Rosie Bolton, Cambridge University 1SKA System level trade-offsSKADS Final meeting, November 2009

System Design choices for the SKA: System Design choices for the SKA: Cost and performance trade-offs.Cost and performance trade-offs.

Rosie Bolton, Paul Alexander, Tim Colegate, Andy Faulkner, Dominic Ford, Peter Hall and

the SKADS System Group.

Rosie Bolton, Cambridge University 2SKA System level trade-offsSKADS Final meeting, November 2009

OutlineOutline

• The SKA Cost/Performance tool

• Populating the tool: Hierarchical telescope design

• Example of design block structure

• Examples of some trade-offs

Rosie Bolton, Cambridge University 3SKA System level trade-offsSKADS Final meeting, November 2009

The SKA Costing toolThe SKA Costing tool

• Concept of “Design Blocks” was introduced in SKADS for D&C 1: Hierarchical Telescope designs made up of discrete blocks and components.

• SKACost tool was developed by the ISPO (Peter Hall, Aaron Chippendale and Tim Colegate)

• Initial SKADS costing based on Excel spreadsheet

• The new SKA costing tool combines the best of these approaches– Advanced cost modelling and uncertainty handling (Monte

Carlo)

– Plug-and-play telescope designs

– GUI for easy browsing and editing

Rosie Bolton, Cambridge University 4SKA System level trade-offsSKADS Final meeting, November 2009

Overview of the toolOverview of the tool

Graphical User Interface

Command-line Interface

Python Interface

Socket-based Interface

Costing/Performance Tool Engine

Telescope Design Data

Design blocks are stored in portable XML format.This could be extended into a full database including other information about components

The engine is written in Python

GUI is the easiest way for non-experts to interact with telescope designs and explore cost trade-offs

Rosie Bolton, Cambridge University 5SKA System level trade-offsSKADS Final meeting, November 2009

Hierarchical designs..Hierarchical designs..

Rosie Bolton, Cambridge University 6SKA System level trade-offsSKADS Final meeting, November 2009

Hierarchical designs..Hierarchical designs..

Rosie Bolton, Cambridge University 7SKA System level trade-offsSKADS Final meeting, November 2009

Hierarchical designsHierarchical designs……

Labour: cost per hour

Tarpaulin: cost per unit area

Wooden poles: cost per unit length

Air con unit: cost per item

Rosie Bolton, Cambridge University 8SKA System level trade-offsSKADS Final meeting, November 2009

Example: parameterised Example: parameterised designs.designs.

• Designs are parameterised – many inputs such as collecting area, number of stations, dish diameter etc.

• Tool supports “Parameter surveys” where successive telescopes can be built by varying one input parameter. This allows cost scaling to be investigated.

Rosie Bolton, Cambridge University 9SKA System level trade-offsSKADS Final meeting, November 2009

Processing boxes in AProcessing boxes in AAAlo stationlo station

Rosie Bolton, Cambridge University 10SKA System level trade-offsSKADS Final meeting, November 2009

Processing boxes in Processing boxes in AAloAAlo station station

• AA lo Station: integer number of processing boards per processing box. More boxes mean shorter analogue cables from antennas.

• “Wrong” number of boxes means inefficient use of processors, so more cost.

• Boxes themselves cost money.

• Example is for a 90m radius station with 1.5m spaced antennas.

Rosie Bolton, Cambridge University 11SKA System level trade-offsSKADS Final meeting, November 2009

Data link cost vs link length Data link cost vs link length (16Tbit/s link)(16Tbit/s link)

• Large data rate link costs from tool show the combine effect of distance break points for different technologies.

• These break have strong implications for cost savings if we change the layout of the Aperture Arrays

Change from short range to

mid range lasers

Introduction of first pre-

amplifiers

Change from short range to

mid range lasers

Introduction of first amplifiers

(80km)

Introduction second

amplifiers (160km)

Rosie Bolton, Cambridge University 12SKA System level trade-offsSKADS Final meeting, November 2009

Impact of changing the distributionImpact of changing the distribution

• Does this matter? Yes. Look at the estimated costs for 250 AA station links, each with 16 Tbit/s. Vary Bmid – distance within which 95% of all stations are placed, cost implications of the order 100 Million EUR.

• 95% within 10km, very few stations out to 180km

• 95% within 100km, remainder out to 180km

Rosie Bolton, Cambridge University 13SKA System level trade-offsSKADS Final meeting, November 2009

Impact of changing the distributionImpact of changing the distribution

• Does this matter? Yes. Look at the estimated costs for 250 AA station links, each with 16 Tbit/s. Vary Bmid – distance within which 95% of all stations are placed, cost implications of the order 100 Million EUR.

• 95% within 10km, very few stations out to 180km: data links total estimate 60M Euros

• 95% within 100km, remainder out to 180km: data links total estimate 140M Euros

Rosie Bolton, Cambridge University 14SKA System level trade-offsSKADS Final meeting, November 2009

SummarySummary• The SKA cost/performance tool developed in SKADS and major

input from an international team – flexible format designs, extensible python code.

• Telescope designs can be investigated by unplugging one design block and replacing it with another.

• Parameterised designs allow the impact of varying any of the inputs to be assessed – with automated output.

• Currently the tool provides a useful starting point for visualising the overall system and making trade-offs.

• Future work must incorporate some more book-keeping e.g. basis of estimates for cost, improved power information.

• The information in the tool will fit naturally into a database, which will be essential as the SKA project moves forward.