I can not think of any reason that the input impedance of 13kOhm between the pos/neg inputs produces such a big change at the output. In any case, the differential voltage between the pos/neg inputs produces a big output. But the output was just 2V or so. This means that the neg input was actually about zero volt. This ensures the principle of the summing amplifier of this kind.
Because the input impedance of the summing node (the additional resister you put at the negative input) is not infinity, the voltage divider is not perfect and shows 10% reduction of the voltge (i.e. the output will have +4.5V offset instead of +5V). But still it is not enough to explain such a small output like 2.3V.
What I can think of is that the earlier stages somehow have the offset for some reason. Anyway, it is difficult to guess the true reason unless all of the nodes around the last stage are checked with the multimeters.
At least, we can remove the voltage divider and instead put a 10k between -15V and the neg input in order to impose +5V offset at the output. This costs 1.5mA instead of 10mA.
[Larisa and Jenne]
We wanted to get rid of the awkward cart that was sitting behind the 1Y1 rack. This cart was supplying a +5V offset to the PZT driver, so that we could use the MC length signal to feedback to lock the laser to the MC cavity. Instead, we put the offset on the last op amp before the servo out on the Mc Servo Board. Because we wanted +5V, but the board only had +5, +15, -15V as options, and we needed -5 to add just before the op amp (U40 in the schematic), because the op amp is using regular negative feedback, we made a little voltage divider between -15V and GND, to give ourselves -5V. We used the back side of the voltage test points (where you can check to make sure that you're actually getting DC voltage on the board), and used a 511Ohm and 1.02kOhm resistor as a voltage divider.
Then we put a 3.32kOhm resistor in ~"parallel" to R124, which is the usual resistor just before the negative input of the op amp. Our -5V goes to our new resistor, and should, at the output, give us a +5V offset.
Sadly, when we measure the actual output we get, it's only +2.3V. Sadface.
We went ahead and plugged the servo out into the PZT driver anyway, since we had previously seen that the fluctuation when the mode cleaner is locked was much less than a volt, so we won't run into any problems with the PZT driver running into the lower limit (it only goes 0-10V).
Suresh has discovered that the op amp that we're looking at, U40 on the schematic, is an AD829, which has an input impedance of a measely 13kOhm. So maybe the 3.32kOhm resistors that we are using (because that's what had already been there) are too large. Perhaps tomorrow I'll switch all 3 resistors (R119, R124, and our new one) to something more like 1kOhm. But right now, the MC is locked, and I'm super hungry, and it's time for some arm locking action.
I've attached the schematic. The stuff that we fitzed with was all on page 8.