HP/Agilent/Keysight CAL?72 comparison

[branadic]

Did a CAL?72 measurement on our Keysight 3458A with option 002 and would like to compare them with numbers from others.

Internal temperature in both measurements was 38.2°C
A=997.134777e-3
B=997.134725e-3
D=10 days

C=((A-B)*1000000)/(A*D)= 0.0052149 ppm per day or 1.9034385 ppm per year

-branadic-

[TheSteve]

Mine drift at 0.04 and 0.116 ppm/day.

[Andreas]

Hmm,

one should also mention the age of the device and if it has option 02 (stable reference).

with best regards

Andreas

[Kleinstein]

If there is no adjustment during an external cal, the CAL72 changes mainly from drift of the resistors in the ADC (U180 ASIC) and is not effected by the reference. So even if one would change the ref. module and before an new cal to 10 V, the CAL72 value would not change much.

So the short term CAL72 drift is a measure on how much the resistors in U180 are drifting. Normally this should not be an important parameter.
It is only because of a possible fault of U180, that is usually first visible as strong CAL72 drift, that makes the CAL72 drift so prominent. So at best one would get an early warning for a likely coming defect.

For normal use the temperature dependence of the CAL72 value should be more important, as it reflects the gain drift before ACAL.

[mimmus78]

Mine unit has a new ADC board ... total drift after 1.5 years is less than 1ppm.

Inviato dal mio ONEPLUS A5010 utilizzando Tapatalk

[dr.diesel]

Mine, both self 002 modded, 1989 vintage:

#1=0.035
#2=0.025

Last 10 day, ppm/day, with a 0.6C temp delta.

EDIT, last 365 days:

#1=-0.0098/day (-3.58ppm/year)
#2=-0.0092/day (-3.36ppm/year)

[retroware]

I have a 1990 vintage unit with the 002 option.  The A3 board was quite drifty so I recently replaced with a new A3 unit from Keysight.  Below are the before and after CAL72 values:

Old A3 (1990 vintage)

6/11/18:        984.809889e-3  31.6 C
6/13/18:        984.809130e-3  31.8 C
6/13/18:        984.808801e-3  31.5 C
6/14/18         984.808158e-3  32.3 C
6/15/18         984.807312e-3  32.1 C
6/15/18         984.806519e-3  34.3 C
6/16/18         984.805934e-3  32.2 C
6/27/18         984.806252e-3  32.7 C
6/28/18         984.806415e-3  32.9 C
6/31/18         984.801381e-3  32.3 C
7/02/18         984.798866e-3  32.0 C
7/04/18         984.798195e-3  32.9 C
7/5/18          984.796966e-3  32.0 C
7/6/18          984.797096e-3  32.4 C
7/7/18          984.797311e-3  32.5 C
7/8/18          984.796825e-3  33.7 C

Brand new (fresh and tested) A3

7/9/2018        983.787396e-3   33.8 C
7/10/2018       983.787486e-3   33.8 C
7/20/2018       983.787471e-3   31.0 C

As you can see, there is quite a difference.

[TiN]

3458A-STD - "golden" meter, pimped A9, modified A1, LTZ1000CH chip, A3 is 2006.

May 26, 2017 CAL:

CAL? 2,1 = 7.18070682
CAL? 1,1 = 39998.898
CAL? 72 = 997.706708e-3
CAL? 73 = 997.507629e-3
CAL? 59 = 35.3

July 24, 2018 CAL (today  ):

CAL? 2,1 = 7.18071188 (+0.705 ppm / 424 days = +0.00166 ppm/day or +0.61 ppm annual)
CAL? 1,1 = 39998.8613 (-0.917 ppm / 424 days = -0.0022 ppm/day or -0.789 ppm annual)
CAL? 72 = 997.705991E-3 (-0.72 ppm / 424 days = -0.0017 ppm/day or -0.620 ppm annual)
CAL? 73 = 997.508502E-3
CAL? 59 = 33.3

3458A-2 - secondary meter, pimped A9, no other mods, new A3 from 2016.

January 5, 2017 CAL:

CAL? 2,1 = 7.07034233
CAL? 1,1 = 39999.2111
CAL? 72 = 982.327173e-3
CAL? 73 = 982.504696e-3
CAL? 59 = 36.5

July 2018 CAL:

CAL? 2,1 = 7.07031717 (-3.586 ppm / 550 days = -0.00652 ppm/day or -2.38 ppm annual)
CAL? 1,1 = 39999.2704 (+1.482 ppm / 550 days = +0.0027 ppm/day or +0.983 ppm annual)
CAL? 72 = 982.325230E-3 (-1.97 ppm / 550 days = -0.0036 ppm/day or -1.307 ppm annual)
CAL? 73 = 982.492417E-3
CAL? 59 = 35.4

Unless one is buying brand new 3458A from KS, then option 002 does not mean anything much. Especially if it's some unknown history box rotten turned off in random warehouse for years. 

[e61_phil]

Hi,

I attached some measured points from our 3458As between the years.

HP3458A#1 -0.07ppm/day
HP3458A#2 -0.02pppm/day
(calculated with the formular from the service note. I think one should use a linear regression with all data points).

HP3458A#2 has the 4ppm option, but the stability from both units over many years of calibration is the same. Both meters are quite old HP ones.


@branadic: Think about the number of digits, please . As you can see in the graphic attached, the drift isn't linear, it is not even monotonic. If I would choose a 10 day intervall (like you) and use the data from day 20 to day 30 for example. My values would be much smaller.