Hi All,

I am developing a small, low power SDR from the ground up, planned
phases include:

a) hardware development DSP and RF up-convertor
b) algorithm development FM encode and decode
c) algorithm development SSB encode and decode
d) hardware development RX down-convertor, and IF gain, with AGC
e) hardware development RX front-end
f) hardware development TX amplifier
g) hardware development LO generation (initially a signal generator)

Phases a) and b) are complete and working to my satisfaction, phase c)
is a couple of days away from completion, so it's time to start
thinking about d)

The IF is base-band (DC to 5kHz) I+Q, and the DSP on board ADC is
capable of 10 bit resolution. I am planning to move to a 16 or 20 bit
consumer audio codec in due course. Assume a 2V ADC reference in each
case, so an LSB change for a 10 bit system will be around 2mV while
and LSB change for a 16 bit system will be around 30uV.

The down-convertor is an RF2703 which has 24dB gain and 24dB noise
figure in the planned configuration.

My question really revolves around noise figure and gain for the IF
and RF front ends for use with 10 and 16 bit ADCs, and for use at
14MHz and 144MHz, also options where to put the AGC.

First issue really is 'what is appropriate performance' for the front
ends ? At 144Mhz, -122dBm for 12dB SINAD seems like an appropriate
number (is it ?). What should I be attempting for 14MHz ?- this band
is much noisier.

Second issue is about gain and noise figure for the RF and IF stages -
any thoughts on how to calculate these figures will be most
appreciated.

Finally AGC - where, how, options, pros and cons.

If these aren't the correct questions for me to be asking, then what
are the correct questions ?

Thoughts, pointers and help all appreciated.

Many thanks, Mark

On Wed, 29 Dec 2010 09:50:00 -0800, MarkAren wrote:

Hi All,

I am developing a small, low power SDR from the ground up, planned
phases include:

a) hardware development DSP and RF up-convertor b) algorithm development
FM encode and decode c) algorithm development SSB encode and decode d)
hardware development RX down-convertor, and IF gain, with AGC e)
hardware development RX front-end f) hardware development TX amplifier
g) hardware development LO generation (initially a signal generator)

Phases a) and b) are complete and working to my satisfaction, phase c)
is a couple of days away from completion, so it's time to start thinking
about d)

The IF is base-band (DC to 5kHz) I+Q, and the DSP on board ADC is
capable of 10 bit resolution. I am planning to move to a 16 or 20 bit
consumer audio codec in due course. Assume a 2V ADC reference in each
case, so an LSB change for a 10 bit system will be around 2mV while and
LSB change for a 16 bit system will be around 30uV.

The down-convertor is an RF2703 which has 24dB gain and 24dB noise
figure in the planned configuration.

My question really revolves around noise figure and gain for the IF and
RF front ends for use with 10 and 16 bit ADCs, and for use at 14MHz and
144MHz, also options where to put the AGC.

First issue really is 'what is appropriate performance' for the front
ends ? At 144Mhz, -122dBm for 12dB SINAD seems like an appropriate
number (is it ?). What should I be attempting for 14MHz ?- this band is
much noisier.

Second issue is about gain and noise figure for the RF and IF stages -
any thoughts on how to calculate these figures will be most appreciated.

Finally AGC - where, how, options, pros and cons.

If these aren't the correct questions for me to be asking, then what are
the correct questions ?

Thoughts, pointers and help all appreciated.

Many thanks, Mark

Front end band pass filters are really kind-a mandatory on any SDR as
people in high RF environments will simply have IM and a/d converter
overload. If you use AGC, the 100KW broadcaster or passing 500w aircraft
radio will close and reduce headroom diminishing weak signal sensitive in
the desired band.

Most all the good SDRs I've seen have several fixed freq band pass
filters designed around common high power bands. But, what would rock
would be a wide band (1Mhz or so) variable band pass filter with steep
slopes on the front end before the first nonlinear device that could be
center freq shifted. This would-could reduce the parts count and cost
since you would only need one filter and a D/A converter to S/W control
the shift.

Another idea would be a log amp on the front end to reduce the dynamic
gain of signals over a certain voltage without changing the gain too much
of weak signals. The signal could be inverse logged after the A/D. The
down side is a non-linear signal feeding the converter which will hurt
EbNo of digital signals.