Another Look at the Spurious Output of AD936x PLUTO Transceiver
Since I have access to a 7.5 GHz spectrum analyzer, I decided to take and share some more data showing the output spurious performance of the AD936x transceiver. My DUT is an AD9363 set to AD9361 mode for additional frequency output on the LOs in a HamGeek PLUTO Plus. To have login access to set the wider frequency mode, I'm using the 0.38 firmware supplied by the vendor that actually uses the default credentials, unlike the stock/QSPI firmware shipped on the device.
This exercise is partly to compare with the one published example that gets thrown around from a 2 GHz span on a Sigilent spectrum analyzer showing plenty of spurious responses from a ~140 MHz fundamental. I've seen it referenced multiple times, but the version I have at hand is from K0SM's talk (slide 32, page 16) for the Rochester ARA. It's shown with a probably 144 MHz fundamental on a 0-2 GHz span with no markers. That gets the point across that filtering after the PLUTO's balun is key with the third-order output looking only -6 dBc!
My setup two 10 dB SMA attenuators with an SMA jumper between them (set the external gain factor to -20.3 dB) into a Tek RSA607A. That's not enough to characterize much on the out-of-band operating (the 10 GHz alias in/out functionality), but does give plenty to see how lower frequency outputs in-band perform. The output was set to FM for simplicity using SDRoxide at 0 dB TX gain and 100% drive for each case.
 |
| FM 446.0 MHz measured across 0.25-5.25 GHz span with select markers |
 |
| FM 1296.4 MHz measured across 1.0-6.0 GHz span with select markers |
Trying to get TX output power of the fundamental didn't work out quite right with Flat-Top windowing. The spurious output measurements used Kaiser (RBW) windowing at 1 kHz RBW. Output frequency drifted as the SDR warmed up, but accuracy against the spectrum analyzer isn't great due to it also only having TCXO accuracy. No good view of the sky for GNSS discipline of the internal reference in that lab. Still, once warmed up, the analyzer is at least decent on stability to show that the SDR kept drifting during testing. The offset I initially recorded between the two was about 2.7 ppm. Enough to need a spectrum scope to find each other!
Working through some numerical data, let's go in frequency order for the measurements I made.
| Alias Order | Measured Level | Frequency |
|---|---|---|
| 1 | 6.41 dBm | 0.44375 GHz |
| 2 | -44.27 dBm | 0.89375 GHz |
| 3 | -4.11 dBm | 1.3375 GHz |
| 4 | low | skip |
| 5 | -13.69 dBm | 2.23125 GHz |
| 6 | invisible | skip |
| 7 | -17.14 dBm | 3.125 GHz |
| Alias Order | Measured Level | Frequency |
|---|---|---|
| 1 | 5.26 dBm | 1.29375 GHz |
| 2 | -38.96 dBm | 2.59375 GHz |
| 3 | -10.40 dBm | 3.8875 GHz |
| 4 | -46.24 dBm | 5.1875 GHz |
| Alias Order | Measured Level | Frequency |
|---|---|---|
| 1 | 0.92 dBm | 2.31875 GHz |
| 2 | -40.67 dBm | 4.6375 GHz |
| 3 | -30.05 dBm | 6.9625 GHz |
I could have used extending the 1296.4 MHz measurements to the 5th alias expected at 6482 MHz, within the measurement capabilities of the analyzer, but missed that option. Still, that's easy enough to filter out when the third-order alias is a more critical target before the power amplifier stages. Remember to filter all the spurious products for those lower frequency fundementals, especially if any of the gain stages before the final PA are wideband! A 6 GHz MMIC will still provide unwanted power even if the 1.33 GHz is well controlled on the 446 MHz signal but your filter starts leaking after 3 GHz.
Comparing against the 144 MHz example, at least (for this use case) the third-order alias power grows ever more distant from the fundamental with increasing fundamental frequency. So, good filtering is more critical for 70/144/222 MHz implementations and grows less stringent headed for 2.3 GHz. With any luck, there's not much to measure for a 5.76 GHz third-order alias!
There's still more for me to learn about taking these characterization measurements, but this has been a good start to familiarize myself with the equipment and get an idea of what I'm up against for filtering. Fortunately, being -11 to -16 dBc on the third alias is an easier filtering task than the traditional LO/IF needs for microwave transverters using 28/50/144 MHz rigs. How to implement that filtering is an upcoming exercise for me to make more use of these AD936x-based SDRs. And then there'll be the challenge of testing at full output without smoking anything!
Comments
Post a Comment