Tektronix TDS 620 still failing after recap.

[shakalnokturn]

Hello all,

When I bought this TDS 620 it was failing:
++Processor
++Acquisition
++Attn/Acq Interface

Opening it revealed the usual leaky SMD electrolytics, so it went through the usual process of removing the old caps by twisting them off, cleaning pads, dishwasher with soap, oven, fitting new caps, cleaning and repairing tracks in spots that had heavier corrosion.
There may still be cut tracks but the corrosion did not look as serious on this scope as on the two previous TDS (500) I had worked on.

I also dumped and socketed the DS1245Y NVRAM. (File attached.)

Since the recap I have absolutely nothing working any better, bedsides the previous fails I also now get a fail most of the time on:
++Cal Initialisation

I got hold of the A10 board schematics on Håkan's resource page: http://www.hakanh.com/dl/docs/hardtofind/TDS644A_A10(Acq).pdf
(Thanks Håkan!)

Some probing around last night revealed that the power supply values mostly look good, I measured across all the SMD electrolytics with a DMM.
Voltages may have been a little low at C590, C1710, C1603, C1693, C1016. I have not yet checked whether there was a good reason for this. (Low value series resistor?)

On the screen both traces give what looks like noise, not exactly the same on both though.
The V/Div and offset controls have no effect whatsoever, also the "TRIG'D" LED never comes on.

The offset control not working got me to start probing at the AD667 DAC. (It is outputting the various control voltages and they move around when adjusting V/Div and offset. I'm not sure the AD667 reference is spot-on, I measured 9.988V at TP901, however this should not stop the scope from working, possibly make it fail cal.

I followed through the offset controls to the U1400 preamp where they look OK. (Approx +/-1.7V maximum.)
The input gets through the FE attenuators (I'm not saying they work perfectly yet, just trying to get a trace on screen...) then to the preamps U1300, U1400.
The preamp differential outputs still carry the signal to the "FISO" CCD (U850), then the output to the ADC is just plain garbage, similar enough to the crap displayed on the screen.

This is my first experience with a CCD based scope and I'm a little lost at this point as to what to expect on different pins of U850, to check if it is defective or just missing some clock/control line.

I'll get a couple of photos up later.

[andy2000]

It would be helpful to see the error log for more details on the errors.  From experience, it's probably caused by more open traces, but other problems are possible.  My last TDS544A had a bad clock driver IC which was causing a bunch of seemingly unrelated errors. 

Sometimes the electrolyte can seep inside components.  Also, they are more than 20 years old, so problems not related to the leaking caps aren't unlikely. 

You need to verify the continuity of all traces that were exposed to electrolyte, even if they don't look corroded.  Pay particular attention to traces that go under components to connect to a pad. 

[shakalnokturn]

I've been checking over traces in the corroded spots, haven't found anything wrong yet.
I have started investigating the trigger section:

The +/-TRIG(x) outputs work on all four preamps when selected as trigger source, they all reach U1551 and U1552, there is no control over TLA or TLB adjusting trigger level though.
Checked all the links to U931 and even replaced it just in case, no luck.
There is variation on TLA, TLB during start-up so the most likely is that the software is blocking the trigger due to one of the self-test failures.

I've attached a screenshot of the error log, hopefully someone can make a little sense of that.

[rf+tech]

shakalnokturn

Thanks for recently linking to this thread from “TDS 620A Acquisition failure.” I have an observation to share that may be of benefit.

Your error log provides a clue from the following lines:

Code: [Select]

   Dac Range Test, **a4 <= exp <= e4 – actual – 84
   Dac Range Test, **a6 <= exp <= e6 – actual – 86
   Dac Range Test, **a5 <= exp <= e5 – actual – 85

The actual value low-byte matches the expected, but the high-byte is always stuck at the same value.

The expected high-byte values written out in binary,

Code: [Select]

  a 1010
  b 1011
  c 1100
  d 1101
  e 1110

  8 0100

Comparing the expected binary values of a through e, to 8 in binary, shows the MSB appears to be stuck low or missing. If the missing bit were 1, then all three results would be c4, c6, c5 and within the expected range.

I had similar errors and patterns (f in the high bytes of expected values) that were tracked down to open data lines. Hairline open circuits at the solder mask opening, on the edges of IC pads, in close proximity to leaked electrolytic capacitors. Ended up removing many ICs for visual inspection with a microscope, and repaired many tracks.

The visible clues are darkening of copper tracks *under* the solder mask from electrolyte ingress. Failure to remediate all such instances of damage now, will leave the scope susceptible to future failures from trapped electrolyte under the solder mask. The dishwasher treatment cannot prevent such further damage.

