In the previous article, I covered the Arduino controller for the transceiver, and in this one, I was hoping to describe the direct digital synthesis and buffer board – but since I shorted the DDS board, and the buffer wasn’t working quite as I’d hoped, I’ll need to revisit that…. Undeterred, this article focuses on the bandpass filters of the transceiver, and how I’m going to test them.

[See the full list of articles in this series HERE]

In the original circuit, the receiver was fed by a bandpass filter, constructed from two Toko KANK3333R coils. (Note: these are no longer manufactured, but compatible coils are available from Spectrum Communications). The transmitter output led into a low pass filter made from two T50-2 toroidal inductors.

The receiver input is filtered to attenuate strong out-of-band signals that might desensitize the receiver, making it harder to receive the in-band signals. Considering the noisy, unfiltered spectrum around the 20 metre band at M0CUV, shown below, there is a definite need to reject anything loud! The peak at 13.557 MHz is some local QRM.

The transmitter is filtered to attenuate harmonics generated by the VFO/buffer/PA.

In this revised project, since I’m replacing the VFO with a DDS module which generates harmonics – see page 9 of the AD9850 data sheet – there’s an even greater need for output filtering. I mustn’t transmit out-of-band, but also don’t want harmonics appearing in other amateur bands.

To simplify the design a little, rather than having separate receive and transmit filters, I chose to have a single bandpass filter just after the antenna input. This is switched between receive and transmit by opto-isolated relays. Since this is a three-band transceiver, I will also be switching the appropriate filter in, also by opto-isolated relays. The three filters are to be built as swappable boards, so I can explore the whole amateur HF spectrum.

This article doesn’t cover the detail of the filters, or the switching arrangement: it’s about testing the filters, to ensure that their passband covers the amateur band, and that the filter drops off as sharply as possible out of that band.

The approach I’m taking is to generate a broadband noise signal, and inject this into my “spectrum analyser” … Assuming that this shows a nice noisy signal across the band in question, and a little off the ends of the band, I will compare the noise with the filter in- and out-of-line, and adjust the filter to ensure good coverage of the band, and rejection outside it.

Since I can’t justify an expensive spectrum analyser, I’m using a somewhat cheaper solution – an RTL-SDR R820 dongle, preamplifier, and upconverter kit from CosyCave – which cost around £26 if I recall.

The noise source is is connected to the dummy load port of my KP4MD Dummy Load/40 dB Attenuator, and its attenuator port connected to the Preamplifier-Upconverter-Dongle chain.

Note: I was initially thinking of building the switchable attenuator from Hans Summers’ spectrum analyser project, but the 40dB from the KP4MD project was sufficient, cheaper and easier to build.

I read N5ESE’s article on using a noise source in this way, which led to this part of the project.

The noise source is a circuit by Tom N0SS, now sadly SK. His page on the project may be found in his Elecraft Kits page, with build instructions in PDF here. I have a cached copy of this file here.

The circuit of the noise source is shown below:

Construction

I built this on good old prototype board, using the rough layout shown below, and mounted it in an Altoids-style tin, with switched 9v battery power and a BNC socket output connector:

Testing

Measuring the noise on my old oscilloscope, I saw:

I’m using the excellent RTLSDR-Scanner software on Mac OS X to use the RTL dongle as a poor man’s spectrum analyser. The necessary pyrtlsdr wrapper software needed by it is cloned from its GitHub repository, and for the Python 2.7 I have installed on my OS X Lion system, I had to checkout this commit (if memory serves – I think the authors are trying for mixed Python 2.7/3.0 compatibility, and the HEAD commit isn’t working on 2.7 for me… so, wind back a bit.) This program is cross-platform, so should work well on all platforms, given the right dependency setup.

If you’re using Linux, you may also find Claudio Giuliani’s RTL-SDR Wide Spectrum Analyser project worth investigating.

The noise measured from 1 to 30 MHz (with noise source connected directly to SDR chain, and output level reduced) looks like:

The blue trace is what’s received with the SDR terminated with a 50Ω dummy load; the green is the noise signal.

Conclusion

I’m hoping this will provide a low-cost method of testing filter response.

Acknowledgements

Next…

I’ll either be rebuilding the DDS/Buffer amplifier board, or winding toroidal inductors for the filter construction.

Advertisements