From: <Chris.Phillips_at_email.protected>
Date: Wed, 5 Jul 2006 20:15:22 +1000 (EST)

Hi guys,

Haystack is developing a digital backend based on the Berkeley "ibob" card.
I think this could be very interesting possibility for places like NZ,
Ceduna for geodetic or high data rate compatibility etc. Below is a draft
of a "requirements" note I would like to send to Haystack soon. This is
for compatibility with the existing LBA DAS.

Cheers
Chris

------------------------
                        **DRAFT**

        Potential LBA requirements for Haystack DBE

Intro

Haystack observatory are developing digital backend as a replacement
for the current Mark4 BBC rack. The system is based around the a
Berkley sampler card and iBOB processor. The DBE would be ideal
replacement for the LBA DAS for stations such as NZ and New Nortia (as
it will be difficult to build any more DAS). They would also be a way
of getting data rates of up to 1 Gbps at Hobart, Tidbinbilla and
Ceduna (in mode compatible with ATNF antennae recording at 1
Gbps). With move to K-band geodesy, the DBE would allow ATNF more
useful participation of LBA antenna for geodetic experiments (more
flexible assignment of channel frequency spacing) and space craft
tracking.

Berkley Casper Project

The DBE is based on the Berkley "Casper" project. They are designing
FPGA modules which can be flexibly linked together. More importantly
they are working on high level software to make it "easy" to make new
FPGA configurations. It looks like this could be an ideal environment
for groups such as AUT and UTas to develop new specialised hardware
and moderate cost ($$ and time).

Summary of DBE specifications

 - 2 x 500 MHz input IFs
 - FPGA based polyphase filter bank
 - Channel widths 0.5 MHz - 32 MHz
   - Fixed separation of channels
 - 2x VSI outputs
 - Selectable recorded output channels (up to 32x 32 MHz)
 - Maximum output data rate ~4-8 Gbps (Mark5b able to handle 2 Gbps)
 - 4 x 500 MHz version possible with twice output datarate

LBA requirements (for compatibility with existing LBA DAS and LBADR recorder)

 - 2,4,8,16,32 & 64 MHz bandwidths
 - Data rates up to 1 Gbps (specifically 4x64 MHz)
 - VSI clock rate no higher than 32 MHz
 - Ability to flip entire band to allow input IF USB or LSB
 - In 1 Gbps modes able to output 512 Mbps per VSI output (32 MHz clock)

Note that for 1 Gbps modes with the LBA data the data is recorded as
two 16bit data streams sampling four 64 MHz channels. Within each 16
bit data stream 8 bits correspond to four 2bit times samples of the
first 64 MHz channel then the same for the second channel (ie the data
is multiplexed in time by a factor of four). If the data was to be
correlated on a software correlator a different arrangement of bits
should not matter and potential the cross bar on the Mark5b DOM
(output module) could be configure for slightly different recording
modes.

Notes:

For narrow band low data rate observations (say 4x 4 MHz or 8x1 MHz)
it would be highly advantageous to be able to have individual channels
with non integral channel separations (e.g. 0.5 MHz steps with 4 MHz
channels). [This make be a problem due to nature of polyphase filter
bank design]

If the DBE could be configured as a spectrometer with 16-32k channels
across 500 MHz or multiple 16-64 MHz channels each with ~1024 spectral
channels if would be highly attractive [this probably exceeds the
processing potential of iBOB card].
Received on 2006-07-05 20:15:43