EEVblog #565 - Tektronix TDS3054 Oscilloscope Repair - Part 2

[EEVblog]

Part 2
Dave follows a more methodical troubleshooting procedure in order to track down the fault in the Tektronix TDS3054 oscilloscope.
Measuring the low and high frequency differential outputs of the hybrid front end, the voltage rails and other parts. And has the magic smoke escaped?
And he couldn't resist playing with his new Flir E8 thermal imaging camera .

[Monkeh]

I think I'd probably fix that diode and try swapping ADCs.

[Erwin Ried]

The "Repair" in the title is usually misleading on the eevblog 

[sync]

I would measure the resistance of the diode (anode and cathode) to the -2.5V rail. Maybe there is a dead short.

[vikpc]

im think CP140 is Coupler

[trackman44]

I think I'd probably fix that diode and try swapping ADCs.

I would repair the diode first. Then, if that didn't work, reflow all the pins on CH3 ADC. There is the small possibility that a trace on the PCB is cracked, so a continuity check between the hybrid module and ADC, or the ADC and glue logic chip is in order.

Will

[EEVblog]

There is the small possibility that a trace on the PCB is cracked, so a continuity check between the hybrid module and ADC, or the ADC and glue logic chip is in order.

Checked that, all ok.

[marshallh]

The diode pair is probably for input clamping on the diff pair going to the ADC. It is a last ditch defense against some transient blowing the ADC inputs.

I could be wrong, but the fact that it got nuked means somebody probably did something stupid with that input and boogered the ADC itself.
If the ADCs are non-RoHS as they should be, you can swap them with hot air and some swearing.

[vaualbus]

At this point the chances go get it repaired are very few, Although it still be possible to repair it. As far it could also be a calibration data corruption of the channel 3. You could let the scope run spc and see what happen. You need to still check the second most important thigh for the adc, the adc clock. The adcs are syncronized to a clock generated in the board, so if the clock is not perfect it could get some error. It also could be a memory corruption like in my tds540. The diode are securely protection diode. There are more chance to repair the hp dsa at this point. It more probably to be a software problem, try to restore the initial condition and check the nvram production date. It work fine for more than ten year but then fail and could corrupt the calibration data that contain. Also check the bypass capacitor.  On my tds540 (it have a similar custom ceramic hybrid for the front end) just for the bad cap the scope not pass spc. I wish to see more videos on both repair. You should also consider the problem to be a board problem like resistor or capacitance.

Probably the center chip is the custom trigger logic and hybrid.

[EEVblog]

And for those who have asked, no it's not the voltage rail ripple / decouplers. Looks exactly the same as the other channels.

[Jon86]

As a quick last attempt, maybe stick one of the SOT-23 packages from another channel into channel 3? The voltages on those pins seems really quite strange.

[mikeselectricstuff]

Scope on the diode on a good channel should tell you if it's for power or for signal clamping

[wn1fju]

I'll relate my recent experiences fixing my Tek TDS350 scope.  Even though it is a generation (or two) earlier than the TDS3054, perhaps what happened to me may offer a hint.  I had a very flaky second channel that completely disappeared at V/div settings < 50 mV.  I stuck an ohmmeter across the BNC input jacks and measured the expected 1M on channel one, but only about 600K on channel #2.  On a hunch that maybe there was some voltage present at the jack, I switched the DMM to volts, only to find about 7 VDC coming out of channel #2's BNC.  I'm still not sure how since the hybrid preamp daughterboard has Vcc = 5V.  Anyway, the 7 volts dropped down significantly with even the slightest load (100K) on it.  So my conclusion was that hybrid board #2 was toast.  Fortunately, I was able to cheaply procure another one which ultimately fixed the scope.  And yes, great fun removing the 45-pin input board!  But what I really want to report is that at least in the TDS350, nothing really works correctly until ALL the calibrations have successfully been made (Compensation, Vert, Horiz, Trigger).  For instance, even though I thought channel #1 was fine, the variable V/div control was non-functional, the baseline traces were all over the place, the trigger point was incorrect, etc.  Even after I cal-ed the Vert, Horiz and Trigger (with the replaced input hybrid #2), it still didn't work right.  Only when the final SPC calibration was done did the scope work satisfactorily.  While the subject TDS3054's fault may indeed be hardware, keep an eye on the calibration if and when you fix it!

[vikpc]

And for those who have asked, no it's not the voltage rail ripple / decouplers. Looks exactly the same as the other channels.

but it marked as CP not a CR... and this CP140 is cracked... this is diffecence from other channels.. i try track wires on your video on both sides..

unfunfortunately youtube HD is not enought for tracing....

[Co6aka]

Very unlikely the problem here, but you freeze-misting the thing reminded me of a "one for the books" experience...

The short version: I had an insidious electronic device in for repair several times; each time it worked perfectly on my bench, and would then work for a couple of hours to a couple of days when it was back with its owner. I finally tried freezing it, and lo and behold it would go into reset, but NOT not when it was just sprayed (i.e. "still frozen") As soon as the frost cleared away, "CLICK" -- off it would go. But then after a while it would start working again, and I found that thawing it out with the hair dryer caused an almost immediate "fix." That's when it finally clicked it my mind -- the "reset on clearing frost" feature should have snapped me out of my stupor. So then I tried exhaling on it (I have dog breath, obviously) and of course "CLICK!" it reset. Eureka!!!

