Keithley 2002 8.5 digit DMM review and teardown

[TiN]

Reversed schematics of ADC module of 2002 ... Pretty big difference compared to K2001.

PNG-pic, 500KB
PDF

Any idea how it works? 

Probably an integrating ADC ?

You mentionned you used "ECAD" for schematic capture, can you give a link to the website of this tool, please ?

Yes, it's integrating ADC, but not as usual/widely used schematic.

As for schematics tool - it's commercial Cadence Orcad. ECAD stands here for electronics-CAD

[branadic]

would you mind measuring the BE voltage of the temperature sensing transistor Q2 of the LTZ1000A, to determine its stabilization temperature? That's pin 6 -7.

Tryed this today but without success. Couldn't work out where to find pin 6 on the topside of the pcb and I've to admit that I'm not prepared to take the device completely apart.
Would have helped if there were of photo of the pcb with backlight, to turn out the nessecary measurement point.

[janaf]

Great post!

And my first post here 

And thanks for making an posting the schematics. Great work.

I'm really interested in how they implemented the ADC. After a look at the schematics, I can not really say I get any grasp except some blocks. but then I'm kind of new to ADC topologies.

Input anyone?

PS Pin 6 of the LTZ reference should most likely be connected to a precision resistor pair, with a ratio of about 1:13, i.e. 1K / 13K or similar.

[TiN]

Hi all,

I got into 2002 again, recovered schematics of voltage reference section..

would you mind measuring the BE voltage of the temperature sensing transistor Q2 of the LTZ1000A, to determine its stabilization temperature? That's pin 6 -7.

It's 0.498322 VDC.

Schematics:



----

Some extra photoshoot:

If someone forgot how 2002 looks like...

.

Lovely custom VFD, same as 2001,24xx,26xx,700x series equipment from KEI.

.

A09 firmware and MEM2 hacked memory option

.

Calibration dates and 10V from EDC MV106.

.

1K0000 VPG VHP202 0.01% resistor.

.

Self-diagnostics test.

.

.

.

Back side same as 2001, and top side internals without cover.

.

ADC board removed, digital board is exactly hardware same as 2001.

.

.

Voltage reference output measurement from LTZ1000 and setup.
VREF Voltage is ~7.13491 VDC. To do more precise measurement it would require more complicated probing with all covers and shields on to ensure proper temperature stability.

.

R264 and R265 voltage measurement.

.

Emitter of Q204 (heater node of LTZ1000A) and MECCA to system common voltage delta.

.

[TiN]

Alright, I'll check 13k/1k and 1.1k ratios with my LTZ ref's, to check if that's a typo in Keithley's manual.

I don't want to desolder resistor from instrument, yet 

[Dr. Frank]

Hello TiN,
It's difficult to measure the resistance, when both R's are built in the circuitry.
Therefore the ratio can be better measured by the voltages across both R268, R269, as I proposed a bit ago.

I assume you measured 0.4392V on R265 or (+) OpAmp, (as shown in the photograph), which is virtually identical to the Ube voltage or (-) Opamp.
Therefore, the ratio is 7.1349V/0.4392V = 16.3.

That would mean the LTZ1000A is also running on >90°C.
That would also fit to the bad longterm stability, like the 3458A.


I don't understand right now the difference between the direct measurement on Ube (0.4983V), and the indirect one on R265 (0.4392V). Are you sure about the first value?

Frank

[TiN]

It may be difference in temperature, as everything was running open, just like on photos.
Both measurements were taken within 5 minutes from each other, so it may be due to difference of LTZ die temperature.

[Dr. Frank]

Well, measurements directly on Ube are very critical, as it may disturb the regulation.

That's a very sensitive node, and maybe you caused sort of oscillation by your DMM, or injected strong AC noise by the measurements leads, originating  from switching power supplies in your lab.

Frank

[Dr. Frank]

It may be difference in temperature, as everything was running open, just like on photos.
Both measurements were taken within 5 minutes from each other, so it may be due to difference of LTZ die temperature.

