High Frecuency design

[geoorge]

Hello i'm making a digital scope using vhdl on a FPGA, but i dont want to make the same old example of converting a/d and then directly to memory.

how can i break the ghz, working on the frecuency domain? or there is a technique to make it

thanks

 

[VisiGoth]

Here is a wild idea (I hope that is what you are looking for!)
Try building a microscan front end. If we assume that a practical digital scope has to come up for air once in a while (it will need to stop collecting data at some point.) I am assuming you do not have a display ability that can show 100% duty cycle GHz samples. There will probably be some gaps in time. So, make the best use of the gap time. Build a chirped RF source and mix a bandlimited signal down to your A/D converter. You would be compressing a wide RF band down into a narrow IF band for the bandwidth limitations of the A/D converter, thus cheating the Nyquist limit for short duration of time. Then you would signal process out the chirp and be left with an approximation of the original data.

Or if you just want to be fancy and want to avoid that nasty RF black magic, then you could build a ganog f A/D converters. Perhaps try 17 converters that are interleaved 16 at a time (always leaving one out in a random like pattern). This will increase your BW! (Well kinda, it is hard to cheat mother nature).

Or, instead of a TDMA method above, try FDMA or even better CDMA! With FDMA you would bandlimit the input to each converter, then signal process splice them together. This is a frequency domain version of the time interleave version above.

The CDMA method would chop or modulate the input different, but orthogonally on each A/D input. Each A/D would be sub Nyquist sampling, but if you are very clever you could unfold the multiple A/D converter data into a broad spectrum exceeding the Nyquis limit capability of each individual A/D converter.