Broke my TDS3000B

[perdrix]

A bit ago I reported a problem with ripple on calibrator trace if my TDS3014B (up-hacked to TDS3064B): https://www.eevblog.com/forum/testgear/tds3000b-trace-ripple/msg3492464/#msg3492464

I replaced the tantalum caps on the main board and that problem is gone, but I've broken it - I got a bit of solder splatter on a few pins of what I suspect is the main CPU (the one marked EX388/ADG360C/MM9595-AUW/156-7644-01) and on two pins of one of four ICs marked ES336AC/ADG361D/MM9916-VJG/156-7645-01 which I failed to notice and powered it on in that state. 

I've since removed the splatter but now any horizontal sweep rate faster the 400uS breaks up into a mess, and autoset doesn't work right.  I'm thinking I've probably killed it 

Of course it happily passes the self test.

Before I put on my scrap pile and weep over the loss of a nice scope has anyone any thoughts - I've carefully re-checked for any further solder bridges - none found.

David

[YetAnotherTechie]

have you cleaned all the fux residue? even underneath? solder can stick in the pins up to the package, an be hard to see except at certain angles.

[sicco]

The MM9595 is not the main CPU. The PowerPC XPC860 is. But this MM9595 does autonomously do the graphics, and gets the traces of the four channels plotted, even if the PowerPC is halted.

Suggest you check for shorts once again, in neighboring pins, beep it through with a multimeter and needle probes... Some pins are interconnect because they’re power supply, GND etc. But the bulk of the ~400 pins are not...

Before you scrap it, consider selling as-is for parts on eBay. Adventurists (like me) may bid more than you’’d expect!

[perdrix]

I've just put it back together after yet another cleaning session, but with no change ☹

However I do have something to report which may be useful:  If I bring up the acquire menu, and select "Fast Trigger" (500 points), then the calibrator wave form is display correctly down to a sweep speed of 20us/div (400us/div in "Normal Resolution" mode). If I go down 10us/div it all breaks up into a mush again.

I'm sure this means something useful to someone who knows more about how this beastie works.

PS I beeped out the pins as suggested, and no shorts to be found ...

David

[perdrix]

Here's what I see when I switch from 400us/div (normal calibrator square wave) to 200us/div - oops:



SPC fails as well

All channels fail in the same way

David

[Jwalling]

Does the sample rate change between those 2 sweep rates?
Also, do you know whether any of the shorts were to a rail?

[perdrix]

In normal resolution mode the break up occurs between 400us/div and 200us/div.   If I bring up the acquire menu, and select "Fast Trigger" (500 points), then the calibrator wave form is displayed correctly down to a sweep speed of 20us/div. If I go down 10us/div it all breaks up into a mush again.

So the break point to using the FISO would be at the point where the real time sample rate would exceed 2.5MS/s

I'd not be a bit surprised if one of the shorts had been to power/ground as lot of pins were shorted together (large solder splat).   I'm wary of probing that IC with power on as that's a pretty fine pitch IC with a lot of pins (I think 100 per side).

D.

[Jarek]

Hi!
Have you solved the problem?
I have the same fault.

[perdrix]

Sadly not ...

David

[Jarek]

Its pity.
Have you replaced this ADG361D and it's not working?
You may have a datasheet of this chip or any schematic oscilloscope ?


Jarek

[Jarek]

Hi
I found a fault in the oscilloscope.
I measured the voltage on the relays (foto)  and it was -2.5V on all of them (between the ground and the coil of the relays), except for one on the first channel where it was -1,6 V .  (When I switch DC to AC coupling, I can not hear the relay click (if is  50 R and same on 1M) for  Ch 1)
I found a place where it is connected  it was ADG361D.
I desoldered the chip and now  it is probably fine on one channel (without the fas triger and up 400us/div  )
An error shows up at startup, but then it works
I do not know if only this one chip is damaged. I have to buy it only there is a problem where?!

[Jarek]

Little updates.
I am trying to recreate the connection diagram of these relays because there is no schematic diagram anywhere.

[Jarek]

I found an article on a Chinese website about the TDS3034B

[perdrix]

Sadly I don't sling the lingo ...

Any chance of an English version?

