Here is ci11's tests with a better (fixed frequency, lower THD and lower noise) oscillator, one of Victor's famous ultra low THD 1KHz units. Victor's oscillator produces less than -140dB THD, as noted here by a fellow who owns the current model Audio Precision distortion analyzer:
http://www.diyaudio.com/forums/headphone-systems/300491-neurochrome-hp-1-ultra-high-end-headphone-amp.html#post4918641It beats the built-in variable frequency oscillator in the AP so he uses it for the lowest possible THD for a fixed frequency.
Victor's oscillator uses relatively few parts (relative to the variable frequency KH4400B) and is direct output, no resistive dividers to generate Johnson noise. The oscillator uses two TL431's for a power supply and in this case I have them embedded in a power supply creation of mine. Kind of went nutz for low noise.
This setup is using 5 9V batteries inside a fully shielded (copper spray inside) Hammond case. The oscillator boards are individually fed with a constant current sink.
The lowest Victor's will go is about 366mV(rms), so I have added my own resistive divider on the output, as shown in the photo. A 10K in series with a 5.1R, both 1% 50PPM/C metal film, for a 1960:1 voltage reduction, or 366mV(rms)/1960 = 187uV(rms) = 264uV (peak) = 528uVp-p, as shown in the attached photo (529uVp-p in the photo). I had to keep the total resistance around 10K to keep from loading the oscillator, ideally it would be lower to reduce Johnson noise. But to that end you might notice in the photo the resistors look a little frosty.
I've sprayed the resistors and their leads down with technical freeze spray to temporarily reduce the Johnson noise as low as I can.
The net result is the 2nd attachment! This is done using the USB stick and the scope "camera" button trick. Notice significantly less noise in the signal than with the variable-frequency KH4400B before. The bulk of that noise previously was the output resistors and (old - LM318!) chips in the KH oscillator. The scope is in high definition acquisition mode again. I couldn't get it to trigger this time using edge triggering. I tried a few options and oddly enough "video" triggering did the job, which is what is used below.
For more photos of the test that were too many to post, see this Google Drive link:
https://drive.google.com/drive/folders/0B67cJELZW-i8VUowRlh1Q3ZtNXM?usp=sharing