Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.naic.edu/~astro/general_info/correlator/correlat.htm
Дата изменения: Mon Oct 2 23:18:02 2006
Дата индексирования: Sat Dec 22 20:29:14 2007
Кодировка:
User's Manual - Arecibo autocorrelators

Arecibo Spectral-Line Autocorrelators


The Observatory has two spectrometers: the interim 8192 lag correlator, and a new 64 K lag correlator. Both are described here.

The current autocorrelator is dubbed everywhere through NAIC documentation as the "interim correlator" as it uses the first, trial batch of Canaris-NAIC chips. These restrict its operation to a 50 MHz maximum bandwidth. A general listing of its capacities follows. It has proved to be an exceptionally reliable and versatile spectral-line correlator, so much so that we have been slow to upgrade it to use the full speed of the Canaris-NAIC chip.With the arrival of the X-band receiver, however, there is a new demand for 100 MHz bandwidths, so the interim correlator is being superseded by a more powerful machine that makes full use of the capacities built into the mature Canaris-NAIC chip. The capacities of this second-generation machine are showcased in the second half of this document.


1.    The Interim 50-MHz Correlator

 


Maximum available Bandpass with this Correlator is 50-MHz

It has the following filters:--

     analog filter    50 MHz
     digital filters   25, 12.5, 6.25, 3.125, 1.563, 0.781, 0.391, 0.195 MHz

 

Table 1: Configurations from existing 4 digital filter boards

 

  
| Config | MaxBW |sbc/Pol|Boards |lags/subcorrelator |
| |per sbc | | Used |Resln (km/s @ 1420 MHz)|
---------------------------------------------------------------------
| 9-level | 25 MHz | 2 | 4 | 2048/(2.6 km/s) |
| 9-level | 25 MHz | 4 | 4 | 1024/(5.2 km/s) |
|3-level interleaved| 50 MHz | 2 | 4 | 4096/(2.6 km/s) |
|3-level interleaved| 50 MHz | 4 | 4 | 2048/(5.2 km/s) |
| 3-level | 25 MHz | 4 | 4 | 2048/(2.6 km/s) |
| 3-level Stokes | 25 MHz |4 (tot)| 4 | 2048/(2.6 km/s) |
---------------------------------------------------------------------

Notes :

  1. Double Nyquist sampling can be used with all configurations except interleaved, but decreases the maximum bandwidth by a factor of two.
  2. 3-level, double Nyquist, 12.5-MHz bw and below will give 4 sub-bands with better resolution than the corresponding 9-level configuration.

     

9-level operation achieves 96.8% of the signal-to-noise of analog correlation, whereas 3-level achieves 81%.

One set of filters connects to each board. There are 8 canaris chips per board in the 50-MHz correlator, each board having a maximum of 4 sub-correlators (or sbc). This permits the interleaved modes of operation, and the obtaining of full Stokes parameters, as well as 9-level operation.

Differing correlator boards can operate at different bandwidths from one another, and (if needed) with differing blanking parameters; orthogonal polarizations are usually processed on the same board by default, though the polarization processed by each board is a user-settable parameter. The correlator can be configured by using the correlator widget in the gui interface to the observing system, and/or by editing a correlator file directly offline.

Each correlator board contains an on-board computer (a 68040 microprocessor), which makes each board an individual, independent correlator. A master cpu treats these as devices to be read in parallel. This allows the possibility of real-time interference excision. Note, however, that the interim correlator can never be used as a pulsar machine. The minimum dump time for spectral-line usage is about 1 Hz.

Notes on usage. The redshifted HI and OH users often dump spectra every second, to facilitate the excision of rfi offline, with minimal data loss. When the redshift and velocity width of a galaxy are well known, it is possible to observe it with up to four different resolutions simultaneously, with one resolution being obtained from each board. Both the bandwidth, and the number of lags to be stored, are adjustable parameters, so the user can retain 32 lags for instance instead of 1024 if that is what is wanted, or choose to use just one (or 2, or 3) boards of the correlator rather than all four.

The data generated by the interim correlator is in an Arecibo format, which can be read into our home-grown reduction system, ANALYZ, or into IDL. Routines exist for reading the data into AIPS++, or translating it into CLASS format.


Correlator Home Page



2.    The 100 MHz Spectrometer (WAPP)


Bill Sisk's design for the 16-chip, 100 MHz correlator board is the fundamental module around which the WAPP pulsar backend was created. This board also supports all of the existing bandwidths in use with the interim correlator. It uses digital filters for those smaller than 50 MHz, and analog filters for the 50 & 100 MHz bandwidths (a 200 MHz capacity is under discussion). It does not support a double Nyquist mode. This machine provides spectra with up to four times the resolution of the interim correlator.

