Micro­power

at the end of the Read routine. This means that the Read routine can be called from any program, even if it uses special tables for input/​output routines.

The routine then scans the keyboard and the serial input to find either the four consecutive FF’s which mark the start of each block of tape data, or the four ‘escapes’ from the keyboard which will abort the Read routine. When the block start is found the next four bytes, which indicate the loading address, the block length and the block number, are loaded into HL and DE. They are then displayed on the screen by routine 6C, which simultaneously adds the four bytes together, returning with the value in the C register. The next byte received from the tape is compared with this checksum in C, and if the values are not identical a qestion mark is printed on the screen and subsequent data is ignored until the next ‘start of block’ marker is found.

If the checksum for the block header is correct Nas-Sys 3 checks to see if an argument has been entered with the Read command. If it has, this argument is added to the loading address, so that the data can be loaded to a different address from that specified in the write command. In Nas-Sys 1 this facility is not available.

The data is now read from the tape; if the command letter entered is R (i.e., we are doing a Read) the data is loaded into memory; if not (for example, if we are using the Verify command) data is not stored. In either case the bytes are summed into register C as they are received from tape. When a number of bytes equal to the block length specified in the header has been received, the total in C is compared with the next byte – the data checksum. If the two are not identical, a question mark is printed; however, the faulty data bytes or bytes which caused the checksum error have been loaded into memory.

If the checksum test is passed a full stop is printed and the routine then checks to see if the block just loaded was the last block. If not, the routine looks for the next ‘start of block’ marker. When the last block is detected the input/​output addresses are reset and the routine terminates by jumping to the subroutine which flips the tape LED.

The Read routine has two faults. Firstly, it loads faulty data; if you are trying to load a tape which is producing a lot of read errors you cannot load a program correctly by reading the tape repeatedly, even if you have several copies of the same program on the tape, because blocks which have loaded correctly are corrupted by faulty reads in subsequent passes. Of course, you can overcome this by copying the program to a different location and then recopying blocks which initially read incorrectly as error-free reads are obtained. However, this is a fiddly task, and in any case you can’t use this method if the program is longer than half the available memory. A second fault is that blocks can be missed entirely without an error message being produced if one of the FF’s in the ‘start of block’ marker is misread.

In order to overcome these faults I use a slightly different Read routine, which puts the tape data into a buffer and only transfers it to the correct memory location if the checksum tests are passed. But where can you locate the