Work in Progress!
Updated 25 July 2014
Important Note: After posting a question about the ANAN-10 and PureSignal to the OpenHPSDR mailing list, I learned something about the ANAN-10 amplifier/filter board that I didn't know: in transmit mode, the board shorts the RX signal line to ground. That, of course, causes a lot of attenuation when you are trying to inject a sample signal into the receiver. My original tests were made with that situation, and are therefore largely meaningless. I've updated this page to show results after removing the ANAN-10 internal RX coax. For historical completeness, here is the original page.
This page documents what I've learned getting PureSignal optimized for my HPSDR station. My first testing is with a TAPR Hermes board interfaced to the Apache Labs ANAN-10 enclosure and 10W amp, and then with my TenTec 418 100W amplifier. This is mainly a notebook, so things don't necessarily flow in sequence.
But first, does PureSignal work? Yes.
These two screen shots show the output of the Hermes + ANAN-10 amplifer on 20M with 10W (calibrated) output:
From this, we can see that the ANAN-10 output is pretty good anyway. But PureSignal makes a great difference: the 3rd order products go from -35dB to -52dB. Actually the products are so close to the noise level we really don't know how low they are.
The next two shots show Hermes+ANAN-10 driving the TenTec 418 to nominally 100W PEP, also on 20M. The drive level has been adjusted to 65 to avoid overdriving the 418; even a lower setting might be appropriate.
Now, we see more typical responses for a 100W, 12V, amplifier. The 3rd order products without PureSignal are -25dB; with PureSignal they are -36 (averaging upper and lower).
Lab notes: Test setup was rig output connected to Bird 43 wattmeter with Coaxial Dynamics 87014S -37.5 dB (nominal) coupling slug (however, actual coupling at 20M is more like 47dB). That slug drives the Rigol 815 spectrum analyzer. The output of the Bird goes to a 30dB, 100W power attenuator, with an additional 10dB attenuator at the output. That -40 dB signal is fed to the RX IN SMA connector on Hermes. The TX attenuator in PowerSDR is set to 2dB for the 10W tests, and 7dB for the 100W tests; this was based on trying to get into the mid "green" range on the PureSignal tab.
However, there've been reports that the crosstalk in the ANAN-100 radio (essentially, the same as my system but with a built-in 100W amplifier) is at a high enough level to cause issues. Does that crosstalk cause problems or misleading results? So, I did some testing with my station.
First, I wanted to know how much signal bleed-through there is between the transmitter and receiver, before adding an external signal tap.
The ANAN-10 amplifier/filter board provides 3 BNC antenna connectors but as far as I can tell, you would need to use an external relay keyed from PTT to use one of those connectors for the RF sample input. however, you can feed the sampled RF into to the "RX IN" SMA port on the Hermes board; this is in parallel with the RX output from the amp/filter board. The parallel connection can cause some issues, but it seems to work well enough for initial tests.
To measure the crosstalk, I connected the "ANT 1" BNC connector to a dummy load, and measured the absolute strength of the signal present during transmit at the "RX IN" SMA connector by connecting a spectrum analyzer to it.
I set the HPSDR drive setting to 80 on each band, and I'm using the "Tune" function with tune level set to equal drive level; for this test I used a CW signal rather than two-tone SSB. The PA gain settings had previously been calibrated so that a drive level of 100 provides 10W +/- 0.2W on each band.
From this table we learn:
- The crosstalk on the Hermes board itself is inconsequential, except perhaps at 10M and 6M.
- There is significant crosstalk from the ANAN-10 amplifier into the Hermes receiver, which increases with frequency.
- There is almost no crosstalk from an external power amplifier. All the crosstalk is from the ANAN-10 PA.
Therefore, the signal we bring from the output of the final amplifier needs to be some number of dB greater than the crosstalk level, or else the PureSignal algorithm will, at best, be correcting the ANAN-10 PA and not the external amplifier's nonlinearities.
What signal levels are available from the samplers or couplers I have lying around? First, I have two sampling slugs for a Bird 43 wattmeter: one a 50dB Bird slug, another, a Coaxial Dynamics 87014S directional sampler. However, both of these are designed for frequencies above the HF bands, and their coupling falls off at lower levels (first the Bird, then the CD slug):
Then, I tested a ClearRF LLC model RF-S2K sampler designed for HF use. This is a really nicely-built unit, by the way. Here's the sample level set to as high as I could get it:
The RF-S2K has an adjustable coupling level, but its effect is highly frequency-dependent:
I'm not sure what the sample level control actually does, but at 20M and above, it doesn't provide a smooth ramp. Above about 7MHz, maximum CCW provides the most coupling, but the sample falls off rapidly below that. When measuring at 20M, just a few degrees from max CCW, there's a dip followed by a gradual rise as you approach maximum CW.
At this point in the page, I used to have some comments about the RF-S2K's ability to work with my system. I'm removing those for now because the new information I've learned about the ANAN-10 signal path means that I was losing a large amount of RX sensitivity, and that I needed far more coupling than should be required. After I do some new tests, I'll update this with how the RF-S2K works, and I suspect the results will be different than I previously reported.
By the way -- these tests were done with a dummy load. I haven't yet investigated whether the mismatch of a real HF antenna could cause the sampling levels to vary. Also remember that the SMA RX input on the ANAN-10 is simply paralleled across the signal path from the Hermes board to the ANAN-10 amplifier/filter board. So the load at that port isn't likely to be 50 ohms non-reactive.