hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
how much time does the helix mp3 decoder take to decode for one frame on sam7s
Started by ●May 17, 2006
Reply by ●May 17, 20062006-05-17
If you look at the values you gave for the lpc it is not possible to do
it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at
least 30Mhz. So 27.5Mhz is not enough. You also want to do other things
with your AT91 :-)
if you are only doing this as a hobby project I think so because you
want to take a mobile phone lcd you could spend the 1$ more for the 256
and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or
vs1011 from vlsi. With the vs1003 you could also do midi and wma files.
Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to
decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
_____
> Terms of Service.
_____
it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at
least 30Mhz. So 27.5Mhz is not enough. You also want to do other things
with your AT91 :-)
if you are only doing this as a hobby project I think so because you
want to take a mobile phone lcd you could spend the 1$ more for the 256
and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or
vs1011 from vlsi. With the vs1003 you could also do midi and wma files.
Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to
decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
_____
> Terms of Service.
_____
Reply by ●May 18, 20062006-05-18
hi Andreas
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
v\:* {behavior:url(#default#VML);} o\:* {behavior:url(#default#VML);} w\:* {behavior:url(#default#VML);} .shape {behavior:url(#default#VML);} If you look at the values you gave for the lpc it is not possible to do it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at least 30Mhz. So 27.5Mhz is not enough. You also want to do other things with your AT91 J
if you are only doing this as a hobby project I think so because you want to take a mobile phone lcd you could spend the 1$ more for the 256 and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or vs1011 from vlsi. With the vs1003 you could also do midi and wma files. Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
---------------------------------
YAHOO! GROUPS LINKS
---------------------------------
---------------------------------
Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
v\:* {behavior:url(#default#VML);} o\:* {behavior:url(#default#VML);} w\:* {behavior:url(#default#VML);} .shape {behavior:url(#default#VML);} If you look at the values you gave for the lpc it is not possible to do it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at least 30Mhz. So 27.5Mhz is not enough. You also want to do other things with your AT91 J
if you are only doing this as a hobby project I think so because you want to take a mobile phone lcd you could spend the 1$ more for the 256 and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or vs1011 from vlsi. With the vs1003 you could also do midi and wma files. Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
---------------------------------
YAHOO! GROUPS LINKS
---------------------------------
---------------------------------
Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.
Reply by ●May 18, 20062006-05-18
Sorry I havent implemented the helix code for a s256.
Trying it out would be the way to see if it works.
Pay attention to the DMA. It works only on 4 Byte boundaries. For
example:
If you want to transmit data through the SPI you have to use 2K instead
of 512Bytes for a sector buffer in fat. So there is a loss of 75% of
RAM.
Using the SSC for I2S is limited to 16bit per channel through the
possible configuration of the SSC. (There is an application note from
Atmel).
If you run the code from RAM you could use the table from the helix
homepage to see how much time it needs because there is no vendor
specific memory controller involved.
For 48Khz 320KBit its 30Mhz. 58% of CPU usage for the AT91SAM7S. And you
will need about 59K of ram. So there is 5K RAM left for your program.
Not very much.
I am also interested in implementing the helix code in an AT91 but
havent got the time to try it out.
Trying out would be the best thing to see how fast it is and how you
could mix arm and thumb code. Thumb code has theoretically single cycle
access because the memory controller of the atmel is fetching two
instructions from flash each time in thumb mode. With lots of jumps this
is not a big advantage but it has to be tried out.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von nader chinichian
Gesendet: Donnerstag, 18. Mai 2006 08:51
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take
to decode for one frame on sam7s
hi Andreas
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more
than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in
s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one
frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
If you look at the values you gave for the lpc it is not possible to do
it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at
least 30Mhz. So 27.5Mhz is not enough. You also want to do other things
with your AT91 :-)
if you are only doing this as a hobby project I think so because you
want to take a mobile phone lcd you could spend the 1$ more for the 256
and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or
vs1011 from vlsi. With the vs1003 you could also do midi and wma files.
Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to
decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
_____
Yahoo!
/evt9666/*http:/messenger.yahoo.com> Messenger with Voice.