Hello TiN,

no that's definitely not the case!

First of all, there's a temperature stabilisation on the LTZ1000, which makes it nearly independant from outer temperature variations. The stabilization temperature may change by about 0.1..0.3 °C per °C external change only.

The observed difference of 0.498V directly on Ube and OP(-) vs. 0.439V on OP(+) is 59mV, which would be a difference of almost 30°C of the chip temperature!!
That can't be the explanation at all, as that would require an external RT change of at least 100°C.


Looking closer on your photographs, titled "R264 and R265 measurements", I discovered non appropriate cables, that you are using.

You use Kelvin clamps with 4 cables, but on a DCV measurement!
That means, that the two cables connected to Ohm Sense are hanging totally free in thin air, that means they act as antennas for all kind of AC disturbances, especially from switch mode power supplies.

If you measured the 0.498V on the sensitive node Ube the same way, that's no wonder, why it's so different to the 0.439V on the same OP. The AC noise, coupled in over these cables, is rectified by the BE diode and added to the DC bias. Maybe also, that those long cables let the OpAmp oscillate, as soon as you connect to the OP(-).

To avoid disturbances, switch off all switch mode PSUs on your bench, and use a pair of short, drilled twisted and/or shielded cables, being convenient for sensitive DCV measurement.

Measure at OP(-) (Ube) and OP(+), directly after each other, and both values have to agree within << 1mV.
( A residual injection of AC noise is expected).

It would be very kind, if you could repeat this measurement, please.

Frank

[TiN]

Alright, you got it.

I'll open meter up and make a video for tube 

[TiN]

Here are correct results, measured by cal'd K2001 with pair of Fluke 87V probes, twisted together.
The only switching power supplies in room is my workstation and monitors, which are around meter away + LED lights.
I repeated measurements without LED lights on - result was same.



Meas points. I probed all voltages on schematics just directly with probe touching leg of LT1013 SO8.



Video (values little differ because I shot it two times, so there was some time delay).

[Dr. Frank]

Hello TiN,
thanks a lot for re-measurement!

Now it's clearly proven, that the divider has a 14.3 ratio, which may be realized by a 13.3k / 1k divider.
This sets the stabilization temperature to about 70°C.
That should normally be good for about 1..2 ppm/a stability.


Therefore, the one year stability of the Keithley 2002 in the 20V range not only depends on the LTZ1000 reference, but additionally on range determining / gain resistors.
You may further re-engineer this instrument, to understand, how it operates exactly in the 2V / 20V ranges.

In contrast to that, the 3458A depends ONLY on its LTZ ref concerning its DCV longterm stability, as it features true autocal, which compensates for any gain drifts.
As no other DMM has that autocal function, all other DMM need ultra stable resistors for ranging/gain to achieve their high stability in the different DCV ranges... and also for the other modes.

Frank

[wiss]

Reversed schematics of ADC module of 2002 ... Pretty big difference compared to K2001.

PNG-pic, 500KB
PDF

Any idea how it works? 

Where is the integrator? What is the voltage-rating of C832?

Or do we have multiple integrators? U815 - U817?

What is up with U813? non-inverting input goes where?

[TiN]

Where is the integrator? What is the voltage-rating of C832?

Or do we have multiple integrators? U815 - U817?

What is up with U813? non-inverting input goes where?

C832 and C833 are just thruhole ceramic caps for decoupling.
According to KI 2002 service/repair manual they are CAP, 2200P, 10%, 100V CERAMIC

U813 is fixed now, thanks for reporting.


PNG-pic, 1000KB

PDF

[TiN]

Revive, revive!

Updated top post for faster photo loading, moved them to static serve, rather than repository engine post.

On of our members shared older firmware A06, and tadaaaa, MEM2 option contents.
It's DS1248Y NVRAM. I got it before and tried it, but was no luck... Hmmm.... Time to open 2002 up again and try again...

