Fluke 5700a repair

[veedub565]

A wrong manual is a very real possiblity.

I think this may be the case, or at least open to interpretation.. The truth table that shows relay positions for the A8 switch matrix shows K15 as Reset for (AC) 2.2mV, 22mV, and 220mV.

According to the theory of operation though, all K15 does is switch between external and internal sense. I checked this and confirmed K15 is in the set position at all times. Unless external sense is selected on the front panel, and then K15 goes to Reset.

There's no reason for K15 to be reset just because (AC) 2.2mV,22mV, 220mV ranges have been selected. So I may have been chasing a bit of a red herring here trying to work out why A11 was causing A8 not to set K15 correctly. This makes more sense as I just can't see how that could ever really happen anyway.

If anybody is able to confirm on their 5700a that A8 K15 should be set all the time (unless ext sense selected) then that would be great. Only on special rare occasion do I get the opportunity to test with another working 5700a


It still makes not much sense really because I still have the problem A11 causes 4x A8 Relay errors (3x A8 relay error and 1x A8 MBD relay error). A8 is 100% tested and known good. A11 tested in a working 5700 produces the same errors, so also 100% A11 is the problem.

It would be nice to know what the 5700a is doing when it steps through the self diag. What is it attempting to do when it flags these errors. And what is it expecting to see. We know A8 is working correctly but possibly DAC is not seeing what it expects and self diag thinks A8 is the problem. In reality whatever it expects to see is not there or incorrect for some other reason (not A8 Relay problem)... Argh 5700a shennanigans !

[veedub565]

Another update for this long running repair job.

I haven't had much luck disassembling the firmware to see what's happening while running self diag. Instead I ran the self diag and when it flagged up the A8 Relay error I checked the position of all the A8 & A11 relays. Every relay on A8 is in the "set" position except for K32 which is "reset". What this means is that "PA COM" is switched to the "RCL" line

On A11 K5 and and K6 are "set" this means the "RCL" line is switched to the + input of the ADC amp, and "REFCAL" is switched to the - input.

I checked the configuration of A11 U32 and it's configured so that "ADC COM SEL" is enabled.

So at the point where I get the first A8 error, it seems that the ADC is looking at "PA COM" on the "RCL" line and comparing it to "ADC COM" on the "REFCAL" line. And for some reason it doesn't like what it see's and thinks there is an A8 relay switching problem. We know the A8 Switch Matrix is perfectly fine as it works in another unit, so the problem is with A11.

The voltage across the inputs to the ADC amp is 0.003v and I checked that the ADC amp has the correct gain of x11 and that this voltage gets to the input to the ADC ok. It may be coincidence, but when set to 10VDC output I get 9.997VDC... 0.003v out! I'm not quite sure what this means (if anything)


I know the ADC must be working because, well, lots of other things would fail if it wasn't. One thing I did notice is that the ADC reference should be 6.4v +/-0.2v and I measured it at 6.20v it's just borderline in/out of specification. I wonder, if the reference is slightly out of spec, the ADC isn't going to work properly, isn't going to self calibrate or zero properly. Possibly (and I haven't checked this) the ADC reference is 0.003v outside the max spec (6.203v).

I don't know though, I mean a spec of +/-0.2v is pretty large, presumably the A11 DAC is calibrated with the ADC reference at whatever voltage it's at within that range. Changing the zener diodes that control the ADC Reference voltage could throw the calibration totally out and still not solve the problem.


**Edit**
One more thing, resistors R88,89,90 on A11 DAC could have an effect on the ADC voltage reference. These resistors were changed from the original carbon comp to metal film. I checked they were all exactly the right value, but the calibration would have been performed with the old carbon comp resistors in place at whatever value they had drifted to. Replacing these resistors may have altered the reference voltage and thrown the cal out resulting in the output voltage error, totally unrelated to the A8 relay error. I do still have the original resistors so I could try putting them back in and see if it makes a difference. May have been best to leave these critical value resistors alone.

