| Author |
Message |
|
|
 Posted: Mar 24, 2012 - 10:52 PM |
|
Joined: Jun 16, 2006
Posts: 617
Location: Sarasota, FL
|
|
I just tested my Xprotolab at 64MHz and everything seems to work!  (Besides that all the time bases need to be at half). This is the device with the ATXMEGA32A4U. The ADC worked (I could see waveforms up to 2MHz). The DAC worked, the frequency of the output waveform was double.
Maybe I could offer an "overclocked" firmware...
The micro did not run beyond 72MHz. |
_________________
www.gabotronics.com
|
| |
|
|
|
|
|
|
Posted: Mar 27, 2012 - 05:57 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
|
|
|
|
|
|
Posted: Apr 18, 2012 - 10:50 PM |
|
Joined: Sep 04, 2007
Posts: 356
Location: Oxford (England)
|
|
I offer the following data:
I've just been running an XMega192A3U at 64Mhz (external 16Mhz crystal, PLL x4).
Only limited testing - just DMA'ing pixel data to the USART.
Everything seemed perfectly OK  |
|
|
| |
|
|
|
|
|
|
Posted: Apr 19, 2012 - 10:10 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
|
|
|
|
|
|
Posted: Apr 20, 2012 - 01:01 AM |
|
Joined: Jun 16, 2006
Posts: 617
Location: Sarasota, FL
|
|
| Well, good that it can do 64MIPS under some (or most?) circumstances... |
_________________
www.gabotronics.com
|
| |
|
|
|
|
|
|
Posted: Apr 21, 2012 - 12:40 AM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
Yes, so far only the ELPM command has caused a problem, and only when crossing what seems to be the 128k address in program memory.
And it's not a total failure, more like a skipped cycle. very odd, but not so good for my video routine!
Brad |
_________________
I Like to Build Bikes : http://www.AtomicZombie.com
I Like to Hack Stuff : http://www.LucidScience.com
|
| |
|
|
|
|
|
|
Posted: Apr 21, 2012 - 09:03 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
|
Code:
Update
After exhaustive tests using an ATxmega256A3, I have found the highest stable overclocking frequency to be 50Mhz. 54Mhz begins to fail on the ELPM instruction when crossing the 128K barrier.
Without using ELPM, 70MHz seems to pass all tests, using every other instruction, and 100% of program as well as internal SRAM.
None of my testing did any EEPROM writing though.
Brad |
_________________
I Like to Build Bikes : http://www.AtomicZombie.com
I Like to Hack Stuff : http://www.LucidScience.com
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 08:06 AM |
|
Joined: Jan 10, 2003
Posts: 533
|
|
I can confirm that XMEGA128A1’s EBI works fine @100MHz (core @50 MHz), maybe more, but I didn’t tested it.
BTW, in your project do you use external SRAM? If yes, in which configuration, and how fast can you access it? |
|
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 04:58 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
Thanks for the report.
Yes, in one version of my project, I control 512K external SRAM, but since the EBI is too slow, I wrote my own access routines in assembly. I can get 25MHz bandwidth from an external SRAM with the XMega running at 50MHz.
Brad |
_________________
I Like to Build Bikes : http://www.AtomicZombie.com
I Like to Hack Stuff : http://www.LucidScience.com
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 07:01 PM |
|
Joined: Mar 16, 2010
Posts: 98
Location: Finland
|
|
| What is the point of overclockin XMega instead of using some other processor cabable of the speeds needed? I'm running XMega at 2 MHz and I'm surprised how much it can do even at that speed. |
|
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 07:30 PM |
|
Joined: Jul 18, 2005
Posts: 62354
Location: (using avr-gcc in) Finchingfield, Essex, England
|
|
| Video generation is the microcontroller equivalent of climbing Everest. Why do it? Because it's there. |
_________________
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 07:40 PM |
|
Joined: Dec 29, 2010
Posts: 466
|
|
|
clawson wrote:
Video generation is the microcontroller equivalent of climbing Everest. Why do it? Because it's there.
Great answer!
Can some one post the highest clock achieved without an eeprom failure?
I suspect this will be the limiting factor for most of us.
