Solartron7061

[Dek]

Got the burst measurements uploaded from memory using the GPIB interface.
The commands have to be issues in a specific manner when doing this manually.

Set instrument range eg 0.1v
Clear the memory history
Set up the instrument to trigger in burst mode - set no. samples required.
Set the GPIB to ++auto 2
send TRI to start the burst mode.
Set to ++auto3
Send Du (DUmp memory contents)
Send ++Read
memory contents are loaded into the serial monitor screen.

Graph below.
Sample rate is 1Khz, the sine wave period is  30ms or 33hz
I did the same on Range 1 and range 10, but the signal is very "digitised" so not very helpful but  is a similar but 'blocky shape'.

Would be interesting to see what your instrument shows Grizewald, performing the same test 
So that is something to think about.
Next I'll swap out the ICL7650 and do the same test.

[Dek]

Took out the ICL7650 and fitted the LTC1052
Ahh, that's disappointing, no difference (1/2 expecting that ) graph below.



While I was in that area, I noticed an additional earth lead that Grizewalds 7061 doesn't have.
Looks like an addition ??



So removed that hoping that would solve the noise problem.



Ahhh, that's disappointing

Dek.

[Kleinstein]

A 33 Hz signal is really odd. The amplitude is also quite high - not just a few µV, but mV range.  As it gets less visible in the 1 V / 10 V range this would be signal coming in at or in front of the input amplifier.

The first point is where is 33 Hz coming from ?
Tthe ADC forcing signal should be more like > 1 kHz. The 7650 internal chopping should be some ~500 Hz and mains should be 50 Hz. I would expect the display update in sync with mains, e.g. 50 Hz or 100 Hz, but not 33 Hz. Aliasing from some F Chopper or F_sample +- 33 Hz would be strange.

For changing the 7650 to an 1052, in this case the more relevant one may be the one at the input amplifier.

The second point is how could there by so much signal at the input, that should not be there. Even the 50 Hz / 100 Hz signal should and is considerably lower amplitude. So it much be something relative strong.
2 mV at 33 Hz may already be just visible with a scope - so one could check that the signal is really there. It would be amplified at the ADC, so well visible at some 200 mV there. It may be with the check, just in case the actual signal is at a different frequency and due to aliasing.

[Dek]

Quick check on set the range 0.1  using 1 ohm resistor.
Seeing exactly the same 33hz  noise.
(The frequency is assumed that the meter is actually taking the samples at 1khz ).


 

[grizewald]

Would be interesting to see what your instrument shows Grizewald, performing the same test 

That might be a bit tricky right now....

This is my 7061:



That extra earth lead that you have is very perplexing. I've scoured the maintenance manual for DP901 and can find no reference to it apart from the component layout for the analogue board. The pin is connected to the power supply ground and seeing as the whole analogue board is supposed to be floating, I'd say connecting it to earth is probably a very bad idea!

[grizewald]

I've ordered a complete set of replacement electrolytic capacitors from Mouser and will be replacing all of them. Some of them, the smaller values in particular, were borderline when I tested them with my ESR tester, so I may as well bite the bullet and re-cap everything while I'm there.

Next task is to get the boards under my microscope for a proper inspection. The analogue board is fairly dirty with dust and metal particles from the case locking screws. Something has to explain the offset I get on the current range and possibly some of the noise I see. Looking at the bottom of the board, I can see some parts on the guard traces particularly where the solder resist is missing. An uninsulated guard and a dirty board is a recipe for picking up noise. So I'll have a really thorough inspection under the microscope before I start cleaning and repairing the solder resist.

It looks like my analogue board has been "repaired" before and whoever did it wasn't very particular about cleaning the flux off the board afterwards.

My replacement capacitors will be here next week seeing as I ordered from Mouser, so hopefully the meter will be reassembled by the weekend and I'll run the same test as you did. I'm pretty sure I'm not going to see anything like what you're seeing there as I doubt the meter could be calibrated with that much noise.