PC-to-Phone calls for ridiculously low rates.
_____
> Terms of Service.
_____
Trying it out would be the way to see if it works.
Pay attention to the DMA. It works only on 4 Byte boundaries. For
example:
If you want to transmit data through the SPI you have to use 2K instead
of 512Bytes for a sector buffer in fat. So there is a loss of 75% of
RAM.
Using the SSC for I2S is limited to 16bit per channel through the
possible configuration of the SSC. (There is an application note from
Atmel).
If you run the code from RAM you could use the table from the helix
homepage to see how much time it needs because there is no vendor
specific memory controller involved.
For 48Khz 320KBit its 30Mhz. 58% of CPU usage for the AT91SAM7S. And you
will need about 59K of ram. So there is 5K RAM left for your program.
Not very much.
I am also interested in implementing the helix code in an AT91 but
havent got the time to try it out.
Trying out would be the best thing to see how fast it is and how you
could mix arm and thumb code. Thumb code has theoretically single cycle
access because the memory controller of the atmel is fetching two
instructions from flash each time in thumb mode. With lots of jumps this
is not a big advantage but it has to be tried out.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von nader chinichian
Gesendet: Donnerstag, 18. Mai 2006 08:51
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take
to decode for one frame on sam7s
hi Andreas
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more
than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in
s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one
frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
If you look at the values you gave for the lpc it is not possible to do
it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at
least 30Mhz. So 27.5Mhz is not enough. You also want to do other things
with your AT91 :-)
if you are only doing this as a hobby project I think so because you
want to take a mobile phone lcd you could spend the 1$ more for the 256
and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or
vs1011 from vlsi. With the vs1003 you could also do midi and wma files.
Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to
decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
_____
Yahoo!
/evt9666/*http:/messenger.yahoo.com> Messenger with Voice.
PC-to-Phone calls for ridiculously low rates.
_____
> Terms of Service.
_____
Reply by ●May 18, 20062006-05-18
hi,
thanks a lot for helping me.
you write that i need about 59 k of ram but
in helix site it write:
13446 ROM + 23816 RAM + 2*4608 (out put buffer) = 46.5 kbyte
so why do i need for the (59 - 46.5) byte that left?
about the tesing:
i buy a at91sam7s64 board from olimx so i can't test it on it.(low ram)
i also order LPC2148 board from olimex only for 74 $.(good price)
the at91sam7x256 board is too high 250$ so i must sure is it usefull for this
project to buy it or use lpc2148 for it.
please if you test it on an at91 tell me the reasult for one frame decode.
with this test can compaire LPC and SAM7.(no bench mark and no good information about speed)
if you see any code that full implement mmmc+mp3 decoder+DAC(i2c) send for me.
i find two examples that did it with lpc2148.that one only decode one
frame and get the time for it(simple test)
secend only read from flash some frame and play it.(not real time)
i didn't see any note about limiting on DMA in datasheet.
where did you find this note about DMA?
thanks.
Andreas Siebel wrote:
v\:* {behavior:url(#default#VML);} o\:* {behavior:url(#default#VML);} w\:* {behavior:url(#default#VML);} .shape {behavior:url(#default#VML);}
Sorry I havent implemented the helix code for a s256.
Trying it out would be the way to see if it works.
Pay attention to the DMA. It works only on 4 Byte boundaries. For example:
If you want to transmit data through the SPI you have to use 2K instead of 512Bytes for a sector buffer in fat. So there is a loss of 75% of RAM.
Using the SSC for I2S is limited to 16bit per channel through the possible configuration of the SSC. (There is an application note from Atmel).
If you run the code from RAM you could use the table from the helix homepage to see how much time it needs because there is no vendor specific memory controller involved.
For 48Khz 320KBit its 30Mhz. 58% of CPU usage for the AT91SAM7S. And you will need about 59K of ram. So there is 5K RAM left for your program. Not very much.
I am also interested in implementing the helix code in an AT91 but havent got the time to try it out.
