I've just done some test on TDS3014B hacked to TDS3064B using RIGOL DSA815TG. Tracking gen in Rigol isn't perfectly flat but I've tried to calibrate it using DSA815 itself. I've used zero span option and compared output level at higher frequencies to output at 50MHz. I did separate comparison for each higher frequency switching between this frequency and 50MHz setting (If you let DSA815 run for a while the level drifts by as much as 1dB. It also changes when DSA is doing self calibration in the background from time to time but those drifts seem to be in SA part not in TG. TG is not flat but it is fairly stable in time). Tests were done on channel 1 of the scope with internal 50
termination, after calibration. Level was measured using cursors on the screen in peak-peak value.
Results are:
Freq: Scope level [mV p-p] unadjusted dB level Sig gen correction corrected dB level
50 MHz 652 0 0 0
100 MHz 622 -0.41 -0.3 -0.11
200 MHz 610 -0.58 -0.2 -0.38
300 MHz 550 -1.48 0.1 -1.58
400 MHz 516 -2.03 0 -2.03
500 MHz 462 -2.99 -0.75 -2.24
600 MHz 434 -3.54 -0.8 -2.74
700 MHz 386 -4.55 -0.7 -3.85
I'd guess that I'm no more than 1dB off on my measurements. 3dB point appears to be above 600MHz. Above 750MHz something strange began to happen with the trace. It looked like there was additional low frequency modulation (similar to aliasing but it couldn't be it). I'm not sure what was it.