[Dek]

Would be interesting to see what your instrument shows Grizewald, performing the same test 

That might be a bit tricky right now....

This is my 7061:

What kind of an excuse do you call that !

Good luck with the recap.
I haven't has the analogue board out yet. When I changed 2 of the PSU caps I managed to do it with the board tipped up.
With the noise problem I'm having, it might be a good idea to have the board out, do a thorough inspection and change the remainder of the caps before going much further.
Oh, I've also the mains filter to fit. (currently running on a in-line jobbie).

Hummm, just spotted the dip switches on your board are set differently to mine. Isn't that something to do with the mains frequency? (now reading the manual)
<edit> They are OK

[grizewald]

I think that's how my configuration switches were set from the beginning. Apart from the line frequency setting, it seems to ignore the rest of the switches if the scanner board isn't installed. The very top one isn't even connected to anything according to the circuit diagram (which has the switch block upside down compared to the orientation on the board).

I spotted a few things when I went over the analogue board last night. I'm even more convinced that it was repaired at some point as I can't imagine it coming from the factory like this:

[Dek]

I'll check mine later to see if the solder blob is a standard factory option 

From Kleinsteins comment about the 33hz noise source, I decided to check the burst sample rate as we are assuming the rate is 1000 samples per sec.
I injected a 50hz 0.1V ac signal with the meter set on DC range 0.1 4 digits.
Trace below show the result.
One 50hz cycle has the same number of sample points as the noise traces above.
The conclusion = the noise is not 30hz but must be 50hz.
This is a relief as the source will be easier to trace.
With my earlier focus on the Glug death problem, I only briefly looked at the analogue PSU - time for some new capacitors I think.
Dek.

[Kleinstein]

Even if the frequency does not match to 100% the 50 Hz part should ideally be suppressed by some 60-80 dB. So those 200 µV of 50 Hz should not cause so much noise. However having quite some 50 Hz background where it should not be indicates a problem.
So checking the supplies is probably a good idea. There may be ripple somewhere.

[Dek]

Recaped the floating PSU - No change

Before the recap, when checking the 33hz / 50hz noise, I have my oscilloscope measuring the input to the meter to monitor the signal from the sig gen.
I removed the sign gen but still had the scope attached to the input and could see 50hz. I'll re do this test to get the values.
Not sure if this is real or just the scope leads picking up mains hum.

Dek.

[Kleinstein]

The 200 µV of 50 Hz Signal should be hardly visible on the scope. However after amplification at the ADC input the signal should than be very visible.

A possible point of failure would be the input protection, e.g. a series resistor to limit the current that could be high resistance. Poor relay contacts may be possible too. Noise could also come in from a common mode voltage, that is input negative side to case. Normally the shielding suppression should better, but it can still help to have an eye on the case shields.

[grizewald]

While I was in that area, I noticed an additional earth lead that Grizewalds 7061 doesn't have.
Looks like an addition ??

I was doing a little scouting around the Internet for more pictures of the insides of the 7061 and found something
interesting regarding your extra wire.

I found this picture on www.pa4tim.nl and it shows an early 7061 with PCBs 3 and 5, just like yours:



It seems that the extra wire from the analogue board's PSU ground going to a spring copper finger in the middle of the case which connects to the brown conductive shielding was indeed a feature of your revision of the meter but was removed on the later models.

I wonder why?

[Dek]

I was doing a little scouting around the Internet for more pictures of the insides of the 7061 and found something
interesting regarding your extra wire.
I found this picture on www.pa4tim.nl and it shows an early 7061 with PCBs 3 and 5, just like yours:
It seems that the extra wire from the analogue board's PSU ground going to a spring copper finger in the middle of the case which connects to the brown conductive shielding was indeed a feature of your revision of the meter but was removed on the later models.

I wonder why?