SO what was the problem? Well, that insidious device's owner lived at the oceanside, so humidity had deposited a dusting of salt on the PCB, and with a tiny bit of humidity/dampness it became conductive. I cleaned up the PCB real good and applied a conformal coating, and that was the end of that.

Anyway... The "noise" on the screen looks like digital crud getting into the ADC somewhere. Perhaps a bad bypass MLCC somewhere? Or a bad ground? Also, it seems the noise switches polarity at neg-something-Volts -- did you notice that?

(GO TO 06:20 IN YOUR VIDEO. Notice how the noise is only either above or below the injected sine wave.)

[Hydrawerk]

The main ASIC is too hot. 100°C is a high temperature. It might be dangerous to operate it without a cooling fan near. And I would expect a heatsink, too. Heatsinks are very common at most scopes today. Agilent DSOX2000, Rigol DS2000, GW Instek GDS-2000A...

[jonwilhelmjr]

This might not be relevant but I noticed something on the HF noise. As a novice comment, Why is the noise respective to a voltage? Why is the noise on the top of the waveform until some negative voltage? BTW is that 1KHz sin AC or DC coupled, does it matter? Then on the square wave, why is the noise not consuming the entire screen? Is there a DSP used in this device, if so what is going into it and coming out. I understand the ADC converts volts at some time to binary, is the Analog signal OK going in to the ADC. Is the AC/DC coupling working right. And last thought how does the waveform look for 1x verse 10x on the scope probe or the setting in the scope. Where i am getting at for this is when the relays will click in and out based off the voltage. More importantly what I am getting at is when does the problem occur based off: sin wave, voltage, time, trig points, complex waves, square wave, saw tooth, or i.e. run a full test or hell cal test (need to do this any how for the other channels).

I can't wait to hear what the more knowledgeable thinks.

Thanks,
Jon

[parbro]

I would take a look at U330 near the blown diode. The chip is marked A00A and appears to be an LMC7101 5-pin SOT. Maybe whatever took out the diode screwed up the op amp. When you heavily sprayed the channel 3 ADC with coolant the scope showed an offset followed by reduction of high frequency noise.

[free_electron]

Good, so the hybrid is OK

it looks like the supply voltage for the input stage is out.
so the first stage is running into its limits. hence the 'hair' you see

Parbro may be on to something. i'd check that 5 pin sot23 part too….

[calexanian]

This is a good replacement for the TV shows being on mid season break! Tune in for the next episode of will it be fixed before something happens and it breaks even more!  O0

[AndyC_772]

There's a major clue in the first video, in that the trace disappears when you switch from 10k points mode to 500 points mode. What possible reason could there be for that to happen if the fault were with the analogue front-end?

For what it's worth, the dc offset in these scopes is known to be temperature sensitive, so I'm not at all surprised you saw an offset when you blasted the board with freezer spray.

Check on one of your other, working TDS3054s; with no signal applied, the dc level you see isn't 0V until the scope has properly warmed up. I've seen it out by about 1/4 of a grid square on my TDS3034B, but it soon drifts into spec once it's warmed up. It's why you have to wait until the scope has been on for a while before running SPC.

Have you tried running SPC?

[hans]

I noticed in the first video the trace disappearing when the sample rate was below 2.5MSPS. So with a shorter sample length, the sample rate will be lower because otherwise it will run out of memory before the screen is full. Therefore you had to zoom in more to see the trace again at 2.5MSPS+

How does the 'phosphor' display work on this scope? Is it similar how it's done on the recent scopes, where a ginormous sample buffer is used to store x thousands of small waveforms and that's redrawn with a HW accelerated display engine?

Does the front-end change any mode at these instances? Maybe Tektronix is doing some convoluted things like mixing the LF/HF bands at the hybrid which is going wrong.

Are we also certain about the same function of those SOT23 parts on each channel? We would be seeing the same voltages or configuration, I'd presume, on each channel. If we measure 3.3V/2.5V on 1, it may as well be a tiny 2.5V programmable zener diode reference for all 4 channels, while the others do -2.5V or something.
I don't think it would be a logical placement though.

[babi]

Hi Dave.

I suggest trying these:
-Feeding both a positive and negative DC to the channel 3. It seems it does something different to positive and negative input signals.
-Feeding a sweep from low frequency to the highest frequency you can make. This could reveal whether the problem is with the high frequency components or the low frequency components. I think the problem is with low frequency ones.
-Changing the coupling to GND (I think most scopes will not capture anything in this mode, they just show a straight line on the screen). That has to show a clean signal if the Memory and the Processor are doing fine.

And if you show us the FFT domain of channel 3 while feeding different frequencies, we may find something similar in the frequency domain. I mean if the problem is with the ADC clock, it could appear in the FFT while we feed different frequency components.

[hikariuk]

The main ASIC is too hot. 100°C is a high temperature. It might be dangerous to operate it without a cooling fan near. And I would expect a heatsink, too. Heatsinks are very common at most scopes today. Agilent DSOX2000, Rigol DS2000, GW Instek GDS-2000A...

It's probably too hot because it's outside its enclosure and not getting air blown over it by the fan.

[SeanB]

Shorted diode possibly was there to protect the 2 power rails. Possibly the short took out an inner layer trace or a via that supplies the HF side with power so it now runs only on a single rail. You really need a pinout of the chip, or trace the 2V5 rail to see which pins it connects to on the other channel and see where the faulty one differs. You probably will find 1 or more pins that are now different.