>
>     --- In [email protected], "Adnan Yusuf" <[EMAIL PROTECTED]>
> wrote:
>     >
>     > my question is , is it possilbe to setup two soundcards to work
>     > simltaneously??I was thinking of using two soundcards to take the I
>     > and Q parts.I will be writting my own software in matlab.Its just
> the
>     > windows detecting both cards at the same time that I am concerned
>     > about.I am doing this because I dont think my soundcard has stereo
> mic
>     > input.
>     >
>     No, I don't think it is possible, due to timing differences between
>     the two cards. But I don't understand you. If you plan to add a second
>     sound card to your PC, who or what prevents you from adding a proper
>     stereo card ?
>     73 Alberto I2PHD
>

Hello Adnan,

A while ago there was this interesting idea by Phil Covington where only one
audio channel is required for implementing a kind of QSD; called RMSD.
The idea was posted as a SoftRock SuperLite idea where only a single switch
does all the mixing and image rejection. At the time of posting, questions
where raised how the image rejection would work and the idea was maybe
misinterpreted and not fully understood. Did anybody solve the puzzle?

73, Guido

---

Here are some snippets of the original idea that was posted by Phil:

"Here is an idea for a SoftRock SuperLight receiver: Instead of using the
QSD circuit to do quadrature sampling use only 1/2 of it to do real signal
sampling. For the real mode sampling detector (RMSD) use a switch like the
SPDT FSA3157 and 180° sampling. Feed the audio to either the left or right
channel of the sound card.

The disadvantage is that you will have cut the possible rx bandwidth in half
- for example, at 96 kHz sampling you will have ~48 kHz bandwidth instead of
~96 kHz with I/Q sampling.

The advantages are that you will have eliminated the I/Q balance problem,
used only 1 switch and 1 op amp, and can use 1x LO (instead of 4x). This
should also extend the useful range of the RX up into 10 meters or better.

The software needs to mix the real samples with a quadrature oscillator to
generate the I/Q in the digital domain but that is only a simple
modification to the software. You could also take advantage of the extra
audio channel by connecting two SoftRock SuperLights to one sound card -
maybe 1 for 40 meters and 1 for 20 meters?

It would be interesting to look at doing this with the AD7760 2.5 MSPS 24
bit ADC. Running the AD7760 ADC at 384 kHz would give you ~192 kHz bandwidth
and should come close to its 120 db DR rating."

"Should be SuperLite... not SuperLight...

It will work!

With the QSD you get I and Q which go to the left and right audio channels
of the sound card. In the software (whether it is PowerSDR, Rocky, etc...)
there is a quadrature software numerical local oscillator that takes the
left and right channels and mixes them, respectively, with the quadrature
software LO so you can tune +/- your center frequency. The output of this
mixing process in the software is still an I and a Q signal which is
frequency translated. In the case of the RMSD's real signal if it is fed in
on the left channel, you would modify the software so that the left channel
is feeding both inputs to the software quadrature LO. The result is an I and
Q output which has been frequency translated. From that point on, processing
is the same as with the QSD.

It requires that the software is modified to feed the same channel to both
inputs of the software quadrature LO - the advantage is that you do not have
I/Q imbalance problems caused by the hardware. The I/Q is generated in the
digital domain and remains in perfect balance.

This can be tested with an existing SoftRock if you are able to change the
software's source code (in PowerSDR you can if you have VS 2003). Just feed
one channel to the sound card (either right or left). In the software, you
need to feed that channel to both of the software quadrature LO's inputs.
The tuning range will be 1/2 the sampling rate now ( ~48kHz for a 96 kHz
sampling rate, ~24kHz for a 48 kHz sampling rate, etc...) but that is the
price we pay for using only one channel and real sampling.

The reason we get bandwidth equal to the sampling rate in the QSD (96kHz
sampling rate = ~96kHz tuning range) is because we are using two channels
(two ADCs). If the input to the two channels is not in perfect balance, you
get images and that is why there is need for I/Q balance correction in the
software - the I and Q channels are affected by the imperfect hardware, both
in the QSD and in the sound card (differences in gain and phase between the
sound card's left and right inputs)."

"I should clarify that the LO would be offset by approx 1/2 the tuning
range. For 96kHz sampling we get ~48kHz bandwidth. The software LO is
centered at 24kHz and varies from 0 to 48 kHz (+/- ~24 kHz)."

"All that has been said about the QSD, DC receivers, quadrature mixers,
etc... is true... but the RMSD does not work in the same way that the QSD
does. It is not 1/2 of a QSD.

The 1/2 QSD suggestion using a SoftRock was to suggest an experiment to give
a hint of how a RMSD might operate. It was a bad example, obviously, because
everyone is focused on it instead of the new idea.

A good friend of mine wrote me and said I have to stop giving away ideas for
free... not sure how I feel about that but this time I will heed his advice.
Here are a few hints (maybe someone will figure it out ;-) ).

The RMSD is slightly more complex than 1/2 of a QSD. It is less complex than
a complete QSD circuit. The RMSD 1x LO does not imply that the center
frequency is at the 1x LO frequency. The RMSD avoids the image problem by
separating the images (not symmetrical around 0 Hz) by a very large amount.
It uses the similar idea (but not the same) of undersampling in a high speed
ADC with a very large input bandwidth. Another hint is my mention of the
higher speed Analog Devices 24 bit ADC - the RMSD works better as sampling
rates increase- higher than the current sound cards give (but still works
well at 192, 96, and somewhat less at 48 kHz).

Doing the above mentioned experiment with a SoftRock and Winrad (as Alberto
showed is possible) the images that you see on your computer screen make the
solution visible for those who give it serious thought. Think about what
those images mean mathmatically and how they could be manipulated
mathematically to separate out the images while preserving dynamic range and
sensitivity. This is how the RMSD came about - by doing the above."

"Here is a link to an interesting ADC by Analog Devices:

http://www.analog.com/UploadedFiles/Data_Sheets/AD7760.pdf

It is a 24 bit, 2.5 MSPS ADC. Here is the circuit idea:

http://www.philcovington.com/HPSDR/TEMP_STUFF/AD7760QSD.pdf

It uses 1/2 of a QSD (not quadrature sampled) to mix the frequency band of
interest down to zero Hz with a bandwidth of something like 250 kHz. The IF
would be offset to something like 125 kHz. The ADC samples the 1/2QSD output
and results in a real data stream. A FPGA or CPLD mixes the real data stream
with a complex NCO to generate a quadrature data stream which is sent over
USB to the PC where the remaining processing is done. The advantage is a 1 X
LO and no I/Q balance problems. One disadvantage is that your sampling rate
should be 4 times the desired bandwidth. For narrow band applications this
should be no problem because the AD7760 will sample up to 2.5MSPS. A 500
MSPS rate into the PC should allow a bandwidth of almost +/- 125 kHz from
the LO frequency."

Reply via email to