I'm not sure I understand your measurement. How are you measuring a 10pS long event with a SR830 lock-in which has a minimum time constant of 10uS? Still, there is a common mistake with lockins which might cause what you see.
A lock in amplifier makes a phase sensitive measurement of an AC signal. It measures in-phase and quadrature signals, which are usually referred to as X and Y respectively. The stanford lock-in you linked also calculates R and θ from X and Y, where R is the magnitude of X+jY theta is the argument. If you have a signal which passes through zero, then R will look like the absolute value of the signal.
The SR830 can be configured to provide X, Y, R or θ on the BNC outputs by pressing the one of the buttons under the LCD displays. Probably you have connected the left hand output to the computer, and accidentally selected R instead of X.
As an aside, the SR830 is a digital instrument. It digitises your signal, then does the actual detection and filtering in the digital domain. It then uses a DAC to convert back to analog, and you're using the NI ADC to go back to digital. You'll get better accuracy cutting out the second round of analog and reading straight from the SR830 using serial or GPIB.
I have only a passing familiarity with pump-probe, but here are a few tips for using lockins in general and the SR830 in particular:
- A lock-in is a really powerful tool, but it's complicated and it's quite possible to shoot yourself in the foot with it. If you can spare the time to really understand the theory of how it works, that will pay off. But if you don't then at least get familiar with what X, Y, R, θ really mean.
- A lock-in measures two things. That can be X and Y, or R and θ, but always record both of them. It makes debugging much easier. If I'm right, then you'll see an abrupt 180 degree change in θ where your curve gets reflected, which is a clear indicator something's wrong.
- I don't know the details of your experiment, and I don't want to contradict your supervisor, but R is rarely what you want to measure. People often measure R because they aren't sure how to best configure the phase reference. You'll probably find that measuring X instead avoids the funny reflection feature and gains you a factor of 1.5 in signal to noise.
- The SR830 has the ability to scale and offset it's outputs. Handy if that's what you want, but confusing if it's turned on when you don't. That could be responsible for the 4V offset. Check there are no "expand" or "offset" lamps on the leftmost LCD or just below.
- You need a stable reference. In your comment you note that you might not have one, if so it's hurting your signal to noise and may cause all sorts of weird behaviour. The square wave from the chopper should go into the "ref in" BNC in the bottom right, and either "pos edge" or "neg edge" not "sine" should be selected. There should be no red "unlock" lamp and the frequency displayed above should be completely stable.
- If you're going to try to use serial (and I suggest you do, if practical) you can use a standard USB-RS232 converter, and a 9-pin to 25-pin serial cable. You'll need to set the visa read terminator to '\r' and send the string 'OUTX 0' to the instrument at the start of each session.