Help me improve this power splitter

[EggertEnjoyer123]

You can definitely reduce the space between the resistors further.

Here is an example of a board I found online. As you can see, the resistors are placed as closely as possible.

(You can also squeeze a little more performance by soldering the resistors upside down).

[khs]

Here a little bit smaller divider similar to the divider described in the S-4 manual on page 2-16.

It's not very to difficult to make. But soldering the three 16.5 ohm resistors requires a little bit patience.

I've made it for a little bit more relaxed TDR measurements to avoid to ruin my SD-24 sampling head with it's +-2V max input voltage. The 7K sampling heads like the S-6 has +-5V instead but it can be destroyed quite easy too if there is no absolute care to avoid electrostatic charges.

The used S-4 has +-5V max input voltage too, but with two HP 8493A 10 dB dividers the input is quite save. Due to the divider there is no need  for a loop thru (and hard to get) S-6 head.

There should be no problem to touch the DUT, because every voltage spike must thru the low impedance 50 ohm divider. Without dividers the spike would reach the sampling bridge first and destroys it and then it is absorbed by the 50 ohm termination. But this would be too late. The 50 ohm termination cannot protect the sampling bridge if it's behind.

The S-4 / S-52 / 7S12 runs in a 7854 mainframe, so it's easy to average the signal and make cursor measurements.

As shown the risetime is about 85ps, it's limited by the 12 GHz bw limit of the attenuator. There is no effect removing the divider regarding the risetime.

The DUT is a more or less 8 cm long 50 ohm cable, so the setup should be work fine down to 2 cm wires.

[tszaboo]

So I have the V2 boards. They are a bit more pain to assemble, soldermask is invented for a reason.
There isn't a huge difference between the Y and the Delta configuration. Y is slightly more wiggly and better SWR, Delta is smoother. Both are within 1dB. The Delta can be easily improved to higher frequencies if better resistors are used, up to the point when the SMA is the limiter. Since it needs 50Ohm resistors, not 16.6 so RF specific parts numbers might be used.
SWR is 1.25 for the Y and 1.35 for the Delta between 400-4000 MHz.
I consider this complete. I do a bit of cleanup and release it on github.

[tszaboo]

Files:
https://github.com/tszaboo/RF_SPLITTER_V2

BOM cost:
~4 EUR for PCB (5 PCBs JLCPCB)
~1 EUR for connectors (unbranded)