-
A very cheap 3458A
Posted by
vtp
on 10 Jul, 2016 17:58
-
The meter is one of the early units from around 1989 and upon arrival of course it did not work. It gave the typical RAM TEST 1 HIGH error message meaning that at least one of the NVRAMs had dead battery. Further, it also complained about 202 - HW FAILURE: SLAVE TEST (DC BOARD) meaning that the DC board had something in it that it did not like.
NVRAMs (from 1988) were easy enough to take care of, I used some old but still good ones from an another unit that had it's NVRAMs replaced with new ones, for now. CALRAM was ok, it had datecode from 2009 so obviously it has been replaced before. FW revision was 3, I put in version 8 in new EPROMs. The meter was also completely stripped down and cleaned of all the dust that had accumulated over the last 27 years.
Then to the DC-board error, root cause was a U1 (TI CD4094B) that was outputting only ones to the next in chain. Great, that ensured that all ohms ranges were selected at the same time and no surprise, ohms current source was faulty with CA3082 that had pin 15 (common collector) open and Q307 blown up.
Took some work to find out those faults but easy enough to fix, replacing those three made the self test pass.
At the same time I also removed all electrolytics as they had datecodes from 1988. That turned out to be unnecessary as all measured ok with only minor difference to new ones. Seems that the old electrolytics were made very well and they had also epoxy sealed bottoms so no leakage. Nevertheless, I put in new ones as the olds were off the board. The two Y-capacitors on outguard power supply board got replacements too as these had their casings cracking already. Line filter and fan got replaced too.
On the AC-board fuses F701 and F702 were 1/16A types. CLIP states these to be 1/8A types so I replaced them. I think this was a factory screw up as they definitely were not replaced before and even the spare fuses F001-F003 were also 1/16A types. F702 was also blown and it had a normal 20mm fuse bodged on it.
Has anyone encountered the same issue with those fuses?
The faulty DC board had obviously been replaced before as it was from 2007. The peculiarity in this instrument is that CALRAM has about 7.18V for the zener but I remember measuring it to be about 7.15V, albeit with a Fluke 87 while debugging the board.
The difference also shows in measurements, they are off about the same amount relatively. The meter is now hooked to a Fluke 732A and seems stable so far except for the gain error.
Now, what could have possibly caused the reference to change operating point?
In this case, I think nothing. The DC board has probably been changed from a newer unit where it failed. Then whoever changed the working DC board from this unit to whatever they had sent the fixed unit for cal and this unit made it's way to surplus store.
Well, thanks for the newer DC-board with VHP101 40k reference resistor. It looks like this meter needs only proper adjustment and calibration now.
Why is this message in metrology section? Plan is to see if this was really fixed and stays that way. That "SELF TEST PASSED" does not mean much - stability is of more concern.
-
#1 Reply
Posted by
TiN
on 11 Jul, 2016 12:16
-
vtpWelcome. I liked your though train, well done. I assume you read
my worklog 
perhaps.
Nevertheless, I put in new ones as the olds were off the board.
Right there, earned a star. Not worth time to discuss reasoning for replacing 20 year capacitors in gear.
I'd watch out for mains EMI filter assembly. Some of them are Schaffner manufactured with issue, which make them
time-bombs and fire generators!.
For zener voltage difference, it's likely that reference module (little A9 PCBA in far end of DC board) was swapped, and unit was not recalibrated since than.
If your 732A is in cal and you know it's output value, you can switch meter to DCV 10V range and run calibration at 10V output. Since your unit's calibration is invalid, you don't risk anything much at this point.
To get ball-park of stability - you should run unit at least a week 24/7 to get everything stabilized, and then run
SN18 test procedure over another week (daily checks). This will give you approximate stability figures for DCV, which should be way <1ppm. Good meter/ADC should show <0.4ppm span over a week.
Here's comparison of 4 units (one with good and bad A3 ADC PCBA) you can reference to from our member here.
P.S. If it's not a secret, how cheap was very cheap?
-
#2 Reply
Posted by
Dr. Frank
on 11 Jul, 2016 12:17
-
As the analogue board A1 and the nv-CALRAM have been changed at different times, the date codes not being necessarily the exchange dates, I assume that the repair in total was not successful already at these dates.
So no new calibration had been done, and the 10V calibration value, you read from the nv-RAM, may refer to another reference board, which may have been exchanged on a different instance.
Anyhow, it's not possible, that the LTZ reference may change its value by that amount, only in case of a defect, and to lesser amount only, if someone "pimped" the oven temperature to lower values. But latter should be easily visible on the LTZ board.
You might also compare the 10k calibration to an external reference; if this value also disagrees strongly, then definitely, several boards may have been mixed.
