I have tried to achieve that with a different type of device, the "138M-4.4G USB SMA signal source/signal generator/simple spectrum analyzer" (
https://www.ebay.de/itm/165015769420). This device only features one ADF4350 instead of two ADF4350/ADF4351 in the NWT4400.
As such you can either use it as a signal generator or a spectrum analyzer, but not both simultaneously as a tracking generator.
I tried to implement aTG by synchronizing both devices. The result, while functional, is useless due to the much increased sweep time required.
Imagine you want to sweep over 500 samples. Instead of launching one command and getting 500 samples, you need to launch 500 individual (!) commands, receive 500 times the single sample, obtained while first setting the signal generator to each of the 500 frequencies. Add a short waiting period for the generated signal to settle.
It is just useless.
The alternative is, as already mentioned, using a noise source like this:
https://www.ebay.de/itm/285099911097Instead of sweeping the signal generator through the selected frequency range, you just generate a noise signal over the whole bandwidth.
The possible measurements are in theory the same, you won't be doing phase measurements anyway with a TinySA.
The reality is not as bright, though, as the noise source is not so stable, the amplitude varies considerably over the whole specified frequency range, but it is the cheapest and most viable option.
Of course, there are better noise generators, but prices increase exponentially with higher specs.
Regarding TG, you can only use an external device, if it can be hooked up to the swept frequency signal - some spectrum analysers offer this signal. Doing so by cheating the FW into using individual measurements is, IMHO, not a viable solution.
I leave this link for further reading:
https://www.ednasia.com/use-a-low-cost-noise-source-as-a-replacement-for-a-tracking-generator/Kind regards,
Vitor