On Thu, 2007-07-26 at 18:35 -0700, Duane - N9DG wrote:
> --- "Dr. Gerald N. Johnson" <geraldj@storm.weather.net>
> wrote:
>
> > There is a fundamental difference between the TenTec and
> the various ham
> > SDR that have open source software. The many SDR use PC
> audio cards for
> > the A/D and then do ALL the computation in the PC external
> > to the radio. Sometimes only a certain few audio cards will
> work fast
> > enough.
>
> The trend is to shift away from consumer sound cards for this
> purpose. The HPSDR Janus board and Flex-5000's being cases in
> point. FWIW it hasn't been the "speed" of the sound card that
> defines their usability for radio IF applications, but
> instead their own dynamic range, and noise flatness over
> their spec'd range of frequencies.
>
> The big payoff for DSP processing on an external PC is the
> user interface possibilities that are opened up. The signal
> processing (DSP) part of doing SDR via a generic PC HW is
> like smashing ants with a sledgehammer. The extra CPU
> horsepower is more likely to be burned up via the OS and the
> other UI things that you may attempt. But PC horsepower is
> cheap, so who cares unless you need very low power
> consumption and/or high portability. I really do hope that
> that upcoming panadapter add-on at a minimum has the ability
> to provide an I/Q out that can be fed to other external PC
> based programs. This will allow 3rd party SDR programs to
> become the "eyes" to a primarily HW radio. I'm already see
> talk of doing this very thing with the K3. And you can bet
> that it will be happening pretty quickly with the K3.
>
> > And then the SDR of the current market are direct
> conversion radios with
> > RF stage, a LO with quadrature outputs and a couple mixers
> > with low pass filters and gain stages having probably no
> more than 150
> > KHz bandwidth. The software controls the LO in large (maybe
> 10 to 50 KHz
> > steps) and ALL the fine tuning, filtering, and detection is
> done in the
> > DSP software running in the attached PC. The simple rock
> locked radios
> > neglect the LO tuning and cover a big chunk of band all
> with the DSP
> > software in the computer.
>
> This is not true in all cases. In the case of the SDR-1000
> the radio itself is fully capable of tuning in very fine
> steps. But there are good design reasons to not normally run
> it that way. Since it uses a DDS LO there are certain
> frequencies when generated will have higher spur content than
> others. All DDS's do this. Traditional wisdom has been to
> follow a DDS with a PLL of some kind to "knock down" the
> spurs. However since these spurs are entirely predictable you
> can simply choose to not use the specific frequencies that
> are most spur prone. This is precisely what the
> PowerSDR/SDR-1000 combo does with "spur reduction" turned on.
> As it works out the LO tuning step size is ~3.1 kHz.
>
> > The ultimate radio performance depends on the dynamic range
> of
> > the RF and mixer and tremendously on that of the A/D
> converter.
>
> More often than not it is the A/D alone that is the limiting
> element. The QSD I/Q has proven itself to be capable of very
> high dynamic range.
>
> > In TenTec radios, the LO is controlled to the finest of
> frequency steps,
> > the RF is bandpass filtered then at the IF its filtered
> more with the
> > "roofing" filter, then converted down to a 15 KHz IF where
> the direct
> > conversion Q and I process is done and then the fine
> filtering, very
> > finest of tuning (if the roofing filter allows), and
> detection is
> > accomplished in the DSP.
>
> I have surmised that the reason that TT chose the 2.5 kHz
> tuning step for the RX320, Peg, Jupiter, plus a few others
> was to minimize phase noise from the MC145170 PLL used. It is
> interesting to note that the K2 also uses the MC145170 PLL
> that is also tuned in comparatively coarse steps with other
> means used for the fine step tuning,
>
> > So the software does a great deal more radio control before
> the DSP and with the use of narrow "roofing"
> > filters the DSP hasn't all that much to do.
>
> Don't think it changes the "amount of work" for the DSP one
> way or another. Processing 24 bits at 192 Ksamples per second
> is the same whether the sampled passband is wide or narrow.
> What the roofing filter does do is to reduce the odds of the
> signal being mangled before it hits the A/D.