WAPP stands for Wide-band Arecibo Pulsar Processor. We have four WAPP boards, each of which can be treated as an independent, 16 K lag, 100 MHz bandwidth autocorrelator. The four boards together constitute the new WAPP-based correlator. These support both spectral-line and pulsar observing modes, and are configured via the revamped GUI observing interface, CIMA. The WAPP board uses the 100 MHz Canaris-NAIC chip, and is interfaced to a PC through a 16-bit bus and a commercial (EDT) PCI card.

The capacities of each WAPP board are:--

I.    Bandwidth

       195 kHz to 100 MHz in factor-of-two steps

II.   Lag configurations (Spectral Line)

A.   195 kHz to 50 MHz BW

   1 channel, 3-level auto, 16384 lags
2 channel, 3-level auto, 8192 lags/chan
1 channel, 9-level auto, 4096 lags
2 channel, 9-level auto, 2048 lags/chan
3-level Stokes, 4096 lags/product
9-level Stokes, 1024 lags/product

B.   100 MHz BW

   1 channel, 3-level auto,  8192 lags
2 channel, 3-level auto, 4096 lags/chan
1 channel, 9-level auto, 2048 lags
3-level Stokes, 2048 lags/product

Thus the Maximum number of lags per output product (or channel) =

16*1024
-----------------------------------------------
{log_3(N_level)}^2 * N_interleave * N_output_products]


where:-

N_level is 3 or 9 [i.e. log_3(N_level) is 1 or 2 respectively];
N_interleave is 1 (for BW <= 50 MHz) or 2 for 100 MHz BW;
N_output_products is 1 or 2 (channel autocorrs) or 4 (Stokes);
& log_3() indicates log to the base 3.
The configurations of the 64 K spectrometer at its largest bamdwidths are:--

--------------------------------------------------------------------
|  Config           |       Total sbc x #Lags (Resolution)
|
--------------------------------------------------------------------
| | For 50-MHz Bandwidth | For 100-MHz Bandwidth |
--------------------------------------------------------------------
| | Next-Generation Correlator |
| 3-level | 4 x 16k (3 kHz) | 4 x 8k (12 kHz) |
| 3-level | 8 x 8k (6 kHz) | 8 x 4k (24 kHz) |
--------------------------------------------------------------------
| 9-level | 4 x 4k (12 kHz) | 4 x 2k (49 kHz) |
| 9-level | 8 x 2k (24 kHz) | -- |
--------------------------------------------------------------------

(Footnote: To expand on the terse Configuration heading: "Total sbc" means the available number of sub-correlators for standard dual-polarization operation from our current one pixel system, so 8 x 2k is to be interpreted as meaning both polarizations for four frequency chunks: this CANNOT be reconfigured to give eight chunks of a single polarization.)

Currently (March 2003) the WAPP correlator writes its output as CLASS FITS. This will shortly be superseded (for most users) by output written in Single Dish FITS, with routines in IDL for processing.

Blanking of the WAPP is different from the interim correlator. In the interim correlator, when blanking is used, the integration time is preserved. In the WAPP, the integration time is not preserved. Thus in the WAPP, you always complete the full set of sources, however the integration times are reduced by the blanking.


III.   Pulsar Modes

50 MBytes/s max for correlator board >50 MBytes/s max for EDT/computer 38 MByte/s max data rate to disk

The best resolution for 100 MHz bands, using 16 bits/lag, and summing the polarizations is given by the formulas,

i = integration time in microsecs
#lags = (i - 1)* 6.25
Data rate to disk = 12 MBytes/s
Resolution = 100 MHz/#lags
    Integration time       Max lags    Resolution
100 microsec 616 162 kHz
64 microsec 392 255 kHz
32 microsec 192 520 kHz

IV.   Direct Sampling

    (a) Two channels
(b) 195 kHz-50 MHz bandwidth, in octave steps
(c) data rate must be less than disk data rate, 35 MByte/s
(d) To reach the disk data rate, burst sampling can be implemented if
requested

Two channels, continuous sampling

Bandwidth(MHz)  Data Rate Mbyte/s     Bit Packing
50 25 1
25 25 2
12.5 25 4
6.25 12.5 4
3.125 6.25 4
1.56 3.12 4
0.78 1.56 4
0.390 0.78 4
0.195 0.39 4

The interim correlator will continue to be available for at least a year in parallel with the 64 K correlator. It is anticipated that in due course the interim correlator will be redeployed for other purposes (perhaps as a radar decoder, perhaps as an rfi monitor, perhaps ??).

 


Murray Lewis
last updated 26 February 2003