Using the same RG58 cable and changing the generator to 50Ohm resistance (from High-Z) and including an in-line 50 Ohm terminator, gives me results well within the instrument's specifications.
Some clarity would be nice on this - I'm trying to "get an RF clue." I was under the impression that the High-Z function emulates a 50Ohm termination... Would anyone care to educate me on what my misunderstanding is?
Thanks to all those that have participated in the thread.
As far as I understand it all the "High-Z mode" does in the DG4062 is adjust the amplitude value, it doesn't change the 50 ohm impedance of the generator. If you set High-Z mode and then ask for 2Vpp you'll get 2Vpp at the inside point of the 50ohm source impedance as the assumption is there is no voltage drop across the 50 ohms as it is connected to a High-Z that passes no current. If the mode is set to 50 ohms and you set 2Vpp then the voltage at the inside of the 50 ohm internal impedance is set to twice this (4Vpp) to allow for the fact that there is a 50 ohm load so the voltage at the output will be one half (half the voltage across the internal 50 ohms and half across the load 50 ohms).
When you set high Z mode the source impedance was still 50 ohms. You then connected it to the 1M/16pF input of your oscilloscope. The problem was the 16pF capacitance, at 1MHz the impedance has a magnitude of around 1k much higher than the internal 50 ohms so the measured voltage is close to the value you set of 2Vpp. At 160MHz the impedance has dropped to only about 60 ohms which is similar to the internal 50 ohms (they are different phases of course, one being real and the other capacitive) so the voltage will have almost halved.
With the 50 ohm pass through load the load starts at 50 ohms (the 1M + 16pF being high impedance) and at 160MHz it is down to about 1/sqrt(50^-2+62^-2) or about 39 ohms which is not that far off the original 50 ohms (I've somewhat simplified the fact that you have different phases, though I've allowed for it a bit in calculating the 39 ohms).
The most accurate results will be obtained with a scope with a proper 50ohm input impedance.
I've not touched on the effect of the cable. With the 50 ohm pass through the 50 ohm cable is terminated in something close to 50 ohms so is well matched so still looks like 50 ohms at the generator. Without the pass through the 1M/16pF impedance will be transformed by the cable but as the wavelength even at 160MHz is still quite long (1.875m in air, maybe one third of that in the cable depending on the dielectric constant) for most of the frequencies the unmatched cable will look like extra capacitance at the generator and will look inductive between the generator and the oscilloscope. It is complicated to work out the effect accurately, you could do it on a Smith chart. The point is that where the cable is not well matched it will have an effect, where it is well matched (50 ohm impedance each end) it should just introduce a slight time delay.
I hope all the above makes sense.