It is necessary to scrape away all solder mask over affected areas, thoroughly clean exposed copper tracks of all corrosion darkening with a fiberglass scratch pen, clean with IPA, then reseal cleaned areas with a suitable overcoat, such as Circuit Works CW3300G or CW2500. The latter product is an epoxy suitable for securing tiny soldered jumpers.

CW2500 is formulated to withstand soldering temperatures, it does not soften or melt like two-component epoxy from big-box stores. CW2500 is not cheap; measuring equal volumes is not easy due to dissimilar viscosities of the components, mixing must be done for 2.5 minutes and requires curing at 150F.

An example of some damage to U1067 (ten defects - all on one side) and how I successfully repaired them on a TDS640 Processor board. Those jumpers are #38 copper wire (.004” diameter), rolled flat, secured with high melting point solder and fixtured with CW2500. ICs were reinstalled with conventional 63/37 alloy solder. Some bending of leads was required to contact pads.

Hint: to better assure equal quantities of CW2500, warm the thicker component B so it flows out into a puddle, then measure out an equal puddle of component A - which flows almost too freely.

RF+ Tech

[shakalnokturn]

Thanks a lot for your useful information rf+tech.
I can't say I'm much good at making sense of the TDS error logs.

How is the DAC test carried-out? Is one of the MUXed outputs sampled by a dedicated ADC on the CPU board, or is it checked through the FE and scope channels acquisition system?
I haven't been able to follow this through on the schematics I have.

I wouldn't have expected a stuck MSB to allow for what seemed to be the normal +/-1.7V output, based on previous measurements and your pointing to the error log I would have thought the DAC was ok but a possible stuck MSB on the ADC side of the self-test loop could be the trouble.

The scope is in storage at a friend's place for now, I'll check on the DAC getting all data lines on the next occasion.

[rf+tech]

Acquisition Processor U600 (schematic A10-11) sends 8-bit control data to DAC U900 (schematic A10-12). Follow the analog output from U900 pin 9 (Vout) to U903A to TP912. Continue on to schematic A10-13, U933 CD4051. Analog voltage from MUX goes through each Probe Control Buffer, to schematic A10-16. ProbeOffesets then go to the Attenuator board. Self-test commands the DAC to run the analog voltage up and down, looking for a response via ADC U750 and U700.

That’s just one path. Refer back to A10-12, DAC U900 analog output also goes to analog switch U908 where four additional paths are found.

Which ever path, the DAC provides a stimulus, the ADC measures the response and pass/fail is decided, probably by the A11 Processor board (could be the Acquisition Processor). Any IC attached to the 8-bit data bus, in close proximity to electrolytics, presents an opportunity for concealed trace damage.

Easy enough to check data bus line continuity to each device, hint - make a spreadsheet. This is how I confirmed which ICs were affected in my TDS640. To locate the first open circuit, I resorted to injecting 100 MHz RF at 200 mVrms at one IC location onto a known open circuit data line, then sniffed the RF radiating from the board with a spectrum analyzer. Signal went under U1067, but could not be found coming back out, where it should have continued on to additional ICs.

I have to wonder how many TDS failures remain unresolved, that may be explained by this failure scenario I’ve documented.

Whenever you resume working on your TDS620, also run through the A10 Diagnostics -> A/D Converter procedure, looking specifically at the ECL clocks and for data aligned with clock rising edges. Bit by bit, confirm all inputs to ADC U750 and U700 are working. Get the board under a microscope with strong light and check for further corrosion. Especially between IC pins - if corrosion is seen, remove the IC.  With the IC on edge under a microscope, look closely at the epoxy, where the lead enters. Check the underside of the leads at the body, for signs of corrosion ingress into the IC body. As cheap as the off-the-shelf parts are, replacement is cheap insurance. There’s a boat-load of TL072s  and 074s used, any leakage at the inputs can introduce offsets that cannot be nulled during SPC (attenuator DC balance) and/or calibration.

Do post back to this thread, I’ll be interested in what you discover.

One more thing, do build a serial console adapter. With this, you'll be able to see complete log line details, that otherwise run off the scope display, and capture the complete boot sequence.

RF+ Tech

[Smoky]

My Tektronix TDS420 suffered from the same problems relating to the corroding of the copper traces as a result of the leaking electrolytic capacitors. I repaired many breaks by doing continuity tests to locate the problem areas. the scope came to life and has been running fine for the past two years or so.

I was given this Huntron Tracker from a retiree not too long ago. Recently, I began to read about how it works, and it seems as though it can be very useful to locate these sort of problems on a PC board very quickly by comparing signatures from points on a known good board against the same points on a faulty one. I turned the machine on and pressed a button and what you see on the screen alternates, showing the unique signal coming from each of the two probes. When the two responses aren't the same, you've found a fault.

I have a Tektronix project coming up where I do have a parts machine with extra boards. We'll see how well it works.

If you do this sort of work often, it may be a worthy investment.