Trying out would be the best thing to see how fast it is and how you could mix arm and thumb code. Thumb code has theoretically single cycle access because the memory controller of the atmel is fetching two instructions from flash each time in thumb mode. With lots of jumps this is not a big advantage but it has to be tried out.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von nader chinichian
Gesendet: Donnerstag, 18. Mai 2006 08:51
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi Andreas
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
If you look at the values you gave for the lpc it is not possible to do it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at least 30Mhz. So 27.5Mhz is not enough. You also want to do other things with your AT91 J
if you are only doing this as a hobby project I think so because you want to take a mobile phone lcd you could spend the 1$ more for the 256 and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or vs1011 from vlsi. With the vs1003 you could also do midi and wma files. Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
---------------------------------
Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.
SPONSORED LINKS
Microcontrollers Cpu chip Arms Development tool Microprocessor
---------------------------------
YAHOO! GROUPS LINKS
---------------------------------
---------------------------------
Blab-away for as little as 1/min. Make PC-to-Phone Calls using Yahoo! Messenger with Voice.
thanks a lot for helping me.
you write that i need about 59 k of ram but
in helix site it write:
13446 ROM + 23816 RAM + 2*4608 (out put buffer) = 46.5 kbyte
so why do i need for the (59 - 46.5) byte that left?
about the tesing:
i buy a at91sam7s64 board from olimx so i can't test it on it.(low ram)
i also order LPC2148 board from olimex only for 74 $.(good price)
the at91sam7x256 board is too high 250$ so i must sure is it usefull for this
project to buy it or use lpc2148 for it.
please if you test it on an at91 tell me the reasult for one frame decode.
with this test can compaire LPC and SAM7.(no bench mark and no good information about speed)
if you see any code that full implement mmmc+mp3 decoder+DAC(i2c) send for me.
i find two examples that did it with lpc2148.that one only decode one
frame and get the time for it(simple test)
secend only read from flash some frame and play it.(not real time)
i didn't see any note about limiting on DMA in datasheet.
where did you find this note about DMA?
thanks.
Andreas Siebel wrote:
v\:* {behavior:url(#default#VML);} o\:* {behavior:url(#default#VML);} w\:* {behavior:url(#default#VML);} .shape {behavior:url(#default#VML);}
Sorry I havent implemented the helix code for a s256.
Trying it out would be the way to see if it works.
Pay attention to the DMA. It works only on 4 Byte boundaries. For example:
If you want to transmit data through the SPI you have to use 2K instead of 512Bytes for a sector buffer in fat. So there is a loss of 75% of RAM.
Using the SSC for I2S is limited to 16bit per channel through the possible configuration of the SSC. (There is an application note from Atmel).
If you run the code from RAM you could use the table from the helix homepage to see how much time it needs because there is no vendor specific memory controller involved.
For 48Khz 320KBit its 30Mhz. 58% of CPU usage for the AT91SAM7S. And you will need about 59K of ram. So there is 5K RAM left for your program. Not very much.
I am also interested in implementing the helix code in an AT91 but havent got the time to try it out.
Trying out would be the best thing to see how fast it is and how you could mix arm and thumb code. Thumb code has theoretically single cycle access because the memory controller of the atmel is fetching two instructions from flash each time in thumb mode. With lots of jumps this is not a big advantage but it has to be tried out.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von nader chinichian
Gesendet: Donnerstag, 18. Mai 2006 08:51
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi Andreas
thanks for reply.
if you look at the mp3 decoder code there is a lot jump in it.
so lpc flash has 20 Mhz at max speed to read flash in single cycle.
but the sam7 have 30 mhz.so for jump code the sam7 is better.
but for other code the lpc have much faster.
for the table the lpc only have access 20 mhz but sam7 is about 27 mhz.
the helix is combine of all of them.
so only a test can make sure that sam7 can or can't.
another thing about DMA for SSP for DAC and SPI for MMC.
lpc doesn't have them in DMA mode so for those it get cpu usage more than SAM7.
is it true?
i think i read a massege from you in another forum that you run it in s256.
if it's true could you please send your code for me.
beacuse i'm new in ARM i think i have a lot problem to test it.
if i run the code from RAM how much time does it take for decode one frame?
