[alma-config] topography mask

[John Conway]

Hi,

 The suggestion below from Simon sounds sensible to me, with 
possibly the additional cost of the transporters  factored in.

{ Although I think we should also have a file with just the gradients
on ASAP so we can start work on going from
pure geometry to reality ]

The obvious unit of the cost function for pads 
is the $$dollars$$ to contstruct a pad at a given position, 
including the cost of cutting into the gradient, constructing 
foundations given the subsurface geology etc.

However presumably  the unit costs of the transporters will scale
mainly as a  (high power) function of the maximum gradient 
they will be able to deal  with. So a few pads in difficult places
_can_ in fact be a significant cost driver.

The road cost is a complex issue, the total length of roads
in most designs doesn't vary too much, even for spiral designs,
if one takes the shortest routes to pads (see memo 260, figure6c).
Presumably for high gradient sites the the cost per unit length will be
dominated by the last part of the  road leading to the pad, and can be 
included into the pad cost. More detailed road models can be added later. 

Perhaps we could have cost masks produced for three different 
limiting  maximum gradients, with values of 0 for points 
above the maximum gradient and the dollar cost elsewhere. We can 
get an estimate from someone of the cost of a transporter as a 
function of gradient. 

Alternatively just a pad cost distribution over the site  and a gradient
distribution would be a good starting point. 

As I said at the beginning we need the 
former as soon as possible(!). Calculating the cost function is 
difficult but the gradients I assume easy, and with that we can start 
making the arrays more realistic.

 cheers
    John



 

On Wed, 12 Jan 2000, Simon Radford wrote:

> 
> Friends,
> 
> How should we incorporate topography constraints in the configuration
> design process? 
> 
> To date, we have discussed using a binary "go"-"no go" mask computed
> from the digital elevation models. Although simple in principle, this
> will be hard in practice. No single parameter characterizes the ease or
> difficulty of putting an antenna station at a given location. Rather,
> several different characteristics must be considered. These include, but
> are probably not limited to, the overall and local surface gradients,
> the local surface roughness, the subsurface geology and hydrology, and
> road access. Given sufficient money, most of these features (or
> impediments) can be mitigated or circumvented. But how should the cost
> threshold be defined? Is it uniform across the site? Is it uniform for
> all configurations? A binary mask must encode all of these factors.
> 
> Instead of a binary mask, therefore, I suggest we develop and use a
> continuous, "cost" mask, where each pixel is assigned a relative
> construction cost, which avoids the difficulty of defining a cost
> threshold. Then the configuration design task is to minimize this cost
> consistent with other goals, such as imaging, etc. This approach has the
> flexibility to accommodate a few "high cost" stations if they are
> balanced by other "low cost" stations and improve the overall
> configuration quality.
> 
> How to calculate the cost mask is not, unfortunately, immediately
> obvious. But it would appear a tractable problem that might tie in with
> Mark Gordon's work. 
> 
> I welcome your reactions.
> 
> Simon
>