Repair of R&S RTB2004 with offsetprobem of toasted LMH6518

[Darkover]

In this thread I like to show you  a huge problem with the R&S RTB2004 and probably some other scopes and how to repair it.

But history first!

Some years ago I bought a new and shiny RTB2004. After two years I noticed that I had a huge offset drift in one or two channel. I don't
understand the reason, but it was just in warranty, so I send it back and R&S repaired it. Repairing means, they exchange
the mainboard and calibrate it. So you got a more or less new scope.

Another two year later I found the offsetproblem again. But this time I understand the reason. I used a selfmade current probe to
measure current of my DUT. In most cases I use this probes on channel 3 or 4, and more seldom on CH1.

So my current measurement destroyed the offset. But why?

My probe has a OPV at the output (Vcc=3.3V) and delivers 0-2.5V for any current range. I am measuring the current consumption
of some microcontroller board. So I put the GND line at the bottom of the screen and the 2.5V at the top.
Most of the time the scope is now measuring 2V on the top of the screen in DC mode.

The internal PGA inside the scope a LMH6518 has to withstand the offset for some time, but it can't.
TI find it is to much for the PGA. They changed the datasheet later. They did not alow a huge DC offset
anymore. Unfortunatly the developer at R&S did not know it.

I had a diskussion with TI about it:
https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1569255/lmh6518-offset-problem-with-lmh6518

However there are some more discussion about this problem, because I am not the first victim.

This time R&S did not repair it in warranty because they only give 1 year of warranty for there repairs.

So I decide it to repair it by myself! That means replacing the LMH6418

But before, have a look of some picture I did BEFORE repairing. It is important to memoryze it to understand what happens at the end.

The picture show you the huge offset drift before repairing and the that the internal offset compensation stops with an error
after 25%. That is important!

Olaf

[Darkover]

Now, we know the problem, we know what we have to do. We are removing the LMH6518!

Unfortunatly removing is not as easy as you might think, because inside the board there is much copper and
the LMH draws more than 1W of power so it need a perfect connection to the copper.

At first I put the board on a big hotplate at 200°C. Unfortunatly that did not work, because the parts of
the bottom made a higher distance of the PCB it becomes not hot enought and I did not like to increase
the hot plate at more than 200°C.

Than I used a heat gun and heated the bottom side and a 858D (at 370°C, Fan at 5) to heat the chip.
That did not work either, but it destroyed the chip. Now the PCB looks like after an impact of an asteroid.

The next step was using the heat gun at the bottom and a Weller Pyropen at the top. This is a gas heat gun
that is MUCH more powerful than 858D. It can easily toast a PCB to black cole!
But it worked. It was possible to remove the LMH6518 and after cleaning the PCB looks good.

 Olaf

[Darkover]

Okay the board is now fine, we have to solder.

Unfortunatly we can not use the big gun this time, because the LMH should survive!

So I used two different type of solder. A special low mealting solder with wismut inside on the exposed pad,
and normal lead-solder on the outer pads. I put also some lead-solder to some GND point. So I notice,
when the lead-solder mealted at 183°C the wismut solder below the chips is melted, too.

It is preheated with the big heat gun and solder with the 858D (370°C/Fan5) from the top.

That works surprisingly well! Soldering was MUCH and I mean MUCH! easier than desolder the old part.

Unfortunatly I lost one 0402 capacitor. But I measured it from an other channel. It is 250pF and I solde a bigger 0603 for testing purpose, because I hace to buy the 250...

Olaf

[voltsandjolts]

Hmm, poor show by TI for not fixing the LMH6518 with new updated replacement part.

Does using ac coupling reduce risk of damage to LMH6518?

[Darkover]

Okay, everything is in place, but it is working?

After I switched on the scope I made the same offset control screen than I made before I replaced the chip.
It looks the same! CH1 has a smaller offset, CH3 and CH4 hat a huge one. CH2 looks okay.

I started the offset calibration in the setup menue. This time it tooks much longer time!
But it stoped at 67% and told me CH2 bad! Because I expect that every Channel needs 25%
of the work load and 67% is calibrating CH3. I am guessing that this is a firmware bug (V3.0),
perhaps a copy&paste error.

So I expect now that CH1 and CH2 are okay and CH3 and CH4 have still the offsetproblem,
because I only replaced the LMH from CH1 until now!

But at least at my testscreen CH3 and 4 looking good now. 

I will use the scope now for some time and write again if something happens. 

