Fluke 8505A vs HP 3456A

[kj7e]

I knew I would need to address the power supply caps, ran for two days before one shorted out.  Ill have new caps ordered in the morning.



I greatly suspect its been many years since this particular instrument has seen any operation. 

[bitseeker]

That was fast. But, better now than in the middle of a project.

[lowimpedance]

I'll bet they made one long groan at that first power up 

Some HP3456a service notes I uploaded 2012 that I found at work, thread here;
https://www.eevblog.com/forum/testgear/hp3456a-service-notes-for-owners-info!/msg144820/#msg144820
Once all the caps in the PS are replaced the meter will most likely be still chugging away long after we are gone  .

Enjoy

[kj7e]

Got to love the old manuals.  This showed up today,



Caps ordered.  I followed TiN's lead and up'ed the size a bit;
https://xdevs.com/fix/hp3456a/#psu_fix

[bitseeker]

That's a pretty good trick: They revised it a year after it was already printed.

[kj7e]

That's a pretty good trick: They revised it a year after it was already printed.

I wondered about that too, but before I even saw it.

New caps, fixed and improved;


(Found this guy NOS on eBay)


(Both meters where cold)


Edit, Digikey part nubers for Rubycon 105C caps with correct lead spacing, a bit extra capacitance and voltage rating;
Index   Quantity   Part Number   Manufacturer Part Number   Description   Customer Reference   Backorder   Unit Price   Extended Price
1   3   1189-3012-ND   50ZLH1000MEFC16X25   CAP ALUM 1000UF 20% 50V RADIAL      0   1.62000   $4.86
2   3   1189-1422-ND   50YXJ2200M16X35.5   CAP ALUM 2200UF 20% 50V RADIAL      0   1.93000   $5.79
3   2   1189-4026-ND   63PX220MEFC10X16   CAP ALUM RAD      0   0.73000   $1.46
4   2   1189-4028-ND   63YXJ100M10X12.5   CAP ALUM 100UF 20% 63V RADIAL      0   0.52000   $1.04
5   1   1189-3906-ND   35MXG5600MEFCSN22X35   CAP ALUM 5600UF 20% 35V SNAP      0   4.23000   $4.23
                     Subtotal   $17.38

Now to let it run for a few weeks.

[bitseeker]

Looking good, inside and out.

[cellularmitosis]

I am impressed - 8.4ppm or so on the native 10v ADC in 27 years - that is some 0.31 ppm per year in average. It is a pity they truncated the 7th digit (see the manual p. 8-12 Table 8.5) from the ADC (while they did include a 7th digit from GPIB on the less stable 3457A and 34401A.)

Do you know where this truncation happens?  I took a look at the manual but couldn’t find anything obvious.  There is a serial link between the inguard processor and the outgaurd processor.  I wonder if that could be spied on to get more resolution?  The theory of operation section describes something about how a comparator check is only performed every 5th clock cycle, so it seems there’s two more bits worth of resolution hiding in there somewhere...

[alm]

If there is enough noise in the signal, you might be able to extract the extra resolution by just averaging readings. That's how Datron extracted an extra digit of resolution on some of their 7.5 digit meters. That would be easier than trying to hack the internal communication. Obviously this will slow down your effective sample rate.

[cellularmitosis]

I figured out the seventh digit 

https://www.eevblog.com/forum/metrology/hp-3456a-7th-digit/

Also, my 3456A was also inoperative on arrival (-4000... on test).  As many others have experienced, my 33V rail was low (around 7V).  I checked all of the caps on the power supply board with an ESR meter and all looked ok except for C7, which had failed open circuit.  Just replacing that cap (47uF, 50V) was enough to get the meter operational again.

[AG7CK]

Do you know where this truncation happens?  I took a look at the manual but couldn’t find anything obvious.  There is a serial link between the inguard processor and the outgaurd processor.  I wonder if that could be spied on to get more resolution?  ....

Sorry - I never saw this post before today. It never occurred to me that the rounding was done outguard (because the manual talks about readings per second / transfer speed / reading size in bits between the inguard and outguard processors. Hence I never suspected they send unused information to the outguard processor).

Very good work. Thanks a lot.