开云体育

Dears,

to avoid cross posting I just add the corresponding link to my issue and my question.

See below.

/g/qcx/topic/qcx_qsd_some_questions_and/105115946

Appreciate your opinion and experience.

vy73
Markus
DL8MBY

Markus,

The QSD integration capacitor values will affect the bandwidth of the QSD filtering property, but not affect its sensitivity or signal handling ability.

The precise value of the bias voltage on the mux is unimportant. Anything close to centered between ground and supply works fine. Only small signal swings should be allowed to exist here. Large signals that might bump against the rails would completely overwhelm the high gain amplifiers that follow.

JZ KJ4A?

On Sun, Mar 24, 2024, 2:57?AM <markus@...> wrote:

Dears,

to avoid cross posting I just add the corresponding link to my issue and my question.

See below.

/g/qcx/topic/qcx_qsd_some_questions_and/105115946

Appreciate your opinion and experience.

vy73
Markus
DL8MBY

Hi JZ, Markus

A transceiver such as QCX+ which is all analog, uses a larger value QSD capacitor for lower bandwidth. We only need enough bandwidth to be able to detect 700Hz well. On an SDR like QMX (and mcHF) we need smaller capacitors for larger bandwidth. QMX samples at 48ksps for example.?

The bias point of the QSD actually DOES make a difference to IP3 (strong handling signal capability). 2.5V midrail is NOT necessarily optimum. However:

1. It isn't clear from a theoretical perspective WHY this should be the case (at least I don't think I've ever seen any explanation let alone a convincing explanation). But it is measurable nonetheless.

2. The difference (improvement) is small.

3. The optimum bias voltage also varies by operating frequency.?

4. IP3 is difficult to measure accurately for most people.

5. Different optimum bias voltages are found for different MUX switch part numbers.?

Ideally to squeeze out every last dB of performance one should have a Digital to Analog Converter providing the MUX DC bias voltage, a measurement of IP3, and a lookup table of values per band. It would take an extremely high performance receiver before it would be worth worthing about such things: for anything less, there would be easier ways to improve it. Lower hanging fruit.

73 Hans G0UPL

On Sun, Mar 24, 2024, 2:02?PM John Z <jdzbrozek@...> wrote:

Markus,

The QSD integration capacitor values will affect the bandwidth of the QSD filtering property, but not affect its sensitivity or signal handling ability.

The precise value of the bias voltage on the mux is unimportant. Anything close to centered between ground and supply works fine. Only small signal swings should be allowed to exist here. Large signals that might bump against the rails would completely overwhelm the high gain amplifiers that follow.

JZ KJ4A?

On Sun, Mar 24, 2024, 2:57?AM <markus@...> wrote:

Dears,

to avoid cross posting I just add the corresponding link to my issue and my question.

See below.

/g/qcx/topic/qcx_qsd_some_questions_and/105115946

Appreciate your opinion and experience.

vy73
Markus
DL8MBY

Hans - I agree with everything you said here from my own experience, but I?thought you might be interested in the published linearity data from TI.

I found this document useful in troubleshooting in correcting IMD problems I encountered in that part when use for transmitting. The bias DEFINITELY needs to be lowered for symmetrical low distortion transmit mixer output at full output - but it hardly matters at all for receiving at normal signal levels.

Here's a link to the comprehensive TI document with the curves for most of the interesting multiplexers:

? ? ? ? ? ? ?

Hope you find this useful...

M

Hi JZ

?

In a transceiver context, someone interested in obtaining the best possible RX IMD performance from a QSD based receiver would, I think, be compelled to consider the additional contributions circuits other than the QSD itself provides.

Yes; and everything is trade-offs.??

?

A transistor T/R switch, or a TX PA which is not fully disconnected from the RX, adds nonlinearity, too. So would protective diodes or dissipative PIN RX protectors.? Even the performance of passives (magnetic core inductors, ceramic capacitors) might need to be considered. An interesting paper in the Microwave Journal discusses IMD due to passive elements that one might normally just neglect.

?

PA3AKE's entire long website on H-mode receiver design is extremely interesting, see:

The H-mode mixer is not exactly the same thing as the Quadrature Sampling Detector but is quite closely related. Particularly interesting is the discussion around the IMD inflicted by the toroids and capacitors, and how in the extreme that becomes the dominant factor limiting performance (not the mixer front end).?

?

There appears to be little or no relevant literature discussing the IMD performance of practical QSD implementations. I would be delighted to see such a paper.

Dan Tayloe's presentation on NC2030?

The absolute ultimate performance, complexity, low cost, small size, and low current consumption don't go hand-in-hand. If cost, physical size and current consumption were no object at all, I'd say:

• Larger sized toroids to avoid IMD arising in the powdered iron cores themselves.
• More advanced band pass filtering than the single series-resonant LC in QDX/QMX/QCX.
• Metal screening of RF stages.
• Relay switching is closer to perfect than PIN diode switching.
• Possibly superior multiplexer switches for the QSD could be found.
• The LM4562 is about the 2nd best performance op-amp that I know of in this application; LT6231 is best but considerably more expensive. Note that input noise is not the only characteristic; many op-amps that look better on paper perform worse in the circuit due to compression at the strong signal end, that reduces the dynamic range.?
• Switched attenuators could be used to optimize the already huge dynamic range window for the band and conditions.

73 Hans G0UPL