TDS744A: 1M input impedance measures: Ch3 = 160Kohm, Ch4 = 706KOhm [resolved]

[John_ITIC]

Hi, I'm working on refurbishing a TDS744A. It failed SPC so I replaced the hybrid relays on Ch3 and 4.

Afterwards, SPC passes 100%. All channels display a 10 MHz sine wave perfect in 50 Ohm mode.

However, there is a DC offset when in 1M input impedance mode. When measuring the input impedance with my ohm meter, I see:

Ch3 = 160Kohm, Ch4 = 706KOhm

Ch1 and Ch2 both measure 1MOhm in 1M input impedance mode.
All channels measure 50 Ohm in 50 Ohm mode.

Ch3 and Ch4 both have DC offset in 1M impedance mode. Both have malformed calibration signal waveforms. See attached images.

I did not have this issue before the hybrid relays were replaced. I'm quite sure nothing was overheated or damaged. Original relays were purchased from Mouser.

Where is the 1M input termination actually located? I was not able to locate any on the hybrids and none of the manuals I have found on the internet shows an actual 1M terminator.

Thanks,
/John.

[BlownUpCapacitor]

How are the relays replaced? If they're soldered, might just be some solderball.

These more modern Tek service manuals aren't nearly as thorough as the older ones, so I have no idea what the hybrid looks like. Maybe provide some images?

I would first think bad contacts, but bad contacts normally increase resistance; in this case, we have a decrease.

Currently, I think solder balls are of suspect.

[TERRA Operative]

Sounds like a leakage path to ground somewhere.
I'd inspect with a magnifying glass or loupe etc, and also wash thoroughly with isopropyl alcohol too.

I'm away from home for a few days so I can't check my notes, but I did have the relay functions written down.
I'll check when I return.

[John_ITIC]

Thank you both for your suggestions. I have not seen anything unusual in my visual inspections under microscope.
All relay solder joints are perfect and clean. No visible damage to the hybrid circuit board.

The hybrids used in my TDS744A are H2462G's: https://w140.com/tekwiki/wiki/H2462
The repair procedure followed was described here (thank you Terra):

In the above tekwiki link, there is a block diagram of the hybrid. I see no control line for 1M input impedance configuration. So, if not 50 Ohm, terminated in 1MOhm on the hybrid? The same tekwiki link has a detailed description of the M753 amplifier used on the hybrid board. The initial picture has a block diagram of the x1, x10 and x10 resistive dividers used. Those resistive dividers must be the 1M input terminators.

https://w140.com/tekwiki/wiki/M753#/media/File:1990-08-23_New_Attenuator_scheme_m753.jpg

I have confirmed that the below are the DC offsets on Ch4 for the different gain settings. The x1, x10 and x100 are the different resistive divider inputs on the M753 amplifier.

x1 (10V/div, 5V/div, 2V/div) : 40mV offset, 700 KOhm input impedance
x10 (1V/div, 500mV/div, 200mV/div) = -240mV offset, 225 KOhm input impedance
x10 (100mV/div, 50mV/div, 20mV/div, 10mV/div, 5mV/div, 2mV/div and lower) = -13mV offset, 680 KOhm input impedance.

I will take the unit apart again to see if I can identify any resistive dividers on the hybrid board. Perhaps I can confirm the above resistances.

Thanks,
/John.

[John_ITIC]

I am suspecting that soldering the relays in may have heat damaged the three 1M voltage divider resistors. This post lead me to do some more experiments:
https://www.eevblog.com/forum/testgear/tektronix-tds520b-attenuator-swap/

Sometimes, input impedance of CH1 is 170ohms in 1Mohm mode and 38ohm in 50ohm mode as if a 170ohm resistor was in parallel with the input. Apart from this, attenuator behaves correctly.
The strange thing is that this fault disappears when I apply some DC voltage (12V for example): after that, the input impedance is back to normal (1Mohm or 50ohm) until powered off.

This lead me to try this experiment myself. When I apply some DC voltage to the 1M input (some 19V), I'm seeing that the input impedance rises several hundred Ohms. I can even go over 1MOhm. I then measure the input resistance with a couple of DMMs, one show around 1MOhm while the other show some 800 KOhm. The values drift towards each other as I swap meters. It looks like the meter voltage is different and that the input resistive divider reacts to this change in meter current.

I'm seeing the same behavior on both Ch3 and Ch4 that I swapped out the relays on. Ch1 and Ch2 are stable across temperature and across multimeters (both show 998 KOhm input resistance). Ch1-2 do not get affected by applying a DC voltage.

So, has anyone seen similar issues where soldering iron heat has permanently affected the temperature coefficient of resistors in such a way?

Note: I'm also seeing on Ch3 and Ch4 that setting the coupling impedance to GND still measures 1.7MOhm on Ch3 and 17 MOhm on Ch4.
Both Ch1 and Ch2 measure infinite resistance when set to GND.

