Re: [softrock40] General Coverage SDR?
- At 01:33 PM 4/1/2008, you wrote:
I'm thinking about how to simply bring a ham-band SDR to
general-coverage while minimizing trade-offs in performance. The idea
is based on using a low-noise high input impedance broadband pre-amp
with output impedance that matches what follows to the QSD etc. This
will isolate the antenna impedance (which varies widely for
general-coverage receivers using typical ham-band-optimized antennas)
from the follow-on effects in a typical ham-band SDR like the Softrock
which employs inverting op-amps after the QSD. In SDR's like the
Softrock, receiver, gain is roughly determined by the op-amp feedback
resistance divided by antenna impedance. Plus, receive bandwidth is
affected by the antenna impedance as well (see Gerald Youngblood's
(AC5OG) 4-part QEX papers, "Software-Defined Radio for the Masses).
I like the Norton amplifier but the output impedance reflects into the input impedance so a small attenuator pad on the output of the amp will stabilize the impedance it sees and the gain will not vary on you. The antenna impedance no longer will affect the audio gain.
The antenna impedance vs. gain issue may be mitigated by using a
differential (instrumentation) amplifier instead of an inverting
op-amp, and today it is possible to make a differential amplifier with
good noise performance, 1nV/rtHz or less. Not as good as the best
traditional op-amps available today in terms of noise but more than
sufficient for HF and maybe even VHF/UHF work. But with a differential
amp, you still have the effects on receive bandwidth dependence on
antenna impedance, which many may rightfully argue doesn't matter as much.
The main problem is that low noise differential instrumentation amplifiers ten to be very expensive, although that is not an issue if it's for a personal high performance receiver. I recently purchased a differential input software controlled variable gain amplifier for use in my own design receiver it solves several problems at the same time. The SDR-1000 and 5000 use differential op-amps after the detector.
Yes I know, SDR purists will balk at the idea of putting a pre-amp in
because it will limit the inherent dynamic range of a SDR without a
pre-amp. But a pre-amp (hi input impedance or not) with good-noise
performance, bandwidth over HF (at least) and sufficiently high
dynamic range is possible these days, at sufficiently low-cost. And it
may remove the antenna impedance from affecting receiver gain and
The trick is to have controllable gain, on the low bands no gain is desired, so in those cases attenuators on both the input and output of the amp are needed. Once you hit 20M and higher the noise starts going down on the band, while the detectors noise is going up. In those cases the amplifier is a desirable thing, but again with the ability to control it.
So the question is, can the reflections be
"shunted" to ground by a broadband directional coupler?
A duplexer where audio goes through but RF is shunted to a load to have the RF input have a decent SWR would be desirable.
Then there's the question of input filters. Ideally a preselector or
some other form of band-pass filter is needed. I would argue that
preselection may be done in the differential amplifier's feedback path
instead of at the input from the antenna.
A QSD will respond to odd harmonics of the LO clock, those signals must be kept away from the detector or you will be listening to signals that are really not on the band that you think you are listening to. By the time it gets to the audio it's too late to figure out which signal is real and which one is a down-sampled harmonic.
One last thing comes to mind, what about just "padding" the antenna
input with a passive-resistive attenuator? 50+j0 is about 0.9 nV/rtHz,
1kOhm is about 4 nV/rtHz; just for comparison. Hmmm...
My own experiment that I almost have all the parts for consist of a variable gain attenuator 0 to -30dB in 1 dB steps followed by a Norton amplifier and output pad to set the impedance into the QSD and then a duplexer followed by a software controlled low noise Differential Instrumentation Amplifier. The IA has a maximum gain of 32X so the RF amplifier will supply some gain when needed. The undecided item at this time is the A/D portion of the receiver but I'm starting to lean towards a high speed base band 16 bit A/D followed by a decimator in a FPGA chip, then into the PC via a USB port.
- --- In firstname.lastname@example.org, "adibene" <i2phd@...> wrote:
> --- In email@example.com, "Terry" <wb4jfi@> wrote:
> > If I understand, HF should have a dynamic range of 130dB or so.
> > translates to 22 bits, I believe. If appropriately dither iscurrent
> > applied, and subsampling to demod audio for SSB or CW, the
> > crop of 14 to 16 bit A/D devices should just barely get us there.range
> > Static crashes and other anomolies that go beyond this 130dB
> > will still overload the A/D, although only at the peak excursions.Hey Alberto. I've heard that number a few times, including Joseph
> Not sure, but 130 dB look to me a bit too many. No analog radio on
> the market, AFAIK, has such a big dynamic range, but they have no
> problems to cope with HF signals.
> 73 Alberto i2PHD
Mitola's book (Software Radio Architecture). He suggests that 130dB
is the dynamic range for HF-RF. He suggests an HF-IF (in the .2-
10MHz range) sould be 72-120dB. Most other places that I've heard it
probably reference back to him, although I think I saw it derived in
at least one other independent book somewhere.
Most radios are narrow-band, with filters, AGCs, switchable amps &
attenuators, and other gadgets to translate the level of the
currently received signal into a narrower dynamic range that the
radio can handle.
I think he is suggesting that the 130dB is the minimum for directly
feeding an A/D converter without overloading, but still reaching the
HF noise floor.
Obviously, we can use the same range-limiting (or maybe it's better
to say range-adjusting) tools if we are using an A/D receiver
approach. Sticking a preamp and/or attenuator inline will help, as
long as it can be switched out (or the gain adjusted with an AGC
loop). Applying filters to remove large unwanted signals (such as
the US AM broadcast band) is another possibility. I know hams have a
hard time (with normal radios) trying to work 160M with nearby AM
broadcast stations, and often resort to filtering.
But, to a SDR "purist", sticking filters, preamps, attenuators, etc,
may all be considered band-aids to mask the real issue - not enough
dynamic range. And, they all potentially hamper optimum reception in
one manner or another. Like in a traditional superhet, these band-
aids may be necessary for a while yet.
I forget where to find the noise floor on the various ham bands.
But, I'm totally guessing that it's around -128dBm on some HF band.
To need 130dB of dynamic range with that small a MDS signal means the
MAXIMUM signal would be over 0dBm! That's a ton of signal. But, I
hear some people who live near AM broadcast (US), or SW broadcast
(europe) transmitters may get that strong of a signal. Running
multiple stations nearby (field day, etc), may also create this
There was a QEX article in Sept/Oct 2002 (The DX Prowess of HF
Receivers) that indicated the BDR of a K2 is 133dB (20kHz) and 126dB
Feeding an antenna directly into an A/D input may be folly for a
while yet, especially during electrical storms. I don't think
anybody plans to have an A/D with that high a dynamic range!!