Ok they are! You're right!
241 values in my table T are repeated values.

Anyway this is non very important as the line

[m,i]= min(abs(T(:,1)-x(n)));

looks for the best index i in T that gives x(n)

Thus in theory nothing change even if I skip the step where
I cancel duplicates.

% cancel duplicates

[h,i] = sort([ QQ , -1, 1e6]);
w = h(find(diff(h)>1e-6));
w = [ w(2:end)];

% store in T non duplicate outputs with the corresponding PSG volumes

T = [];
for i=1:length(w)
j=find(II(:,1)==w(i));
T = [T; II(j(1),: )];
end

% T has 608 different values !
% Sorry :-)

Quote:

I did all that in the article on the MAP that I updated yesterday.

So... Wich of the tables is better? I must say, that I can't understand the SNR system that well...

I think at least the one on the MAP is sub-optimal , but not by much.

In effect, they will likely both sound similar. Although if ARTRAG takes a base range of 1.45, his output will probably be louder.

I’m still working on using the real error, instead of counting the number of times the deviation passes a threshold. If both my and ARTRAGs calculations are correct, we should end up with the same values.

Btw, the MAP one is sub-optimal in the sense that it might not use the most optimal range (that is, the one with the least amount of errors). The values in the table are all correct though.


~Grauw

Ok, next add keyclick bit and start again

Unfortunately I don't think keyclick volume is standard.

I don’t think so either...

Quote:

@@Artrag: do you want to try my table ? As far as I can understand it should give an SNR almost identical to the one of a 8bit PCM player.

I tried the table and and I don't find differences, only replay speed.
It would be better if we make testing with the same 8bit-mono-unsigned WAV file.(20kb max.)

webs.ono.com/WYZ/8BITart.bin

@Grauw
My samples play louder, but also should also have a lower distortion
As the PSG levels are spaced not uniformely the range you choose for
play the output becomes very important.
AR

@WYZ
Do you want a sample?

Yes, because I'm testing the table with my own 8 bit sample.
But I don't understand this:

db 0,0,0 ; 0 0,0000
...

why 00h 8bit data means 0,0,0 volume in PSG instead of 80h?

Send me an email I'll send you my sample

db 0,0,0 ; 0 0,0000

means that the 8bit PCM value 0 corresponds to the
output level 0,0000 which in turn corresponds to 0,0,0 on the 3 channles

ATRAG: I know.

I found the problem why calculating errors didn’t work well, I’ve been declaring my ‘errors’ variable globally instead of locally .

From the updated routine, it seems that a output volume range of 0 ... 1.328 has actually got the best signal-to-noise ratio (read: the least distortion).

Below is a list with cumulative errors for each range, with increments of 0.01 (I’ve actually been testing with 0.001 increments). Divide an error by 256 to find the average error per value.

I think that as you can see from the table, the distortion is pretty much irrelevant, because as you can see in the generated list it stays within acceptable limits for pretty much all volume ranges between 0.5 and 2.0 (out of 3). Which is nice, because then you can base your range on the actual output volume you want.


