On Tue, 14 Nov 2023, jmr via groups.io wrote:
If you look at the RLC D-1814 photo I posted yesterday, you'll see that it
is not all that simple. The construction is VERY rigid -- it is made from a
heavy solid block of copper (or brass? can't tell as the thing is heavily
silver plated) with just small holes for the connectors' central pins into
that small cavity. And that is not all -- it has a clearly visible hole
between the resistors. It is definitely not just decorative or venting or
machining artefact. It was a THROUGH hole to the other side but has some
slug (or screw) in it from behind that was used for some tuning, then excess
from the bottom was cut off and sealed/fixed by solder. Looks like a HARD
solder, not regular Sn/Pb one.
That direct connection method using three resistors and three connectors soldered together would be a quick way to do it, and it should be possible using 1206 SMD resistors and SMA/3.5mm connectors. I think some care would be needed to make sure it was rugged to withstand repeated use with a torque spanner. It would have to be very stiff in construction to avoid stress being transferred to the SMD resistors.? As you say, any inductance in the launch connection will degrade the VSWR up in the GHz region. It will be possible to compensate this but it would be quite fiddly to add gimmick compensation using metal shapes. However, I think it would be possible to do it although I'd recommend the use of a decent stereo microscope during construction. It does also have the advantage that it avoids the extra loss in any PCB traces.
I opted for the PCB method because I wanted the same footprint as the 3.5mm connector version of the 11667. It's nice to handle and is very rugged. It's also easy to include any compensation in the PCB artwork. The Rogers PCB does add some loss due to substrate and metal losses but this is negligible up to about 3GHz.
One other thing to consider is that a key performance parameter is the source VSWR of the two output ports when used in a closed loop levelling system. The VSWR needs to be very low. I measured mine using a virtual balun in my CAD SW. I exported a three port s-parameter file of the splitter using the VNA and then attached a perfect balun in the simulator. This allowed me to drive the outputs differentially and create the virtual ground at the common node of the two resistors. This (kind of) simulates it being used in a closed loop system as it creates a virtual ground.
Here's the plot of VSWR up to 3GHz when I drive backwards into the two outputs differentially. The result is very good indeed. I don't think it's enough to just measure the input port VSWR with the output ports terminated. There are probably better ways to measure the source VSWR but the differential method at least gives me a good indication that the splitter will provide a low VSWR to the DUT when used in a levelling loop.
---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*
Sergey Kubushyn