarding variation in ambient temp a few degrees C
around 23, and for long-term drift?

than the specs would lead you to believe, but I have no practical experience
with these units.

https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts

I use my 731B for both 10v standard, and to dial in microvolts. Since I'm
always switching the unit, does anyone know if it would be a good idea to
treat the contacts with Deoxit Gold?

On Sat, Mar 21, 2015 at 7:54 PM, Bill Gold wpgold3637@att.net wrote:

Richard,

Some really good information has already been posted.  Here is what I have
found.

I posted a comparison of the 731B against the Geller Labs SVR-T here:

http://www.eevblog.com/forum/testgear/geller-labs-svr-and-svr-t/msg282582/#msg282582

There were within 1ppm of each other over 48 hours and 3.8 deg K.

I'm currently running a comparison against the reference from an HP 3458A,
using an Agilent 34461A in ratio mode.  A screen capture follows if the
list lets it through.  There is only 5 hours here as I just updated the
34461A firmware (it adds another digit to the scaling limits on the trend
chart and is definitely worthwhile).

One thing about the 731B is that it if the NiCd pack is dead or fully
charged(!), there will be much more noise on the output.  I designed a
'high-side' switch to switch between AC and battery power and it improves
noise when the NiCd is fully charged.  (The original design is a diode
switch and if you have a NiCd pack which has a high voltage when fully
charged, the half-wave rectified AC that is used to charge the NiCd pack
leaks through to the voltage reference power supply.)

A big WARNING.  If a 731B still has its NiCd pack installed, replace it.  I
had Batteries Plus make me a replacement pack.  About $30 as I recall.  If
the NiCd pack isn't installed, you need to remove the diode that goes from
the battery pack to the voltage reference supply, or you'll get unfiltered
half-wave rectified AC on your voltage reference power supply and it does
get into the output!

My 731B seems to be doing quite well IMO, being within 1ppm of the new at
the time Geller SVR-T.  It has also been reading 9.99996 to 9.99997 on the
34461A for the last year (as long as I don't turn on the 8568A that is next
to the 34461A... the 8568A blows warm air into the 34461A and then all bets
are off).

Orin.

[image: Inline image 1]

On Sat, Mar 21, 2015 at 1:02 PM, Richard Moore richiem5683@gmail.com
wrote:

A couple of years ago, I sat my 735C, a (at that time) recently Agilent calibrated 3458A, and two 731B's on a bench and have left the references on, with battery back-up for any power failure.  I have turned the 3458A off and on a number of times in the interim.

When I set all this up, I adjusted the 735C to 10.000000 VDC as best I could using the then recently calibrated 3458A and adjusted the 731B's to match the 735C using the 3458A as a 'null' meter.

Tonight, after the 3458A had been on for several weeks, I did the 'AutoCal All' on the 3458A and read each of the references.

The 735C measures about 10.000016 VDC, and the 731B's measure about 10.000005 VDC each, within about 2 uV of each other.

In the next couple of months, I plan to send the 3458A back to Agilent/Keysight for recalibration and see what the meter measures 'as received' and 'as left'.

We'll see how close it is to being 'in tolerance' 'as received'.

Hope this helps.

Joe

-----Original Message-----
From: volt-nuts [mailto:volt-nuts-bounces@febo.com] On Behalf Of Orin Eman
Sent: Monday, March 23, 2015 12:36 AM
To: Discussion of precise voltage measurement
Cc: richiem5683@gmail.com
Subject: Re: [volt-nuts] Fluke 731B

Richard,

Some really good information has already been posted.  Here is what I have found.

I posted a comparison of the 731B against the Geller Labs SVR-T here:

http://www.eevblog.com/forum/testgear/geller-labs-svr-and-svr-t/msg282582/#msg282582

There were within 1ppm of each other over 48 hours and 3.8 deg K.

I'm currently running a comparison against the reference from an HP 3458A, using an Agilent 34461A in ratio mode.  A screen capture follows if the list lets it through.  There is only 5 hours here as I just updated the 34461A firmware (it adds another digit to the scaling limits on the trend chart and is definitely worthwhile).

One thing about the 731B is that it if the NiCd pack is dead or fully charged(!), there will be much more noise on the output.  I designed a 'high-side' switch to switch between AC and battery power and it improves noise when the NiCd is fully charged.  (The original design is a diode switch and if you have a NiCd pack which has a high voltage when fully charged, the half-wave rectified AC that is used to charge the NiCd pack leaks through to the voltage reference power supply.)

A big WARNING.  If a 731B still has its NiCd pack installed, replace it.  I had Batteries Plus make me a replacement pack.  About $30 as I recall.  If the NiCd pack isn't installed, you need to remove the diode that goes from the battery pack to the voltage reference supply, or you'll get unfiltered half-wave rectified AC on your voltage reference power supply and it does get into the output!

My 731B seems to be doing quite well IMO, being within 1ppm of the new at the time Geller SVR-T.  It has also been reading 9.99996 to 9.99997 on the 34461A for the last year (as long as I don't turn on the 8568A that is next to the 34461A... the 8568A blows warm air into the 34461A and then all bets are off).

Orin.

[image: Inline image 1]

On Sat, Mar 21, 2015 at 1:02 PM, Richard Moore richiem5683@gmail.com
wrote:

I would never adjust any pots on a working fluke voltage standard. When the
pot moves you introduce more drift.
Just keep track of differences.
By the same token I do not turn off my 3458As as long as the line power is
up.
There was a time I had a rack of 15 3458As running waiting to be sold.
We have a power outage about 1 time per year long enough to trip our julian
date and time clock. We use it to track power up time.