[Kleinstein]

The absolute value of the reference votlage should not effect the ADC function very much and the calibration should take care of this.
The calibration should be using the main DAC reference to calibrate the ADC's full scale.
Especially a change in the reference should not effect the zero point / offset.

The resistors R88,R89,R90 at least in the plan (PDF file) I have are low tolerance precision resistors that do effect the ADC. They should not be carbon composite, but high quality types. There are some versions that look like carbon composite but are quite a bit better.
Wrong resistors could really upset the ADC offset. Chances are there is an calibration compensation step for this, but the range my be limited and wrong resistors may have more effect on the ADC. 

[veedub565]

Thanks that's interesting to learn. I verified 13v reference at 13.02v nice and stable. Also 6.5v reference is good too.

It's possible then that replacing these resistors is what has caused the big drift in output voltage. I'll try putting the original ones back in and see what happens. Originally when I first got this, it did output 10.0000v (but still with A8 relay error) after resistor and cap change then output is 9.997v. I assumed a bad zero cal due to A8 relay error, but there is a chance it is the change in these resistor values, especially if they can really upset the ADC offset. I did replace with modern low tolerance metal film, but still values could be slightly different.

[veedub565]

Argh it seems I have kept all the old resistors except those 3 

The new ones I put in all measure dead on 20.0k and one 50k


Honestly I even contacted Fluke. even though it's not officially supported they offered to look at it £750 investigation charge. And then diagnostic/repair time on top of that assuming they can even fix it and have the parts. And then calibration on top of that. Repair costs could easily spiral out of control, and easily spend £750+ and they could still say can't fix it... I'd pay to have it repaired, but I'm not signing a blank cheque for maybe or maybe not fix it

[Kleinstein]

The 3 resistors work togeter. So the individual resistors do not matter, but just the combined value of R88*R90/R89. So it is only a single value to trim. So there is a chance one could try trimming the value with an added parallel resistor (e.g. 10 M range) to one of the 20 K ones.
This way one could at least check what effect the resistor would have.

PS: I had a look at the service manual description of the ADC. The AZ capacitor C95 is supposed to compensate for an offset error of the DAC and this would include the part of R88,R89,R90. So unless the resistors are far off (e.g. >> 1 %), they should not have a significant effect on the result.

[veedub565]

Well I have tried a few different values in parallel and series with A11 R90 to see the effect. I was able to get the reference voltage at U27B pin 1 almost exactly 6.4v. And also I was able to make it a long way out of spec (5.90v). In all cases, the output voltage did not change at all, even after zero cal.

So yes it does appear the absolute value of the reference voltage does not affect things very much. Changing the value of R90 also has very little effect.


Back to the drawing board for me then

• When I first got this the output was 10.0000v, and I guess most operators don't perform the zero cal. Then it developed A8 relay fault, but output was still good. Then I performed zero cal and output is now bad. So whatever problem the DAC has is causing bad zero cal. Although it seems to think zero cal has performed successfully.

• It also seems to think it has an A8 relay problem, but 100% fault is on the A11 DAC

• output at 10VDC is out by 0.003v and also input to ADC amp is 0.003v

• At the point during self diag where it throws A8 Relay error, it appears to be comparing "PA COM" to "ADC COM", both of these should be 0v (when measured it is 0.003v) without getting into the firmware It's difficult to know what it is expecting to see here

• Following the A11 troubleshooting steps shows no problems, also no A11 self diag errors. So whatever the fault is has evaded both troubleshooting and self diag.

• If A12 is removed then I don't get the A8 relay error, only when A12 is fitted do I get the A8 relay error.  Yet both A12 and A8 have been tested in a working 5700a and both are fine, problem is with A11.

Some or none of these things may be significant, trying to piece it all together to even guess where to look 

[Kleinstein]

I would expect the zero cal step to be repeated from time to time, especially as part of any calibration, but also on a few more extra occasions.
There is a slight chance the old calibration was done without the extra zero step - hard to tell, were and when this was done. With a broken instrument there may be a more sloppy adjustment in an attempt to fix the problem.