If you have a 732A, it's very easy to re-calibrate your 3458A to the uncertainty level of this 732A
Frank
-
#3 Reply
Posted by
TiN
on 11 Jul, 2016 12:19
-
I my case pimping A9 PCBA got original value 7.074225xx to 7.180754xx. Just for reference here.
-
#4 Reply
Posted by
vtp
on 11 Jul, 2016 15:33
-
I should have mentioned that the reference board was also from 2007 and has most likely been changed as a pair with the DC board. I still think someone swapped this faulty DC-board along with the ref board from another meter.
I got bored yesterday evening and already adjusted the meter to my Fluke 732A. That cured the reference voltage difference and "CAL? 02" reports now 7.15407265V.
My 732A is not calibrated as it is fairly new to me but I measured it with my other 3458A (ok, well, in that sense it is now actually calibrated) and used that value for 10 volts. Final cal will have to wait until I send the 732A in office to national lab sometime in the autumn. There is also a brand new 3458A in office, I just unboxed it last week. So, I think I am covered in getting the volt to home lab.
The other 3458A has not been on much and is likely still very near what it was a year ago when the company 732A came from visit to national lab. I then adjusted the 3458A to it within 24 hours.
About the EMI filter, as I mentioned the old one took a flight and there is now a brand new Schaffner there. Whenever I get a new box I always change those filters. And all yellow cased X- and Y-class caps too.
And as for the cheapness of the meter, it cost $95. Add shipping and repairs and it is already at several hundred dollars now. But still a very good deal, assuming there is nothing else wrong with it. If it starts drifting, then, well...
-
#5 Reply
Posted by
TiN
on 11 Jul, 2016 15:50
-
Seems you are set and have nice job to allow playing new toys.
Only thing left would be get 10KOhm standard and transfer resistance into your fixed box as well.
That covers most needs, unless you into AC-measurements, in which case you will need AC SCAL calibration.
And great deal at 95$, even drifting
I'd take two of such any day.
-
#6 Reply
Posted by
macboy
on 11 Jul, 2016 15:58
-
...
And as for the cheapness of the meter, it cost $95. Add shipping and repairs and it is already at several hundred dollars now. But still a very good deal, assuming there is nothing else wrong with it. If it starts drifting, then, well...
What?!
Where does a person get a complete 3458A for $95? Even with faults, it's complete and it powers on, so that's
insanely cheap. In the condition you described, I'd guess $1500 to be reasonable market value in North America (more elsewhere). I don't suppose your source has any more available?
-
#7 Reply
Posted by
vtp
on 16 Jul, 2016 18:56
-
Top unit is the newcomer:
Ever since it was turned on after repairs it has shown a steady 1.45 ppm drift (CAL? 72) per 24 hours. Test continues and it will be interesting to see if the drift settles, though I doubt that. Another interesting examination will be what drifts in it if it continues. ACAL DCV cures the drift every time and reading of Fluke732A 10V settles very close if not exactly to previous day reading.
There is not enough data to draw any conclusions yet.
Bottom unit was turned on as a reference a day later and since has shown less than 0.02 ppm drift - likely just noise.
Stacking effect on top unit "TEMP?" has been below +0.5 degC relative each morning. I take the "CAL72?" results each morning as the meters have had the room undisturbed to themselves over night.
SN18A refers to serial number range US28041400 to US28032927. Apprently these have been manufactured between 2000-something to 2005 or 2006. The note itself is dated early 2007. Both my meters have SN prefix 2823 and the ADC board is also from that time so SN18 should not apply to these. That of course does not mean the ADC hybrid could not drift.
-
#8 Reply
Posted by
bitseeker
on 16 Jul, 2016 22:22
-
And as for the cheapness of the meter, it cost $95. Add shipping and repairs and it is already at several hundred dollars now. But still a very good deal, assuming there is nothing else wrong with it. If it starts drifting, then, well...
That's
not "a very good deal." That's an amazing deal!!!
Let me know if any more show up. My 34401A and 34410A won't mind having a big brother.
-
#9 Reply
Posted by
Dr. Frank
on 17 Jul, 2016 07:31
-
AN18 is not applicable, as it's an earlier instrument, and because ACAL DCV brings reading back to original value. So the LTZ reference is just fine.
Unfortunately, the A/D converter seems to drift, just like TiNs first converter PCB.
If this turns out to be true, that's a total loss, as only a replacement of this PCB can help.
Frank
-
#10 Reply
Posted by
TiN
on 17 Jul, 2016 08:27
-
I agree with Dr.Frank regarding reference stability. If that would drift, ACAL would not return back to original values.
So replacement of A/D PCBA or U180 A/D converter analog hybrid is the unfortunate option
SN18 is not a direct application here, but just as indicator and one of methods to test drift issue.
S/Ns covered in SN18 not apply to my unit as well, but I just used same method to debug.