~Grauw

Total error: 44,23519 (range: 0,01)
Total error: 16,84636 (range: 0,02)
Total error: 9,64568 (range: 0,03)
Total error: 6,28239 (range: 0,04)
Total error: 4,42793 (range: 0,05)
Total error: 3,39897 (range: 0,06)
Total error: 2,68690 (range: 0,07)
Total error: 2,26966 (range: 0,08)
Total error: 1,91377 (range: 0,09)
Total error: 1,65222 (range: 0,1)
Total error: 1,45297 (range: 0,11)
Total error: 1,29424 (range: 0,12)
Total error: 1,17503 (range: 0,13)
Total error: 1,07904 (range: 0,14)
Total error: 1,00098 (range: 0,15)
Total error: 0,93451 (range: 0,16)
Total error: 0,86804 (range: 0,17)
Total error: 0,82759 (range: 0,18)
Total error: 0,79926 (range: 0,19)
Total error: 0,75884 (range: 0,2)
Total error: 0,70973 (range: 0,21)
Total error: 0,67699 (range: 0,22)
Total error: 0,65442 (range: 0,23)
Total error: 0,63096 (range: 0,24)
Total error: 0,60689 (range: 0,25)
Total error: 0,59186 (range: 0,26)
Total error: 0,57259 (range: 0,27)
Total error: 0,55390 (range: 0,28)
Total error: 0,53223 (range: 0,29)
Total error: 0,52242 (range: 0,3)
Total error: 0,51395 (range: 0,31)
Total error: 0,49644 (range: 0,32)
Total error: 0,48656 (range: 0,33)
Total error: 0,47593 (range: 0,34)
Total error: 0,46670 (range: 0,35)
Total error: 0,46656 (range: 0,36)
Total error: 0,45193 (range: 0,37)
Total error: 0,44630 (range: 0,38)
Total error: 0,43068 (range: 0,39)
Total error: 0,40155 (range: 0,4)
Total error: 0,40912 (range: 0,41)
Total error: 0,39539 (range: 0,42)
Total error: 0,39180 (range: 0,43)
Total error: 0,38540 (range: 0,44)
Total error: 0,37862 (range: 0,45)
Total error: 0,36894 (range: 0,46)
Total error: 0,36619 (range: 0,47)
Total error: 0,37171 (range: 0,48)
Total error: 0,37226 (range: 0,49)
Total error: 0,36292 (range: 0,5)
Total error: 0,35934 (range: 0,51)
Total error: 0,35224 (range: 0,52)
Total error: 0,35325 (range: 0,53)
Total error: 0,33895 (range: 0,54)
Total error: 0,34738 (range: 0,55)
Total error: 0,33155 (range: 0,56)
Total error: 0,31900 (range: 0,57)
Total error: 0,31556 (range: 0,58)
Total error: 0,31418 (range: 0,59)
Total error: 0,31948 (range: 0,6)
Total error: 0,31131 (range: 0,61)
Total error: 0,31412 (range: 0,62)
Total error: 0,30191 (range: 0,63)
Total error: 0,30897 (range: 0,64)
Total error: 0,29906 (range: 0,65)
Total error: 0,31641 (range: 0,66)
Total error: 0,29768 (range: 0,67)
Total error: 0,30547 (range: 0,68)
Total error: 0,30165 (range: 0,69)
Total error: 0,29931 (range: 0,7)
Total error: 0,30129 (range: 0,71)
Total error: 0,30367 (range: 0,72)
Total error: 0,28418 (range: 0,73)
Total error: 0,29206 (range: 0,74)
Total error: 0,28506 (range: 0,75)
Total error: 0,28157 (range: 0,76)
Total error: 0,27310 (range: 0,77)
Total error: 0,27442 (range: 0,78)
Total error: 0,27032 (range: 0,79)
Total error: 0,26229 (range: 0,8)
Total error: 0,26094 (range: 0,81)
Total error: 0,25829 (range: 0,82)
Total error: 0,26112 (range: 0,83)
Total error: 0,25177 (range: 0,84)
Total error: 0,24942 (range: 0,85)
Total error: 0,25672 (range: 0,86)
Total error: 0,25522 (range: 0,87)
Total error: 0,25829 (range: 0,88)
Total error: 0,24838 (range: 0,89)
Total error: 0,24611 (range: 0,9)
Total error: 0,25249 (range: 0,91)
Total error: 0,24920 (range: 0,92)
Total error: 0,26177 (range: 0,93)
Total error: 0,22135 (range: 0,94)
Total error: 0,24428 (range: 0,95)
Total error: 0,25329 (range: 0,96)
Total error: 0,25167 (range: 0,97)
Total error: 0,23510 (range: 0,98)
Total error: 0,24791 (range: 0,99)
Total error: 0,24387 (range: 1)
Total error: 0,25408 (range: 1,01)
Total error: 0,25244 (range: 1,02)
Total error: 0,26981 (range: 1,03)
Total error: 0,23293 (range: 1,04)
Total error: 0,23891 (range: 1,05)
