Repair of R&S RTB2004 with offsetprobem of toasted LMH6518

[Hydron]

OK so I found a half hour to solder in the new LMH6518 and we are back in business - wasn't very hard to resolder (prep and cleaning took a lot more time and effort than the actual soldering); CH4 is the best case but I suspect leaded solder will be fine for them all given good pre-heating, without a need to go with 138C bismuth stuff.

I also did a quick check looking at input vs ADC voltage (only at a couple of input levels, 1.25 and 2.5V) - it appears that ~330mV is full scale high/low on the ADC input, so you need to overdrive it by about 5x to get into the danger zone. Problem is that I don't think there's a way for the scope to tell how much it's clipping by, so probably no way to do a firmware based safety limit (if there was a way to tell how heavily it was clipping there could be a failsafe to change gain and avoid sustained heavy clipping).

I will admit that I expected the ADC input to be a bit closer to the LMH6518 limit - you wouldn't normally leave it clipping by 5x for long, but maybe the damage occurs pretty quickly (fuck TI for not documenting this properly!). I also wonder if there is another factor at play - there is a clipping circuit (some RF schottkys and a DAC/opamp, the autocal routine has a step to calibrate it) that clearly kicks in at some point, maybe it starts sinking more current than the LMH6518 likes as it starts limiting the signal? I'm hesitant to push it further testing this, don't want to have to change yet another PGA, at least while I don't have stock of them or time to do the work!

Anyway I think the name of the game is to try and avoid sustained heavy clipping, which unfortunately means needing to be much more careful than should be strictly necessary going by the datasheet input limits

Edit: abs max is 1700mV, so danger zone is 5x full scale, not 3x I stated originally.

[daisizhou]

suggest you communicate in English。(建议你最好使用英语进行沟通交流)

[Darkover]

Anyway I think the name of the game is to try and avoid sustained heavy clipping, which unfortunately means needing to be much more careful than should be strictly necessary going by the datasheet input limits

I think so, too. I faced with this problem when I use the fullscreen for my measurement, it was my idea to stay away one div away from top and bottom when I do it the next time.

Olaf