[hpsdr] Bandwith and decimation questions

[Jeremy McDermond]

On Mar 22, 2011, at 8:39 AM, Maximo EA1DDO_HK1DX wrote:

> There are some things in my head not really clear.
> 
> Are next facts right?
> LT2208 at 16 bit and 130 million samples per second; 16 x 130.000.000 = 2080 mega bits/second

That's about correct.

> Where decimation takes place, at Mercury or at Ozy/Metis?

My understanding is that the decimation and filtering happens on Mercury.  Check out the Mercury software block diagram at:

http://openhpsdr.org/wiki/index.php?title=Image:Mercury_Verilog%286%29_1.jpg

Notice the CIC blocks to decimate. 

> Because if decimation takes place at Ozy/Mercury it means that are 2080 Mbits/s on the Atlas bus. Is that right?

See Above.  The decimation is on the Mercury boards.

> Then, USB2 from Ozy has a continue bandwith around of 250 Mb/s. But Metis will be around 960 Mb/s.
> Will decimation adapt to the Metis bandwith?

No.  There are three fixed sample rates that will come out of Ozy or Metis:  48000, 96000 and 192000.  On PowerSDR those are, mildly deceptively, set in the Audio tab.  On Heterodyne they're just labeled as "sampling rate" on the control panel.

> I mean, if decimation adapts the output to the existing bandwith. 
> More bandwith plus less decimation means more data on the PC, so the spectrum we see/hear will be better ??

No.  Again, you get three fixed bandwidths on each "receiver" you get from the hardware, so there are no immediate advantages to more bandwidth.

This doesn't mean it's this way forever, though.  Right now, with the "multiple receiver" architecture, a single Mercury can give you two 192k pieces of bandwidth.  Likewise, Joe's diversity work can give you 192k pieces of bandwidth off of multiple boards.

What I'd kinda like to see is an option to get entire ham bands at 2xbandwidth samplerate.  So, for example on 20m I'd like to see a way to get:

Frequency: 14.000 - 14.350
Bandwidth: 350 kHz
Sample Rate: 700 kSps
Data Rate: 16.8 Mbps

I did some "back of the envelope" calculations and covering the entire US ham band allocations simultaneously at two times the bandwidth sample rate is around 361.8Mbps.  If you exclude 6m, you get only 169.8Mbps.  There should be plenty of bandwidth available on Metis to do this.  Whether your computer can handle the datarate may be another story.  I'm hoping to learn enough Verilog to be dangerous one day and to try to implement such a system.

> I have not really clear how it works, thats the reason for my questions.
> 
> Thank you.
> 
> 73, Maximo - EA1DDO

--
Jeremy McDermond (NH6Z)
Xenotropic Systems
mcdermj at xenotropic.com




 1300811116.0