A difference between PA COM and ADC COM could be a contact problem with a card connector or a fault in ground current compensation. The two may have to come together. So when the ground current compensation is working well a bad contact may not cause an error, and if the connection is very good it may not trigger an error even with poor current compensation.
 
It the 3 mV difference measurend with the calibrator (e.g. some bug mode?) or with an external DMM ?
The ADC part could have some offset from the amplifiers (2xOP227 in the amplifier stage and LT1056 as input buffer). This offset would be subtracted digitally and should normally not reach 3 mV, as the LT1056 is only after the gain.

[veedub565]

The zero cal should be performed regularly, but from experience engineers don't often do this like they should.

I did look at the solder joints for the card connector, and all looked ok, but I will try giving the connectors a clean with some IPA.  And also take a look at the ground current compensation circuit, maybe something odd going on there.

The 3mV difference, I set the calibrator to 10VDC output and measured at the output terminals (with a DVM) 9.997V. So it's reading 0.003V too low. Then I ran the self diag routine,  at the point where it fails A8 relay error I measured the input to the ADC AMP (common to TP12) I see 0.000V on the - input (REFCAL) and 0.003V on the + input (RCL) so a 0.003v difference, which seemed a suspicious coincidence. I'm guessing the self diag routine isn't expecting to see any difference at all here, and any problem then A8 relay error.. but that is just pure speculation

[Kleinstein]

For the 3 mV error for the output, is this more more like a 3 mV offset of a 300 ppm error in the scale factor. A measurement at some 1 V or 0 V should give the answer to this.

[veedub565]

For the 3 mV error for the output, is this more more like a 3 mV offset of a 300 ppm error in the scale factor. A measurement at some 1 V or 0 V should give the answer to this.

I'm back at work for rest of the week now so not much time to investigate too deeply at the moment, but I did make a few measurements to help answer above. All measurements taken using calibrated Datron 1281 at 6.5digit resolution, and after zero calibration had been completed.

220mV Range
0V = 0.2889v

2.2V Range
1V = 0.9999v
2V = 1.9997v

11V Range
3V = 2.9995v
4V = 3.9992v
5V = 4.9990v
6V = 5.9988v
7V = 6.9986v
8V = 7.9984v
9V = 8.9981v
10V = 9.9979v


So 0V output, even after zero cal shows some 0.2889v error. And then at 10V error is 0.0021v.


Among other things suggested, I'd also like to investigate the duty cycle circuit and do the math to see what output voltage should be. I'd be interested to see if duty cycle math works out at 10.0000v or 9.9979v. It would be interesting to see if DAC is setting the correct voltage, but somehow between DAC output and terminal posts it's going wrong. Or if DAC is setting the wrong voltage to start with.

If DAC is setting output to be 9.9979v (@10v) and same 9.9979v is measured at terminal posts, then not much wrong with DAC output. And problem may be with ADC side, and maybe reading it's own output incorrectly. .. It's a theory and something to investigate anyway.


Part of troubleshooting guide for the duty cycle says the following

"Check OUT1 from U6. Connect an oscilloscope to U6 pin 16 (common to TP1). Set
the 5700A to 6.5V dc, operate, and set the oscilloscope to 2V/div at 2 ms/div. The
oscilloscope should display a TTL-level square wave with approximately a 50% duty
cycle.
"

I measured 48% duty cycle.... but no spec only approximate so I thought that's ok.... OUT1 is only for course adjustment, so maybe 48% is close enough to 50%. Then OUT2 is fine adjustment. Still, worth investigating and to see how the maths works out

[Kleinstein]

The error mainly looks like an error in the scale factor. The 220 mV is however way off. This could be the divider 220 mV setting itself, but also the part used for the internal 5700 cal steps to measure the divider and offsets.

48% PWM should be close enough to 50%, as the absolute values of the reference voltage has quite some tolerance.