Olaf

[Darkover]

Does using ac coupling reduce risk of damage to LMH6518?

Yes, AC is fine. Only DC is the problem. I guess they can not fix the chip because the power
consumption is to high. I think it is only possible so solved by R&S with a different firmware
that used lower voltage level in DC mode, but we have to pay for higher noise in this case.

But I think R&S did not like to talk about this, because it can mean that many user will knock
on there door with a broken scope.

Olaf

[Hydron]

Try FW version 2.0 if you need it to get through the entire calibration process. Then go and look through the (huge) log to see how much offset was needed for each channel.

As for fixing this properly, TI only recently actually fessed up as to how the problem actually occurs. That said now that it's obvious where the fault is, R&S really should be pulling this from the market (assuming it's the same design with the "new" RTB2x scopes) given it's now a known fatal flaw.

[Xyphro]

Looks like an open heart surgery on an expensive patient :-)

I still wonder how many other scopes use the same frontend IC and if some consider it somehow in terms of frontend settings and saturation checks.
This IC is quite tempting to use compared to making something discrete.

[Hydron]

Looks like an open heart surgery on an expensive patient :-)

I still wonder how many other scopes use the same frontend IC and if some consider it somehow in terms of frontend settings and saturation checks.
This IC is quite tempting to use compared to making something discrete.

I believe at least some Siglent scopes use it too, and I'm sure there'll be others (it's basically what the IC is designed for). Question is how high each design let the output signal get - datasheet didn't give a limit until many years after the IC was released, so this will probably depend on the input range of the ADC used rather than the PGA damage limit.

[Darkover]

https://www.eevblog.com/forum/testgear/rigol-ds2xx2a-and-ds2xx2a-s/

Some Rigol have them, too.

Olaf

[voltsandjolts]

LMH6518 also in the R&S HMO1000 series
https://www.eevblog.com/forum/blog/eevblog-879-rs-1202-scope-bandwidth-hack-investigation/

..and in GWInstek GDS2000A
https://www.eevblog.com/forum/blog/eevblog-475-gw-instek-gds-2000a-oscilloscope-teardown/

[Xyphro]

I think my hmo2024 has it too from what I remember having seen.
What about Batronix Magnova? Just bought one... spec wise it could also use that IC, though it is of course not clear if they operate it also to such amplitudes in their y/div vs. Gain settings. Don't wanna open that one and could not find teardowns :-)

[Xyphro]

Btw: Reducing the front gain scaling by balancing analog vs. Digital gain will not result in 0 risk. Scope inputs are typically spec'd with an abs. Max. input voltage independent of vertical setting.
I guess most of us have seldomly also already driven a frontend willingly to saturation to view small gnd level artifacts on larger scale signals. OK, not often and for long times and of course signal distortions can happen then.

But getting to "a german accuracy level" of spec., a 10V DC input signal with y/div set to 1mV and 0V Y offset is safe in the way typical scopes inputs are spec'd and shall not result in something that stresses abs max ratings of the internal signal chain. Practically speaking this case is not so relevant, but spec wise it is OK and will result in some unwanted permanent drift.

A protection could though be done in software by detecting a long term saturating DC, throwing a warning and then setting the frontend gain to a very low one.

[Darkover]

You are never ready, until you are ready! 

Today I switched on the scope after calibration yesterday and I found it forgot the result of the calibration!

I think the reason is the error at the end of the calibration. I guess it did not store the calibration value
permanently until it will run without an error.

Okay, I expected that I had to replace two more LMHs, but now it is clear!

Olaf

[Darkover]

Just in case someone likes to repeat my steps, here are the stl files
for a Steinel to ArcaSwiss Adapter to connect a Steinel heatgun to a tripod.

Olaf

[Darkover]

I have another important notice for the brave hearted who decide to open the scope.

I am just in the process to remove the mainboard a second time to exchange the LMH for channel 3 and 4.

I found that two screw that hold the mainboard in the area of the channel connector are seized. When I install
the mainboard last time I did NOT use cordless screwdriver! All was done by hand with lower force! And I am not Hulks bigger brother.
But R&S, a company that always had to take care for there money, had used self taping steel screw in aluminum
that is prone for seizing. So it is probably a good idea to use anti seize paste or even better normal screws in
case we have to open the scope more often.

BTW: I am still thinking if it is possible to install a 50R switch a least in one channel......

