开云体育

I recently received my assembled QMX+. It is a very neat transceiver,
well designed and nicely built.

I have played with the USB, reading the IQ samples from the 20-bit,
48-kHz ADC. The QMX+ is in CW mode and IQ is enabled. I cannot find
information on the architecture, but I guess it is like follows.

The antenna signal passes through the bandpass filters to the Tayloe
detector. The frequency to the mixer is the dialed frequency minus the
IF frequency of 12000 Hz minus the side tone frequencey of 700 Hz. If I
dial 7020 kHz the mixer will use (7020000-12000-700) Hz = 7007300 Hz.
The resulting product passes through an amplifier with a lowpass filter
in front of the ADC, about 24-kHz bandwidth. The sampled signal has two
channels, I and Q, real and imaginary part. These signals, which have a
bandwidth of ±24 kHz, are available at the USB port. If I listen to a CW
signal tuned to 7020 kHz, it appears as a +12700-Hz signal on the USB
port.

The CPU in the QMX processes the signal by mixing it with 12 kHz
(complex sine, exp(jwt)). The resulting signal, I/real part only,
(appears at +700 Hz), is then passed through the CW bandpass filter,
nominally adjusted around 700 Hz. I have attached a PDF sketch of my
understanding.

Is this description correct?

I experiment with a configuration where the CW filter is a (IQ/complex,
two equal channels) Gaussian low-pass filter around zero frequency with
a bandwidth of 10-50 Hz. This corresponds to 20-100 Hz bandwith in a
bandpass configuration. Consequently, I mix the USB signal with IF+ST,
before feeding it to the filter. The LP output is then mixed with the ST
frequency to the speaker/earphone. (Note that zero beat is not useful in
the analog implementation due to noise and bias.)


Best regards

Anders, SM5KAE

PS I get some tiny spikes 1 kHz apart in the FFT of the USB data. Are
these generated in my setup? DS

On 21/03/2025 15:18, Anders Helmersson via groups.io wrote:

The frequency to the mixer is the dialed frequency minus the
IF frequency of 12000 Hz minus the side tone frequencey of 700 Hz

Anders,

Presuming the QMX is like the QDX, in IQ mode there is no 12KHz offset.
The QDX manual has a good circuit description.

73 Alan G4ZFQ

On Fri, 2025-03-21 at 16:42:35 +0000, Alan G4ZFQ via groups.io wrote:

Presuming the QMX is like the QDX, in IQ mode there is no 12KHz offset.
The QDX manual has a good circuit description.

Alain,


Thanks for your response.

I found the description of the receiver in CW mode in section 5.22
"AGC system" of the QMX operation manual (1.00_026/27). There it is
described how the receiver signal passes through 25 steps of processing.
From this description, also verified by tests, I conclude that the VFO
is set to the dialed frequency with an offset of 12 kHz plus the side
tone offset (nominally 700 Hz), assuming the the USB output during CW
mode with IQ enabled is the unprocessed 24-bit output from the ADC.

Maybe this description should be placed in a separate section, since it
is not only of interest to the understanding the AGC system. I guess
that a similar description will be needed for the new SSB mode,
describing both Rx and Tx functions.

Since I come from a different background (Automatic Control), I have
some difficulties understanding the rationale behind all the steps in
the receiver path. Why is the side tone frequency added already at the
Tayloe Mixer/QSD mixer?

An alternative would be to run the CW filter in zero-beat mode in the
I/Q domain and mix the sidetone in front of the DAC. Zero beat is not a
problem since there are no low-frequency noise nor bias after the 12-kHz
mixing.

A dual low-pass filter will then be needed, but the order of each filter
would be about half of a corresponding bandpass filter. Also, the
Hilbert transform (used for phase-shifting) can be removed and the
Goertzel filter becomes a trivial moving average. A zero-beat indicator
could be added to the display. The bandwidth of the low-pass filter is
also easy to parametrize using Tustin transformations. If needed an
offset of the center frequency can be added. The subsampling/decimation
and interpolation steps can also be absorbed in the low-pass filter
chain by implementing filter blocks with different sampling rates.


73 Anders SM5KAE