(44.1 khz)
thanks for your help
Andreas Siebel wrote:
If you look at the values you gave for the lpc it is not possible to do it out of the flash.
26ms is maximum. The lpc needs 13ms at 60MHz. That means you need at least 30Mhz. So 27.5Mhz is not enough. You also want to do other things with your AT91 J
if you are only doing this as a hobby project I think so because you want to take a mobile phone lcd you could spend the 1$ more for the 256 and you could run the helix code out of ram.
You could also use a s64 with an external mp3 decoder like the vs1003 or vs1011 from vlsi. With the vs1003 you could also do midi and wma files. Acc is also supported I think.
Ok its a little bit more expensive but it sounds great.
To ways to go and I think both are equal in expenses.
S256+codec+helix
Or
S64+vs10XX
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag von naderus2000
Gesendet: Mittwoch, 17. Mai 2006 19:51
An: A...
Betreff: [AT91SAM] how much time does the helix mp3 decoder take to decode for one frame on sam7s
hi,
i want to know how much time does the helix mp3 decoder take to decode
for
only one frame for 44.1 khz sample rate.(1152 samples) does any one
test it?
i have to fill the 1152 buffer with the raw data in about less than 26
ms for
real time playing.
lpc2138 do it for about 13 ms @ 60 MHz from flash.
i'm confusing to choice s128 or s256 to buy.
if the code can run enough fast from flash(55/2'.5 MHZ in ARM mode) so
s128 enough for mp3 player(lower price than s256)but if the speed
not enough for decoding code for real time so i must buy s256 that run
code
from it's internal flash.i also need some ram for using nokia 6610 lcd
for
my mp3 player.
any help can help me.
thanks a lot.
---------------------------------
Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.
SPONSORED LINKS
Microcontrollers Cpu chip Arms Development tool Microprocessor
---------------------------------
YAHOO! GROUPS LINKS
---------------------------------
---------------------------------
Blab-away for as little as 1/min. Make PC-to-Phone Calls using Yahoo! Messenger with Voice.
Reply by ●May 30, 20062006-05-30
Hi,
I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
with the Helix software. It's true that the SAM7 is quite a bit slower
than the LPC2000 when code is executed from Flash, but that doesn't
mean that you have to put the complete decoder into RAM. It is
sufficient to run only a few critical functions like the imdct and the
polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
for one MP3 or AAC frame, more than enough to play MP3 with high
bitrates (the highest I tried was 195kbs VBR).
The code doesn't take up much space in RAM, but you have to take into
account that you need at least two output buffers (I'm using 3) and
enough input buffer to make sure you always get a complete frame (2kB
works fine). In total you need about 50kB. That leaves plenty of space
for all the other stuff (file system, display etc.).
What Andreas Siebel writes about DMA is not correct. DMA is very
straightforward to use on the SAM7, there are no big obstacles. I'm
using DMA for the SPI transfer from the SD card and for the SSC
transfer to the DAC.
There is some information on my website, unfortunately at the moment
it's only in German:
http://www.mikrocontroller.net/articles/ARM-MP3-Player
Andreas
I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
with the Helix software. It's true that the SAM7 is quite a bit slower
than the LPC2000 when code is executed from Flash, but that doesn't
mean that you have to put the complete decoder into RAM. It is
sufficient to run only a few critical functions like the imdct and the
polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
for one MP3 or AAC frame, more than enough to play MP3 with high
bitrates (the highest I tried was 195kbs VBR).
The code doesn't take up much space in RAM, but you have to take into
account that you need at least two output buffers (I'm using 3) and
enough input buffer to make sure you always get a complete frame (2kB
works fine). In total you need about 50kB. That leaves plenty of space
for all the other stuff (file system, display etc.).
What Andreas Siebel writes about DMA is not correct. DMA is very
straightforward to use on the SAM7, there are no big obstacles. I'm
using DMA for the SPI transfer from the SD card and for the SSC
transfer to the DAC.
