Author Topic: Fluke 8520A  (Read 4322 times)

0 Members and 1 Guest are viewing this topic.

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Fluke 8520A
« on: January 07, 2012, 10:28:40 AM »
Some days ago a friend of mine brought to me a fluke 8520A.
He bought this instrument about 4 years ago, it was functional,
and from then it sat unused in a moisty “lab” for all this time.
Last week we powered up it, and it started regularly but soon
we realized that it had some problems: Vdc measured correctly
but Vac measured a fixed value and Ohm gave an High Voltage error.
I have not much experience in repairing instruments, and
this beauty has the same age than me, but it was not hard
to understand it's problems.
Fortunately the fluke is very well documented, so I downloaded
the service manual. The second time I started the instrument
it simply stopped working, with freezed display. After a reboot
it's display wasn't even lighting up.
Here are the 4 faults I found:
the first was to understand why the display wasn't functioning, the
CPU was reading the memory addresses but wasn't doing what it
should have done. The cpu is an MK3880 from mostek, reading about
it I learned that it's a Z80 clone, so I swapped with an old Z80 and the
 instrument came back to life. This took me a pair of days familiarizing
with the instrument circuits :)
The second problem was the keyboard, some buttons didn't respond.
They were electrically ok, but one of the outputs of a MM74C906 was
open. I didn't want to wait for a replacement so I built a small board that
replaced the buffer output.
Third problem was a 7908 neg. regulator, temporary solution is a 7905
with a led and a pair of diodes to bring it to approx -8V.
Now I was ready to start the self test, but I was wrong: the instrument
was upset because it had too low resistance on it's inputs, it read
between 2 and 3 nanoSiemens and it wanted it to be less than 1nS, i.e. >1GOhm.
I tried to heat up the High Ohms input components but the conductance
worsened. At last I cleaned with some TCE the bottom side of one of the input
relays and the conductance rapidly fell down to 0,3-0,7nS.
The point I cleaned was visually clean and dry :)

Now the input reads always >0,6 nS and raises with temperature, is it normal?
I imagine it's a cleaning problem?

Edit:
please let me know if the images are too big to be comfortable to view, I'll resize them.
Fabio
« Last Edit: January 07, 2012, 11:02:55 AM by muvideo »
Fabio Eboli.

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #1 on: January 07, 2012, 10:37:28 AM »
This should be the voltage reference board and it's schematic,
 I haven't found anything about it yet,
what do you think about it ?

Fabio.
« Last Edit: January 07, 2012, 10:13:57 PM by muvideo »
Fabio Eboli.

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #2 on: January 07, 2012, 11:04:25 AM »
The conductance should be less then 0.1 nS. Did you have the guard button OUT when you were testing this? It should be if you don't have a separate guard wire connected.

Heating up the connectors will cause thermal voltages to be generated, so that will not work.

So you probably have some leakage on the board, contamination trapped under relays or inside switches, or components have gone leaky such as the some of the protection components or input components.

Tracking down leakages in the sub nA range is not easy. Good Luck. If you have a schematic, there may be places you can break the connection between the measurement circuit and the input terminals. That may help isolate where the leakage is.

Richard.

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #3 on: January 07, 2012, 11:53:05 PM »
The conductance should be less then 0.1 nS. Did you have the guard button OUT when you were testing this? It should be if you don't have a separate guard wire connected.

Heating up the connectors will cause thermal voltages to be generated, so that will not work.

So you probably have some leakage on the board, contamination trapped under relays or inside switches, or components have gone leaky such as the some of the protection components or input components.

Tracking down leakages in the sub nA range is not easy. Good Luck. If you have a schematic, there may be places you can break the connection between the measurement circuit and the input terminals. That may help isolate where the leakage is.

Richard.

Hello Richard, thank you I didn't even knew the proper conductance value I should have read.
The guard button was out. For now I cleaned all the socketed relays, and part of the board with TCE.
Maybe this and maybe the nice dry weather today (last days it was raining and wet), the conductance
is slowly decreasing with opened case. This evening I'll close the case and tomorrow I'll see.

