LibreCAL detection and respones spike anomalies

[AnotherPseudonym]

I am trying to use a LibreVNA with a LibreCAL and have two, possibly related issues.
Setup: HW 1 Rev.B FW 1.6.4, Cal FW is 0.3.0 and detected as not a clone (I forgot the specific message, but it seemed OK). I am using two Sucoflex 104PE SMA cables which test out OK on other equipment.

1. If I try to auto detect the ports only port 2 (of the LibreVNA) detects correctly. I've tried using the LibreCAL ports 1 & 2 and ports 3 & 4 with the same results. I can manually select, but it seems odd that it wouldn't detect.

2. There are some spikes in the responses at both low frequency and around 4.3 GHz. The accuracy below 3 MHz also seems to vary a lot between calibration runs.
   These are with manually selected ports and both ports calibrated (nothing manually selected on the LibreCAL).

With LibreCAL through on the connected ports (I understand it won't be a perfect through, but I would expect to see similar curves for S11 to S22 and for S12 to S21):

With a real through:

Are there any known issues that would cause this?
Is there anything I should do to diagnose further?

[jankae]

I don't have any concrete tips right now but am also not quite sure what your screenshots show. Could you please tell me what was connected to the LibreVNA for each one and also what kind of calibration was active and how the calibration was achieved? (Manual kit or LibreCAL).

Can you make normal measurements without the LibreCAL on other DUTs or do you see strange spikes below 3 MHz and 4.3 GHz as well? I think the first step is determining whether you issues come from the LibreVNA or the LibreCAL.

[AnotherPseudonym]

I do see the spikes with other things connected. I don't not seen them generally.

I hooked up a spectrum analyzer while NOT sweeping (triggering ) shows a fairly strong peak at 4.314 GHz.

Looking at port 2 with the SA shows more peaks, but of much lower intensity:

For the setup 100 kHz-6 GHz log, BW: 100 Hz, 2001 pts., LibreVNA port1-Sucoflex cable-LibreCAL port1, LibreVNA port2-Sucoflex cable-LibreCAL port2, after using electronic calibration with manual selections of Port1: Port1 and Port2: Port 2 and that calibration having been run and activated by the GUI, leaving the LibreCAL connected without manual configuration (I assume the MUXes have nothing selected, so probably open-ish), I see (log and linear):

With the above setup, except that the LibreCAL was configured to have Port1 through and Port2 through by using the control buttons, I see:

Replacing the LibreCAL with an SMA FF (which should be a fairly good through, especially at low frequencies), but still with the calibration from above active, and using the same span, I see:

If I deactivate calibration and measure real standards (no-name fleabay short and opens and Mini-Circuit ANNE-50X+ loads) directly connected to the LibreVNA, I see
Shorts:

Opens:

Loads (Mini-Circuits ANNE-50X+ , phase traces turned off):

Sucoflex cable:

[jankae]

Thank you very much for the detailed answer. I'll only respond to a few details right now because I think you need to solve them first before you do further tests with the LibreCAL.

I hooked up a spectrum analyzer while NOT sweeping (triggering ) shows a fairly strong peak at 4.314 GHz.

That 4.314 GHz tone is there even when the VNA is inactive? (E.g. VNA connected via USB but the LibreVNA-GUI is not running). The tone should not be there.

Your uncalibrated S11 data also looks really strange, especially for the open and short standards. -40 dB at the low frequency suggests that something is seriously wrong with that port. S11 should be much closer to 0 dB, similar to your S22 data. Maybe you could open the VNA, look for obvious problems at the port (e.g. mechanical failure, bad soldering) and post a few pictures of it?

[AnotherPseudonym]

Thanks for the reply!

That 4.314 GHz tone is there even when the VNA is inactive? (E.g. VNA connected via USB but the LibreVNA-GUI is not running). The tone should not be there.

Yes, it is there after doing a single sweep and waiting for the sweep to end. It is also there when powering up the LibreVNA without the GUI or anything else talking tothe LibreVNA.

With an e-field probe and brief power ups (less than a minute each) without the lid on, the signal seems to peak around here (the base image here is from your helpful docs):

I took some pictures under a microscope which has independent polarizers on the ring-light as well as the objective (this allows cross polaraization which can be useful, but full cross polarization looks different than what you would normally see). For most of the shots I have it not fully cross polarized.

This is an image near the area circled above. I assume the chip has some thermal compound on it.

Further zoomed in, without and with cross polarization, there may be what appears to be some flux residue, or residue of a solder splash. It's not a solid splash. The blue stuff on the ADL8501 pad toe seems to be thermal pad material.

Here's the Port 1 input cap. The shiny portion of the solder filet looks black because it's shiny, which is good.

Here's the other chip connected to port 1 (obviously not a Qorvo part as in your block diagram, but maybe OK):

Looking around generally, I see some fibers (or maybe flux?). These are near the FPGA.

There's a(nother) solder splash near the middle switch, but it looks adhered to the soldermask, so should not be an issue at present - if I rework the board I'll clean it off with the other fibers.

I have hot air and moderate success rates with mounting QFNs. Thinks it's worth trying to replace the port 1 ADL8501 and/or other components?

[jankae]

not a Qorvo part as in your block diagram, but maybe OK

Yes, that is okay. My latest LibreVNA uses the same part.

It is also there when powering up the LibreVNA without the GUI or anything else talking tothe LibreVNA.

Strange but not necessarily that much of a problem as I thought before. I probed a few of my LibreVNAs and found one that sometimes has a similar peak when not sweeping (although at 4.6 GHz and only -35 dBm). The tone disappears when I enable the LO or the port 1 mixers/amplifier. Sometimes the tone also disappears on its own, I can not replicate it all the time. That particular LibreVNA shows no problem when making VNA measurements.

Maybe there is something oscillating in the ADL5801? The situation seems to be more severe on your LibreVNA. Can you please start the manual control dialog and play around with these checkboxes?

Does the 4.3 GHz tone go away when you check them?

Thinks it's worth trying to replace the port 1 ADL8501 and/or other components?

Something is definitely wrong with your LibreVNA but I am not quite sure what it could be. If you are comfortable with replacing the ADL5801 it is worth a shot. Personally, I would probably try to debug the problem a bit more first but that decision is up to you. Since your S11 measurements are quite strange at low frequencies, you could try setting the source/LO to a low frequency (such as 50 MHz) and probe the signal along its path to port 1. At such a low frequency this could even be done with the high impedance probe from an oscilloscope.

Questions to get you started:

• Does the signal level drop more than expected at any point between the source PLL and the port 1 SMA connector?
• Does the signal level drop significantly between the port 1 SMA connector and the ADL5801 input? (leaving the port open for a large reflection is best for this measurement)
• How high is the amplitude of the 1.IF (62 MHz) and 2.IF (250 kHz) at port 1?
• Does the circuit around port 1 draw more/less current than at port 2? There is no convenient way to measure this but if you have a thermal camera, large differences in temperature between the (symmetric) ports would be suspicious