Tracing the Orders

The tracing of the paths of the orders across the data frame is often a source of difficulty as it is fairly easy for blemishes in the frame to fatally deflect order tracing algorithms from the actual path of the order. ECHOMOP provides a variety of options to help combat these problems.

ECHOMOP order tracing first locates the positions of the orders at the centre of the frame, and estimates the average order slope.

It uses this information to predict the existence of any partial orders at the top/bottom of the frame which may have been missed by the examination of the central columns during order location. Tracing then proceeds outwards from the centre of each order.

At each step outwards a variable size sampling box is used to gather a set of averages for the rows near the expected order centre. The centre of this data is then evaluated by one of the following methods:

Gaussian

Attempts to fit a Gaussian profile Works well for bright object frames.

Centroid

Calculates the centroid of the data. Most generally applicable method.

Edge

Detects the upper and lower 'edge's and interpolates. Is less accurate but works well for difficult flat fields (e.g. saturated).

Balance

Calculates the centre of gravity Works well for difficult data when G and C methods are having problems.

Retrace

Uses a previous trace as a template the trace whenever it cannot be centred. Will normally be used in conjunction with automatic trace consistency checking to improve poorly traced orders.

User

The last resort allows the user to of points close to the order centre. Then functions as for Retrace by using a polynomial through the supplied points to predict order position whenever it is lost.

The trace algorithm will loop increasing its sampling box size when it fails to find an explicit centre. The sample box can increase up to a size governed by the measured average order separation.

When a set of centres have been obtained for an order, a polynomial is fitted to their coordinates. The polymonial degree being is selectable. For ideal data, these polynomials will represent an accurate reflection of the path of the order across the frame.

For real data it is usually helpful to refine these polynomials by clipping the most deviant points, and re-fitting. Options are provided to do this automatically or manually.

When dealing with distorted data it is often necessary to use a high degree polynomial to accurately fit the order traces.

This in turn can lead to problems at the edges of the frame when the order is often faint. Typically the polynomial will 'run away' from the required path. The simplest solution is, of course, to re-fit with a lower order polynomial, however, this may not be satisfactory if the high degree is necessary to obtain a good fit over the rest of the order.

In these circumstances, and others where one or more orders polynomials have 'run away', ECHOMOP provides an automatic consistency checker. The consistency checking task works by fitting polynomials to order-number/Y-coordinate at a selection of positions across the frame.

The predicted order centres from both sets of polynomials are then compared with each other and then mean and sigma differences calculated. The 'worst' order is then corrected by re-calculating its trace polynomial using the remaining orders (but excluding its own contribution). This process is repeated until the mean deviation between the polynomials falls below a tunable threshold value.

The consistency checker will also cope with the 'bad' polynomials which can result when partial orders have been automatically fitted.

The order traces may also be viewed overlaid on the trace frame if DISPLAY=YES is used.