Joy-it JDS6600 - signal clipping

[notadave]

Let me first say that I like my JDS6600, it has nice functionality at a mass product price, but after using it for a few times the CH2 is now defective.

My Joy-IT JDS6600 Rev.15 has an offset and clipping issue with one of the two channels. Thanks to the two channel nature and the rather obvious fault, I was able to trace the issue back to one IC. Sadly all the important ICs were "scrubbed". The tops and labels were removed. The function of the broken IC is to shift the signal from the range of 0 to 840mV down to -350 to +350mV and drive a lower (200Ohm) impedance. The possibly broken SOIC-8 that makes the difference is connected to two SOT-23-6 U4, U6 which provide an input that does not require a resistor in between. The output of the defective IC is phase inverted.

Here is what I measure at the pins:
1. Non-inverted out
2. Some Feed-back port for the unused non-inverted out 1k Ohm
3. The used input
4. -5V
5. 0V
6. Some Feed-back port for the inverted output 100 Ohm
7. The used inverted out
8. +5V
The IC is very likely a differential output OpAmp as is used for driving ADCs (Example, which can not be the same.). One of the pins must act as V_OCM, likely either 5 or 6.

U4 and U6 show at their six pins:
1. Signal In
2. GND
3. 0V
4. Signal Out same as in
5. 3.3V
6. 3.3V
They might act as high input impedance buffers.

Speculation: The IC seems to generate a differential output from a dual differential input, but I could not see that the differential output feature was used to drive anything.
The Pin usage afaik is:
1. Out+
2. In-
3. In+
4. -5V
5. GND (Unused In-)
6. Unused In+
7. Out-
8. +5V
The IC can not be a dual normal OpAmp, it must have an internal feature to remove the offset. I know such ICs are available but can not name one.Instead of replacing the IC, I could replace the entire circuit, replace the function, not the IC. I would have to find a way to wire in a small PCB in such a way that it is 50MHz compatible. It could be done elegantly by stacking a tiny interposer PCB in-between, exposing enough of the original pads to get the heat under.I can think of two ways to remove the offset: to build a 10Hz 2. order lowpass and subtract the result from the input or to go high impedance and make it a HP at 10Hz using 1uF 10V Cap. The first way has the problem of phase shift between the subtracted signals the second way might have trouble at >10MHz due to the high impedance. The IC doesn't seem to do it either way from the outside, I can not see the DC, average on any of the pins, nor do I see a large external capacitor.