That is interesting, at least it's not some hack by a previous owner.
The lid of my 7061 has a piece of PCB that sits above the analogue side, and grounded to the main analogue pcb (not mains earth) by the copper spring.
As to why, perhaps they determined it was better to let the the noise out of the meter than keep it inside  ( well on mine anyway).
Checked the +-36v and +-15v rails again, all bang on with ripple less than 1mv.

I'm now struggling to think of a way forward.

Time to fit the new mains filter, it's doing no good in the mouser bag.

Dek
 
<edit>
Grizewald,  reading your earlier post regarding the recap, I have only changed the larger PSU caps.
Possibly a way forward is to replace all the smaller caps too.

 

[grizewald]

I think having a look at what the A/D circuit is actually doing may prove to be more useful. See what the waveforms on your machine look like compared to the reference waveforms in the manual, both when measuring 0V and when measuring a battery. Also, like Klienstein suggests, track the input through the input protection circuitry to the input amplifier and then through that to see if you can see where the 50Hz is entering the circuit.

I made a start on removing the smoothing capacitors on my digital board last night, but I was a bit tired and ended up destroying a plated through hole to the ground plane. 

So I'll leave the rest of that job until Saturday morning when I'm rested and have natural light to work with.

[Dek]

Hello Grizewald,

I've started tracing the I/P as Kleinstein suggested.

Using my Fluke 187.
2 AA batteries as the input  'reference'
Meter set 10v range 7dig no filter.

Meter grounded to the LO side of the spark gap.
Measure the DV voltage and the AC mV component.
Without any kind of base line as to what level of ac I should be expecting ( I've limited experience on these meters to say the least !) comments are welcome.  :-)

I'll work through the IP amplifier next.

[ Specified attachment is not available ]

[Kleinstein]

The DC readings should reflect the voltage drop due to the resistance and ~ 10 M input resistance of the fluke 187. So the DC part looks normal.

The AC numbers are a little confusing. It looks a little like AC coming in from the probes (e.g. capacitive coupled) and the different reading may reflect the impedance seen at the different points. So the lower impedance points see less AC.  In this concept the readings make sense and relays seem to work and the resistors seem to be no open.

Measuring the real AC without coupling in extra hum could be tricky. The Level seen before was at some 200 µV, so much less than read here. 

It the bootstrap /guard signal looking OK ?  If wrong this could enter through the JFETs.

Another point to test could be input bias current (with a high resistor or drift with a 10 nF range cap).

The gate drive with opto-couplers is a nice solution. However those opto's may age, not turning on all the way and thus possibly not turning off the JFETs all the way.
A lot of AC signal from the AC unit could be a problem too. So it the output from the RMS part OK ( low voltage, not much AC) ?

[Dek]

The DC readings should reflect the voltage drop due to the resistance and ~ 10 M input resistance of the fluke 187. So the DC part looks normal.

The AC numbers are a little confusing. It looks a little like AC coming in from the probes (e.g. capacitive coupled) and the different reading may reflect the impedance seen at the different points. So the lower impedance points see less AC.  In this concept the readings make sense and relays seem to work and the resistors seem to be no open.

Measuring the real AC without coupling in extra hum could be tricky. The Level seen before was at some 200 µV, so much less than read here. 

It the bootstrap /guard signal looking OK ?  If wrong this could enter through the JFETs.

Another point to test could be input bias current (with a high resistor or drift with a 10 nF range cap).

The gate drive with opto-couplers is a nice solution. However those opto's may age, not turning on all the way and thus possibly not turning off the JFETs all the way.
A lot of AC signal from the AC unit could be a problem too. So it the output from the RMS part OK ( low voltage, not much AC) ?

Thanks for the reply Kleinstein,

Measuring the bootstrap WRT 0v (spark gap low side to R9 R10 R11 Node) 3.19VDC 0.30mV AC 100.00 Hz
The guard is interesting! 0v DC 1.89V AC 50 Hz

I pulled the link plug PL6 to isolate the AC converter, appeared to make no difference.