The narrow roofing filter takes work away from the DSP and hobbles the
DSP at the same time when noise reduction is desired. That's because
most noise reduction algorithms work by comparing coherent signals to
incoherent noise. But a 500 Hz slice of random noise isn't all that
random. So the algorithm fails to work well. And the narrow filter
broadens time of the very narrow pulses typical of power line and
ignition noise so that the DSP has a hard time telling those pulses from
desired signals.
>
> However what I think I'm seeing is that the designers of
> narrow roofing filter radios are now bumping into is the IMD
> generated within the roofing filter itself. I'm increasingly
> convinced that this is what the Orion and the K3 have as
> their limiting factor for close spaced (<2 kHz) IMD DR
> performance.
That can be a factor. There surely is a linear limit to the amplitude of
vibration of a quartz crystal. And as likely linearity limits on cores
of any inductors or transformers in the circuit.
A standard problem that shows in the close in intermod of rice boxes
too, is that there is gain after the roofing filter which means signals
within the roofing filter pass band are even stronger at the second
mixer. The fact that most receivers show poorer intermod performance
tells me that the second mixer isn't built to handle a stronger signal
than the first mixer and so the intermod inside the roofing filter pass
band is higher. That should be a relatively simple design change to
increase they intermod range of the second mixer which then demands more
LO power, but its a known solution. I'm sure its complicated by
amplitude and phase noise on that LO signal.
Mixer and amplifier terminations are critical too. A filter that
reflects signals outside the pass band contributes to mixer and
amplifier intermod because the load impedance outside the filter pass
band can be far from the in band impedance. And so the amplifier or
mixer runs excessive current or voltage for the out of pass band signal
which can cause much intermod. Mixers also need to be terminated at the
output image and harmonic image products as well as the LO frequency
else they won't do rated intermod. But phase noise on the LO seems to
make blocking worse than those bad terminations, at least in the FT-736
that I've been attacking receiver intermod on. It has no isolation
between the first mixer and the first crystal filter, but adding a
broadband MMIC amplifier with a P1 of +19 dBm has had not detectable
effect on blocking range or sensitivity, though the added 12 dB of gain
lowers the AGC threshold usefully and makes the receiver "sound alive."
>
> > That makes most of the dynamic range depend on RF hardware
> and allows a smaller dynamic range
> > and narrower bandwidth A/D but exposes the radio design to
> all of the
> > same foibles in dynamic range and close in intermod of
> multiple conversion
> > analog radios that have been fought for decades.
>
> Ah yes, the folly of numerous IF conversions. It is kinda
> ironic that radio started out without any IF stages at all;
> and then reached a peak of 4 or so, and has now retreated
> back to having just one or two. And now the possibility of
> having none is just not that far away.
Before Armstrong's invention of the superhet receiver, there weren't
really many RF tubes and tuning a several stage TRF often ended with it
oscillating and clobbering the neighborhood receivers. In those days
only triodes were used and they were inherently unstable. Fixed
frequency IF stages at least allowed better feedback neutralization of
each IF stage so the IFs didn't oscillate. Multiple conversion came
about because it was easier to make LC filters at lower frequencies,
indeed some of the earliest SSB receivers used a 15 KHz IF just like the
TT Orions.The Hallicrafters SX-96 and SX-88 used 50 KHz as the last IF.
Then when crystal filters came along the filters moved towards the
antenna but most analog local oscillators weren't up to the stability
required before the permeable tuned oscillator of Collins and TT. There
is a rub in moving all the selectivity way up front and that is the wide
band noise introduced by the IF stages and mixers and receivers would
benefit from some analogue filtering at the detector too. There is one
German 2m transceiver kit that does that running single conversion that
gets good dynamic range and super S/N at the audio. so it would appear.
That's a project that I need to finish some year.
>
> Duane
> N9DG
>
> > But TenTec has known how to produce
> > radios with good dynamic range for nearly that long too.
> > --
73, Jerry, K0CQ,
All content copyright Dr. Gerald N. Johnson, electrical
engineer
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|