I do not see anyone posting 3458A in stat mode data: mean, SDEV, High, and
Low, Range, and Number of samples over extended time.
then reverse polarity and do it again.
Also DC cal values before and after Auto cal and Keysight cal would be nice
to know.
Also using Before and after cal data from Keysight to correct values.
Do not adjust the Flukes  but keep a spreadsheet of best guess of its value.
Something like  mean of present guess + before cal data correction + after
cal data correction.
Example:
Say you present guess (guess1) is 9.9999468 cal the meter using 9.9999468
and then read the cal var 1 (I think that is the one for the reference
board)
Your as found data should indicate how far off your reference is.  correct
and call that guess2
After you get the meter back read cal var 1 and then use successive
approximation to when caling to get the same value in var 1. That should be
guess3
So new guess is mean (guess1, guess2, guess3).
If you have better confidence in one guess or less confidence in an other
the values could be weighted.
You never really know what the actual value is  a given time but do know it
is changing.  You can guess at it's value like 9.9999468 is just a guess
even if the real value is 9.99994681 or even 9.9999468 with what we have
you just cannot know for sure. There can alway be another digit.

Anyone want to poke some holes in what I wrote?
Does anyone have a better way to come up with a Cal value for a fluke 10
volt reference without sending it out for calibration?
Where could send it to get a better calibration and what would it cost?

On Thu, Mar 26, 2015 at 8:03 PM, J. L. Trantham jltran@att.net wrote:

--

John Phillips

John,

Thanks for the thoughts.  I basically agree with the idea of not 'adjusting' the standards but when I got them, they all had 'problems', including leaking NiCd batteries that required clean up.  Not knowing how long they had been off and having only calibrated DMM's (3458A and 7081) I chose to get the standards up and running then 'adjust' them to begin a 'new era' of standards, at least in my set-up.

I also have two 732A's that have been brought back to life and they are sitting on now for about 12 to 18 months.  I have not tried to read them in a while.

I have been collecting other gear including Fluke 720A, 721A, 750A, and 752A and a Fluke 845AB.  I also have a Fluke 887AB that I need to get working.  My problem is that work is about 80 to 100 hours per week and I have little time to spend on this.  But, once retired.........

I have left my 3458A's on for extended periods of time (weeks) and have been very pleased that when I turn them back on after weeks off, they come back within a couple of uV to the last readings of the 735C suggesting everything is fairly stable or at least drifting at the same rate and in the same direction.

I'm a bit nervous about leaving the 3458A's on all the time due to the potential for the displays to fade or fail.  So far, that's not been a problem.

As far as calibration of the Fluke references, I would think Fluke would be the best bet but there is the problem of getting them there and back while still powered.  I've been working on a method to do that using SLA batteries or using the Fluke 732A-7003 battery charger/pack.  I think Fluke made a case that would house the 732A and the 732A-7003, both filled with batteries and completely charged, to ship the units in for calibration.  My plan is to construct a 'case' to house two 12V 18AH SLA's and a charger that can be plugged in once the unit arrives at the calibration facility.

I have acquired a couple of dead, industrial strength, UPS's that, once I get some time, I would like to rebuild and power my 'metrology section' with to deal with any power interruptions.  I have a stand-by generator system as well.  However, it all takes time to get up and running and I don't have much of that right now.

Joe

-----Original Message-----
From: volt-nuts [mailto:volt-nuts-bounces@febo.com] On Behalf Of John Phillips
Sent: Friday, March 27, 2015 1:41 AM
To: Discussion of precise voltage measurement
Subject: Re: [volt-nuts] Fluke 731B

I would never adjust any pots on a working fluke voltage standard. When the pot moves you introduce more drift.
Just keep track of differences.
By the same token I do not turn off my 3458As as long as the line power is up.
There was a time I had a rack of 15 3458As running waiting to be sold.
We have a power outage about 1 time per year long enough to trip our julian date and time clock. We use it to track power up time.

I do not see anyone posting 3458A in stat mode data: mean, SDEV, High, and Low, Range, and Number of samples over extended time.
then reverse polarity and do it again.
Also DC cal values before and after Auto cal and Keysight cal would be nice to know.
Also using Before and after cal data from Keysight to correct values.
Do not adjust the Flukes  but keep a spreadsheet of best guess of its value.
Something like  mean of present guess + before cal data correction + after cal data correction.
Example:
Say you present guess (guess1) is 9.9999468 cal the meter using 9.9999468 and then read the cal var 1 (I think that is the one for the reference
board)
Your as found data should indicate how far off your reference is.  correct and call that guess2 After you get the meter back read cal var 1 and then use successive approximation to when caling to get the same value in var 1. That should be
guess3
So new guess is mean (guess1, guess2, guess3).
If you have better confidence in one guess or less confidence in an other the values could be weighted.
You never really know what the actual value is  a given time but do know it is changing.  You can guess at it's value like 9.9999468 is just a guess even if the real value is 9.99994681 or even 9.9999468 with what we have you just cannot know for sure. There can alway be another digit.

Anyone want to poke some holes in what I wrote?
Does anyone have a better way to come up with a Cal value for a fluke 10 volt reference without sending it out for calibration?
Where could send it to get a better calibration and what would it cost?

On Thu, Mar 26, 2015 at 8:03 PM, J. L. Trantham jltran@att.net wrote:

--

John Phillips

volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.