Total error: 0,23978 (range: 1,06)
Total error: 0,24048 (range: 1,07)
Total error: 0,23775 (range: 1,08)
Total error: 0,23987 (range: 1,09)
Total error: 0,24111 (range: 1,1)
Total error: 0,22823 (range: 1,11)
Total error: 0,22688 (range: 1,12)
Total error: 0,22653 (range: 1,13)
Total error: 0,23181 (range: 1,14)
Total error: 0,21896 (range: 1,15)
Total error: 0,21020 (range: 1,16)
Total error: 0,22228 (range: 1,17)
Total error: 0,22460 (range: 1,18)
Total error: 0,22700 (range: 1,19)
Total error: 0,22034 (range: 1,2)
Total error: 0,21695 (range: 1,21)
Total error: 0,22254 (range: 1,22)
Total error: 0,22277 (range: 1,23)
Total error: 0,22397 (range: 1,24)
Total error: 0,21076 (range: 1,25)
Total error: 0,21311 (range: 1,26)
Total error: 0,21453 (range: 1,27)
Total error: 0,22634 (range: 1,28)
Total error: 0,21816 (range: 1,29)
Total error: 0,21343 (range: 1,3)
Total error: 0,20568 (range: 1,31)
Total error: 0,23326 (range: 1,32)
Total error: 0,21189 (range: 1,33)
Total error: 0,21937 (range: 1,34)
Total error: 0,22188 (range: 1,35)
Total error: 0,22015 (range: 1,36)
Total error: 0,21533 (range: 1,37)
Total error: 0,22161 (range: 1,38)
Total error: 0,21932 (range: 1,39)
Total error: 0,22349 (range: 1,4)
Total error: 0,22628 (range: 1,41)
Total error: 0,22279 (range: 1,42)
Total error: 0,22360 (range: 1,43)
Total error: 0,22220 (range: 1,44)
Total error: 0,22260 (range: 1,45)
Total error: 0,21568 (range: 1,46)
Total error: 0,21547 (range: 1,47)
Total error: 0,23134 (range: 1,48)
Total error: 0,23381 (range: 1,49)
Total error: 0,22591 (range: 1,5)
Total error: 0,23088 (range: 1,51)
Total error: 0,23085 (range: 1,52)
Total error: 0,22086 (range: 1,53)
Total error: 0,22554 (range: 1,54)
Total error: 0,22043 (range: 1,55)
Total error: 0,23093 (range: 1,56)
Total error: 0,22873 (range: 1,57)
Total error: 0,23862 (range: 1,58)
Total error: 0,24495 (range: 1,59)
Total error: 0,23777 (range: 1,6)
Total error: 0,22956 (range: 1,61)
Total error: 0,24087 (range: 1,62)
Total error: 0,23827 (range: 1,63)
Total error: 0,25010 (range: 1,64)
Total error: 0,26857 (range: 1,65)
Total error: 0,26796 (range: 1,66)
Total error: 0,26575 (range: 1,67)
Total error: 0,25778 (range: 1,68)
Total error: 0,27983 (range: 1,69)
Total error: 0,27124 (range: 1,7)
Total error: 0,27299 (range: 1,71)
Total error: 0,26943 (range: 1,72)
Total error: 0,27029 (range: 1,73)
Total error: 0,27326 (range: 1,74)
Total error: 0,26310 (range: 1,75)
Total error: 0,27293 (range: 1,76)
Total error: 0,26729 (range: 1,77)
Total error: 0,27023 (range: 1,78)
Total error: 0,27555 (range: 1,79)
Total error: 0,27297 (range: 1,8)
Total error: 0,29826 (range: 1,81)
Total error: 0,28601 (range: 1,82)
Total error: 0,28829 (range: 1,83)
Total error: 0,29094 (range: 1,84)
Total error: 0,29220 (range: 1,85)
Total error: 0,30633 (range: 1,86)
Total error: 0,33192 (range: 1,87)
Total error: 0,28438 (range: 1,88)
Total error: 0,30270 (range: 1,89)
Total error: 0,29994 (range: 1,9)
Total error: 0,29962 (range: 1,91)
Total error: 0,30532 (range: 1,92)
Total error: 0,31269 (range: 1,93)
Total error: 0,31540 (range: 1,94)
Total error: 0,32793 (range: 1,95)
Total error: 0,31110 (range: 1,96)
Total error: 0,32536 (range: 1,97)
Total error: 0,33506 (range: 1,98)
Total error: 0,33640 (range: 1,99)
Total error: 0,33353 (range: 2)
Total error: 0,34135 (range: 2,01)
Total error: 0,34238 (range: 2,02)
Total error: 0,33174 (range: 2,03)
Total error: 0,32812 (range: 2,04)
Total error: 0,32695 (range: 2,05)
Total error: 0,34908 (range: 2,06)
Total error: 0,32552 (range: 2,07)
Total error: 0,31970 (range: 2,08)
Total error: 0,32985 (range: 2,09)
Total error: 0,32639 (range: 2,1)
Total error: 0,33628 (range: 2,11)