D.

[Jarek]

Enter this page in google and translate it.

http://www.edatop.com/test/56802.html

[Jarek]

A little update.
I noticed that David L. Jones on his schematic diagram as U320 AAAG wrote that it was MAX8877EZK33
On a Chinese web site, they suggest it's a switch.
 I found a MAX4516 chip that fits better.

[Jarek]

It is a pity that no one has a schematic diagram of these oscilloscopes.
I used to repair an old IWATSU DSO and also didn't have a schematic and wrote to Iwatsu company and someone from staff sent me the schematic. Now this schematic is on the official IWATSU website.
Maybe someone has a good contact with Tektronix and will ask for this diagram?

[Jarek]

Update

[sicco]

I have a tds3032b unit with exactly the same fault.

Not yet fixed, but here’s something really odd. The 5V to 3.3V TO220 LDO is bad. Only found out after unsoldering and testing just the naked ic. LM3940IT-3.3. 5.6 VDC on its input gives 5.4 V on its output… It does not regulate. WTF…

In circuit it had 3.27V, just a little bit low which is why i took it out.
Looks like that near normal voltage develops because the loading on the 3V3 rail was just enough? Or chips supplid by the LDO act as a zener?

The >400 us/div problem does not disappear when powering the 3V3 rail from a lab bench power supply.
I have it on 2 (both) channels.
From memory, it initially was on just one channel.
Elsewhere I read that some have units with 4 channels that all fail.

Could it be that a bad LDO is the root cause of damaging the TDS3000 ADC chips?

[benj38]

The noisy signal above 2.5Mega-samples-per-seconds is a dead giveaway that you indeed have a bad ADG361. This is not an uncommon fault for these scopes.

Unfortunately, these are unonbtainiums, except for a donor board from another scope.
I once located some new ADG361s on eBay from China, but they were all bad/fake.

Interestingly, depending on the channel the bad ADG361 is in, it may fuck up all channels or just some. When it only affects some, the others usually show excessive DC bias.

[sicco]

Thank you Ben.

UTsource have loads of these ADG361 chips, so I'm taking another gamble there... My hope is that the HF noise is not signal noise but something of a lookup table for correcting each sample with a bias and gain factor. It seems to have a 512 sample fixed pattern in the time domain plots. And that these factors is populated during the SPC calibration. But only if SPC does not fail halfway. So if there's something that prevents SPC to fully complete, on any channel, then the 'HF noise' at >2.5 Ms/s does not go away. Just speculating. But it might explain why 'Chinese eBay ADG361' is a genuine and working product after all, but just needs to go through and reach the finish line for SPC before it's useful again in the TDS3000.

Attached my refinements for the schematics.

I now ended up with a unit where the DAC output for 'level/position' for channel A does not work. Also not after swapping the ADG361 from a donor board. Twice... That might have been a bad donor though...


But I learned a lot. Like how it's SPI coming from/via the trigger chip ADG365 that sets things like the DC bias DAC, the relays, and so on. I think even the time base / sample clock.

Oh and those LM3940 LDO's: they're not faulty as I thought. It's just that unloaded their output goes up to 5V. Apparently they need a few dozen of mA load current before they output the promised 3.3V...
However, in another TDS3000b board I found that they changed to the more powerful LMS1885ACT33. Which has a different pinout. So swapping those from one donor board to another is a bad idea...

edit 19aug2023: TDS3000 schematics / circuit diagrams updated, attached (TDS3012B TDS3014B TDS3022B TDS3024B TDS3032B TDS3034B TDS3042B TDS3044B TDS3052B TDS3054B TDS3064B)

[benj38]

All the scopes I have seen with this fault never pass SPC.

But it might explain why 'Chinese eBay ADG361' is a genuine and working product after all, but just needs to go through and reach the finish line for SPC before it's useful again in the TDS3000

The chips I got from eBay were verified bad/fake by substitution: at the time I have had access to multiple TDS3000 series scopes, and in order to get a definite answer I replaced an ADG361 from a working scope into the noisy scope and it fixed it (immediately, no SPC needed to get rid of the noise and large DC bias), and putting the original bad chip, as well as each of the 4 Chinese chips I had, in the working scope made it exhibit the fault of being noisy above 2.5 Ms/s, and failing SPC.