To everybody:
can anyone explain to me the reference circuit?
How does it function?
Is it an heated zener (maybe fluke proprietary part)?
Fabio Eboli.

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #4 on: January 08, 2012, 01:20:15 AM »
It is not a heated zener. Making a super stable voltage reference from a combination of a thermally bonded forward biased diode and a buried zener is well known technique.  In this case, the diode is replaced by the Vbe voltage of a transistor.  The transistor has a -2.2mV/C temp coefficient.  If you match it with a zener with a 2.2mV/C coefficient, you can then fine tune the temperature drift by varying the zener current. The more current, the greater the temperature coefficient, and this is a very predictable current versus temp coeff. formula. The extra resistor and diode will be some form of extra compensation to get the best drift.

So I think the two unmarked resistors will be selected resistors to match the particular zener IC operating current. Then I suspect they thermally cycle this board, measure the total temperature coefficient from resistors, opamp offset drift, zener drift, and from that they can work out what variation to the output voltage setting (ie zener current) will give a zero coefficient.

I think the reason why the output voltage range is so great is they are only after maximum stability. The seperate meter calibration of the different measurement functions will compensate for the error.

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #5 on: January 09, 2012, 09:14:19 AM »
It is not a heated zener. Making a super stable voltage reference from a combination of a thermally bonded forward biased diode and a

cut...

So I think the two unmarked resistors will be selected resistors to match the particular zener IC operating current. Then I suspect they thermally cycle this board, measure the total temperature coefficient from resistors, opamp offset drift, zener drift, and from that they can work out what variation to the output voltage setting (ie zener current) will give a zero coefficient.

I think the reason why the output voltage range is so great is they are only after maximum stability. The seperate meter calibration of the different measurement functions will compensate for the error.

Thank you for explanation, I'll try to understand better later.

It seems I have another problem, maybe the same that causes the leackage.
After some cleaning the leackage lowered and the instrument now
reads about 0,2-0,4nS when cold.
When the temperature raises the nS reading rapidly surpasses 1nS.
When the instrument is cold it almost passes the self-test, "almost"
because the values of the failed tests are only slightly offrange.
But when the instrument is warming it fails badly 3 tests:
100mV zero, 100mV range and 10MOhm . In the manual
this failure pattern indicates that the DC buffer x64 setting is failing.
Also when the instrument fails the test, the 100mV DC range reads
a fixed -15.104mV value, but it shows a "normal" value in 1V range
that has a different dc buffer gain.
The failures and the input leackage have in common the same
dc buffer gain setting of x64.

For me the suspect component is the dual jfet Q234 marked T502.
It's the suspect because when slightly heated it makes the instrument
fail also when cold, and when swapped with Q227
the instrumet doesnt' fail when heated in 100mV range.
When I try to cool q234 when
the instrumet is hot it returns to normal behaviour for some
minutes.

Now the question: how can I test these jfets?
Where can I search for spares or equivalents?
Fabio.

P.S.
In the nS range, the instrument connects the dc buffer input
directly to panel input trought some relays and a fet in input switch.
This input is also connected to a reference voltage trought some resistors,
these together with the device under test form a voltage divider.
There is also a protection device that I disconnected, without effect.

P.S. 2
I know this poor fluke insn't in the best hands,
and the learning process is painful for patients :)
« Last Edit: January 09, 2012, 09:25:51 AM by muvideo »
Fabio Eboli.

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #6 on: January 09, 2012, 10:09:44 AM »
First a warning. It is possible that Q231, Q232 and Q233 have no gate protection. If this is the case, then if they are ever removed from the circuit, the leads have to be shorted together first before removal, and have to remain shorted till it is back. If one of these is removed without shorting the 4 leads together, then if they are the kind that have no internal gate protection, there is a good chance you have damaged or killed it.

I would check the meter functionality as much as possible of the ranges that work.

Now I think the 100mV range is the only one that uses the x64 gain setting.  A problem with this range could be caused by a problem in the gain switching circuit (Q236, Q237 and Q238), or the DC offset adjustment (R236) may be too far out the the A/D to auto-zero. If the offset it too far out, it will be the x64 range that is affected worse then the other gain settings.

