Teardown : HP 3455A reference module 11177B-01723

[TiN]

RAWR, Voltnuts



EEVBlog member, codeboy2k unleashed a pandora box, caused ebay rage for precision nuts.
Is this going to be waste of $80.75USD, or it's a winner here?

I'm really curious to know what is that mysterious 5-pin reference zener there, marked HP 1902-0926. Yes, it's 5-pin can, as in schematics, not your standard LM399 (11177A have 399).
Almost "wanna-open-that-can" curious type.

Anybody knows?

[max666]

Mine is still at customs. Also I didn't get a B version, someone else snagged them up before I even got a chance (just kidding)
But I'm willing to carefully lift the skirt on mine, if that would allow the experts to identify the die?

[TiN]

A-version have LM399.
And afraid there is no word "careful" combine with opening hermetic welded metal can Words "saw" and "pliers" come to mind  Anyway, all that butchery left for later, if it's ever going to happen.
Want see those ppm's first.

[Zucca]

Anybody knows?

If you don´t know, surely Google also doesn´t.... Best option we have is to go to the HP group here, where 3792 Members are waiting to get exicted for a strange question about the HP 1902-0926.

In the meantime, a X-Ray shot would be nice. 

[daqq]

I don't know, but it sure looks like a rebranded LM399, which is pretty much the standard for this kind of stuff:

http://www.daqq.eu/?p=577

[max666]

A-version have LM399.

Ah, somehow I missed that, never mind then.

[BradC]

And afraid there is no word "careful" combine with opening hermetic welded metal can Words "saw" and "pliers" come to mind

I was thinking a custom chuck and a lathe rather than saw and pliers, but I suppose you can get agricultural if you like

[TiN]

I see, thread without photos is useless.

Daqq, I saw your great article about 3455, got job.
But my module is 11177B, which uses different reference chip. I am familiar with LM399, so it's definately not it. Also schematic does imply it's 5.1V reference zener, not 6.95V.

[TiN]

Here are some initial photos, as usual clickable.



One of benefit from old ages, is the fact that full schematics often was included in manuals. Wish today T&M vendors follow with that approach as well!

10V voltage reference circuitry:
Reference IC is custom HP 1902-0926 5-leg metal TO-99 can. It does run little warm by touch, but not nearly as hot as LM399.
Zener diode is marked on schematics 5.1V, actual measured voltage was 5.2485 VDC. Can have marking 6355 on side, can it be 55 week of 1963 ?
U2 opamp is remarked with HP part number - 1826-0059, made in HONGKONG, date code 35 week 1978.
Another opamp U3, remarked with HP P/N 1826-0009, date code 24 week 1978 is driving heater element inside VREF IC.



Precision resistance circuitry.
* R24 is epoxy encapsulated wire-wound 1.0020K 0.01% 1ppm/°C beast
* R21,R22,R23 are metal hermetic 0.01% 1ppm/°C resistors



Close-ups:



Sorry for bad photos, with wires sticking around, I was desperate to try if this puppy works, so hooked it all up before taking photos. It does take some time to do proper photoshoot, and it's already late after midnight here, so will do that later.

Test setup:
* Linear transformer PSU with 7915 powering -15V input on board. Current consumption according to my Fluke 87V was ~15.8 mADC...Now it's dropped a little to 15.727 mADC as it's few degrees warmer in the room now.
* Keithley 2400 is providing +29.300 VDC and reading +15.38mADC on this rail.
* 10V output measured by one of Keithley 2001 (roughly calibrated it myself, as my primary 2001 with Tek cal is busy at moment gathering data from ESI DB52 bridge).



Output was ~9.99975 VDC when it was powered on as is and let settle for 15 minutes. Little fiddled with R6 trimpot to set 10.000000VDC.
Now after ~1 hour 2001 showing Max: 10.000011 VDC, min voltage 9.999967 VDC.

I will compare this module to my LTZ units and my earlier LM399 10V breadboarie with K2002 to see how stable it is.
Also will compare 1Meg and 1K resistors to my custom 0.01% Vishay's.

[barnacle2k]

A-version have LM399.

Ah, somehow I missed that, never mind then.

That might not be the case, i think on that that both A and B versions exist in LM399 and zener form.
Late revisions use the zener.
Even though 40$ for a LM399 10V reference and a 1k and 1M ohm resistance standard? - not bad.

My units are still in transit, will probably get them late next week.   

On topic:
The units might become more stable by replacing the two OP-Amps with Zero drift ones like the LTC2057 (crazy cheap for it's specs).
It's become my favorite OP for DC circuits, so i bought a whole tube worth.

[Rupunzell]

Few T&M manufactures are going to provide proper docs today as companies  are interested in moving product and recycling them once their expected and designed in life time is reached.

Back when this hp VR was designed and built it was a different world of T&M where instruments are expected to last and be in service for decades, be completely serviceable and not tossed in the bin in a few years to be replaced with a "new revision with buggy software and hardware and crippleware." Gold plated pcbs, Hermetic sealed parts for long term stability and durability and overall good design. All this adds to production cost and subtracts profits from the bottom line.... and if this VR last decades problem free, how many might they sell?

