On Sun, Nov 12, 2023 at 08:22 AM, Jared Cabot wrote:

I'm nothing more than a highschool dropout domestic electrician who dabbled in BASIC 30 years ago and some Windows batch script stuff a while ago

Don't let that stop you from trying.? Your ideas are good, and as you have seen, there are always people here willing to offer suggestions, like how the "big boys" (HP and others) did it.? Learn your VNA, they are wonderful!? There are several groups.io VNA groups, and a number of good application notes on VNAs.

Try this and KISS to start. Drawing attached. Butt the 3 connectors of you choice?
together. Solder body to body. You can bridge the wide opening with brass shim.
Nip the center pin as far back as you can...
Wick your solder to the small axial, say 1/8 W or 1206 film R. If you use a pair of 100 ohm
in parallel, parasitic tend to assist in achieving an improved "50 ohm" R. Measure that?
system and see how well you do.?

I won't even be able to begin to know how to use any simulation software, especially needing scripts etc...
I'm nothing more than a highschool dropout domestic electrician who dabbled in BASIC 30 years ago and some Windows batch script stuff a while ago, my simulation abilities basically end with something like the 'Saturn PCB Toolkit'
For example, I'm still figuring out the 'real world' differences of microstrip and coplanar waveguide...

I always seem to start a project thinking 'it's just a few parts, it can't be that difficult' then soon realising I'm staring into the bottomless pit. Something something not knowing enough to know what I don't know...haha

Maybe I'll just throw some axial resistors in a box and call it good enough. I haven't even figured out how to measure the performance of the thing on my VNA (Or even how to use a VNA) yet.

开云体育

Open EMS is very good I teach it on university. A great alternative to paid software.?

You will find that a significant limitation to performance is the launch to the resistors.
Assume the resistors are perfect. Set the connector launch to a simple model of series L and shunt C as there is a discontinuity at the connector
as soon as you exit the coax structure. I set the L to 500 pH and the C to 500 fF.
That limits the return loss at the output port to ~ 12 dB at 3 GHz.
Be prepared to dress the connector. I file the pins flush to the dielectric and file the body to a flat to create a launcher.
Not perfect but going in the right direction. Your figures, pictures, ?demo way excessive series inductance.

Alan

I have an 85097 driving a 3.5mm F-F 1-26.5GHz module. I don’t have anything else to compare it to, but I could take high resolution pictures of the pcb. If you do the same, maybe we could figure out how they differ.

Is it possible only early ecal modules required a change?

Those splitter PCBs linked to have a serious flaw. They use 3 50R resistors when they should be 16.67R (Zo/3). If they can't get that right what confidence is there that the PCB impedance is correct??

I have just found these edge launch N connectors:



Manufacturer website:



They look intriguing...

The original splitters were made with discrete components without a PCB. To maintain a 50R system they are enclosed in a slpit block with machined cavities whose diameter matches the component dimensions to 50R. For this type of construction there are often tapers to accomodate changes in diameter. Tuning screws were used to compensate for parasitic effects.?

More modern devices may use striplines on PCB, ceramic or sapphire substrates. HPAK use a lot of sapphire. Even striplines need care in the packaging. Often RF absorber will be required to suppress resonances in the enclosure.

As noted by JMR the best DIY approach is to use striplines. Standard N types are difficult to "launch" into a stripline. See for the amount of effort needed to get a good match even at 500MHz. SMAs designed for stripline applications are easier to find.

Robert.

Ok, now you all have me thinking about stripline on rogers material.... :D

I'm looking at RF resistors, and now wondering about wattage ratings.. Is this an important consideration? Assuming my most powerful function generator is a Tek SG5030 which is good for 5.5V peak-peak, do I need to think about this specifically?
Am I right in saying that for 5.5V P-P = roughly 2Vrms, so with 50ohm series resistor in my splitter plus 50ohm termination at the load = 100ohms, I'll be looking at 0.04W? Or is it more complex than that?


Jared

On Sun, 12 Nov 2023, AMiTHA wrote:

OK, I've got to the resistors in that RLC D-1814.

Those are just 2 0603 resistors soldered fron the input N-type connector to
2 output ones.

However, the entire thing is a single monolithic block of copper (or brass?
can't tell as everything is heavily silver plated) and those resistors are
in a small cylindric cavity. The cavity is machined into that block and have
a cylinder on the open side closing it from all sides. The resistors are
almost touching that cavity walls, no any insulation there.

There is a threaded plug closing that cavity under the label (I suspect
original 11667A is same). The cylinder plug closing the cavity is under that
threaded one. It was not an easy task to remove that threaded plug -- it is
sealed with silver-filled epoxy and it is brass (or copper? probably brass)
so it is almost too soft to break that epoxy (it is slotted) when driven
with a big screwdriver.

Will take a couple of photos later today or tomorrow after more urgent
things are done.

Hi,

Likely the tracking performance won't be flat above 500mhz, I would suggest
to use sma connectors, and and smd resistors. Why not look into Weinschel
1670A splitter. It's way cheaper, and gives about the same performance with
the keysight splitter. I have been using this for years now they are very
robust.

On Sat, 11 Nov 2023, 13:38 Jared Cabot via groups.io, <jaredcabot=
[email protected]> wrote:

I'm currently in the process of building my own DIY 11667A power splitter
as the cost of these things from Keysight is insane, and even second hand
is rather too expensive.
I can only test this one to 3.2GHz as that's the limit of my VNA (Siglent
SVA1032X) but that's all I need for now which makes things a bit easier.
I plan to use two 100ohm SMD resistors in parallel, oriented back to back
so the resistive side of the resistors are together to hopefully reduce
parasitics a bit.

I have my diecast box with 3x N-connectors, but I'm wondering if the
resistor placement in the box will have an effect on performance?

For example, these 4 basic connection ideas:



Are any of these (or even something different) better than the others to
save me running through all permutations, or does it not matter?


Thanks!
Jared

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*

You just know the engineer just wrote "ALC" because it is painfully obvious (to him)...and then the writer had to look it up.? Unfortunately, "Ar" comes before "Au."? I am actually surprised that got through.

with three 16.7 ohm resistors, if one has a 50 ohm source, one 50 ohm load and 75 ohm load, the effective load on the source is about 55 ohms.

situation can come up where a 50 ohm power measurement device with a 75 ohm load on an SMB might arise. The result isn’t horrible if there’s some reason not to use a minimum loss pad.

Yes, the label is a bit strange and it should really say automatic level control. It's Keysight's drawing that I've borrowed and I've seen them use this misleading label for ALC in another publication as well.?

It's clear that Jared wants to make a splitter that uses two 50R resistors (as in the 11667A) and that is what I also made 10 years ago using four 100R resistors to make up two 50R resistors. So I posted up some performance plots which hopefully will show that excellent performance is possible even with some cheap SMD resistors. I think the resistors I used cost about ?0.01 each back then. The key to good performance is to experiment with various resistor types and build it on a decent PCB. I used Rogers 4003C at 0.02" thick and this makes it possible to use RF lines that are fairly similar in width to the 0603 resistors. I used high quality (proven) SMA end launch connectors that I knew worked well with 0.02" thick Rogers 4003C. It actually worked really well to beyond 6GHz as you can see in the plots. I also modelled it using Sonnet EM before I built it and from this I had a good indication that it would work well at 1GHz.?

My other device uses three 50R resistors in a classic delta pattern and this was an attempt to get similar RF performance to a HP 11636A up to 1GHz. You can see in the plots that I achieved this, including the phase balance. It's also similar to the 11636A up to 3GHz. I used a very accurate milling machine to make the PCB to try and get the phase performance the same over a wide bandwidth. I again used Sonnet to simulate and optimise the PCB artwork for lowest VSWR over a wide bandwidth.

开云体育

This result a good reason to expend power on two resistors .?

开云体育

On Sat, Nov 11, 2023 at 09:38 PM, Patricio A. Greco wrote:

Please note that the metrology splitters has only 2 resistors… this is becuse are used to build a close to ideal equivalent reflectivity coefficient with a leveling loop.?

?

My homebrew clone of the 11667A splitter uses only two 50R resistors and one of the plots in my previous post is for this splitter. I use it with a levelling loop as in the image below. I've also added a plot of the frequency response of one of my old Agilent spectrum analysers when measured using this setup.?