Perhaps you can take few photos of your A/D A3 PCBA for our reference?
-
#11 Reply
Posted by
vtp
on 17 Jul, 2016 09:09
-
... and because ACAL DCV brings reading back to original value. So the LTZ reference is just fine.
Yes, I am very well aware of that, thanks.
Here's the actual data on "CAL? 72":
11.7.2016 0.994877516 (sometime after being turned on but after DCV cal with Fluke 732A)
...
14.7.2016 0.994871752
15.7.2016 0.994870256
16.7.2016 0.994868840
17.7.2016 0.994867293
18.7.2016 0.994865862 I also would not call the meter a "total loss". It just needs further repairs. Even with a new ADC assembly the total cost would still be reasonable for a 3458A - in known and verified condition. Which the typical units available for example in ebay or other surplus stores certainly are not in.
-
#12 Reply
Posted by
TiN
on 17 Jul, 2016 12:21
-
Noone in sane mind would say it's a total loss, it's just not as candy with A/D replacement need, as the previous one you got is
I was tracking SN18 more-or-less periodically on mine
for quite long time, so you can reference data.
Even w/o replacement you still can use meter for short term measurement and that will need to do ACAL like every 15 minutes to stay in spec.
-
#13 Reply
Posted by
Dr. Frank
on 17 Jul, 2016 13:13
-
Noone in sane mind would say it's a total loss...
I think, I said so.. Thank you very much , TiN!😁 😁 😁
If you order from K.S., they will charge about 2k$, I suppose?
Not everybody gets a couple of this PCB for trial-and-error and for free, I suppose
Frank
-
#14 Reply
Posted by
TiN
on 17 Jul, 2016 13:29
-
Their USA Parts site shows
$1306 for new one.
Hopefully,
plesa will emerge somewhen with hi-res pics of that tricky U180 for us. Thinking of everything I tried on drifty A/D one thing wasn't tested - if baking for few days on high temperature could help it, given theory of leakage in some of switch FETs due age. Perhaps
vtp could try this if have desire and oven capable for +80-90C.
-
#15 Reply
Posted by
Kleinstein
on 17 Jul, 2016 13:59
-
With TiN's drifty meter there was not only the ADC itself drifting, but also the resistors inside U180 used for reference scaling. So this very much points to the resistor part of U180 was drifting. With some of the resistors drifting it's very likely that the drift of the ADC part itself was also due to resistor drift and not something else.
A more or less constant 1.4 ppm / day drift of the ADC may not be that bad, as frequent ACAL can be a workaround and the predictable part could be corrected.
-
#16 Reply
Posted by
TiN
on 17 Jul, 2016 14:39
-
I'm not that convinced about resistors, as drift on my A/D PCBA was present even with external foil resistors instead of ones in hybrid.
-
#17 Reply
Posted by
vtp
on 17 Jul, 2016 15:07
-
Their USA Parts site shows $1306 for new one.
That has interesting description: "TBR ANALOG TO DIGITAL IN-GUARD PCA (UNTESTED FRESH BOARD)".
03458-69503 is referred to as "Rebuild A3 assembly" in SN18A. Putting in that part number gives it as non-orderable and refers to two "refurbished" assemblies, 03458-66503 and 03458-66513 of which the previous is obsolete. So it seems they all are rebuild or refurbished assemblies, not fresh new boards.
It is highly likely I'll get one but that will be later in the autumn. The meter works now otherwise.
...as frequent ACAL can be a workaround and the predictable part could be corrected.
Yes, the meter works absolutely perfectly with ACAL before any serious measurement. As I wrote, ACAL cures the drifted reading approximately perfectly, verified with a Fluke 732A.
The reading in DCV is going upwards in time so I would expect there is a (positive) reference voltage that could be going downwards in the ADC. Or negative reference going upwards. Unfortunately there is no very good internal schematic in the CLIP of the ADC hybrid, will have to study HP Journal article again for the ADC operation.
I should point out that there is just one cheap meter, not two. The bottom one in the picture is my old unit which I have had for several years. It was certainly not as cheap as the top unit.
-
#18 Reply
Posted by
vtp
on 17 Jul, 2016 15:47
-
Thinking of everything I tried on drifty A/D one thing wasn't tested - if baking for few days on high temperature could help it, given theory of leakage in some of switch FETs due age. Perhaps vtp could try this if have desire and oven capable for +80-90C.
I am not sure if putting the board in oven would be a good idea. All products I design or supervise the design of at work get accelerated lifetime testing at high temperature and that certainly makes all TaN (tantalum nitride) precision resistors drift. I would not want to age the ADC board more than it has already been aged.
