< PROBLEMS ENCOUNTERED (contd) = < < = CONTENTS = > > = END >

Conclusion

8.1. Other Ways of Speeding Things Up

Timing Chart

Fig 23 - Pie Chart of Scan Time
Pork Pie

Stepping the Motors

This takes 55 ms for a movement of 100 mm. The speed is restricted by the specifications of the motors.

Thinking Time

This is 5 ms and has already been reduced from 40 ms. It could only be reduced significantly more by getting a faster computer, but this could only speed it up by a maximum of 5 ms.

Lowering and Raising the LVDT

This uses compressed air to lower and a return spring to raise. Lowering takes 50 ms and raising takes 55 ms. Increasing the pressure on both of these to speed things up is likely to damage the sample.

Bedding In

This takes 300 ms. A lot more investigations would have to be performed to find out the cause of the delay in the moving down the last 20 mm of travel. It is unlikely that the time spent investigating this effect would ever pay for itself in terms of increased speed alone

Confirming Steady Reading

This takes 125 ms for the PC to take 100 readings to confirm that the reading from the LVDT has stabilised. This is dependant on the speed of the PC so if the number of readings taken was reduced, this could result in less accurate readings if the PC was changed for a faster one at a later stage.

8.2. Further Improvements

Displaying Roughness Visually

To add a routine to the program to enable a cross section, or even a series of cross sections, on the screen would have been fairly easy and would have looked really impressive. However to show the data in a useful format so that it could be rotated about any axis etc. to check effects such as the successful marrying up of scans using the merge facility, or to view the effect of changing from a blunt to a sharp probe, was a completely different matter. Unless a package can be found that can do this with little or no modification, then adding a routine to visually display the roughness is a project in itself.

Purge Routine

The feeling at Nuclear Electric was that this routine should be expanded to check the points on the periphery of the scan as well as those that have all 4 neighbours, and also, if possible, give the facility to automatically re-check those points that are inconsistent. Both these modifications are possible but as the root cause of the problem that the purge routine was created to fix has gone, it seems unlikely that the purge routine will be needed in future so such modifications would be a waste of time.

8.3. Overall Evaluation

The software meets all the new requirements for the program and has fixed all the existing problems such as the parallelogram effect and probe sticking effect. Many more improvements have been added because of the availability of additional input lines such as checking the limit switches and the remote keypad. The scan is also faster as a result of using a variable speed and faster ADC conversion rate. The DMA (direct memory access) file structure has resulted in a far more flexible merge routine as well as many by-products such as Point Edit. Small refinements such as 'Estimated Time to Finish' have been added where ever I have seen a need.

The result is a product vastly superior to the old BBC system. This was readily acknowledged by the users of the systen when it was recently demonstrated to them.

< PROBLEMS ENCOUNTERED (contd) = < < = CONTENTS = > > = END >