To do a good job at analog design means pushing the limits of what is possible according to the laws of Physics and making a deal with Nature. Different from digits where greater processing power and speed and memory often means increased functionality and generally a better device.


Bernice

One of benefit from old ages, is the fact that full schematics often was included in manuals. Wish today T&M vendors follow with that approach as well!

[codeboy2k]

Nice TiN... good to see that you got it !  I'm looking forward to test results!

I really wanted one for those 0.01 % 1ppm resistors too

Cheers!

[Tallie]

Host server is extremely slow and images are extremely large. I wish you'd host them elsewhere. Thanks for them, though.

[TiN]

I will consider using smaller ones for forum posts, thanks.
Sorry for speed, there times when Taiwan ISPs upload ain't great for US visitors, but currently I have no plans to host data externally.

[codeboy2k]

I usually put my forum images up on imgur.com, because that's what they do best - image serving.

[barnacle2k]

OK TiN you won 

My A unit sports an National branded LM399, mfg year is 86.
The thing is older then me  .


Will do some measurements when the ones for the DMM noise thread are done.

I intend to build a breakout box for my modules, still looking for the card-edge connector HP used.
The cinch 50-30A-10 could be a match, ~3,96mm spacing seems about right.

[barnacle2k]

Ok, i did some initial tests with the unit.

And the accuracy of those resistors is just amazing!
They were all bang on, even after all those years.
My measurement setup was far from perfect, it was just a quick check. (4 Wire only to the edge-connector of the unit)
For example: the 100k measured 100,000.12 Ohms!
The 1k was 1000.XXXX Ohm (other digits meaningless due to 2W)

What was a bit weird was the 10V reference, it was more like 10.037014V not 10.000000V. 
And the adjustment range of the pot seems not far enough to get that to 10.000000V.
Might have to cut some jumper links.

[codeboy2k]

I really wanted those resistors too    My home lab is missing some nice low-tempco precision resistors like that.
Nice to see they are spot on for you.

Your 10V seems far off.  Did you let it warm up for 24 hours?  I would have done that first, since they've been powered off for a long time.

Since you might  have already tried to adjust it, I suggest you power it up and wait 24 hours before trying to make any adjustments or making jumpers. Your unit might have suffered some excessive component drift over the years and might need some time to settle.  Perhaps I'm just being overly cautious but that's what I would do with it before trying to adjust it.  An hour is probably good enough but I'd wait 24 hours

[TiN]

Did you supply both correct +29.3 and -15V to module?
Seems 10V is too off, something should be wrong.

LM399 getting within it's stable state within few minutes, warmup should not be an issue.

[babysitter]

oO(There might be a reason why it ended up on eBay, sometimes...)

[barnacle2k]

oO(There might be a reason why it ended up on eBay, sometimes...)

Both +29.3V and -15V hooked up, again way the module is connected at the moment is not ideal so there is propably Thermal-EMF involved.
As i wrote above i'll get compatible connectors for the card edge connector.
The -15V is not really used on the LM399 variant, its only connected between ground and a resistor.

Attached a Plot of the last 40 Hrs, the grid on the graph is 10µV aka 1ppm of 10V.
I'll start measuring the LM399 voltage directly to check if at least that is fine.

[Vgkid]

that 10v is .37% from ideal, something is certainly not right.

[barnacle2k]

Seems like Mr. Murphy wants to claim participation in this experiment  .

As i was checking the voltages around the LM399 i found that heater-ground was above the ground potential.
Further checking revealed that in the mess of clamps and wires the heater ground connection on the edge connector had slipped off the pad.
There is a diode from heater ground to reference ground on the board (and in the LM399) which prevented damage and enabled operation.

With that connection established i get 10.000000V on the output. (adjusted since i had already touched that pot)

Edit:
I am still unsure if i should tear the modules down and move the good parts on a new
board which better suits my requirements or if i should build a "lab standards box" which these modules plug into.
I kinda want to separate the precision resistors from those performance degrading trimpots.

Things i want out of this:
10V and 20V Voltage standards (With option for the 10:1 and 100:1 divider for 0.1V, 0.2V, 1V, 2V)
4Wire 1K 2k 100k 1M 2M Resistance Standards
10:1 and 100:1 Reference dividers (there is a matched 10:1 resistor set in each module)

[Vgkid]

That is good to hear. Should have bought one.

[barnacle2k]

The saga continues:

I decided to keep the modules as-is not to stress the precious components.
I'll order some connectors and build a box for them.

In my testing i found a problem:
The OP that does the gain to 10V is more sensitive to temperature then i would like.
When sprayed with freeze spray the output went down 1mV and the offset voltage of the OP dropped by 18mV.
I will replace it with a LTC2057, which will involve some circuit modifications.
The current circuit uses offset trimming to purposefully increase the offset voltage of the OP.
I think it, its to alter the bias current through the LM399, though i am not certain. 

On a positive note: when i sprayed the resistor network the 10V output did not move a single digit, indicating excellent ratio tracking.
(meter in 5 digit mode for speed reasons)