The nanoVNA-H connectors are end launch PCB connectors and they have a decidedly spongy feel as 1Nm torque is approached. This was due to flexing of the PCB and was likely to lead to track cracks in the longer term.
Specs for SMA connectors range from minimum of 0.2Nm torque to maximum of 1.7Nm, but 0.6Nm and 1.0Nm are common commercial practice.
Some nanoVNA sellers state:
As the SMA ports are made of cast copper, please not connect hard 50-7 / RG213 and other cables directly to the SMA ports through M-to-SMA connector to avoid damaging the SMA ports. You can use the included SMA cable to connect to the SMA port as shown in the picture below, and then use M to SMA connector.
Clearly Chinese Cheats, they will say anything to make a sale and anything to avoid commitment to quality. These connectors are very unlikely to be copper, but are likely to be a copper alloy: brass. What they also avoid in the above statement is claim for PCB damage due to flexure of the SMA connectors torqued to accepted industry torque for reliable connections and measurement.
Above is a pic of a modification to reinforce the connectors. This article sets out the analysis of just the solder joint within the cross section of the brass pieces.
A side effect is that this modification bonds the ground planes for the input and output parts of the nanoVNA via the brass bar where they have been kept isolated to some extent.
I should note that there has been much discussion online as to whether the noise floor of the nanoVNA is degraded by the shields fitted to the board, and various modifications to the shields. Some of this discussion proposes that the issue is mostly around the mixers and noise loops, and I note that in -H designs prior to v3.3, the mixer power supply was not adequately decoupled. It is possible that electrical connection of the SMA connectors in this way degrades noise performance at some frequencies. No significant change was observed in the noise floor of s11 or s21 channels from 1 to 300MHz (I don’t regard instrument performance to be good above 300MHz). I have not seen credible evidence of degradation of the nanoVNA-H v3.3 build.
If indeed bonding the two SMA connectors close to the instrument increases the noise floor or has other performance impacts as suggested, it questions whether the currents on the exterior of the coax influence measurement (which it should not) and it questions whether two port measurement fixtures or adapters should be attached close to the nanovna.
(See also Reinforcement of nanoVNA-H connectors – performance discussion.)
At first, the strength of the butt soldered joint might seem a simple case of beam analysis where the beam is of cast solder of the same cross section l x w as the soldered joint.
I think that is not a good model, because the very thin solder layer probably invalidates the assumption of a neutral axis in the middle of the section. My initial thinking is that the strength can be worked up assuming the the solder joint ‘hinges’ on the compression side of bending and the bending moment limit is that value that stretches the extreme fibres on the far side to the tensile strength of cast 60/40 solder.
On that basis, we can assume that the pressure within the solder rises linearly from zero at one side to maximum at the other, ie that \(p(x)={\sigma}\frac xl\).
So the total torque when loaded to σ is the sum of the pressure times area times distance for the incremental areas across the section.
\(T=\int^l_0{\sigma}\frac xlwxdx\)
\(T=\int^l_0{\sigma}\frac{wx^2}ldx\)
\(T=[\frac{{\sigma}wx^3}{3l}]_0^l\)
\(T=\frac{{\sigma}wl^2}{3}-0\)
Taking \({\sigma}=50MPa,w=0.0016,l=0.00635\):
\(T=\frac{5e7\cdot0.0016\cdot0.00635^2}{3}-0\)
\(T=1.075 Nm\)
So, it looks promising.
A test piece was constructed and tested to destruction at 1.8Nm.
It is a little stronger than expected, possibly due to the solder used, possibly contribution of the small excess outside the brass cross section, and it is only a single test sample. The experiment suggests that the analysis approach is better than treatment as a long slender beam (which would result in a lower predicted strength).
The final test is that the instrument has been in use for weeks with the reinforcement and connectors torqued to 1Nm with no signs of cracking or sponginess. Beyond this test period, it is my intention to tighten the connectors to just 0.6Nm (common commercial practice).