There is some information on my website, unfortunately at the moment
it's only in German:
http://www.mikrocontroller.net/articles/ARM-MP3-Player
Andreas
Reply by ●May 30, 20062006-05-30
Hi,
could you please tell me how to use the DMA without wasting so much RAM
if you want to do 8bit SPI DMA transfers.
I could get it only working if I put the bytes I want to send on in a
long array which is aligned to a 4 Byte boundary.
Like
unsigned char u8buf[10];
unsigned long u32buf[10];
for(int I = 0; I < 10; I++){
U32buf[I] = u8buf[I];
}
/* now sending */
Andreas
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von Andreas Schwarz
Gesendet: Sonntag, 28. Mai 2006 23:59
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take
to decode for on
Hi,
I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
with the Helix software. It's true that the SAM7 is quite a bit slower
than the LPC2000 when code is executed from Flash, but that doesn't
mean that you have to put the complete decoder into RAM. It is
sufficient to run only a few critical functions like the imdct and the
polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
for one MP3 or AAC frame, more than enough to play MP3 with high
bitrates (the highest I tried was 195kbs VBR).
The code doesn't take up much space in RAM, but you have to take into
account that you need at least two output buffers (I'm using 3) and
enough input buffer to make sure you always get a complete frame (2kB
works fine). In total you need about 50kB. That leaves plenty of space
for all the other stuff (file system, display etc.).
What Andreas Siebel writes about DMA is not correct. DMA is very
straightforward to use on the SAM7, there are no big obstacles. I'm
using DMA for the SPI transfer from the SD card and for the SSC
transfer to the DAC.
There is some information on my website, unfortunately at the moment
it's only in German:
http://www.mikrocontroller.net/articles/ARM-MP3-Player
Andreas
SPONSORED LINKS
Microcontrollers
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=JL1WKF1vUqEgZhFSvLcTwQ>
Cpu
+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=J2LJ4J
ioN5vEM0428aODAg> chip
Arms
p&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=yPOxLkzDXn
5eqYeA1wMVCQ>
Development
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=lhbODg5ptVr7u61IFQfZXQ> tool
Microprocessor
w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sigCHIRUROYao456Telb69zdA>
_____
> Terms of Service.
_____
could you please tell me how to use the DMA without wasting so much RAM
if you want to do 8bit SPI DMA transfers.
I could get it only working if I put the bytes I want to send on in a
long array which is aligned to a 4 Byte boundary.
Like
unsigned char u8buf[10];
unsigned long u32buf[10];
for(int I = 0; I < 10; I++){
U32buf[I] = u8buf[I];
}
/* now sending */
Andreas
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von Andreas Schwarz
Gesendet: Sonntag, 28. Mai 2006 23:59
An: A...
Betreff: Re: AW: [AT91SAM] how much time does the helix mp3 decoder take
to decode for on
Hi,
I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
with the Helix software. It's true that the SAM7 is quite a bit slower
than the LPC2000 when code is executed from Flash, but that doesn't
mean that you have to put the complete decoder into RAM. It is
sufficient to run only a few critical functions like the imdct and the
polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
for one MP3 or AAC frame, more than enough to play MP3 with high
bitrates (the highest I tried was 195kbs VBR).
The code doesn't take up much space in RAM, but you have to take into
account that you need at least two output buffers (I'm using 3) and
enough input buffer to make sure you always get a complete frame (2kB
works fine). In total you need about 50kB. That leaves plenty of space
for all the other stuff (file system, display etc.).
What Andreas Siebel writes about DMA is not correct. DMA is very
straightforward to use on the SAM7, there are no big obstacles. I'm
using DMA for the SPI transfer from the SD card and for the SSC
transfer to the DAC.
There is some information on my website, unfortunately at the moment
it's only in German:
http://www.mikrocontroller.net/articles/ARM-MP3-Player
Andreas
SPONSORED LINKS
Microcontrollers
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=JL1WKF1vUqEgZhFSvLcTwQ>
Cpu
+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=J2LJ4J
ioN5vEM0428aODAg> chip
Arms
p&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=yPOxLkzDXn
5eqYeA1wMVCQ>
Development
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=lhbODg5ptVr7u61IFQfZXQ> tool
Microprocessor
w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sigCHIRUROYao456Telb69zdA>
_____
> Terms of Service.
