hp 3458a 40K reference issues

[jlmoon]

Hello,
I have a 3458a that I suspect is in trouble with the actual values for Constant 1,1 (40 K reference derived).  The allowable range according to the upper and lower constant values is 40.00000.. E+0  to somewhere near 41.000xxx E+0 .. does anyone know where or how the Constant 1, 1(actual value measured) is derived during calibration.  I am aware that this is considered a "External gain adjustment" per the manual and that there are a sequence of cal events that must occur before setting up for the 10K cal ref.  I am about to send to Agilent for a once over (Super-Std-Cal) and hope this does not get too deep.  Any suggestions would be greatly appreciated..

Thank you,

JM

[Dr. Frank]

Interesting point, I've never noticed that before.
But it's quite obvious, how to interprete this.
A good description of the instrument is in the April 1989 HP journal.

My values are those:

CAL? 1,0 = 40,0000000E03   (default value)
CAL? 1,1 = 40,0000000E03   (actual value = cal value)
CAL? 1,0 = 40,0800000E03   (max. actual value)
CAL? 1,0 = 39,9200000E03   (min. actual value)

CAL? 2,0 = 7,20000000E03   (default value)
CAL? 2,1 = 7,20509364E03   (actual value = cal value)
CAL? 2,0 = 7,50000000E03   (max. actual value)
CAL? 2,0 = 7,00000000E03   (min. actual value)


The x,1 values are the real values of the references, which are determined during 10V / 10 kOhm artefact calibration. That means, you unsecure the cal mode, apply an exactly known reference as 10,000050 V or .. kOhm from external standards (732B / ESI SR104), and type in the exactly known value. See calibration manual.

The 3458A then compares very precisely the internal reference against the external standard and calculates, which voltage or resistance values its internal references have, actually.

This calibration constant x,1 is stored in non volatile memory and used  afterwards for each measurement.

If the calibration memory is lost, the default values are used, i.e. 40,0 kOhm and 7,2V.

The internal reference components, that is a Vishay hermetical sealed metal foil resistor, and the Linear LTZ1000A, have a specified tolerance, i.e. min. and max. values.
Those limits were checked during the calibration process.
If the calculated actual value is outside the limits, the components may be faulty, and perhaps the calibration is not accepted, i.e. an error occurs.

The 40k resistor may vary obviously +/- 0.2%. In the CLIP, it's specified 0.1% / 1.3ppm/K, so the limit of 0.2% is wide enough for all possible temperature variations of the reference resistor, i.e. up to 80°C.


Similarly, the LTZ1000A is specified in the LT datasheet 7.0 - 7.2 - 7.5V @ 5mA, so that's exactly what HP has programmed into the 3458A.

If your actual 40k value 1,1 is corrupted, i.e. outside the limits, your cal memory may be defective. During normal operation, there may be no check on that.
Anyhow, then you should already have gotten CAL RAM errors.
If you send in such an instrument, agilent will for sure repair the CAL MEM, by completely updating the PCB (additional 1000 bucks or so).

It's possible to repair that on your own before sending to agilent, look for the numerous threads here and at other places.

PS: Would be nice that you publish your x,1 constants here, so that the situation is clear for the reader also.
And your final findings - would be nice if you post that too, as a feedback.

Frank

[jlmoon]

Hello Frank,

Thank you for the very informative replay.  The reason I have come to this fork in the road is because I figured I would go for broke and verify the overall sanity of my device before I box it up and send it in to Agilent.  To make a lengthy story short, my machine fails as I'm sure numerous others have in the past.. "Ram Test 1 Low" indicating cal constant corruption.  When the curious side haunts me, I have to know the correct answers to the problem... so...., I set off on an adventure by carefully removing the Constant NVRAM along with the working NVRAM's an replacing them with 'old name' Gold Augat' sockets and shiny new DS-1220 / 1230 NVRAMS.  Before I touched the parts, I made sure to duplicate contents - especially the Constant 'U132' part and safely store the .bin image in multiple places.  Then.. all is well .. but still when I power up I get at times conflicting "Ram Test 1 Low" and have seen "Ram Test 1 High" and moving on.. I figure no problem.. so lets go check the various functions for integrity.  That's when I realized that the Cal Constant 1,1(actual value -222.xxxxxx E-3 was way out of the shop [as Dave quotes it ].  This all started when I  first  checked my 7v ref and zero offsets by going through the standard adjustment procedure as stated in the service documentation.  Funny thing about this meter, is it passes self test (for what they're worth!) and doesn't appear to be out of cal that much, even though on start-up the device states "ALL CALIBRATION REQUIRED" . Now you have me thinking about the concept of pimping my box working around that 'possibly broken 40k ref' circuit.  Not that it would make much difference.. but those Vishay Z series parts sure do look nice.. .  I was going to send this box to Agilent today, but think I will study the device a little more and maybe look for the obvious 40K failure.., could be a bad relay contact as well .. those mechanical things are sure to fail after so many cycles and the age of this unit (89-90 model).  This weekend I make an attempt to gather all the constants and post them as accurately as I can. 
.. the lengthy story as short as it will ever get!!


Regards,

JLM

[Dr. Frank]

very brief reply:
I assume the 40k resistor is ok, but you have some problems with the cal ram and the 2x 32kB RAM. or the data bus, or simply a data corruption during programming, which the instrument does not fix. All cal required hints to that.

Perhaps the instrument takes the default value for measurement, so that the reading is logically ok.

If you saved the cal ram content, you might search in the binary for the 40k value inside (floating point value, afaik).
Maybe it's still reasonable, so you might recal the instrument with ordinary 10k and 10V, just to see what happens.

Afterwards, if it's a simple data corruption, you may reprogram the constants or send to agilent.

Frank

[jlmoon]

Frank,

Just a brief note, I sent my 3458a to the Agilent Loveland facility with the 40K ref problem.  They informed me that the problem was due to cal memory failure (20 + year old part)
I did record my cal x,1 constants before calibration but I suspect they are really of no value.  Agilent completely rebuilt the unit revising A2, CPU & Firmware, not to mention Display, Keyboard and Front Bezel... thus bringing it up to spec along with a standards lab calibration.  Currently I am checking my other systems and equipment against it and they look spot on for the most part.    Thanks for your earlier input on this problem
I can still post my constants should someone need them for reference.

JLM