All the readings are fairly repeatable.
If there is a better measurement method I'll try with the instruments I have - 200mhz scope, Racal 1998 counter and DMM I have a HP 3457 I can bring in from work.

Dek.
 

[Kleinstein]

300 µV 100 Hz AC signal at the bootstrap signal is odd, as the real signal measured some 200 µV 50 Hz, though with some harmonics. At least the DC level seems to be OK. 
From the gate drive there is some load from the -36 V to the bootstrap signal. This could be a way 100 Hz could come in.

The guard (terminal) may be just floating and picking up hum.

For a test one could use the scope at the ADC input / amplifier output with the meter in the 100 mV range. With a gain of 100 I would expect some 20 mV of hum visible. One could than try a short at the OP1 testpoint (11.3 mV AC noted there) and maybe a few more points to see how much this reduces the hum. Similar one could check in the current range for the hum visible.
It is still not 100% clear if the hum is more like capacitivly coupled to the input (would have to be quite a bit) or an error from the amplifier itself (e.g. via the supply to the main amplifier), or the ground point for the divider.

There may be quite some common voltage. Is there something like a shield winding in the mains transformer that is not (or wrong) connected ? One could try measuring the common mode voltage / coupling. Ground would be via scope to PE and the interesting point the neg side input, guard and maybe a few case / shield parts if not directly connected.

Could it be that one should have a connection to the guard (e.g. to the negative input).

[Dek]

300 µV 100 Hz AC signal at the bootstrap signal is odd, as the real signal measured some 200 µV 50 Hz, though with some harmonics. At least the DC level seems to be OK. 
From the gate drive there is some load from the -36 V to the bootstrap signal. This could be a way 100 Hz could come in.

The guard (terminal) may be just floating and picking up hum.

For a test one could use the scope at the ADC input / amplifier output with the meter in the 100 mV range. With a gain of 100 I would expect some 20 mV of hum visible. One could than try a short at the OP1 testpoint (11.3 mV AC noted there) and maybe a few more points to see how much this reduces the hum. Similar one could check in the current range for the hum visible.
It is still not 100% clear if the hum is more like capacitivly coupled to the input (would have to be quite a bit) or an error from the amplifier itself (e.g. via the supply to the main amplifier), or the ground point for the divider.

There may be quite some common voltage. Is there something like a shield winding in the mains transformer that is not (or wrong) connected ? One could try measuring the common mode voltage / coupling. Ground would be via scope to PE and the interesting point the neg side input, guard and maybe a few case / shield parts if not directly connected.

Could it be that one should have a connection to the guard (e.g. to the negative input).

With my 2xaa batteries connected to the IP 3.191v

Multimeter connected 0V to TR3 (the 50Hz 5.45mV point) shorting OP1 to the input 0V reduces the AC by about 1/2.
Multimeter connected to 0v and the integrator IP (handy test link ) 3.191Vdc and 7.75mVac noise. Shorting OP1 makes no change to the noise but drops the voltage to 1.0638vdc
(may be a good test to split this link and check circuitry before and after the link to check which section the noise is from.



With scope ground connected to the meter PE, probe to the guard terminal - screen shot above.

Scope ground  0V to bootstrap 3.2vdc but noisy -screen shot below.


PE to the IP 0v measures 4.905
PE to + IP 8.096
Screen grab below.