_____
Reply by ●May 30, 20062006-05-30
Hi.
I've implemented helix for AT91SAM7S256. It works fine.
I did simple tests and I can say that this CPU needs 46 MHz for decoding
mp3 file (44,1 KHz, stereo 320 Kbs) and 42 MHz for 128 Kbs.
Quite lot of power - LPC2138 needs 38 MHz. BUT S256 can run at 105MHz!!!
Yes it works fine and stable. Unfortunately it doesn't want to run at
110 and more :(
So at 105 MHz it decodes mp3 files more then 2 times faster then real
time decoding.
It uses about 28 KB of RAM including output buffer and excluding input
buffer.
upanie
Andreas Schwarz napisał(a):
> Hi,
>
> I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
> with the Helix software. It's true that the SAM7 is quite a bit slower
> than the LPC2000 when code is executed from Flash, but that doesn't
> mean that you have to put the complete decoder into RAM. It is
> sufficient to run only a few critical functions like the imdct and the
> polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
> for one MP3 or AAC frame, more than enough to play MP3 with high
> bitrates (the highest I tried was 195kbs VBR).
>
> The code doesn't take up much space in RAM, but you have to take into
> account that you need at least two output buffers (I'm using 3) and
> enough input buffer to make sure you always get a complete frame (2kB
> works fine). In total you need about 50kB. That leaves plenty of space
> for all the other stuff (file system, display etc.).
>
> What Andreas Siebel writes about DMA is not correct. DMA is very
> straightforward to use on the SAM7, there are no big obstacles. I'm
> using DMA for the SPI transfer from the SD card and for the SSC
> transfer to the DAC.
>
> There is some information on my website, unfortunately at the moment
> it's only in German:
> http://www.mikrocontroller.net/articles/ARM-MP3-Player
>
> Andreas
>
> SPONSORED LINKS
> Microcontrollers
>
> Cpu chip
>
> Arms
> Development tool
>
> Microprocessor
>
> >.
>
>
----------------------------------
Tysiace zdjec swietnych samochodow! >>> http://link.interia.pl/f1944
I've implemented helix for AT91SAM7S256. It works fine.
I did simple tests and I can say that this CPU needs 46 MHz for decoding
mp3 file (44,1 KHz, stereo 320 Kbs) and 42 MHz for 128 Kbs.
Quite lot of power - LPC2138 needs 38 MHz. BUT S256 can run at 105MHz!!!
Yes it works fine and stable. Unfortunately it doesn't want to run at
110 and more :(
So at 105 MHz it decodes mp3 files more then 2 times faster then real
time decoding.
It uses about 28 KB of RAM including output buffer and excluding input
buffer.
upanie
Andreas Schwarz napisał(a):
> Hi,
>
> I have successfully implemented MP3 and AAC playing on an AT91SAM7S256
> with the Helix software. It's true that the SAM7 is quite a bit slower
> than the LPC2000 when code is executed from Flash, but that doesn't
> mean that you have to put the complete decoder into RAM. It is
> sufficient to run only a few critical functions like the imdct and the
> polyphase filter from RAM. My SAM7 with 48 MHz needs about 16-21 ms
> for one MP3 or AAC frame, more than enough to play MP3 with high
> bitrates (the highest I tried was 195kbs VBR).
>
> The code doesn't take up much space in RAM, but you have to take into
> account that you need at least two output buffers (I'm using 3) and
> enough input buffer to make sure you always get a complete frame (2kB
> works fine). In total you need about 50kB. That leaves plenty of space
> for all the other stuff (file system, display etc.).
>
> What Andreas Siebel writes about DMA is not correct. DMA is very
> straightforward to use on the SAM7, there are no big obstacles. I'm
> using DMA for the SPI transfer from the SD card and for the SSC
> transfer to the DAC.
>
> There is some information on my website, unfortunately at the moment
> it's only in German:
> http://www.mikrocontroller.net/articles/ARM-MP3-Player
>
> Andreas
>
> SPONSORED LINKS
> Microcontrollers
>
> Cpu chip
>
> Arms
> Development tool
>
> Microprocessor
>
> >.