Of course, there is a very interesting topic of what actually makes the meter drift. Is it the precision resistor array on the hybrid or something else? If the hybrid has been ok for say 25 years then what made it start drifting now? Sudden moisture ingress? Stress due to some more benign electrical fault within the hybrid? The drift is constant meaning that the fault condition is increasing in effect.
The resistors on the hybrid are most likely TaN as that is highly stable and well known alloy for precision resistors. It is also very resistant to moisture. Any moisture related issue should be very visible in microscope picture of the resistor array.
If it really is a resistor drifting, my bet would go to one or more of the resistors being stressed beyond design parameters. That is how they usually age (and drift).
-
#19 Reply
Posted by
Kleinstein
on 17 Jul, 2016 16:15
-
The resistors in the U180 hybrid are not directly connected to outside. So it is essentially impossible to have them stress to more than about twice the normal voltage, even if OPs brake down. This should not be damaging stress as normal operation is at low temperature rise for high linearity. It also looked like more than one of the resistors where drifting. The one case of overload I could imagine would be a latch-up in U180, causing the whole chip to run hot for some time and somehow not burn it all the way.
Not all the resistors from U180 could be replaced externally - the ones to the switches are not accessible from outside. They are also tricky, as low and stable capacitance may be needed.
There seem to be trouble with some of the U180, causing drift (TiN had 2 of them), and there was a service note on some of the units needing a replacement ADC board to fix drift problems. So it might be some risk to go for a untested refurbished board. The current one does not look that bad.
-
#20 Reply
Posted by
TiN
on 17 Jul, 2016 16:48
-
Original A/D PCBA had error 114, convergence after few samples due to bad U180.
There are always risks with such gear, but who cares anyway.
Reason why I mentioned oven baking, to remove humidity from the chip and see if that changes drift rate. I was swapping and soldering U180's on both boards few times, and results were following chip directly.
-
#21 Reply
Posted by
e61_phil
on 17 Jul, 2016 20:54
-
I keep on wondering since quite some time why is Auto-CAL needed for the 10V range. I think there is some circuit which makes the +/- 10V references for the ADC and this 7V -> 10V can drift.
Why is HP going this way? If you have a superb ADC isn't it possible to measure 140% of the reference voltage? So why not +/- 7.xxV as reference without a drifting 7V to 10V circuit? The inverted reference should be much more stable with a gain auf -1.000.. (two identical resistors on a chip or something like the LTC1043).
-
#22 Reply
Posted by
vtp
on 18 Jul, 2016 08:47
-
Results from the last 24 hours are in and amended above. Day to day drift this time was 1.43 ppm - so it works like a clock.
I think I have had enough of this drift experiment for now.
TiN, instead of an oven to bake the chip I would use vacuum. If there is humidity ingress that should make it evaporate. Better of course if at the same time in vacuum the chip could be coated with something reasonably inert and humidity resistant - assuming that there actually was humidity in and the package was leaking.
-
#23 Reply
Posted by
Dr. Frank
on 18 Jul, 2016 09:24
-
I keep on wondering since quite some time why is Auto-CAL needed for the 10V range. I think there is some circuit which makes the +/- 10V references for the ADC and this 7V -> 10V can drift.
Why is HP going this way? If you have a superb ADC isn't it possible to measure 140% of the reference voltage? So why not +/- 7.xxV as reference without a drifting 7V to 10V circuit? The inverted reference should be much more stable with a gain auf -1.000.. (two identical resistors on a chip or something like the LTC1043).
Please read HP Journal 4/1989, page 8 onwards, how this A/D converter works.
The usual dual-slope technique is in fact independent of the stability of its resistor and capacitor used.
In this case of a multislope technique, it's much more complicated, as due to the speed requirement, the 3458As A/D depends on the stability of the ratio of its resistor network, including the integrating resistors, as well as the +/-12V voltage reference amplifiers. Why they had to choose 12V instead of +/- 7.15V, I also did not yet understand, but anyhow, they needed at least a precise and ultra stable (-1) amplifier, which could not be realized with such simple capacitive inverters, I think. Charge injection is also a crucial aspect.
Frank
-
#24 Reply
Posted by
plesa
on 18 Jul, 2016 11:27
-
Results from the last 24 hours are in and amended above. Day to day drift this time was 1.43 ppm - so it works like a clock.
I think I have had enough of this drift experiment for now.
TiN, instead of an oven to bake the chip I would use vacuum. If there is humidity ingress that should make it evaporate. Better of course if at the same time in vacuum the chip could be coated with something reasonably inert and humidity resistant - assuming that there actually was humidity in and the package was leaking.
Te U180 seems to be not hermetic, there is can glued. Baking on low temperature is better than vacuum.
I can make experiment with TiN ASIC ( measure in vacuum and bake). But give me few weeks. I'm quite overloaded and do not have much time for this kind of experiment.