[time-nuts] OT - Portable Digital 'scope
Jim Lux
jimlux at earthlink.net
Thu Feb 23 23:16:27 UTC 2012
On 2/23/12 2:22 PM, Azelio Boriani wrote:
> Actually, undersampling does use the alias effect to bring down the RF
> carrier. That is, the direct sampling radio concept cannot avoid the
> aliasing: it is exploited to avoid, for example, to sample a 2GHz carrier
> modulated with a 20MHz signal with a 4Gsample/second ADC (by the way, does
> it exist?). A simple 20Msample/second ADC would be enough.
Well.. a 40 MSPS ADC for a 20 MHz wide signal, but yes.
There *are* multi Gsample/second ADCs out there. ADC12D1800, 12 bit,
3.6 GSPS.
ADS5400 12 bit 1GSPS
etc
Yes, to analyze
> an analog signal in real time I doubt you can use this method, if the
> signal is periodic maybe... you can advance the sampling trigger a bit,
> cycle by cycle, and reconstruct the whole signal after this little amount
> has covered a full cycle.
>
That's the way the "sampling head" used for microwaves works (back in
the day, when tunnel diodes were a new and exotic thing, for instance).
Still used to today on scopes where they call it things like "equivalent
time sampling"...
basically a real fast sample and hold, and a not so fast ADC.
There's also clever schemes with fast S/H and multiple interleaved
ADCs.. but accounting for the tracking errors between ADCs is always a
chore.
And, various "sub-band-coding" approaches where you subdivide the
frequency into sub-bands, digitize them in parallel and do fancy math to
recombine it into a single stream of samples. If the channelization and
ADC clocks are cleverly chosen and derived from the same reference, you
can do quite well (& compensate for differences among ADCs)
The high performance radar processing world is full of this kind of
thing (look up STAP: space/time adaptive processing).
More information about the time-nuts
mailing list