So, it seems there is some shunt resistance that is related after all. But where?

Edit: I'm curious how the GND setting is implemented. Per the below diagram, the resistive dividers can either be set to the input or to the cal signal but not be disconnected from both. The relays are 2-way relays (DS1E-M-DC12V), with one common, one normally connected and one normally open terminal.

https://w140.com/tekwiki/wiki/M753#/media/File:1990-08-23_New_Attenuator_scheme_m753.jpg


Thanks,
/John.

[andy2000]

I've replaced the relays on several of these without any problems.  You do need to be careful not to lift the silver pads from the board, but otherwise it wasn't too hard.  Maybe there's flux residue causing some leakage? 

[John_ITIC]

I've replaced the relays on several of these without any problems.  You do need to be careful not to lift the silver pads from the board, but otherwise it wasn't too hard.  Maybe there's flux residue causing some leakage?

No, the hybrids are perfectly cleaned after the soldering job. I did notice that the original Tektronix solder paste was barely melted as I could see the original paste granules under my microscope after the relays had been cut away. My theory is that Tek knows that something is heat sensitive and that they were really careful with the applied temperature.

On the other hand, my theory does not make sense really as I would think that reflowing the whole hybrid board would soak the board with much more heat than a small / quick solder pen dab with heat would do...

Thanks,
/John.

[John_ITIC]

Okay, this issue has now been resolved. It turns out that the solder flux I used (Kester 2331-ZX) had been absorbed into the hybrid substrate. The substrate became conductive, which formed a resistive path in parallel with the 1M input impedance. I found that submerging the hybrids in IPA and scrubbing with a brush did not reverse this unwanted conductivity. The only thing that worked was to soak the hybrids in flux remover, scrub, rinse in boiling hot water, scrub some more and blow dry.

This is the flux I'm using:
https://www.kester.com/DesktopModules/Bring2mind/DMX/API/Entries/Download?Command=Core_Download&EntryId=3679&language=en-US&PortalId=0&TabId=96

This is an "organic acid, water soluble" flux and it is stated that hot water should be used for cleaning. There is no mention of flux remover or IPA. I'm not sure what the ideal cleaning procedure is.

My recommendation is to measure the resistance between the Hybrid's BNC input pad and neighboring GND tabs before soldering back into the A15 attenuator assembly. The input relays are by default in the GND mode (input floating) so there should be infinite resistance. If you have any resistance, even tens of MOhms, it means you still have contamination on the board. The black relay is responsible for the DC/AC/GND setting and it is set to GND when unpowered.

My scope now passes SPC and there is roughly 0.250mV offset between the channels when in 1mV/div setting in 1M input impedance setting. I'm not sure how to actually use the 5mV/div, 2mV/div and 1mV/div setting because the scope will not allow me to select those fine vertical resolutions when I'm plugging passive probes in (min vertical resolution then becomes 10mV/div).

So, I consider this repair successful. Hopefully it will help someone else doing the same repair.

Thanks,
/John.

[andy2000]

I recommend avoiding "organic acid, water soluble" flux unless you absolutely need to use a flux that can be cleaned with water.  It's conductive (as you found out), and corrosive if not cleaned completely off.  It's not very soluble in the solvents used to clean rosin flux off such as alcohol, or acetone.  It's not a good choice for a general purpose flux.  It's meant for situations where the solvent needed to remove rosin flux would cause damage, or for mass production where it's desirable to be able to clean boards with water rather than volatile chemicals that are expensive, or banned. 

[TERRA Operative]

I use standard rosin flux cleaned off with alcohol and never have a problem. I'll be sure to avoid organic fluxes on ceramic substrates from now on though.

Also as a side note, the thermal paste on the hybrid module is important, so make sure you replace it after washing or you'll burn out the chips on the hybrid module!

[bdunham7]

I'm not sure how to actually use the 5mV/div, 2mV/div and 1mV/div setting because the scope will not allow me to select those fine vertical resolutions when I'm plugging passive probes in (min vertical resolution then becomes 10mV/div).

If you connect 10X probes w/ readout or select 10X manually for non-readout probes, the vertical scales are all multiplied by 10.  So with a 10X probe, the apparent 10mV/div setting is actually the 1mV/div setting to the scope.

[fenugrec]

or select 10X manually for non-readout probes,

AFAIK that cannot be done on the TDS7* scopes, probes without the readout pin are assumed to be 1x.

[David Hess]

This is an "organic acid, water soluble" flux and it is stated that hot water should be used for cleaning. There is no mention of flux remover or IPA. I'm not sure what the ideal cleaning procedure is.

That would explain it.  I tested organic water soluble flux many years ago and rejected it because of the cleaning requirements outside of production.  I did not run across the leakage problem, but in retrospect that seems obvious.