Total error: 0,33341 (range: 2,12)
Total error: 0,33423 (range: 2,13)
Total error: 0,33698 (range: 2,14)
Total error: 0,35078 (range: 2,15)
Total error: 0,35550 (range: 2,16)
Total error: 0,35267 (range: 2,17)
Total error: 0,35136 (range: 2,18)
Total error: 0,36378 (range: 2,19)
Total error: 0,36829 (range: 2,2)
Total error: 0,35376 (range: 2,21)
Total error: 0,35989 (range: 2,22)
Total error: 0,36921 (range: 2,23)
Total error: 0,36399 (range: 2,24)
Total error: 0,36999 (range: 2,25)
Total error: 0,36514 (range: 2,26)
Total error: 0,37530 (range: 2,27)
Total error: 0,39190 (range: 2,28)
Total error: 0,40033 (range: 2,29)
Total error: 0,42321 (range: 2,3)
Total error: 0,43598 (range: 2,31)
Total error: 0,44684 (range: 2,32)
Total error: 0,46169 (range: 2,33)
Total error: 0,46422 (range: 2,34)
Total error: 0,46343 (range: 2,35)
Total error: 0,45436 (range: 2,36)
Total error: 0,46151 (range: 2,37)
Total error: 0,46923 (range: 2,38)
Total error: 0,47888 (range: 2,39)
Total error: 0,47170 (range: 2,4)
Total error: 0,47948 (range: 2,41)
Total error: 0,47544 (range: 2,42)
Total error: 0,48373 (range: 2,43)
Total error: 0,48164 (range: 2,44)
Total error: 0,50521 (range: 2,45)
Total error: 0,51061 (range: 2,46)
Total error: 0,51910 (range: 2,47)
Total error: 0,52917 (range: 2,48)
Total error: 0,51526 (range: 2,49)
Total error: 0,51017 (range: 2,5)
Total error: 0,52078 (range: 2,51)
Total error: 0,52193 (range: 2,52)
Total error: 0,52795 (range: 2,53)
Total error: 0,54087 (range: 2,54)
Total error: 0,55451 (range: 2,55)
Total error: 0,58646 (range: 2,56)
Total error: 0,58929 (range: 2,57)
Total error: 0,62763 (range: 2,58)
Total error: 0,65096 (range: 2,59)
Total error: 0,69321 (range: 2,6)
Total error: 0,71855 (range: 2,61)
Total error: 0,74487 (range: 2,62)
Total error: 0,77314 (range: 2,63)
Total error: 0,79779 (range: 2,64)
Total error: 0,80361 (range: 2,65)
Total error: 0,80585 (range: 2,66)
Total error: 0,81701 (range: 2,67)
Total error: 0,82162 (range: 2,68)
Total error: 0,82679 (range: 2,69)
Total error: 0,81629 (range: 2,7)
Total error: 0,81614 (range: 2,71)
Total error: 0,81397 (range: 2,72)
Total error: 0,81675 (range: 2,73)
Total error: 0,81551 (range: 2,74)
Total error: 0,82986 (range: 2,75)
Total error: 0,84290 (range: 2,76)
Total error: 0,85899 (range: 2,77)
Total error: 0,87378 (range: 2,78)
Total error: 0,89772 (range: 2,79)
Total error: 0,92300 (range: 2,8)
Total error: 0,95007 (range: 2,81)
Total error: 0,97228 (range: 2,82)
Total error: 1,00343 (range: 2,83)
Total error: 1,04830 (range: 2,84)
Total error: 1,08120 (range: 2,85)
Total error: 1,11709 (range: 2,86)
Total error: 1,15316 (range: 2,87)
Total error: 1,17638 (range: 2,88)
Total error: 1,20349 (range: 2,89)
Total error: 1,22223 (range: 2,9)
Total error: 1,26128 (range: 2,91)
Total error: 1,25397 (range: 2,92)
Total error: 1,26898 (range: 2,93)
Total error: 1,27263 (range: 2,94)
Total error: 1,28741 (range: 2,95)
Total error: 1,29091 (range: 2,96)
Total error: 1,28453 (range: 2,97)
Total error: 1,28541 (range: 2,98)
Total error: 1,28617 (range: 2,99)
Total error: 1,26241 (range: 3)

(lower = better)

Actually your best result is

Total error: 0,20568 (range: 1,31)

and not 1.328

I'm doing the same research I'll tell you my result very soon
AR

PS
my result says that 1.272 maximize the SNR for the linear patter
of 256 evenly spaced samples

Quote:

db 0,0,0 ; 0 0,0000 means that the 8bit PCM value 0 corresponds to the output level 0,0000 which in turn corresponds to 0,0,0 on the 3 channles

This is a test with your 8bit sample and table:
http://webs.ono.com/WYZ/aleluya8BIT.bin

...And why don't test this table?

WAV - PSG

00h - F,E,E
...
80h - 0,0,0
...
FFh - F,E,E