Analysis and Characterization of Devices Database

[bvacaliuc]

There is a need to understand the behavior of the RASDR devices at the operating points that they are likely to be used. With a broadband noise source of known output power, one can sweep across the RASDR tuning range and gather statistical information of the performance of the receiver across all of its ranges. There should be no outside sources of energy except that which leaks in from the case, or is produced by internal components or the cables, etc. Knowing the bandwidth of the receiver as well as the gain/attenuation settings in the signal path allows one to predict what the response should be and compare that to what the response actually is.

For example, I was surpised by the following observations in which the only variation was sampling clock frequency when looking into an external 50ohm load SMA screwed on to the RX input port:

50ohm load, Fcenter=408.1MHz, 1.5MHz bandwidth, 10Msps

50ohm load, Fcenter=408.1MHz, 1.5MHz bandwidth, 8Msps

50ohm load, Fcenter=408.1MHz, 1.5MHz bandwidth, 4Msps

50ohm load, Fcenter=408.1MHz, 1.5MHz bandwidth, 2Msps

I think that one would make these observations and populate a database (or spreadsheet) with the metrics and then create some kind of a 'heatmap' of the performance changes as different parameters are varied. In a sense, 'visualizing' the behavior of the receiver under some handful of performance metrics. Here are some metrics I am thinking of:

1. DC balance: how far away from 'correct' the average value I and Q signal components are. For a calibrated noise source, they should have the same average value, and based on the bandwidth and gains that average value is a known quantity.
2. baseline flatness: So what I see with the RASDR, is that even looking into a 50ohm load, the receiver outputs some spurs and not all sample rates produce a flat spectral baseline. This is bad, and represents a pretty strong "quality" metric of the receiver system itself. Knowing where the "bodies are buried" is important to using the receiver effectively.
3. SFDR: this one is a bit tricky, as it needs a signal injected in it. But I can do that now. This tests the liklihood of intermodulation effects to interference signals. If we can quantify how the RASDR receiver is affected by signals inside the box as well as what might come through the LNA, we can develop a better filtering plan. I know that the LMS6002D has a pretty bad IP3 spec (-1dBm). Some references: IP3 Demystified, Cascaded 2-tone 3rd-order intercept point.
4. sensitivity: this would be the minimum signal level detectable by the device at a particular frequency, and naturally would be a function of the overal noise level in the receiver electronics itself.
5. NF/Noise Temperature: Probably should also characterize the device this way, as it is in common use. Some referenecs: On Noise Figures and Noise Temperatures, Three Methods of Noise Figure Measurement, Fundamentals of RF and Microwave Noise Figure Measurements

Each observation would be indexed by the following attributes:

• Device Serial Number
• Date of Observation
• Center Frequency
• LPF Bandwidth
• Sample Rate
• LNA Selected (LNA1, LNA2, LNA3 or 50ohm)
• LNA Gain mode (Max, Mid, or Bypass)
• VGA1 Gain
• VGA2 Gain
• Total received power (sqrt(I_I + Q_Q) in dBm)
• Calibrated noise source power (in dBm)

[bvacaliuc]

I was surpised by the following observations in which the only variation was sampling clock frequency when looking into an external 50ohm load SMA screwed on to the RX input port. See rasdr2-017-0013-10msps-1600MHz-calibration-terminations.pdf converted from original .png snapshots via png2pdf. I have gathered the statistics into a spreadsheet, rasdr2-spur-observations.xlsx.

In these tests, I had the following setup on my RASDR2 device: REFCLK, CW OUT terminated with an SMA 50ohm. Terminal block not plugged in. PC connected via USB2 to a powered USB3 hub. RASDR2 connected to the powered USB3 hub. The RX connector always had an SMA connector on it: Either: a 50ohm load, a special SMA with an open circuit or a special SMA with a short circuit. These I got when I obtained my miniVNA Tiny from mini RadioSolutions.

I performed tests at two frequencies (1600MHz and 1599.5Mhz) and two LPF bandwidths (3.84MHz and 5.0MHz) with four combinations of terminiation of the RX input: HiZ (RASDR2 internal), 50ohm, open-circuit SMA and short circuit SMA. I did this to minimize to the best of my ability any external signals going into the system. We have already shown (at the SARA annual conference 2016) that RASDR2 when operating does not radiate more than -80dBm across a broadband spectrum, so there is reason to expect that the converse is also true.

From these tests, I observe that there is SIGNIFICANT energy entering the system when the internal HiZ termination is used. This is comparable to the thermal noise produced by the 50ohm termination. Moreover, there is a strong signal at 1597.450MHz. I believe this to be a 4th order harmonic of the 400Mhz clock frequency used by the USB3 FX3 chip. The other spurs I think are mixing products and intermodulation due to the incredibly noisy environment that exists in the case.

My next step is to conduct the same observation but with the FX3 clock frequency changed in the firmware to something else. If the theory on the clock has merit, then we should see a corresponding shift in the 4th harmonic pickup.

After than the same tests with another RASDR2 unit, and then one in which has the RF shield soldered on it.

After that, I plan to fit a copper mesh in between the DigiRED board and the MyriadRF board to see how much of the clock I can suppress.