Nov 1, 2011 RN - 1 Jet Propulsion Laboratory California Institute of Technology Implementation...

Nov 1, 2011 RN - 1

Jet Propulsion LaboratoryCalifornia Institute of Technology

Implementation Issues and Choices for VLBI data Acquisition System in DSN

Robert Navarro

Stephan Rogstad

Joseph Trinh

Chris Jacobs

© 2011 California Instituteof Technology. Governmentsponsorship acknowledged.

9th US VLBI Technical Meeting

November 1-2, 2011

Nov 1, 2011 RN - 2


Key features of DSN VLBI Processing (DVP) System

• An IF Switch will handle up to 12 IF inputs from DSN antennas.

• Two IF inputs, each covering up to 500 MHz of bandwidth.

• Accepts DSN IF input band of 100-600 MHz. Good for S band (2.3 GHz), X band (8.4 GHz) and Ka band (31.2 GHz)

• Uses JPL IF sampler module, CASPER ROACH board for Digital Processing and Channelization.

• Interfaces to JPL Deep Space Network monitor & control infrastructure.

• Records up to 32 upper/lower or 16 complex channels. Channel max BW is 32 MHz

• Mark5C disk packs used for data storage

• VEX files used for input

• Data stored on disk in VDIF format

DSN VLBI Processor (DVP)

IF SWITCH

IF_IN1

IF_IN12

IF1

IF2

ROACH

(CHANNELIZE

AND FORMAT)

MARK5-CRECORDER

JPL IF Digitizer

(1280 MHz sample clock)

DataDisk pak

10GigE

DVP Control Computer

(DVP M/C)(Vex I/F)

CC&S

ROACHTRANSISTION

MODULE

(Fiber to copper)

Fib

er O

pti

c

LV

DS

DSN Station Network Monitor/

Control

Freq & Timing100MHz

1PPS

Nov 1, 2011 RN - 3


VEX files on the DVP

• Vex files contain configuration information needed to run the DVP and point the DSN antennas

• VEX files are delivered to the DVP manually, or through another DSN subsystem

• DVP Script building software (vbuild.py)

– A Python script on the which executes a VLBI pass on the DVP

– Automatically triggered by delivered VEX files

– Searched for relevant blocks in VEX files and parses

– Produces a DVP command script and a DSN antenna pointing file

• VEX file blocks used

– $EXPER – Getting the experiment name

– $STATION – Configure which IF switch antenna input to use

– $SOURCE – Antenna pointing and scan labeling

– $FREQ – Configuring the channel frequency, bandwidth and bits

– $IF – Determining the RF to IF downconversion oscillator value

– $PHASE_CAL_DETECT – to determine whether phase-cal will be present for monitoring

– $SCHED – for determining when to start and stop scan recordings and when/where to point the antenna

Nov 1, 2011 RN - 4


DVP Script Builder Outputs

• DVP Script File

– A set of commands, including some time stamps

– Executed on the DVP using the SCRPT command

– Each command is a separate script or executable that runs in a BASH shell and performs a specific function (ie: CHAN for channel config.)

• Antenna Pointing File

– Sent to a DSN Network Monitor and Control (NMC) computer through a shared file system and used to automatically point the antenna

– Points the antenna to RA, DEC of a source at the specified time

– Format of each line:

<ascii_time_stamp> <RA (hr)> <DEC (hr)> <RA (deg)> <DEC (deg)> <source name>

Nov 1, 2011 RN - 5


DVP Channelization & Sub-band filtering

• Wideband 1280 MHz real to seven channel 160 MHz wide channel filterbank completed. Channels centered at 80, 160, 240, 320, 400, 480 and 560 MHz.

– FPGA (XC5VSX95T) Usage: 77% Slices; 70% DSP48s, 57% BRAM

• 32 upper/lower or 16 complex sub-channels can be formed

– Max real channel BW= 16 MHz; Minimum BW = 500 Hz

– Max complex channel BW = 32 MHz; Minimum BW = 1000 Hz

• Sub-band filtering used two stage filtering.

– 1st stage is CIC filter that decimates from 2 to 10,000

– 2nd stage is Fir filter that decimates from 1 to 16

– Fir filter has 16n taps, where n = decimation factor.

– Sub-band channels are complex. Changed to upper/lower representation using Hilbert transforms

• Processing organization means upper/lower channels always occur in contiguous pairs.

