Re: [Discuss-gnuradio] Re: Welcome and brief update

[Dave Emery]

On Sat, Jan 11, 2003 at 09:34:49PM -0800, John Turner wrote:
> Occam's Razor
> "one should not increase, beyond what is necessary, the number of entities
> required to explain anything"

        Yes, I admit closely related to KISS...

> 
> It sounds as if there are a number of highly diverse applications being
> developed. The topology I put forward was meant to form a platform roughly
> consistent with wideband radio receivers of the hobbyist variety, refer to
> ICOM and Kenwood units. These radios are suitable for general monitoring of
> signals. Their typical radio specifications are in the 50 to 60db range,
> i.e. 60 dB selectivity and image rejection, 50 dB intermodulation rejection.
> The radios typically demodulate AM/FM/SSB, some even include analog TV
> tuners, but they are not intended for commercial applications. The ultimate
> goal of the hardware being to allow a wide spectrum of applications to be
> explored without undue cost or effort.
> 
        I wonder where you get that impression about typical hobbyist 
receivers ?   I agree for example that the adjacent channel selectivity
of a typical high end hobbyist receiver (say the ICOM R-8500) might
not even be 60 db.   But that is the adjacent channel.   The radio no
doubt will work just fine with a signal  80 or 100 db higher
nearby in frequency (eg on the same band within a few hundred khz or
so).   This is because ultimate IF selectivity (except maybe at a few
images of lower IF frequencies) is much much higher than just 60 db.
And the rf dynamic range is really larger than just 60 db.

        But a cable modem tuner converting 6+ mhz around a desired
signal outputs at its IF that nearly 100 db stronger signal (say a
Nextel base station nearly overhead in the 800 mhz band) a few hundred
khz away for the ADC to convert to bits.   And that means the ADC has to
be able to digitize that IF with its 80 or 100 db more powerful signal
in it without smearing IM products all over our poor weak signal or
running out of dynamic range altogether.

        And unfortunately many portions of the RF spectrum of interest
look like that these days - with local nearby and rather powerful
transmitters not far away relative to the bandwidth of a cable
tv tuner from weak signals of interest.

> I made a general assumption that the host PC has limited processing
> capabilities, i.e. limiting the maximum effective sampling rate of 500 to
> 1MSPS. I'm curious, what can a PC realize in terms of processing benchmarks
> at sampling rates of up to 20MSPS? And are there typical formats that are
> desirable FM/AM/SSB, IEEE 802.11, Bluetooth, GPS, OFDM, CDMA.

        One can do a lot with multiple 2-3 ghz floating point capable
processors largely running out of cache... 
        
        And PCs only get faster...

> 
> I wouldn't dismiss CATV tuners out of hand. They have come along way with
> the advent of inexpensive frequency synthesis and the requirements of DOCSIS
> cable modems. The conversion phase noise is specified as dBc/Hz instead of
> residual FM (close to 90dBc/Hz at 10kHz offset). The conversion phase noise
> of the tuners will limit the demodulated C/N of the signal, my guess is to
> around 40dB. This may limit the performance of phase sensitive systems that
> require high C/N ratios but these are pretty rare. Phase noise also has the
> effect of limiting adjacent channel selectivity performance (aka reciprocal
> mixing). The phase noise mentioned is consistent with 50dB selectivity at 25
> kHz offsets. As for intermodulation and dynamic range they are required to
> function in a multicarrier environment, an 860 MHz cable plant has 110
> carriers, and the tuners have intermodulation products typically down 55dBc.
> The third order intercept point approaches that of mobile grade LMR
> equipment. Again, this performance is consistent with the hobby level
> wideband receivers.

        I am impressed by the specs you cite.  They have gotten a lot
better lately.

        And it depends on your application - narrow band weak signals or
broadband signals,  narrow band signals near powerful signals or not
etc.   Clearly your figures do match what the PLL synthesizers and front
ends do at best on many VHF/UHF hobby scanners.

> 
> As for signal filtering, careful selection of the sampling frequency to
> match the selectivity of the tuner eliminates spurious responses. At the ADC
> input the signal would be wideband, i.e. 6 MHz. since the bandwidth is wide
> relative to the expected signal bandwidth the effects of amplitude and phase
> would be minimal. Note: The ADC would need to match the dynamic range
> requirements, i.e. 12 bits is sufficient.
> 
        I fully agree that using a wideband TV tuner for narrowband
signals results in almost no delay or amplitude distortion across the
signal - my comment was only directed at using it with truly wideband
signals (basically those it was designed for).

        But see above for my comments on real life dynamic range with
real signals in a 6 mhz tuner bandwidth.
 
        And of course AGC in a wideband intermittently keyed carrier
environment gets interesting as you either back off gain to prevent 
overflow and have poor SN and worse quantizing noise and distortion
or somehow ride gain in a way that works when that pager transmitter
lights off 500 khz away.

> The incorporation of an FPGA allows features such as filtering at the sample
> rate and digital down conversion to be realized to balance the processing and
> data transmission requirements. Note: As USB 2.0 takes hold there may be a
> migration path to allow higher sample rates.
> 
        I agree completely here.


> Again, this isn't meant to be the ultimate in receiver / tuner performance.
> My interpretation from the postings is that a simple and flexible hardware
> platform is required.

        I agree that it is a good start and that making it available cheap
would go a long way in the gnu radio project.

        I think I share the views of another poster that maybe 1 ghz 
ethernet has interesting properties too.