>
>
----------------------------------
Tysiace zdjec swietnych samochodow! >>> http://link.interia.pl/f1944
Reply by ●May 30, 20062006-05-30
--- In A..., "Andreas Siebel" wrote:
>
>
> Hi,
>
> could you please tell me how to use the DMA without wasting so much RAM
> if you want to do 8bit SPI DMA transfers.
> I could get it only working if I put the bytes I want to send on in a
> "long" array which is aligned to a 4 Byte boundary.
The actual DMA transfer is quite easy to do:
*AT91C_SPI_RPR = receivebuffer;
*AT91C_SPI_RCR = nbytes;
*AT91C_SPI_TPR = transmitbuffer;
*AT91C_SPI_TCR = nbytes;
*AT91C_SPI_PTCR = AT91C_PDC_RXTEN;
*AT91C_SPI_PTCR = AT91C_PDC_TXTEN;
while(! (*AT91C_SPI_SR & AT91C_SPI_ENDRX));
*AT91C_SPI_PTCR = AT91C_PDC_RXTDIS;
*AT91C_SPI_PTCR = AT91C_PDC_TXTDIS;
The initialization:
pSPI->SPI_CR = AT91C_SPI_SPIEN | AT91C_SPI_SWRST;
pSPI->SPI_MR = AT91C_SPI_MSTR | AT91C_SPI_PS_FIXED | AT91C_SPI_MODFDIS;
pSPI->SPI_MR |= ( (SPI_MR_PCS<<16) & AT91C_SPI_PCS );
pSPI->SPI_CSR[SPI_CSR_NUM] = AT91C_SPI_CPOL | AT91C_SPI_BITS_8;
>
>
> Hi,
>
> could you please tell me how to use the DMA without wasting so much RAM
> if you want to do 8bit SPI DMA transfers.
> I could get it only working if I put the bytes I want to send on in a
> "long" array which is aligned to a 4 Byte boundary.
The actual DMA transfer is quite easy to do:
*AT91C_SPI_RPR = receivebuffer;
*AT91C_SPI_RCR = nbytes;
*AT91C_SPI_TPR = transmitbuffer;
*AT91C_SPI_TCR = nbytes;
*AT91C_SPI_PTCR = AT91C_PDC_RXTEN;
*AT91C_SPI_PTCR = AT91C_PDC_TXTEN;
while(! (*AT91C_SPI_SR & AT91C_SPI_ENDRX));
*AT91C_SPI_PTCR = AT91C_PDC_RXTDIS;
*AT91C_SPI_PTCR = AT91C_PDC_TXTDIS;
The initialization:
pSPI->SPI_CR = AT91C_SPI_SPIEN | AT91C_SPI_SWRST;
pSPI->SPI_MR = AT91C_SPI_MSTR | AT91C_SPI_PS_FIXED | AT91C_SPI_MODFDIS;
pSPI->SPI_MR |= ( (SPI_MR_PCS<<16) & AT91C_SPI_PCS );
pSPI->SPI_CSR[SPI_CSR_NUM] = AT91C_SPI_CPOL | AT91C_SPI_BITS_8;
Reply by ●May 30, 20062006-05-30
Thanks a lot..
I see that you are using the fixed CS configuration
I used the variable CS configuration till now. Maybe this is the cause
why it wont work. I will try this configuration when I find some time.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von Andreas Schwarz
Gesendet: Dienstag, 30. Mai 2006 19:09
An: A...
Betreff: Re: [AT91SAM] how much time does the helix mp3 decoder take to
decode for on
--- In A..., "Andreas Siebel" wrote:
>
>
> Hi,
>
> could you please tell me how to use the DMA without wasting so much
RAM
> if you want to do 8bit SPI DMA transfers.
> I could get it only working if I put the bytes I want to send on in a
> "long" array which is aligned to a 4 Byte boundary.