ODD ROM for A06
EVEN ROM for A06
MEM2 NVRAM dump

I saw one of 2002's on bay have A10 firmware, so mine is not latest..

His meter photos (it's made in 1997, 10 years older than one I have):







Digital board is exactly same as mine Rev K in both 2002 and 2001 with B-firmware
A/D board is same revision B too
Analog board is little different, his is Rev E, while mine is E-E5.

Differences so far spotted by eye:

* On digital board, mine does not have CR627,CR628,VR602,VR603 diodes on GPIB port (cost-down to save few pennies? )
* R346 resistor assembly is different, using 4 ceramic substrate resistor networks, while mine just have two resistors glued together. Newer 2001 have similar network, while older ones (all mine) are just blue resistor network there.
* Optoisolation is same as 2001, while mine had wonky PCB with IL717 isolators.

[ManateeMafia]

I installed a new DS1248Y and didn't get any error messages. The secret init command worked and I guess it is good for another 20 years. I think you may have a defective rtc in your nvram.

[TiN]

Thanks for info. Probably my 1248's are fake or RTC faulty. I have two, both doing same stuff, bought from ebay from two different sellers.
They work fine as 1245, keeping data, so internal battery likely be okay...

[TiN]

Another DS1248Y come, seems I got real lucky with fakes... different seller, same part, same result?



Here are "DS1248" chips I got and real DS1243 which was in 2002 as STD option.



You can spot DS1315 Phantom Time Chip in 1243, while nothing on "1248".



Tried to debunk it, but it's not so disassemble without nasty acids, so I just broke thing apart with pliers. There were two CR 3V batteries, SRAM die and PCB from looks of it.

Further destruction reveals DS1215 in it, so there is a phantom controller actually there.



To be exactly correct, chip does store it contents and works as DS1245 well, so it's not completely useless. I don't have programmer which can support DS1248, and not sure if it's worth the effort to borge some CPLD gear to try read registers in these DS1248's to see what's wrong

Data commands and diagrams both both are same at first glance, so not sure where is a caveat just yet. 

[ManateeMafia]

The DS1215 and DS1315 datasheets show a simple interface to SRAM. I wonder if you can reverse engineer the board to see if there is a design flaw? What are the odds one or more parts are also counterfeit? It looks like they normally have three lines printed below the DALLAS name, where your new parts have four.

I have attached a pic of the old and new DS1248Y for comparison.

[slurry]

I have attached a pic of the old and new DS1248Y for comparison.

Why would a serious manufacturer leave out the trashcan-symbol (batteries inside) on a later productions?,
it not like the rigorous environmental regulations for Electronics have been eased up and suddenly makes it OK to toss batteries with household garbage? - no way 

Based on only that i assume it is fake.

[ManateeMafia]

The top nvram is the new replacement. That is fully functional sourced from Mouser.
The bottom one is dated 1996 and actually still functions.

[slurry]

ah, 1996..
the toss-it-anywhere-you-like-era before trashcan-symbols 

[TiN]

Bingo, TL866 got new software 6.13 which finally supports DS1245 NVRAM.



Now I finally was able to erase and programm my DS1248Y to get MEM2 and RTC going on 2002.




Steps, NVRAM dump and details for those who interested here, as usual.

[VintageNut]

That is a truly spectacular result. Congratulations for making the memory upgrade work. I want one very much.

I used a 2002 to (attempt to) linearize my Fluke 332A using the 20V range of the 2002. When I moved to the 200V range of the 2002, it appeared that the 2002 was more than 25ppm off measuring 100V.

There appeared to be almost perfect agreement for the 2V range compared to the 20V range but not good agreement for the 200V range and 1000V range.

Do you have a way to compare the 20V range to the 200V range? The best way is a 100V source and a Kelvin Varley Divider.

[TiN]

I don't have anything stable enough for high voltages just yet, if not counting 2400 SMU for that.