Oh..and another hint. There is an expensive high quality  and high density connector to connect the lcd to
the mainboard. It has a locking metal bracket. Turn it bei 180degree and than use it to pull the connector, so it
is very easily removable.

Olaf

[Darkover]

Okay, I removed the parts from CH3 and CH4.

This time it was easier, because I had some experience and the parts are closer to the border. So the board sucks less heat.
Even the removed LMH6518 looks good now. Perhaps I can sell them at ebay as untestet.

I think soldering the new PGAs will happen next week. I orderd some Yorkshire tea from Internet for my mental health, I guess that works better than German beer!

 Olaf

[Hydron]

Ooh happy to see that CH4 is probably the easiest, still have to find time to fix mine (again), thankfully that's the channel that went this time.

Fingers crossed yours is all fixed once you put the new ones on, I pinged the R&S guy about the issue too btw.

[voltsandjolts]

Keep CH1,2,3 as AC coupling only, because CH4 easy to replace

[tooki]

Unfortunatly we can not use the big gun this time, because the LMH should survive!

So I used two different type of solder. A special low mealting solder with wismut inside on the exposed pad,
and normal lead-solder on the outer pads. I put also some lead-solder to some GND point. So I notice,
when the lead-solder mealted at 183°C the wismut solder below the chips is melted, too.

Oh, no…

(FYI, it’s “bismuth” in English.)

You must never mix bismuth and lead solders! Even small amounts of lead combine with bismuth and tin to make an alloy with an extremely low melting point (worst case, 95°C!!!). Additionally, the bismuth-lead-tin combination has terrible mechanical properties, making the solder much more brittle.




Overall, your technique is not very good. The reason you’re resorting to low-temp solder is because you’re not preheating the board properly. Rather than a single point of hot air on the underside (which means one spot getting very hot, but the rest of the board acting as a big heatsink), you want weaker heat all over. 1 minute of preheating is not nearly enough IMHO. When I work on heavy multilayer boards, I like to preheat the board from the edge, moving around the edge of the board around and around, so that the edges can’t act as huge heatsinks. The center gets heated as it steals heat from the edge. This can take a solid 3-4 minutes to do, but avoids overheating one spot. Alternatively, an oven is a great way to preheat the board uniformly. Then, after you secure the board to work on it, keep gentle heat on the bottom.

Also, your copper foil should be stuck down, so that hot air doesn’t sneak underneath and blow small components away; gaps can actually end up focusing heat exactly on the components you’re trying to protect!

[Darkover]

Keep CH1,2,3 as AC coupling only, because CH4 easy to replace

Aehem yes, this comes to my mind, too.

Olaf

[Darkover]

Even small amounts of lead combine with bismuth and tin to make an alloy with an extremely low melting point (worst case, 95°C!!!).

I know that, but I did not care so much about it. And I did not mix it with lead. There is pure bismut/tin in the center. lead is only outside.

Additionally, the bismuth-lead-tin combination has terrible mechanical properties, making the solder much more brittle.

Of course, but this is also true for bismut/tin. This is the reason why I only used it on the EP.

Alternatively, an oven is a great way to preheat the board uniformly. Then, after you secure the board to work on it, keep gentle heat on the bottom.

There are always better ways, but I am not in my company. It has to work on my private desktop!

Also, your copper foil should be stuck down, so that hot air doesn’t sneak underneath and blow small components away; gaps can actually end up focusing heat exactly on the components you’re trying to protect!

This did not happen, because the copper tape is glued on the PCB. All part stay where they are. Oh and the Weller Pyropen did not blow very much. It has no fan like a normal heatgun. I think it is more radiant heat.

Olaf

[daisizhou]

You're missing a BGA soldering machine. Replacing it with an LMH6518 would be a piece of cake, and the soldering result would be perfect.

You can use 559 soldering fluid, which can lower the temperature and make chip removal easier.

[Darkover]

You're missing a BGA soldering machine. Replacing it with an LMH6518 would be a piece of cake, and the soldering result would be perfect.

It is not my idea to make a living from it. 

You can use 559 soldering fluid, which can lower the temperature and make chip removal easier.

Nobody, at least not me, know what a 559 is. We only know a 555.
But removal can not be easier because you has to deal with the solder used by R&S when they build
the board. For the next time it will be easier, because of the solder I had lying around in my fridge.

Olaf

[daisizhou]

To be frank, your operation is very dangerous and can easily cause PCB bulges, that is, insulation layer bulges


Perhaps each region has a different name, so I can only send a picture. I hope you understand what I'm saying