Monday, 15 January 2018

Polyphase harmonic rejection mixer: AirSpy HF+

Polyphase harmonic rejection mixer: AirSpy HF+


Can you get excited about a new mixer, usually boring devices that haven't changed in decades? Yes, the new polyphase harmonic rejection mixer in the AirSpy HF+ is almost as revolutionary as SDRs and will have a major influence on their design.

The big advantage of a polyphase harmonic rejection mixer is that it acts as a RF filter for the selected signal, as well as suppressing harmonics and other aliases of the mixing process and local oscillator. It means that the mixer can virtually be connected to the antenna. Typically, a polyphase harmonic rejection mixer converts down to an ADC at base-band. It seems they can be used for both RX and TX.

The post covers how the AirSpy HF+ works, and gives references to what I have been able to find out about polyphase harmonic rejection mixers. They are new and still covered by recent patents. A link to a PowerPoint gives general technical details of the mixer.

AirSpy HF+

The AirSpy HF+ is rather unique for modern SDRs as its main purpose is to cover the HF bands, although it does cover VHF as well, although it only covers 200 kHz. And costs just $199. Most new SDRs start at VHF and go to daylight, well 3 or 6 GHz! They are intended for wide band mobile phone type applications, with coverage up to 30 MHz. The new LimeSDR (and $99 mini) and transverter ($299) covers up to about 10 GHz, but has limited RF band-pass filtering.

The unassuming appearance of the HF+ is shown in Picture 1 and the basic architecture of the HF+ is shown in Picture 2, clipped from

Picture 1 AirSpy HF+

Its maker's description: "Airspy HF+ achieves excellent HF performance by means of a low-loss preselection filter, high linearity LNA, high linearity tunable RF filter, a polyphase harmonic rejection (HR) mixer that rejects up to the 21st harmonic and multi-stage analog and digital IF filtering.

The 6 dB-stepped AGC gain is fully controlled by the software running in the DSP which optimizes the gain distribution in real time for optimal sensitivity and linearity. Harmonic rejection is a key issue in wide band HF receivers because of the large input signal bandwidth of the input signal. The output of the IF-filter is then digitalized by a high dynamic range sigma delta IF ADC for further signal processing in the digital domain."

Picture 2 The basic architecture of the HF+

Polyphase harmonic rejection mixer

The way the new mixer works is not simple, it uses multiple phases of the local oscillator to use phasing to reject its harmonics, but at the same time, and because it is to a 200 kHz base-band, it rejects everything else too.

The big advantage is not needing a large number of band pass filters like a direct sampling SDR; the IC-7300 has 15!

The best explanation I have found is a slide show; It is also subject to patent; One of the authors wrote the slide show.

Presumably the mixer is an analogue CMOS device, but I have not found one. And the RF cover on the HF+ is too hard to remove! The HF+ site cites ST Microelectronics as a collaborator, but I have searched their site with no success; It may still be in development or proprietary intellectual property.

The new mixer is not entirely new, as stated in the patent, it relies on existing harmonic rejection mixers and other patents.

Performance of AirSpy HF+

The HF+ is still very new, I only received mine in the last couple of weeks. The HF+ gives some performance results. There have been a number of comparative reviews against other SDRs, such as the new $99 RSP1a, by radio amateurs and shortwave listeners. However, there has not been a full technical review by the ARRL or RSGB.

However, with the limited testing the HF+ seems to have a high dynamic range and superior ability with weak signals near large signals, as would be expected from the design.


The polyphase harmonic rejection mixer of the Airspy HF+ is a significant development in radio design and is likely to rival other technologies over the coming years.

Friday, 5 January 2018

Modifying cavity filters for DATV TX or for repeaters

Modifying cavity filters for DATV TX or for repeaters


I am currently doing further work on using notch cavity filters for DATV DVB-T transmitters. My earlier efforts were with what I had at hand and not knowing the solution; I (re)discovered that notch filters clean up DVB-T TX very well. However, it was at low power, 10 W, and high losses, >6 db because of the six cavities in a mobile duplexer. Here, I will report on modifying high power >100 W individual filters. In the next post I will report on using them and determining is just one pair are sufficient. The other goal of this post is to show how easy it is to modify older commercial filters for DATV or repeater use.

Modifying cavity filters

Old commercial filters are relatively easy to modify as the only thing that changes is the coupling loop, provided they are on frequency (not too hard to change that too!). Notch filters are the simplest as they use a single simple coupling, just a loop of metal. Old commercial filters are usually made very well, often silver plated. On UHF, they are relatively cheap; $100 for a four cavity duplexer.

Other than the coupling, the RF design of a cavity filter is simple, a quarter wave resonator (antenna) in a box, usually a cylinder. With a notch filter, the cavity is connected to the TX coax line with a single coax "T". The cavity absorbs the RF at the resonator's resonate frequency; an antenna in a box! The impedance is determined by the ratio of the cylinder to the resonator, like coax, about 3:1 for 50 Ohm.

