TAPR High Performance Software Defined Radio

[Terry Fox]

Thanks Frank for the response.  I agree with everything.

One of the big advantages (IMHO) of SDRs is that they can display the
spectrum of a whole band segment, while still having the selectivity to
pinpoint individual signals.  Pouncing during contests is just one
application of this.  Seeing general band conditions or activity is another,
and finding the signals at all when downconverting from UHF or microwave
bands is yet another.  In the old days, wider first IFs, with narrower audio
filters (or lower IF filters) would have been one way to accomplish the same
thing.  The concepts are not new, just the implementations!!

Designing these new SDRs becomes more of a trade-off than ever.  How many
A/D bits, what speed A/D, how many stages in the CIC and FIR filters, how
the DSP is done, and many other things, make up the overall receiver
performance.  Of course, what is considered an important parameter in
measuring that receiver performance is also paramount.

Regarding sample rate, I believe Phil Convington had such a choice when
designing the QS-1R.  He could have used a less expensive version of the
LTC2208, which sampled at 16-bits but at a slower sample rate.  I think he
said it was about $20 cheaper.  He decided against it, not only due to
wanting to cover 6M, but also the higher sampling rate allowed him to
achieve slightly more processing gain on HF.

For a 16-bit A/D, the only reason you will need a preamp is that the input
noise figure of most A/D converters is horrible.  Putting a preamp in front
of the A/D helps in that area, and if the preamp gets overloaded, switch in
an attenuator.  I think that is something that Phil has found, and the
reason for his upcoming addition of a preamp board for the QS-1R.

The approach that the local guy here has taken with using a much-cheaper
12-bit A/D, with a VGA in front of it, also seems interesting.  With the
addition of a switched BPF before the VGA, you only have to worry about real
strong signals on the same BPF band that you are listening to.  Then, as
long as you don't have a very strong (overdriving) signal in the same filter
band as a very weak signal of interest, things should be OK.  As you point
out about Europe, this is possibly OK in the US, but likely a problem in
Europe.  His goal is to have this board cost less than $100 (not including
the Digilent Nexsys II FPGA board).  This looks to be a very interesting
"entry level" design, compared to the HPSDR and QS-1R.  I hope he allows the
design to be propogated.  He has put up a web site:
http://sdrtrack.drupalcafe.com/ to show the design.  Included is a pix of
his board plugged into the Nexsys II, and some other info.  He has shared
this with some GNU people, so I don't think it's secret work.

BTW, he is actually working with the GNU Radio SDR software.  Yesterday, I
mistakenly said Linrad.

Well Frank, I've also gone on and on, but I think that is part of what this
is all about, an exchange of ideas, comments, and thoughts.  I think we
should do more of this, and look forward to more emails.

Also, if I'm wrong about any of this, others should feel free to chime in.
I'd rather be corrected than believe in something  that's not correct.

Thanks for the kind words.
73s
Terry


> -----Original Message-----
> From: Frank Gentges [mailto:fgentges at mindspring.com]
> Sent: Wednesday, January 07, 2009 10:58 PM
> To: WB4JFI
> Cc: Tacos; Terry Fox
> Subject: Re: TAPR High Performance Software Defined Radio
>
>
> Terry,
>
> I am impressed with the newer hardware and how we can get this
> hardware/software in the home as hams/experimenters.
>
> First, the 130 dB dynamic range you quote is a good number for HF, at
> least in the 1970s when was looking at that.  The number of 500 kW
> broadcast stations created a number of high level signals to deal with.
>   Essentially, the front end had to not overload for even an instant
> (sample period?) to realize all the dynamic range.
>
> However, if the clipping can be limited to just a few samples, you can
> let them happen if you are not digging down to the 130 dB down stuff.
>
> Now with Radio Moscow pretty much gone and the fact we will be using our
> receivers on this side of the Atlantic, we can ratchet down out top end
> peaks that would have caused the clipping.  Right now I would be happy
> with a 66 MHz sample rate at 16 bits and put up with a few clipping
> points. Put that same receiver in Europe and all the strong signals and
> it is not so great.
>
> But lets step back to an earlier day.  Consider the older vacuum tube
> receiver front ends and IFs.  With a Hammarlund SP600 you would get two
> tuned stages plus a tuned mixer before the first IF.  As you moved away
> from the tuned frequency you did not have to be far before the front end
> would start seriously rejecting the signal enough that the IF would
> never see enough of the signal to be a problem.  As you got even closer
> to the tuned-to frequency the IF would have several stages to again
> reject the signal.  The problem was as much with the first RF amplifier
> overload and clipping as further down in the chain.
>
> With the R-390A they found it was better to eliminate that first RF
> stage.  It lost the tuned circuit but it also lost the amplifier and its
> clipping on really strong near in signals.  That turned out to be a
> favorable trade off, especially when cost was a factor.
>
> Now we find ourselves with a wide open front end getting our dynamic
> range with more bits, and therefore linearity, in the A/D and accuracy
> in processing.  Rather than tune the front end to the exact frequency we
> now are putting a bandpass filter like a 1/2 octave wide to reject off
> frequency signals.  Nearby signals that pass unattenuated through the
> bandpass filter present a problem with overload and its clipping.  A VGA
> will avoid clipping at the expense of noise figure and weak signal
> signal performance.
>
> I also want to agree with your comment earlier that oversampling has
> processing gain.  As faster A/Ds become available, we can choose to
> sample faster as an alternative to the number of bits of precision.
> This is often overlooked as we seek to get the fastest sampling rate
> available and costs increase exponentially.  Newer hardware may present
> inexpensive alternatives.
>
> Well, here I go on and on and on.  An interesting and insightful email
> Terry.
>
> Frank K0BRA
>
<snip>