I have the guard terminal not connected, just 4 wires.
The manual states that the standard Solartron cable and connector set does not use the guard pin
Dek

[Dek]

Some progress:-

Took out the link TL101
With Fluke multimeter  from 0v to TR3 and see the previously measured 7 to 10mv AC
7061 meter set on the 0.1v range and IP socket shorted (excluding pin 5 guard)
The 7061 display fluctuates about ~2 counts of the last 7 digit (approx 0.000,000.5)
Temporarily re-connect TL101 and the display fluctuates 2 or 3 counts of the 5th digit.  (approx 0.000,035.1)

Same test with meter set on range 1
TL101 link open, the display fluctuates 2 counts of the 6th digit. (approx 0.000,006.4)
Connect TL101, display fluctuates 6 counts of the 5th digit.        (approx 0.000,056.1)

Same test range 10
TL101 link open, the display fluctuates 1 count of the 5th digit. (Approx 0.000,024)
Connect TL101, display fluctuates ~3 counts of the 5th digit.  (Approx 0.000,047 )

Range 100
TL101 open 1 count on the 4th digit
TL101 closed 3 counts on the 3rd digit.

[Kleinstein]

7.7 V AC at the integrator input would be a real problem.
Even with 14 mV RMS at the input a gain of 100 should only cause 1.4 V. So there seem to be something really bad with the amplifier.

The reduction of the DC voltage to about 1/3 with OP1 shorted is the expected value, so the resistors seem to be OK.
With the AC part only going down by a factor of 2 (and not 3) is a little odd. So AC wise there seem to be some extra impedance - like some 30 nF to ground, so not just a little, more like part of the filter at the amplifier.
Opening the link to the integrator may make some sense,  though I would not expect much noise from the integrator (should be rather high impedance anyway).

Besides the 50 Hz hum, there also seem to be quite some higher frequency background of a few MHz.

The ICL7650 used in the main amplifier can no be used along - it would need some extra support / bootstrapped supplies, that could go wrong.  One could also check at lest the normal amplifier function with something like a 150 Hz  100 mV or so square wave signal to the input. Chances are things are somehow not working perfect there.

The noise tests with the open close link suggest that there is quite some noise coming from the amplifier. Some in expected for the 100 mV range, but not much for the higher ranges. It is odd to see different noise in the different ranges also with the link open. In this case there should normally no much effect of changing the gain.

[Dek]

7.7 V AC at the integrator input would be a real problem.
Even with 14 mV RMS at the input a gain of 100 should only cause 1.4 V. So there seem to be something really bad with the amplifier.

Hi Kleinstein,

Sorry my typo should be 7.75mV AC (scope shows this as noise and also spikes that align with the rising / falling edge of the glugs)
The guard appears to be floating, no continuity to PE or any other part of the board, so is acting like an antenna when looking at the trace on the scope.

Dek.

[grizewald]

I have the guard terminal not connected, just 4 wires.
The manual states that the standard Solartron cable and connector set does not use the guard pin
Dek

Don't believe the manual!

I don't know how they got this wrong when writing the manual, maybe some miscommunication between the engineers and the technical writers, but the guard is most certainly used. In the standard Solartron cable with just + and - inputs, all five wires are used. The +V and Ohms Hi wires are joined just before the wires exit the shield, as are the -V and Ohms Lo wires. The guard is bonded to the -V Ohms Lo pair just before they exit the shield. Furthermore, the guard input is connected to the transformer from the analog board on pin 4 (yellow wire) of the connector at the rear of the analogue board. The yellow wire goes into the transformer windings, suggesting, as Klienstein correctly assumes, that there's a shield winding inside the transformer.

Not only that but the guard goes to a track which runs around the entire circumference of the analogue board which is connected to the test point labelled "guard" at the front left of the analogue board. Note though that the guard wire from the inputs is not connected to the rest of the meter unless the meter is powered on and RL6 is energised.

I hope you haven't been chasing ghosts caused by the guard input being left floating!

Oh, and the guard is connected to the screen of the measurement cable at the far end (furthest away from the meter) The shell of the input connector is earthed at the meter, but the shield is not connected to the shell.

[grizewald]

For some bizarre reason, the leads are all described in detail in the 7081 user manual but not at all in 7061 manual. Here's the relevant description of the two terminal lead from the 7081 manual.