Nov 1, 2011 RN - 6


Wideband Overlapping Filter Channels (7)

64 FIR taps per channel

Nov 1, 2011 RN - 7


Sub-band Filter performance

• Filter performance mainly affected by FIR filter coefficients in second stage filtering. But for FIR decimation of 1, CIC effects are noticeable.

• Aim is to get 0.1 db ripple in passband and at least 40 db attenuation in stopband.

Wide ChannelSelect(7-1

MUX)

Complex Mixer

High Resolution

NCO

CascadeIntegrate

Comb Filter

FIRFilter

IQ

cos sin

I

Q

Sub Channel Processing32 MHz to 1 KHz(16 Instances)

CascadeIntegrate

Comb Filter

FIRFilter

Decimate By

1 to 10000

Decimate By

1 to 16

Wide Channel

Filterbank(1280 MHz to Seven 160 MHz channels)

-+

++

U

L

A/D Samples1280 MHz

8 bits

Nov 1, 2011 RN - 8


Sub-Channel Filter Performance

Filter response measured in hardware by sweeping sinusoid through band

• Complex, upper and lower responses plotted

• For these sub-channels, first stage CIC filter dec=5 and FIR filter dec=1

• Fir filter compensates for CIC droop.

• Bumps outside passband are from first stage CIC filter

• Looking at ways to improve response for this case by using CIC dec = 2, FIR filter dec = 5/2.

Nov 1, 2011 RN - 9


Sub-Channel Filter Performance

Filter response measured in hardware by sweeping sinusoid through band

• Complex, upper and lower responses plotted

• For these sub-channels, first stage CIC filter dec=5 and FIR filter dec=2

• Fir filter compensates for CIC droop.

• Bumps outside passband are from first stage CIC filter

• Complex rejection in stopband better than 60 db.

• Upper/lower filters have spike in frequencies opposite transition region.

• These spikes still attenuated 20 db and are in transition region.

Nov 1, 2011 RN - 10


SubChannel Output

• Actual data from tone buried in noise shown below. Sub-channel is 16 MHz upper / 16 MHz lower.

• Cutoffs of Upper and Lower bands apparent at edges.

Nov 1, 2011 RN - 11


VDIF usage

• Would like to use VDIF, not Mark5B emulation mode as main mode for DSN VLBI recorders.

• Would like to support both upper/lower and complex channels.

• Single Channel Data Threads are more flexible, but more difficult to implement in firmware. Buffer space one key issue.

• Multi-Channel Data Threads easier to implement and take less FPGA memory, but limit all channels to same data rate, bit representation.

• If single channel data threads used, then a hybrid data thread with an upper/lower channel would be more convenient . It would allow packet size to be same as complex channel without requiring twice the memory.

• What Payload Frame size? Some where between 2000 and 8000 bytes?

Nov 1, 2011 RN - 12


More VDIF UsageMulti-Channel Frames & Frame Rates

Complex BW Bits/sample Samples/Frame Frame Size Data Frames/Sec

32 MHz 1 1250 5000 Bytes 25600

32 MHz 2 625 5000 Bytes 51200

8 MHz 1 1250 5000 Bytes 6400

8 MHz 2 625 5000 Bytes 12800

2 MHz 1 1250 5000 Bytes 1600

1 MHz 2 625 5000 Bytes 3200

Possible Frame sizes and rates for 16 complex or 32 upper/lower channels

Nov 1, 2011 RN - 13


More VDIF UsageSingle Channel Frames & Frame Rates

Complex Bandwidth

Bits/sample Data Format

Samples/Frame Frame Size Data Frames/Sec

32 MHz 1 Real 16000 2000 Bytes 200032 MHz 2 Real 8000 2000 Bytes 400032 MHz 1 Complex 8000 2000 Bytes 400032 MHz 2 Complex 4000 2000 Bytes 80002 MHz 1 Real 16000 2000 Bytes 1252 MHz 2 Real 8000 2000 Bytes 2502 MHz 1 Complex 8000 2000 Bytes 2502 MHz 2 Complex 4000 2000 Bytes 500

Possible Frame sizes and rates for 1 single channel or 1 upper & 1lower channel

Nov 1, 2011 RN - 1 Jet Propulsion Laboratory California Institute of Technology Implementation...

Documents

Transcript of Nov 1, 2011 RN - 1 Jet Propulsion Laboratory California Institute of Technology Implementation...