TEKTRONIX DPO 6GHz scope issue - broken BGA ball connection?

[ZBIQ]

Hi guys,

Currently I'm resurrecting a high-end scope (especially back in the days) TEKTRONIX DPO 70604C. After replacing its HDD and ATX PC board it finally boots (it fails to do so from time to time, but whatever...).

Right mow, however, I'm facing acquisition board issue. On the first channel, after switching the full sampling rate (from 12,5 GS/s to 25 GS/s) I see large spikes appearing. This tells me, that some additional sampling circuit kicks in and starts to show bad behavior. There is 320 ps of separation between the spikes, thus it is 25 GS/s divided by 8. Makes sense - there are multiple ADC for a single channel for sure.

I tracked it a bit more and found a chip under glued heatsink - see arrow in the picture. This is probably the CH1 multiplexer. After pushing it, especially on one of the crones as my pencil shows, or by spraying some freezer - the spikes stop and scope behaves as new. Is this a broken ball behavior?

If it is, is it worth to try heating up and pray for the IC to reach the PCB? I am quite afraid about doing so, because:
1) The board is about 32 layers and 5 mm of thickness.
2) The thermal addhesive underneath the heatsink is solid as a rock - very hard and quite impossible to remove. I tried tu cut it a bit with a dental floss but without much luck.
3) The suspect IC is surrounded by other delicate ADC ICs.

Is it worth trying to re-solder this IC?

Best!
Mateusz

[hanakp]

How strong mechanical force is needed to "fix" the IC? If it's not too big, I'd just install a spring that pushes the IC against the cover. It won't last forever (solder tends to creep), but you'd "fix" it for a few months or even years.

[ZBIQ]

I would say approx. 1,5 kg. Quite a bit. but not dramatic.

If I could only remove this heatsink... Are there any 0,1 mm OD diamond wires on the market to cut this epoxy?

[mikehank]

That IC is an A2D HFD204 ASIC to be specific.  If you are having spikes during acquisitions it certainly will need to be rebelled.  You can actually measure the problem by probing the gold traces.  If the A2D is open you will get 60 or 28 ohms i cant remember.  Compare to the other A2D's.  BTW the chip that the traces are gong to is the track and hold.  Please ask more questions if needed.

[ZBIQ]

Hi mikehank,

That's a lot of information for me in a single post! Have you used to work with or service those scopes?
My questions are mainly about the servicing procedure i.e., how to remove those heatsinks properly and what PCB preheating is required? I wonder if 20x20 cm2 heater will be good. If no, due to potential PCB stresses, I will look for some BGA guru nearby.

Cheers,
Mateusz

[mikehank]

Hi ABIQ,

Yep I've worked on these scopes for years fix dozens maybe hundreds of board/scopes.   To remove the heatsink is actually pretty easy.  Get a heat gun and a small screw driver.  Place the screw driver between the fins of the heat sink.  Begin to heat the heat sink from above while wiggling the screw driver.  Once the epoxy  gets to a certain temp the heat sink will start to move and come off.  If you are heating for a long while stop and take a break and let the board cool.  We don't want to flow the balls under the BGA.  It takes a little practice but it not to difficult.

[ZBIQ]

Great!

Would you consider using a small 0,6 kW preheater just under the faulty IC? Or should I find some real reworking preheat to cover most of the acquisition PCB?

Edit:
One more thing. I measured the Rdc of the golden traces incoming to the broken IC. They are always 28 Ohms, even if the spikes are still visible (I switched of TEK and quickly measured resistance on the chip still being hot and faulty).

Cheers,
Zbiq

[mikehank]

Hi,

You don't really need to pre-heat the board for the heat sink removal. 

Removing a BGA hell yes, you will need a BGA replacement machine and a trained guy.

[ZBIQ]

Thanks, I've already started to look for a professional reworking guy.

Anyway, from what I noticed, the TEK hardware is prepared for much higher bandwidth and sampling rate, maybe even up to maximum in its 70000, i.e., 33 GHz. So I was wondering if it is possible to squeeze a bit more of scope's analog bandwidth. Of course, sampling rate will be limited, but working in the Equivalent Time sampling mode maybe could show much higher frequency response. Has anybody ever tried to hack such scope?

Mateusz

[ZBIQ]

Hi Folks!

I started to dig a bit more into DPO debugging and I have a short question. Where I can find extended selfcal/test files of the DPO 7k series? One of the colegues pointed, that I should look for c:/TekScope/calibration, but there is no such folder in my device.
My scope is running one of the latest software versions on win7, so probably files architecture has changed...

Cheers,
Mateusz

[ddrl46]

Hi Folks!

I started to dig a bit more into DPO debugging and I have a short question. Where I can find extended selfcal/test files of the DPO 7k series? One of the colegues pointed, that I should look for c:/TekScope/calibration, but there is no such folder in my device.
My scope is running one of the latest software versions on win7, so probably files architecture has changed...

Cheers,
Mateusz

Did you enable "show hidden folders" in explorer? From what I can remember from the Windows XP units it should be there.