To see if the gain switches are working on the 100mV range, put 100mV DC on the meter input, and see if you get 6.4V  on TP206.

It is possible that calibrating the offset may fix the conductance range, but it may be that there is leakage in any one of the many components connected directly or through switches to pin 3 Q227. If it is leakage, then something to be able to locally heat or cool parts may help isolate the problem. Trick is not to cool parts to the point you get condensation.

Definitely Q227 and Q234 are not common components at all. Hopefully they are fine, but they may have to be sourced from Fluke. There is a good chance they are custom made IC's for Fluke.

Richard.

« Last Edit: January 09, 2012, 11:11:41 AM by amspire »

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #7 on: January 10, 2012, 12:47:42 AM »
First a warning. It is possible that Q231, Q232 and Q233 have no gate protection. If this is the case, then if they are ever removed from the circuit, the leads have to be shorted together first before removal, and have to remain shorted till it is back. If one of these is removed without shorting the 4 leads together, then if they are the kind that have no internal gate protection, there is a good chance you have damaged or killed it.

Fortunately I didnt' touch these mosfets :)

I would check the meter functionality as much as possible of the ranges that work.

Now I think the 100mV range is the only one that uses the x64 gain setting.  A problem with this range could be caused by a problem in the gain switching circuit (Q236, Q237 and Q238), or the DC offset adjustment (R236) may be too far out the the A/D to auto-zero. If the offset it too far out, it will be the x64 range that is affected worse then the other gain settings.

Checking better, the x64 gain is used on 100mV and 10Ohm ranges. I was wrong, this setting isn't shared with
nS measurement, that use x1 gain.
The DC offset and ADC auto-zero is something I should indagate, i'll try to understand better, it's hard for me.

Definitely Q227 and Q234 are not common components at all. Hopefully they are fine, but they may have to be sourced from Fluke. There is a good chance they are custom made IC's for Fluke.

Richard.

Sorry I miswrote the part name, it is intersil IT502, is a dual cascoded jfet.
I tried measuring 100mV dc with the voltage ranges and 6,8Ohm with the resistance ranges.
When the instrument is cold it reads reasonable values at all the ranges.
When the instrument warms up it reads as before at all ranges but 100mV and 10Ohm,
 these ranges both use the x64 gain.
In 100mV and 10Ohm it reads about 70mV and about 10Ohm instead of
correct values. The transition from correct to wrong value and reverse is abrubt, like a switch.
With both correct and wrong values the amplifier gain seems to work correctly, it brings the
100mV to 6400mV before and after the error.
The error whows up heating or cooling Q234 or U211.
I think the instrumet was repaired back in 2003 or later:
U211 and U210 are OPA445 with a datecode 32Z, dip8,with the 8pin socket adapted to
a TO5 pattern. They can be seen in the photo analog_lr.jpg I posted before,
low to the left of "AC Converter" aluminium box.
In the service manual the original part is listed as LM143 package TO5.
After all it's fun to analyze these schematics, I hope I don't make too many damages :)
Fabio.
Fabio Eboli.

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #8 on: January 10, 2012, 10:00:41 AM »
The key thing is to work out if the input stage is fault for the 100mV/100ohm range, or just needs offset calibrated.

Put it in the 100mV range, and monitor the voltage on  TP206. A oscilloscope may be better, as the input is being switched between normal and autozero mode (unless you can turn autozero mode off).  Anyway, see what happens when it you heat Q234/U211 and it stops working.

If the problem is in the amplifier section, you will see a dramatic change on TP206. If you don't then the problem is probably something to do with the fact the DC offset in the amplifier stage just got too much to cope with in the x64 gain mode.

The IT500/1/2/3/4/5  family looks like a nice device, but probably hard to source now. You can use IT500 or IT501 instead - they are higher spec'ed versions. I did find this http://www.alibaba.com/product-gs/518279530/IT500.html but I suspect they don't actually stock any. I still hope that you do not need to replace them.

Richard


Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #9 on: January 10, 2012, 05:37:50 PM »
In case you are interested, it looks like http://rcfreelance.com may have one IT501 chip instock (the higher specification version) and Intersil USA may have 7 IT502 chips.

