Miscellaneous - AGAIN AY. An article about the correction of sounding music. coprocessor.

<b>Miscellaneous</b> - AGAIN AY. An article about the correction of sounding music. coprocessor.

More about AY3-8910/12


(C) 1999, (R) soft
__________________________________________


   I do not know where to put the editor in chief
this article, "Scrap" or "Miscellaneous"
gist of the article remains the same. Problem
"Falsehood" musical co-processor on shovels Spectrum. 
Unfortunately, as an avid zhelezyachnik, I have this problem 
with no soldering could not solve. Sometimes the problem is the 
incompatibility of software is not decided perepatchivaniem 
megabytes of code, but simply by replacing the offending gland. 
That's what I did ... 


                  *


   I read the third issue of "ZX-Power", and
I was literally shocked the two articles. That -
"Fake editors" by JAM and unfinished article "Trouble with AY" 
by TAN. Immediately I say that in both articles, I noticed some 
inaccuracies (let me not be offended authors ;-). Article JAM'a 
stated that frequency of the clock in the company "Spectrum

2 + "related to the frequency of the processor as
fAY = fCPU / 2, which is not correct. TAN
correctly observed that it is related to the frequency of color 
subcarrier PALa 4.4336190 MHz. If it is divided by 2.5, we 
obtain 1.773448 MHz, which are plant on a sound coprocessor. 
Here again, little inaccuracy. The thing is that the frequency 
of 4.4336190 MHz obtained by dividing into two

standard frequency PALovskogo quartz
8.867238 MHz. From this comes not a divider ratio of TANa, you 
have to share frequency of the quartz into five, rather than 
the color subcarrier frequency to 2.5. 

  In general, I have the idea to modify the
a AY such that it fully conforms to corporate clock frequency. 
First, I verify that the Whether the specified frequency (in 
the original Spectrum I certainly did not look really ;-)

used in the "Spectrum 2 +? Had
There were several unusual way. For
background I have a formula for a frequency
notes on the AY (note = FtaktAY/16/delitel), and
also several programs that work with AY.
As we know, each player ringtones for AY is plaque frequency 
dividers for each note within seven or eight

octaves - it all depends on the specific type
player. To test I took
five types of programs (more precisely, even four
;-). This is music. Editor ProTracker 3.1,
music player for the AY of the game "Batman" well-known firm 
Ocean, mus. demo "FXSOUND" by Frantishek Fuka and music. Player 
of the game "Monty Free". In general, all four programs on the 
principle of playing melody is not significantly different. The 
most important thing - I needed a table of divisors. In muses.

Editor ProTracker those tables turned out to be two (from the 
old SoundTrackera and native). They are switchable to the 
editor (sorry, that is not downloadable !!!). And that's what 
happened. The player "Monty Free" full table was not, and notes 
were obtained by dividing the basic twelve subgroups most low 
(first) octave or two, until the desired octave. 

   I figured this is the most flagrant conversion because the 
frequency of notes do not appear to integer hertz, and to two 
decimal places and by dividing the clock frequency does not 
receive the AY the desired frequency with sufficient accuracy.

While in the program "Monty Free" is used
quite the correct method of forming subgroups of music, but not 
in this case. Ideally, as is done in the EMR (Musical 
instrument) - synthesizers? It is worth twelve calibrated 
generators frequency. Then each frequency is divided by

two, thus obtaining all the required octaves and frequency. In 
this case, specifically with AY exact frequency of the notes 
did not get in each case, there will be some

error after rounding the coefficients. division. Here's an 
example as I was checking the correctness of a divider for the 
notes. It turned so that out of five received the tables of 
divisors of two fully matched - SoundTracker'a and Batman'a 
(for convenience I will call them so), others differed somewhat 
from each other. First, I "grabbed" for Player Frantishek'a 
because was his a longtime fan and trust him 100%.

