Threshold Detection

[rroonnaakk]

Hello,
I am working on my fourth year project and are building a UWB transever. Well I am working on the A/D block and was wondering if anyone could help out here... I need a 1GHz (Need a decision every 1nS) threshold detection. Due to the speed concerns I dont think I could use comparators.... we have 10 correlators and need a a A/D conversion on each line... So basically I need something that will convert voltage levels into Digital. a Quantizor maybe.... Could someone point me in the right direction...

 

[Russlk]

What kind of A/D? one bit or 10 bit? How many volts is the signal? What does UWB mean?

 

[rroonnaakk]

Well I guess I should have clarified all that... but here it goes....
UWB - UltraWide Band.
The A/D is lets say 16 bits
but basically I jus need a A/D converter that will make decisions every nanoSecond. and will convert voltage levels to 16bit digital signals.
The voltage levels come from a correlators... and can be modified by changing the multiplication coiffecients in the correlator.. So lets assume any value we find convient.

 

[Russlk]

Sorry, I have no idea how to do a conversion in 1nS, but am sure it requires an integrated circuit, it would be impossible with discrete components.

 

[Roff]

Have you considered that, with a one volt peak-to-peak signal, you are hoping to resolve 15 microvolts at 1gigasamples/sec? I think this is way beyond the present state of the art. According to Maxim's web site, the **broken link removed** is their highest performance A/D. It's 8 bits at 1.5 gsps. There may be faster ones out there, but keep in mind that 16 bits isn't twice as "good" as 8 bits - it's 256 times. There are schemes (such as in digital scopes) that achieve sample rates as high as 20 gsps, but they are probably also 8 bits.

 

[jem]

Just adding my $0.02 worth to this discussion...

I think you should review your architecture if by what you hear, your current one is technically close to impossible (at least with off-the-shelf components). It might be possible with SiGe, or HEMT technology, but certainly not to college students.

Could you conceivably stagger the ADC's. Give each one, say, 10ns to do a conversion, and read each one in turn while the others are still converting? Maybe you could divide your 1GHz UWB into ten 100MHz bands, and downconvert each higher band to 0-100MHz. Just a couple of random thoughts.

Jem

 

[Roff]

Just adding my $0.02 worth to this discussion...

I think you should review your architecture if by what you hear, your current one is technically close to impossible (at least with off-the-shelf components). It might be possible with SiGe, or HEMT technology, but certainly not to college students.

Could you conceivably stagger the ADC's. Give each one, say, 10ns to do a conversion, and read each one in turn while the others are still converting? Maybe you could divide your 1GHz UWB into ten 100MHz bands, and downconvert each higher band to 0-100MHz. Just a couple of random thoughts.

Jem

A/D interleaving, or multiplexing, is a technique for getting higher conversion rates, but I think A/D matching becomes a big problem. Not impossible, but big.