Noise performance test results

Results of extensive noise floor testing of the dual detector VNA v4.0 and v4.1 boards

This page has some information about the noise tests I completed on the dual detector boards. I first tested a v4.0 revision board, then v4.1 one which was of an improved layout. A set of tests was suggested by Paul, N2PK, to evaluate the noise characteristics of both detectors on the new board and if and how they influence each other. For example, how the noise floor of one of the detectors changes if the second detector becomes driven with the RF DDS signal? Is there crosstalk between the channels? How much? Because the detectors share same PCB space, that would be a valid concern.

I used Paul's set of DOS-based utilities, which he kindly made available to me, including his program that can sample both channels simultaneously. The data files then were imported to MS Excel and the charts were built. I publish them here to keep you informed about the work done and let you decide if you see any problems with the new board layout in this context.

Noise testing of the board turned to be an exhaustive experience. Before and during each test a calibration sweep would need to be done. With the detectors outputs that low any small thing such as physical board disturbance or moving the cables around may affect the readings. Another thing was I worked with the lowest sampling rate to maximize the dynamic range, and multiple sweeps through 1 to 60 MHz were taking significant time, especially if averaging was used. I found the thermal drift to be the biggest factor bringing instability in readings. I did not have any heatsinks installed on either DDSs or Valpey-Fisher oscillator. The latter was the guy who generated most of the heat. Now imaging running a test that takes several sweeps taking a few minutes each. By the end of the test you would be 15 to 20 minutes away from your calibration point. That would not be much of a problem for most of measurements, but with the noise tests we talking -100 to -120dB levels, pretty sensitive conditions. Take this into account when analyzing the charts.

Test one was done on the Detector 1. Its input was kept open. The output was monitored while sweeps were done with the second detector input open, and then fully driven with the RF DDS signal. This test utilizes single ADC sampling process. Ideally there should be no change in the Detector 1 noise floor.

Test two was done on the Detector 2. Its input was kept open. The output was monitored while sweeps were done with the first detector input open, and then fully driven with the RF DDS signal. This test utilizes single ADC sampling process. I think thermal drift caused the yellow line shift (sweep 2) from the pink one (sweep 1), which was closer to the calibration point.

Test three was done on the Detector 2. First, both inputs were kept open. Then Detector 2 output was monitored while two sweeps were done with the first detector input connected to the reflection bridge (some signal injected in the Detector 1). This test utilizes simultaneous ADCs sampling.

-Update March 10th. - A comparison test showed that a unit built on the classic boards (noise floor figures provided by Paul, N2PK) beats the above noise figures from 10 to 15dB for the Detector 2 (for the 'no averaging' option ). The classic configuration has Detector 2 built on a separate board. Possible factors contributed to the higher noise for the single board layout are: it was tested in the open air (no enclosure),it did not have any heatsinks installed on either DDS chip or Valpey-Fisher oscillator, it did not have the ground plane bridges over the SMD components on the back of the board, the detectors share same power circuits, and it is a single board after all.

I will try to run another set of tests this weekend, but in the enclosure and may be with the heatsinks mounted, to see how much a differential design is susceptible to the common mode ambient noise.

-Update March 14th. - One more set of tests has been completed, this time with the VNA mounted in the Hammond enclosure. Noise figures are somewhat better now. Below are the pictures for the Detector 1 & 2. For each detector two sweeps were taken with the other detector open, then driven with the full amplitude RF DDS signal.

You can see that after mounting the board in the enclosure the noise level dropped by approximately 10dB down below -110dB mark (the blue curve) for the most of the frequency range, with the other detector being quiet. The result seems to be very satisfactory to me.

I've mounted almost all connectors, including the SMA connectors, on the board, and will post the latest board picture so you could get an idea how it all looks like when assembled.

-Update May 4th. - Now the layout has been updated to board revision v4.1. I've assembled one that arrived recently from the PCB house and here is how the noise level of the new v4.1 board looks like:

I think one can see a clear improvement by a few dB versus v4.0 board. The VNA was not in the enclosure neither it was shielded by any means at the time of the test. One could expect a little more improvement after mounting the board in the enclosure.

-Update May 11th. - I launched the second v4.1 board and it also came out OK. I ran the usual noise tests for both detectors and then superimposed the charts for the first and the second board. The result is below.

It can be seen that both detectors curves are pretty much consistent to each other and to the other board detectors. Now, after making two v4.1 boards from the first manufacturing batch alive, I am quite comfortable with the layout and will place another manufacturing order for the people who could not have the boards kit from the first batch. You can reach me at the email address below if you want your name included in the list for the second round. Thanks and have a good time assembling your VNA.

Contact: miv@makarov.ca