> 
> JT
> address@hidden
> www.valhalla-systems.com
> 
> ----- Original Message -----
> From: Dave Emery <address@hidden>
> To: <address@hidden>
> Sent: Saturday, January 11, 2003 8:21 PM
> Subject: [Discuss-gnuradio] Re: Welcome and brief update
> 
> 
> > On Sat, Jan 11, 2003 at 03:00:18PM -0800, John Turner wrote:
> > > Hi Folks,
> > >
> >
> > > I believe the following topology represents a good value performance
> > > tradeoff, between $500 to $1000 unit cost.
> > > TV Tuner offering 50 to 860 MHz range and 60dB dynamic range performance
> A
> > > channelized output of 40 MHz.
> > > ADC converter operating at 70 Msps, the idea is that the IF is
> undersampled.
> > > Digital Downconverter as a sample rate converter
> > > Sample rate to USB converter
> > > I believe the Downconverter and the USB interface could be built into an
> > > FPGA.
> > >
> >
> > I hate to inject a note of reality, but isn't the objective of
> > any receiver design (what we are talking about here I think) to receive
> > some specific signal or set of signals ?
> >
> > Specific signals impose specific requirements on a SDR receiver.
> >
> > Some are narrowband (tens or hundreds of hz at most), some
> > moderate width (2.1, 3,2, 5 khz, 11.2 khz, 15 khz,  20 khz, 30 khz),
> > some relatively wideband (50khz, 200 khz, 350 khz, 1.25 mhz), some quite
> > wideband (6 mhz), and some very wideband (18 mhz, 27 mhz, 36 mhz, 50
> > mhz, 100 mhz...)
> >
> > And some are found in environments with nearby signals in the
> > passband of the tuner at comparable levels within say 20-40 db (CATV
> > signals on a cable for example or signals on a C or Ku band satellite
> > transponder) whilst others (weak signals at close to the noise floor on
> > HF or VHF (2 meter weak signal work) may be up to 110 db or more  below
> > the strongest signals in the passband of the tuner.
> >
> > Some wideband signals (ATSC TV) take quite a bit of processing
> > horsepower, whilst others (FM broadcast) may be able to be demodulated
> > with simpler less cpu intensive algorithms.
> >
> > In general for narrowband signals LO phase noise and phase
> > and frequency stability is important - reciprocal mixing makes all
> > signals downconverted by a LO as bad in respect to frequency and
> > phase stability and noise sidebands as the LO is.   So for narrowband
> > work one needs really clean stable synthesizers.   And long term
> > frequency accuracy and tuning resolution may become important too
> > with narrow band signals.
> >
> > And clean, low phase noise LOs are important with weak signals
> > in the presence of strong signals too, as the phase noise sidebands from
> > a noisy LO downconverting nearby strong signals can land on top of the
> > weak barely discernible signal and cover it up.
> >
> > But this has to be balanced off against cost (crummy LOs are
> > cheaper) and wide tuning range and quick lockup.
> >
> > TV tuners are designed to work with wideband signals where LO
> > stability and phase noise has not been a major issue but cost and quick
> > tuning is.   And TV tuners have generally not be required to work all
> > that well in very high dynamic range environments, though they have
> > gotten a lot better in recent years.
> >
> > And of course for many receiver applications there are other
> > considerations too - overall noise figure of course determines minimum
> > detectable and usable signal, and for a lot of real world applications
> > maximum sensitivity is vital to getting useful results as at least
> > some of the signals of interest are quite weak.
> >
> > And very critical in modern rf environments (particularly at
> > VHF and low UHF and increasingly at higher frequencies too) is dynamic
> > range (3rd order intercept).   Many actual signals in the modern world
> > are found in dense rf environments with powerful nearby transmitters
> > on adjacent channels and if ones rf front end overloads or generates
> > significant intermodulation products one may be unable to hear a signal
> > of interest at all even if it is well over the noise floor.
> >
> > And for any SDR there is a tradeoff between analog filtering
> > ahead of the A/D and use of an A/D with high dynamic range (lots of
> > bits and few mixing products).  Fast, high dynamic range A/Ds cost
> > more (but are coming down in price fast), whilst super high dynamic
> > range A/Ds exist for narrow band signals that are very cheap.
> >
> > And finally, of course, for wideband signals one often cares
> > about the amplitude flatness and phase linearity of the tuner passband.
> > If it isn't flat more complex demodulation and equalization algorithms
> > may be required to correct for delay or amplitude errors induced by
> > the tuner.
> >
> > So simply designing an abstract architecture ignores some
> > of the considerations involved with real signals and if one wants
> > to obtain useful and even competitive performance one has to decide
> > what sort of things one is targeting and what one is not.
> >
> > For example, if some GNU radio hardware is directed primarily at
> > cheaply allowing a PC to tune in FM broadcast and NTSC or perhaps
> > especially ATSC TV signals, one hardware architecture may suffice,
> > whilst if one intends to use it as a useful scanner or communications
> > receiver another may be needed.  And if one intends to use it to decode
> > GPS or other specialized signals highly specific requirements may exist
> > in order to get useful results at all.  And obviously if one intends to
> > build a device that will interoperate with existing systems, one needs
> > to try to meet minimum acceptable performance standards for devices used
> > with those systems.
> >
> > --
> > Dave Emery N1PRE,  address@hidden  DIE Consulting, Weston, Mass.
> > PGP fingerprint = 2047/4D7B08D1 DE 6E E1 CC 1F 1D 96 E2  5D 27 BD B0 24 88
> C3 18
> >
> >
> >
> > _______________________________________________
> > Discuss-gnuradio mailing list
> > address@hidden
> > http://mail.gnu.org/mailman/listinfo/discuss-gnuradio
> 
> 
> 
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/discuss-gnuradio

-- 
        Dave Emery N1PRE,  address@hidden  DIE Consulting, Weston, Mass. 
PGP fingerprint = 2047/4D7B08D1 DE 6E E1 CC 1F 1D 96 E2  5D 27 BD B0 24 88 C3 18