Up to now, this project is in the "early breadboard state", but initial tests with an ugly dead-bug style construction showed a surprisingly good suppression of the carrier (138.55 kHz) and the unwanted sideband. Using I2PHD's JASON or DL4YHF's Spectrum Lab to produce audio signals with quadrature output (I+Q), the "wanted" signal was S9 + 20dB on my TS850 while the carrier and the "unwanted" sidebands produced no S-meter reading in all cases (only weakly audible!).

Two outputs signals from the soundcard ("left" and "right" channel) provide an audio signal with a 90° phase difference between them. By convention, the "right" output lags the "left" output by 90°. When building the prototype, I didn't care much about where to connect "I" and "Q" to the IC pins... if the mixer produces USB instead of LSB, or LSB instead of USB, simply reverse the two audio inputs. The same is true if you swap the inverted and non-inverted buffered audio signal in one of the two mixer branches.

The audio buffer uses a TL084 quad audio OPAMP to produce four audio signals:

• AI+ = "Audio, in-phase, positive polarity"
• AI- = "Audio, in-phase, negative polarity"
• AQ+ = "Audio, quadrature phase, positive polarity"
• AQ- = "Audio, quadrature phase, negative polarity".

The four buffered audio signals are fed into two single balanced mixers, realized with a 74HC4066 (four analog switches). Both mixers are fed with the same L.O. frequency (138.55 kHz in my case), with a 90° phase difference between both L.O. outputs.
The L.O. is shown in the next chapter, it actually produces 4 outputs, quite similar to the audio signals: In-phase, quadrature-phase, and for both of these inverted and non-inverted outputs.

The outputs from both mixers are simply added via the two 1-kOhm-resistors, which are also part for the first lowpass filter. The combined output only contains the "wanted" signal, and some higher harmonics, which can all be filtered out with simple lowpass filters.

Local oscillator with quadrature outputs

There are many ways to build a local oscillator with quadrature output, including DDS, programmable devices like microcontrollers, etc. But since the 136 kHz amateur radio band is so narrow, it easily fits into the bandwidth of a soundcard (even if the soundcard only runs at 11.025 kHz). So, instead of changing the L.O. frequency, we change the audio frequency to "QSY". The L.O. only needs a fixed frequency which should be a few hundred Hz away from the upper of lower edge of the 136 kHz band.
Fortunately this can be easily achived with standard components, and a standard crystal frequency: 4.433619 MHz. The crystal frequency is divided by 8 using a simple counter (a 74HC161 is only one of many possibilities). The oscillator itself may be placed in a small "oven" if ultimate frequency stability is required.

The result is a signal near 554.202 kHz. This drives another divider (using the two J/K-master/slave flipflops in a 74HC107). This final divider produces four (!) output signals near 138.5506 kHz:

• I+ = In-phase output from L.O., non-inverted
• I- = In-phase output from L.O., inverted
• Q+ = Quadrature output from L.O., non-inverted
• Q- = Quadrature output from L.O., inverted

The inverted L.O. outputs are only required if a (74HC)4066 is used as the mixer. Other analog switches, usually "analog multiplexers", may not require complementary switching signals (but only "I" and "Q" from the L.O. instead).

To use this circuit in a "high-power" transmitter, good shielding of the mixer, and some lowpass-filtering of the audio signals may be helpful. Any transmitted RF which enters the analog mixer ports may degrade the unwanted sideband cancellation.






More construction details, and possibly a nicer photograph of the first prototype may appear here in future, after some cosmetic surgery...
The two IC's at the top are the 74HC161 and -HC107, in SMT, soldered on a copper-clad board. A sharp knife was used to cut the "layout" into the copper.
The two IC's at the bottom are the 74HC4066 and the TL084, both "dead-bug style" (with the legs in the air, the resistors and capacitors soldered directly to them). The audio I+Q signals are connected at the right side. The "RF" output (136.x kHz) is in the lower left corner.

Further reading

A much nicer drawing of a similar image-cancelling mixer for the LF bands was made by John Andrews, so if you are interested in the subject don't miss this one:

http://www.w1tag.com/Phasing.htm