Since the discussion on Siglent's counterpart to this generator turned into the direction of evaluating in how far an I/Q modulator "hack" could be possible, and the DSG800A series offers much more functionality than Siglent's products (internal baseband generator), I decided to take a closer look at Dave's teardown photos of the DSG815.
It's actually quite amazing how simple the operational principle of this generator is. As always, the difficulty is located in the details. Generally speaking, the DSG800 works in the upper frequency range (probably >300MHz) as a plain PLL VFO with several, switchable frequency dividers and a whole bunch of configurable low pass / band pass filters. The whole unit is full of switching diodes and non-reflective switches (HMC284). The first LO is split into three ranges and is almost a carbon copy of the corresponding section of the DSA815 spectrum analyzer. The reference frequency for the HMC704 PLL is generated by an AD9781 DAC (and not as I initially mistakenly indicated in my scheme, supplied directly from the 10MHz reference oscillator). This way, frequency and phase modulation is possible. The second channel of the DAC is used to generate the LF output signal. The DAC clock is supplied by the highly stable Z-COMM CRO3640B-LF VCO (3.64GHz), controlled by an ADF4106 PLL, divided by eight, hence it's
405MHz 455MHz (already forgot how to properly divide by two...
).
To generate lower frequencies, a minicircuits ADE-12MH mixer is used to mix the 1LO output with 910MHz from the aforementioned synthesizer.
In my scheme, I used orange for the signal path on the visible side and magenta to approximately show what's going on on the hidden side of the PCB. I may have missed a filter or two on the hidden side, but the working principle should be covered failry accurately.
It gets obvious that the I/Q add-on board contains more circuitry than just a few ESD ptotection devices and some interconnections, it's got to contain the complete I/Q modulator, and as it seems also considerable circuitry to generate the I/Q baseband signal. So no easy hack, if possible at all...
But a possibly more interesting conclusion on the circuitry may be that it should be possible to output much higher frequencies than 1.5GHz since it's possible to route the oscillator signal trough the instrument without passing any frequency divider. Moreover, all the semiconductors used should be able to handle up to approx. 4GHz (I didn't believe that initially before I did the reverse engineering). So we may actually be up to a surprise if we compare this 1.5GHz version to one of the higher-frequency specimen of that model range... Another amazing detail: The PA is a NPTB00004A, capable of >5W at 4GHz!
So I'ld say this unit contains a very good hardware and there may actually be a chance for "improvement" of the entry models. What kind of calibration would be necessary afterwards is written on a different page, though...