Re: [Discuss-gnuradio] Resistance? Capacitance? Inductance?

[Tom Rondeau]

On Wed, Feb 27, 2013 at 11:38 AM, Marcus D. Leech <address@hidden> wrote:
> On 27/02/13 10:08 AM, Tom Rondeau wrote:
>>
>> On the other hand, one of the major areas of work that we are still
>> pursuing lies in the RF front end. We have wideband systems. Ettus has
>> produced a number of daughterboards that cover multiple GHz of
>> spectrum, which is fantastic. But through that, we suffer a bit on the
>> amplifiers and filters needed for some kinds of communications tasks.
>> What Ettus has done is produced very good IP3s, NFs, gains, etc over
>> these large bandwidths, but that doesn't exactly compare to having a
>> filter and amplifier specific to a small bandwidth for something like
>> cellular communications. Or even, for that matter, antennas for
>> various waveforms.
>>
>> Even today's wideband RFICs tend to have a lot of tweakable/tunable
>> parameters to meet specific needs of different areas of spectrum. Are
>> there software solutions that could be used to automatically adjust
>> these parameters? Or an RLC matching circuit? Some of this, I know,
>> requires advances in the materials and components to make any sense,
>> in other cases the feedback loop could be a bit long to make any
>> significant impact. But it's fun to think about. Goes back to my
>> dissertation days, actually :)
>>
>> Tom
>>
> In some sense, what we're talking about here is the difference between
> Software *Controlled* Radio, and
>   Software *Defined* Radio.
>
> A chain of DSP blocks applies a series of mathematical transforms to a
> digitized signal, in similar ways
>   to the way a series of R/L/C/Gain components do to an analog signal.
> One can think of the R/L/C approach
>   as performing rough *approximations* to a transform that is defined in
> strict mathematical terms.  The DSP
>   approach, in general, is able to achieve a higher "fidelity" of those
> transforms with a higher degree of
>   flexibility and reconfigurability than could possibly be achieved with
> analog hardware. Although, at some
>   considerable cost--a simple FM demodulator chip is $0.35 in bulk,
> whereas the amount of compute-gear you
>   require to do the same thing is considerably more costly.  But DSP/SDR
> doesn't require that you break out
>   the soldering iron and parts bin every time you want to
> tweak/repurpose things.
>
> Now, having said that, the notion of having some kind of "tracking"
> filtering isn't a bad idea, the problem comes
>   down to implementation, and the cost trade-offs involved.
> Considering things like daughter cards covering
>   30Mhz to 4.4Ghz, exactly how many "cuts" across that bandwidth do you
> make, and how much are people
>   willing to pay for additional dynamic range implied by band-limiting
> at the RF end of things?  The technology
>   is mostly there -- GaAsFET RF switches, LTCC filter modules, saw
> filters, dielectric filters, etc, are all out there.
>   But almost any "hard" decision made by the manufacturer in such things
> is likely to be "wrong" in enough cases
>   that perhaps all of that should be done externally to a daughtercard,
> with provision for a generic switching
>   interface (like the existing GPIOs on many Ettus daughtercards).

Well said.

> In an ideal world, you wouldn't need much front-end filtering.  Your
> gain stages would be uber-linear up to ridiculous
>   input powers, and you'd have a 24-bit ADC sampling at several Gsps.
> We aren't anywhere near there yet.

Yeah.... that's a hard one to swallow. The near-far problem in some
bands makes even what you're considering here problematic. ~140 dB
dynamic range does sound pretty good for most things, though :)

There are other proposed solutions out there that take this concept
even farther but require much greater investment in hardware cost
(we're talking multiple ADCs, DSP units, etc.).

But we're still going to need front end filtering and amplifiers for a
while longer. Oh, and let's still not forget the antenna, though there
are decent solutions for narrowband signals over large bandwidths
(that is, good performance over a large bandwidth with varying group
delays along the way, so you can only get away with narrowband signals
or clever correction algorithms).

Tom


> --
> Principal Investigator
> Shirleys Bay Radio Astronomy Consortium
> http://www.sbrac.org