Doubt over accuracy test
IN HIS letter in the August issue, M Campbell refers to the consecutive application of sin, cos, tan, arctan, arccos, and arcsin - to get back to the original value - as a check for the accuracy of a calculator in handling trigonometric functions.
That artificial test is not a good one as it can give poor results with calculators which are satisfactory for handling complex trigonometric functions in real problems. Campbell quotes a half percent error on Casio calculators and 33 percent error on Sinclair calculators. Using a starting value of pi/4 (45 deg.) my 100-step programmable Novus calculator does not even complete the sequence as an intermediate result is outside the permitted range of the argument for the next function.
When that test is applied to the ZX-81 your correspondent states: "You will be amazed and aghast alternately at the results".
You may be amazed, because the ZX-81 has a much better calculating ability than most desk calculators, but you should not be aghast unless you overlook the argument and result ranges for which the functions are valid.
For example, sin pi/4, sin 3pi/4 and sin 9pi/4 all give the value 0.70710678 (SQR 2)/2) but arcsin 0.70710678 gives pi/4 (0.78539816) because the result of arcsin must be one value and is therefore limited to the range -pi/2 to pi/2.
If the sin cos sequence is started with 3pi/4, it will finish with pi/4.
I SHOULD be grateful if you would let me know if there is any software suitable for recording family trees designed for ZX-81 plus 16K RAM. I may be wrong but I think there may be more to such a program than a simple recorded list.
We have not seen any programs for family trees but perhaps readers might be able to help.
Program points cleared
I READ your magazine with interest and always try the programs in it. Only about four out of every five work; for instance, Tim Crossley's program for Snap contained several errors. First, I believe line 110 should read: 110 IF INT(C/2)=C/2 THEN LET B=18.
Also if Bar is printed after the word cherry, you get a new word - BARRRY. Inserting the line:
120 PRINT AT .O.0,B;A$; " " (3 spaces)
instead of line 120 deals with the problem.
One of the problems for any magazine is the general printing of the programs. I know that the quality of some printers is not 100 percent but I am sure bad lines could be printed again, in clear letters. Take for instance Tim Crossley's program again; I could not tell what line 110 read - the first + , if it is a plus, which I believe is "=", is very confusing.
Black Jack error code
IN PUBLISHING my letter regarding Black Jack in the June issue, an error crept into lines 20 and 50. They should read:
LET CA=INT (RND*13)+2
Adding 16K can be easy
I HAVE read that if a program has been SAVED by a computer with only 1K RAM and subsequently a 16K RAM has been added the 1K program could not be re-loaded without removing the memory expansion.
I have found, on my ZX-81 at least, that is not true. You simply load the program, run it once, 'break' it and run it again. It should work perfectly.
I have bought every issue of your magazine and I think it is great.
Plea for notes on listing
I LIKE your magazine very much; I find it very helpful. There is still a great deal which I find very baffling but that is probably my fault. I have tried some of your programs but the clarity is not very good.
In a program there must be notes on what the lines do, so if there are variables which can be changed to alter speed and numbers, I can do so. It helps if I could be told whether to make them smaller or larger; I find that an easier way of learning programing than wading through handbooks and their useless examples.
I SOMETIMES have problems obtaining Sinclair User. Could you inform me where I can buy it?
The best way is to take an annual subscription. Alternatively you can order it from your newsagent, quoting our ISSN number, 0262-5458.