A lot of delicate soldering and de-soldering was involved. Lots of fun 

@sicco, in any case, please don't forget to report if the chips you get from UTsource are any good!

[sicco]

The UTsource ADG361 chips arrived today. I think they are ok.

I carefully soldered two in the TDS3032b that had this noise issue. Soldered by hand, just to make sure I would not overheat these delicate ic's. Carefully double checked that there were really no solder shorts between each of their 100 pins.

Some success, but not yet 100%.

The TDS3032b now boots up fine. Initially it shows the noise again below 400 us/div. And it has no calibration and huge offset errors. But now, if I do SPC, that noise disappears. And the offsets nearly all disappear. But SPC does still fail. At the end. Much further in the process as what I had before. If I look ate the system diagnostic port (serial out at 38400,n,8,1 on the 100 pin connector pin ...) then it says:

Class 3 Level 3  InterpolatingSearch:  Ch1 found invalid value.
 16:12:55 08/29/2023 (thread: rtl)
Class 3 Level 3  findZeroData:  couldn't find plusFineVal. - channels (bitOr'd) 0x1.

The issue now is that CH1 has still an offset of about .5 div. CH2 much better.

Other than that the scope seems fully functional, all the way up to 1 ns/div as it's a TDS3052b now. But since the SPC didn't make it to the finish line, the calibration results are not stored in the scope flash it seems. So after a reboot, it is ugly again, unless I run SPC first.

I can trim the CHA offset away if I go into Utility->Devel->Subsystem DAC Ctrl->CH1 and then adjust Coarse or Fine Offset. But that correction isn't persistent. And diappears already (resets to 0) as soon as I change scale or coupling mode or position...

I swapped the CH1 opamp and CMOS switch. No luck.

So I'm now at the point of either swapping ceramic input module, or swapping again the ADG361...

No datasheet obtainable for these ADG361's, but I think I found the pedigree for its family. TI website, ADC07d1520, ADC08B3000 and the likes. Interleaving to double the sample rates. So two ADC inside each, both with differential analog inputs. Probably a circular memory or so inside that gets read out at a much lower rate after the acquisition is frozen. The differential inputs have are low ohmic internal termination resistors it seems. Those appeared gone in my old/bad ADG361 chips...

To be continued...
 

[sicco]

Update: found a workaround. A little bit cheating maybe, but an enormous relieve to see it reaching the end of its self calibration procedure.

After finding out the hard way that the remaining offset issue was not in the CH1 ceramic attenuator board, I decided to use a bit more brute force to get rid of the offset error that apparently is present in one of the two ADG361D'a that I scored at UTsource. Or which I may have induced in some other way while soldering or handling.

The fix: inject just a little bit more offset, in the right direction, on the input of Ux30, the LMC7101 OPAMP that routes the offset compensation signal from various sources. I had to pull it down. I did that with two 1Mohm resistors in series, and then tied that to -2.58V supply rail. As per red additions in the diagram attached. With that the SPC came much further. After locking the metal shield with the two added resistors inside the metal shield CH1 section, it managed to run SPC all the way to the end. And save the new calibration settings. So that next time the scope boots up it no longer shows the noise as in the beginning of this topic.

Learnings: quite likely I've been swapping ADG361 IC's way too often, and declared units that were good after all as faulty. Others I think have labelled ADG361's from China / eBay as counterfeit - but they probably were good genuine working parts! Any ADG361D ic seems to have a unique fingerprint, which is that noise pattern as pictured in first post above by Perdix in 2021. The fingerprint is I think like an offset per sample, a number for each physical S/H amplifier or something like that. With 512 or so S/H capacitors inside the chip. Halfway SPC the scope identifies this pattern. Much earlier in SPC, it checks already DC offsets, gains etc, and when something is not OK already then, then SPC does not continue to the fingerprinting steps. If it manages to fingerprint OK, it still has many more checks and calibration to do. Only if those all pass ok, only then the SPC finishes with the 'success' message, and the new settings and fingerprints get stored in the scope flash in a file somewhere inside.

[Jarek]