The mechanical design is more complex, particularly with a variable length resonator to change frequency. The Q should be as high as possible, which is why many are silver plated brass, although can be copper plated or aluminium. The adjustment screw is an non-magnetic, low thermal expansion alloy of steel, Invar, with finger stock for a very good connection to the movable part of the resonator. There are "tricks" with the couplings to get good results without high cost. Some cavities use a capacitive "hat", to change frequency, as is done with antennas.

Couplings are mechanically simple but very complex for RF. There is virtually nothing in textbooks, most of it is proprietary, but most types are covered in: Black magic!

The key point of resonators here is that the closer to the resonator, 2-3 mm, the higher the coupling and the deeper the notch. However, as coupling increases, losses increase.

Making modifications

I have made a new coupling for a pair of large aluminium cavities, 150 mm diameter and about 400 mm long. The process of doing it is fairly easy, remove the original coupling, a loop soldered to an N connector. Unsolder the end of the loop attached to the connector pin and cut the earthed end to allow the new coupling to be soldered to it.

Make a sketch of how the coupling is mounted in the cavity and measure all the critical dimensions, particularly the connector center pin to the resonator and the same for the earth point. A small measure can be made by cutting a rectangle of grid paper. Then do a 1:1 drawing of the location. The new notch coupling is about 20 mm parallel to the resonator and 2 or 3 mm from it. The coupling can be made from a strip of copper about 5 mm wide and 1 mm thick, or a larger diameter piece of copper wire. The coupling is bent with a pair of long nosed pliers so that it matches the drawing. See Photo 1 of my drawing.

Once the coupling is accurately bent, solder it to the connector and adjust the shape as needed. The only part that is critical is that the piece of the coupling closest to the resonator must be parallel.

Photo 1 Sketch of new coupling, as described. I was originally going to solder the earth  leg to the coax connector, OK if PTFE, but soldered it to a tag I cut from the old coupling instead. Both arrangements are drawn. The top plate was 10 mm thick, making things a little awkward.

Photo 1.5 The modified loop. The earth is soldered to part of the old coupling rather than to the connector as originally planned. The earth screw is a bit corroded, I should clean it.

With the resonator screwed back in place, its RF response can be shown with spectrum analyser. Spectrum analysers for DATV can be improvised using an SDR and a noise source for about $200 vs >$1500 for a Chinese one (which are very good). See

Photo 2 The response of the new coupling, a sharp asymmetric notch and about 22 db deep with less than 1 db loss. It initially was about 20 db, but bending the coupling closer to the resonator, a small increase was obtained.

Notch filters are limited to about 25 db. For repeater cavities, I would chase that, but it is not that critical for a DATV skirt/splatter filter.

For a DVB-T filter, a sharp rectangular response is desired. Notch filters have it on the high frequency side, but a shallower response on the low frequency side. As a DVB-T signal is a 7 MHz wide rectangle, made up of nearly 8000 carriers, another notch filter is needed on the high frequency side, but the response reversed. This can be done with a quarter wave length cable between the coax T and the cavity.

 For initial DATV testing. I will only do one side, so I can compare it directly with the unfiltered response on the other side of the signal.

Other UHF cavity filters

I bought a four cavity repeater duplexer a couple of days ago that I might use if I need two cavities per side for DVB-T.

I connected up one of the cavities and had a look at how it worked. Wow! An excellent pass reject cavity for a 70 cm amateur repeater. I opened one cavity and was surprised by two things. First that it was copper plated brass (not silver) that was still working well after about 30 years. The second, was how far the coupling loops were from the resonator, >20 mm. This was significant for me as I had struggled with pass reject cavities for 2 m. I tried to put the coupling near the resonator, as per notch cavities, but may have introduced too much induction with long wires. The other problem is when the connectors are opposite each other from the resonator, common with pass-band cavities.

Photo 3 The test one I have been discussing earlier on the right and the old cavity just noted on the left. Size matters for cavity filters as the surface area is proportional to Q, as well as power handling; the bigger the better.

Photo 4 The response of the old pass reject cavity, a huge 50 db! Great for a repeater but no use for a DATV skirt/splatter filter.

Photo 5 The assembled duplexer, a Motorola T1500 series, and unassembled cavity .

Photo 6 A close-up of the resonator and coupling loops, note the large spacing from loop to resonator. It is configured as a pass reject with a variable piston capacitor between coupling loops. May be original but looks like a modification; not mentioned in the 1983 Motorola brochure. Mounting the capacitor can be mechanically difficult as it must be insulated and accessible for adjustment outside.

Photo 7 The pass reject response can be changed to notch, by unsoldering a wire from one coupling loop, then using a coax T on that connector. A very disappointing 10 db because the loop is not closely coupled, being so far from the resonator. The response can be improved by making a new loop that is 2 - 3 db from the resonator, as described in the main article. The cavity can be converted to pass-band by removing both wires and increasing coupling.


It is relatively simple to modify used cavity filters suitable for use as a DATV DVB-T filter. The next step is to set it all up to see how well one high power cavity will work.