Richard

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #10 on: January 10, 2012, 10:20:47 PM »
The key thing is to work out if the input stage is fault for the 100mV/100ohm range, or just needs offset calibrated.

Put it in the 100mV range, and monitor the voltage on  TP206. A oscilloscope may be better, as the input is being switched between normal and autozero mode (unless you can turn autozero mode off).  Anyway, see what happens when it you heat Q234/U211 and it stops working.


The gain value is visible only with oscilloscope, the adc samples at 200Hz.
In the DC buffer ther is also offset compensation, trough the mosfets.

If the problem is in the amplifier section, you will see a dramatic change on TP206. If you don't then the problem is probably something to do with the fact the DC offset in the amplifier stage just got too much to cope with in the x64 gain mode.


TP206 doesn't change. So maybe you are right about the offset compensation, I'd read the calibration manual, maybe
there are maximum levels accepted.


The IT500/1/2/3/4/5  family looks like a nice device, but probably hard to source now. You can use IT500 or IT501 instead - they are higher spec'ed versions. I did find this http://www.alibaba.com/product-gs/518279530/IT500.html but I suspect they don't actually stock any. I still hope that you do not need to replace them.

Richard


Thank you Richard, for now I hold, I'll wait for some spare parts, the repaired ones and
the two OPA445. When I'll have these parts I'll continue the work.

Fabio.
Fabio Eboli.

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Fluke 8520A
« Reply #11 on: January 11, 2012, 12:29:29 AM »
Fabio,

If you couldn't see any change to TP 206, but the meter stopped working on the x64 ranges, then it is very possible that there is no fault in the amplifier section for the 100mV and 100 ohm ranges.

Now since the A/D works fine on all the other ranges, it is probably OK to.  So there is a good chance that it is just the input amplifier offset needs recalibrating. I do not know this meter, but assuming it is a ramp-type A/D, there are ramps that have to cross zero within the correct range of time for steps like auto zero. If it doesn't, then the whole A/D ramp sequence probably has to be aborted.  That is why you can get that sudden change to the reading, if something like the input amp offset is to high.

Richard
« Last Edit: January 12, 2012, 09:29:02 AM by amspire »

Offline danielroca

  • Newbie
  • Posts: 2
Re: Fluke 8520A
« Reply #12 on: January 12, 2012, 07:55:37 AM »
Hi I'm new to the forum and I'll like your opinion if I should buy a Fluke 8520A from eBay for 45$us, It's really worth it to have this kind of equipment in a lab.

Thanks for your help

Offline muvideo

  • Frequent Contributor
  • **
  • Posts: 381
  • Country: it
Re: Fluke 8520A
« Reply #13 on: January 12, 2012, 05:52:56 PM »
Hi I'm new to the forum and I'll like your opinion if I should buy a Fluke 8520A from eBay for 45$us, It's really worth it to have this kind of equipment in a lab.

Thanks for your help

Hello, I don't have experience with other "high end" multimeters, I'm used to handheld multimeters,
my main instrumet is an old 4.5digits. I dont' know if it is worth 45USD in USA.
If it was 45eur here and it was in functional state I'd bought it in a heartbeat.
Compared to my handheld this fluke is fast, with very high input resistance and capable
of 4wire resistor measurement with .1mOhm resolution, that is very nice if you, like me,
use ofter low value current sense shunts. It has only Volt and resistance modes.
Consider also that probably it's more than 30years old, so maybe it can need often
servicing. It is very well documented, fluke has all the user/calibration/service manual
on their site. It has no battery backup memory so no worry to lose the calibration.
The active components are mainly 74Cxx logic IC, discrete fets and few quality opamps.

Fabio.
Fabio Eboli.

alm

  • Guest
Re: Fluke 8520A
« Reply #14 on: January 12, 2012, 06:16:47 PM »
$45 is not a bad price in my opinion, assuming it's working. Wether it's a good buy for you is impossible to tell based on the limited information you provided. See this recent thread on the same topic:
http://www.eevblog.com/forum/index.php?topic=6084.msg82030#msg82030


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf