开云体育

I did the SMA build and the cal.
The SMA variant is up there with the best BNCs (+/- 0.04dB) with one notable difference.
Less deviation within the stated flatness range. So, overall the amplitude curve is smoother and contains no sudden resonances, albeit small ones which don't affect the peak to peak results.

It is possible that the signal flow changes made to accommodate the new 4-40 clamped housing could have had some impact on this, but I suspect the SMA integrity helped a bit. See the Excel data.

Hi Ancel,
I have a question about this latest version. I looked at the pictures and
the edge launched SMA connector looks "odd" on two accounts:

1) As near as I can tell the PC board must be about half as thick as the PC
board the SMA connector was designed fort. All the edge launch SMA
connectors I have ever seen are designed for a 0.064" (0.162cm) PC Board so
it mates directly to the PC board surface (top and bottom) allow for tight
coupling to the ground plane. It is hard to tell for sure from the photos
but it appears you are using PC Board that is about 0.032" (0.081cm)

2) It is usually good RF practice to run the copper ground plane right up to
the SMA connector. Your copper stops about 0.1" before the edge the SMA
connector is on. That would cause some reflections, particularly at very
high frequencies in the GHz region since there is no microstrip at that
point. This gap may not be important at 1GHz but it just seems like good
practice to avoid dielectric gaps in the microstrip.

Dennis Tillman W7PF

Good observations, Dennis.

The FR4 is 0.8mm, so there's a little standoff with the SMA launcher. No 0.8mm SMA launchers I know of.

The PCB perimeter gap you see is 0.050", not 0.100". Now since the SMA launcher is grounded at all 4 points the net 'inductance' increase derives from 0.0125" ....which is too small for significant phase effects at 1.050 Ghz.

I built another 'experimental' SMA unit. But I couldn't get the performance as good as the first. I got +/- 0.07dB over the range vs +/- 0.05dB. I fab the boards (PCB trace etch tolerance) and then there's component & connector tolerances and soldering thickness variations (which affects inductance a little).

What I am seeing is a little bit of VSWR 'ripple' across the span, equivalent to around one complete sine wave; (+/- .02dB worth.)

But I suppose I am splitting hairs as there's always a bit of performance difference between parts.

The positive outcome though is I reduced the component count by one using PCB capacitance parasitics to do the 'tuning' and got a bit more ground plane % coverage out of it as well as a more solid mechanical assembly.

I'll build one more SMA unit to nail down the dimensions of the parasitic cap, but it's around 0.25pF based on the permittivity of the 0.8mm FR4.

I'll put up the extra SMA units on Ebay, as tuned units. There doesn't seem to be any fundamental performance gains in using the SMA vs the BNC at 1Ghz. There is miniaturization and better build convenience by not converting the BNC connector into a launcher.

As it stands the stock PCB design can use both the BNC launcher mod. and the stock SMA Launcher with no alterations or 'nibbling' etc. Similarly, the latest compact enclosure accommodates both varieties.

I'm taking pics as I go along to do the build blog for the DIY page.

Ancel

I should have said, no SMA launchers I have used are 0.8mm.


I need the 1.6mm version though, as the PCB design accommodates both the BNC mod and the SMA launcher, the 0.8mm clearance is required to maintain the double sided ground plane perimeter. Otherwise, I'll need more vias, which is avoidable with this approach.