My hopes were justified. I took the first coefficient from the 
table and substituting in the formula with a frequency of 
1.773448 MHz. Ie 1773448/16/4031 = 27.497023, or obtained

almost note la (A) zero (lower) octave
(Ideally 27.50Gts). This is the maximum rate from the table 
notes (Kmax = 4095). Next I printed out a table and wrote a 
note next to each subgroup, which corresponds to it. I took the 
plate frequencies of notes from zero until the seventh octave 
and began testing the division factor, but now I've shared 
this: 1773448/16/261.63 (note to the fourth octave) = 
423.653633. (That is no longer divided a factor, and the 
frequency of the notes.) In rounded to 424, the number of 
consistent divider in a table. Next: 1773448/16 /

/ 587.33 (note re the fifth octave) = 188.7192
or 189 - is correct. In general, I'm sick of it (conversion on 
a calculator) and I have written in haste on the program from 
BASIC, which posed the coefficients divider

every note and mark rate, if
it could be rounded or more or in
lower side (ie the decimal value
is 5). Oddly enough, but a couple of factors of 84 notes per 
unit do not coincide with my settlement. But in general the 
error was very small fraction (on hearing about

;-) Not distinguishable). In Frantishek Fuka table starts with 
the notes an octave la zero. In Batman, Protracker and Monty 
Free - with notes To zero-sharp octave. The table ProTracker'a 
I got a hitch. In general there is the division factor obtained

for a clock frequency of the order of 1,447 MHz AY
(???), But do count, I never could.
For me, it remained the only mystery.

   Theory - the theory that calculates - not
calculates and the result is the same - the problem
counterfeit notes. I took up the scheme. That's what
I got:



   Scheme was simple and inexpensive
for details. Generator, assembly, and DD1.1
DD1.2 generates a frequency of a quartz resonator. Trimmer set 
the exact frequency of oscillation. If C1 is not set, then the 
scheme will be and so to work. But quartz is shunted resistors 
R1 and R2. Obtained underestimated by several tens of kilohertz

frequency. Therefore it is better to place C1. DD2 counter 
operates as a frequency divider to five. The scheme needs some 
adjustment - To establish the exact frequency oscillation 
frequency ZQ1, rotating engine C1 (Controlled by either a 14 or 
an 8 foot meter). If not the frequency, the capacitor C1 put in 
the 12-22 pF (Ideal quartz still do not). I

turned out to rebuild the frequency up
-10 Hz. But, of course, such accuracy is
not particularly necessary. All the same, the error rate notes 
will be rounding up a divider integer value. Printed circuit 
board generator size 20x30 mm mounted on top of board

audio coprocessor. After testing the
muses. Editor (of course, with a table of SoundTracker'a) and 
try to podygryvaniya favorite compositions were seen quite 
exact match actual notes. Well, naturally! Theory crossed with 
the practice ie as it should be. Finally I was able to

normally play up to tracks "Two Winters Only" group "My Dying 
Bride". Generally, those who are engaged in music, such a 
guileless modifications will allow more productive use of its 
Spectrum for simple arrangements. I recall that some five years 
ago I and a friend tried to SoundTrackerom to make arrangement 
of the song with a guitar. In editor had written a simple 
rhythm box and something like a flute. Now, Flute He does not 
fit into the guitar. I could not understand why. Understand much

later, even more accurately, I am someone suggested.
As it turned out, all because of the mismatch
frequency. Now that would be the track to check. ;-) Well, 
maybe I'm not wasted time to address this, to put it mildly,

Gluck?


  Okay! It's all good. And if not all
have the opportunity to climb inside of his
Spectrum with a soldering iron? I see here two
out of the situation.

1) As has been rightly said JAM (I subscribe!) - To write new 
music. Editors normal table of divisors under 1.75

MHz. Although, in my opinion, it is easier fixed a
case in the editor (eg ProTracker'e). It is better to be able to
podgruzki plates with the disc. As said
Rst7: "Drivers - rulez! ;-).

2) Purchase of corporate "Spectrum 2 +" ;-)
This, of course, a joke. The first option is more
fits.


   Well, if you do not want koryachit new and
old software (music demos, toys, etc.) -
Gather the above scheme and enjoy the "brand" does not sound 
fake. 


   Special thanks to Chief
editor of the "ZX-Power" for publishing an article.
The main thing is that the idea has not disappeared. ;-) And 
the theme seems to be not yet dried out. With interest I will 
answer all the questions and objections to the following 
specified address. 



       Sincerely, Vladimir Bakum

              aka (R) soft

         2:461 / 1024.6 (7/11/1999)
__________________________________________