[veedub565]

The 220mV divider is on A8 which works ok, I'd say in that case more likely problem with part used for the internal 5700 cal steps to measure the divider and offsets. ADC ic itself must be ok otherwise lots of other things would fail self diag.

Maybe have a look at the parts A11 plays in self cal and zero steps.

[veedub565]

Another update on the progress of this repair

I've methodically tested each diode/zener and marked them off as tested. I've also been methodically removing each transistor from circuit and testing them one by one. The 9th one I puled tested bad !

Q23 The Peak Atlas DCA component tester identifies this as a faulty component. I checked with a DVM and the gate is open circuit from the source & drain. I considered that maybe the Atlas DCA can't identify this part properly. So I pulled a couple more with the same part number and it correctly identified those as a N Channel JFET, and identified the Gate terminal.

So it seems Q23 is indeed faulty. I'm not sure what effect this would have on the circuit, or how it would produce the faults I'm seeing, but it can't be good. So I now need to source a suitable replacement part, and see if this cures at least one of the problems I've got (after many false promises I'm not holding a celebration party just yet)

I did already do a check of all the transistors (one of the first things I checked months ago), and checked for shorts. usually they fail short circuit, it's a bit unusual for this to have failed open circuit.

[Ugur]

Hello,
Can you check U2 in the A11?

[veedub565]

Yes of course, U2 Neg Offset Circuit and Shunt Linearity Control.  Is there anything in particular you want me to check ?

I need to replace Q23 first before I can do any measurements. I was going to replace U2 actually before I found Q23 was faulty. 

[veedub565]

Q23  I checked with a DVM and the gate is open circuit from the source & drain.

I must have made some mistake yesterday, I rechecked this morning with the DVM and read the following

Drain to Gate = 168ohm
Source to Gate = 258ohm
Drain to Source = 320ohm

it does still read faulty on the component tester

[Kleinstein]

At least in the plan I have Q23 is a MOSFET and not a JFET. The gate should thus be isolated. With an open gate the drain source reading can be rather random, depending on the gate charge.

[veedub565]

On my schematic I have Q23 as J2086, but the parts list shows it as J2907-TR3 and the part I pulled out is marked J2907 (this is on the A11 DAC card)

[Kleinstein]

OK- the part numbers repeat on different PCBs.  In the DAC part Q23/Q24 are used as diodes. In this use the JFET type is not that critical.
It looks like for protection between different nominal grounds. At least for testing it should be OK to use another low leakage diode there (e.g. BAV199 or transistor BE junction) - for a 1st test even 1N4007 could do.

Used as a diode the part could actually even be a BJT, but doubt the current part is a 2N2907.

[veedub565]

Thanks for the explanation, I've got a 2N3819 JFET handy would that work? it's only 25v not 60v but if it's not that critical might be ok.

I'm hoping this solves the problem I have with this card, but I can't really see how it would. Especially if it's just there for protection. This one appears to have gone low resistance Gate to Source/Drain so acting more like a resistor than a JFET, maybe having some strange effect. Anyway it can't be good and needs to be replaced. 

[Kleinstein]

The 25 V rating should not be a problem, as it is used as back to back diodes, so never more than 1 V reverse voltage and likely relevant with less than 100 mV.
The 2N3812 may not be very robust, but should work.

[veedub565]

Great I'll give it a try

[Ugur]

Q23 part code J203 N-Channel Jfet.
You can also take out the Q24 and check to be sure of the part, it's the same part.
By the way, it just occurred to me.
There are two large relays just behind the front panel of the device, these relays can be removed from the board after opening their tabs.
Check the resistance of the contact terminals of this relay. (in open and closed state)
should be zero ohms.
Remove the cover and clean the contact ends and common ends with fine sandpaper, then clean with oily contact spray and try again.
This could be causing the error you are experiencing.

[veedub565]

Well Q23 was definitely faulty, replaced with another JFET, but it hasn't made any difference at all to the fault 

Back to checking out every single part one at a time, you never know I may find another failed FET somewhere.