The actual DMA transfer is quite easy to do:
*AT91C_SPI_RPR = receivebuffer;
*AT91C_SPI_RCR = nbytes;
*AT91C_SPI_TPR = transmitbuffer;
*AT91C_SPI_TCR = nbytes;
*AT91C_SPI_PTCR = AT91C_PDC_RXTEN;
*AT91C_SPI_PTCR = AT91C_PDC_TXTEN;
while(! (*AT91C_SPI_SR & AT91C_SPI_ENDRX));
*AT91C_SPI_PTCR = AT91C_PDC_RXTDIS;
*AT91C_SPI_PTCR = AT91C_PDC_TXTDIS;
The initialization:
pSPI->SPI_CR = AT91C_SPI_SPIEN | AT91C_SPI_SWRST;
pSPI->SPI_MR = AT91C_SPI_MSTR | AT91C_SPI_PS_FIXED |
AT91C_SPI_MODFDIS;
pSPI->SPI_MR |= ( (SPI_MR_PCS<<16) & AT91C_SPI_PCS );
pSPI->SPI_CSR[SPI_CSR_NUM] = AT91C_SPI_CPOL | AT91C_SPI_BITS_8;
SPONSORED LINKS
Microcontrollers
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=JL1WKF1vUqEgZhFSvLcTwQ>
Cpu
+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=J2LJ4J
ioN5vEM0428aODAg> chip
Arms
p&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=yPOxLkzDXn
5eqYeA1wMVCQ>
Development
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=lhbODg5ptVr7u61IFQfZXQ> tool
Microprocessor
w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sigCHIRUROYao456Telb69zdA>
_____
> Terms of Service.
_____
I see that you are using the fixed CS configuration
I used the variable CS configuration till now. Maybe this is the cause
why it wont work. I will try this configuration when I find some time.
-----Ursprgliche Nachricht-----
Von: A... [mailto:A...] Im Auftrag
von Andreas Schwarz
Gesendet: Dienstag, 30. Mai 2006 19:09
An: A...
Betreff: Re: [AT91SAM] how much time does the helix mp3 decoder take to
decode for on
--- In A..., "Andreas Siebel" wrote:
>
>
> Hi,
>
> could you please tell me how to use the DMA without wasting so much
RAM
> if you want to do 8bit SPI DMA transfers.
> I could get it only working if I put the bytes I want to send on in a
> "long" array which is aligned to a 4 Byte boundary.
The actual DMA transfer is quite easy to do:
*AT91C_SPI_RPR = receivebuffer;
*AT91C_SPI_RCR = nbytes;
*AT91C_SPI_TPR = transmitbuffer;
*AT91C_SPI_TCR = nbytes;
*AT91C_SPI_PTCR = AT91C_PDC_RXTEN;
*AT91C_SPI_PTCR = AT91C_PDC_TXTEN;
while(! (*AT91C_SPI_SR & AT91C_SPI_ENDRX));
*AT91C_SPI_PTCR = AT91C_PDC_RXTDIS;
*AT91C_SPI_PTCR = AT91C_PDC_TXTDIS;
The initialization:
pSPI->SPI_CR = AT91C_SPI_SPIEN | AT91C_SPI_SWRST;
pSPI->SPI_MR = AT91C_SPI_MSTR | AT91C_SPI_PS_FIXED |
AT91C_SPI_MODFDIS;
pSPI->SPI_MR |= ( (SPI_MR_PCS<<16) & AT91C_SPI_PCS );
pSPI->SPI_CSR[SPI_CSR_NUM] = AT91C_SPI_CPOL | AT91C_SPI_BITS_8;
SPONSORED LINKS
Microcontrollers
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=JL1WKF1vUqEgZhFSvLcTwQ>
Cpu
+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=J2LJ4J
ioN5vEM0428aODAg> chip
Arms
p&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sig=yPOxLkzDXn
5eqYeA1wMVCQ>
Development
s&w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.si
g=lhbODg5ptVr7u61IFQfZXQ> tool
Microprocessor
w2=Cpu+chip&w3=Arms&w4elopment+tool&w5=Microprocessor&c=5&s&.sigCHIRUROYao456Telb69zdA>
_____
> Terms of Service.
_____