Also is there a feasible way of increasing the internal voltage ? |
|
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 08:41 PM |
|
Joined: Mar 16, 2010
Posts: 98
Location: Finland
|
|
|
clawson wrote:
Video generation is the microcontroller equivalent of climbing Everest. Why do it? Because it's there.
OK. So there is no technical or economical reason to use XMega instead of a faster CPU? |
|
|
| |
|
|
|
|
|
|
Posted: Apr 22, 2012 - 09:59 PM |
|
Joined: Dec 29, 2010
Posts: 466
|
|
|
jmaja1 wrote:
clawson wrote:
Video generation is the microcontroller equivalent of climbing Everest. Why do it? Because it's there.
OK. So there is no technical or economical reason to use XMega instead of a faster CPU?
If you don't get it, read the title of the website you're in. |
|
|
| |
|
|
|
|
|
|
Posted: Apr 23, 2012 - 09:13 AM |
|
Joined: Mar 16, 2010
Posts: 98
Location: Finland
|
|
|
MarioRivas wrote:
If you don't get it, read the title of the website you're in.
Yes I get it. I was just interested are there also some real benefits for using XMega instead some other CPU for that purpose. |
|
|
| |
|
|
|
|
|
|
Posted: Apr 23, 2012 - 10:31 AM |
|
Joined: Jan 10, 2003
Posts: 533
|
|
|
AtomicZombie wrote:
Thanks for the report.
Yes, in one version of my project, I control 512K external SRAM, but since the EBI is too slow, I wrote my own access routines in assembly. I can get 25MHz bandwidth from an external SRAM with the XMega running at 50MHz.
Brad
Can you be more specific about EBI, please? I use EBI with SDRAM module (XPlained 128A1), and LD instruction from SDRAM takes 6 cycles, and what is worse, it depends on refresh cycle. I plan to use SRAM 10ns, so EBI can work without wait states, and according to XMEGA A datasheet in LPC2 mode, read/write operations should take 4 CLPKER2 cycles (2 CLKPER cycles), and even better in no CS mode. To me it is enough, but I don’t know if I can trust the datasheet, and I don’t have a functional circuit to test it.
And, BTW, how you can get better timing then 2 cycles/access without EBI? Do you use some external components, like counters or so? |
|
|
| |
|
|
|
|
|
|
Posted: Apr 24, 2012 - 01:40 AM |
|
Joined: May 28, 2006
Posts: 716
Location: Toronto, Canada
|
|
|
Quote:
but I don’t know if I can trust the datasheet
but
Quote:
I can confirm that XMEGA128A1’s EBI works fine @100MHz (core @50 MHz), maybe more, but I didn’t tested it.
George. |
_________________
www.sofgel.ro bootloader for mega and xmega
www.elsofgel.com XME development board
|
| |
|
|
|
|
|
|
Posted: Apr 28, 2012 - 02:51 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
|
TFrancuz wrote:
Can you be more specific about EBI, please? I use EBI with SDRAM module (XPlained 128A1), and LD instruction from SDRAM takes 6 cycles, and what is worse, it depends on refresh cycle. I plan to use SRAM 10ns, so EBI can work without wait states, and according to XMEGA A datasheet in LPC2 mode, read/write operations should take 4 CLPKER2 cycles (2 CLKPER cycles), and even better in no CS mode. To me it is enough, but I don’t know if I can trust the datasheet, and I don’t have a functional circuit to test it.
And, BTW, how you can get better timing then 2 cycles/access without EBI? Do you use some external components, like counters or so?
For most of my projects, 4 cycles would be a lifetime! I usually use external SRAM when doing VGA projects, so I need pixels out of the video memory at 2 cycles, or even 1 cycle. For two cycles, I can push an 8 bit address like so...
Code:
inc ADRLO ;1
out PORTA,ADLO ;1
This was giving me fCLK/2 as a pixel clock. I have also experimented with tying the LSB to the system clock and switching it with an external logic switch such as an HC157, and this gives a pixel clock equal to the AVR clock.
Things get more difficult when needing a sequential address of more than 256 since only one port can be pumped at fCLK/2. Because of this, I have not been able to create a bitmapped VGA system beyond 256 horizontal pixels.
Here is some of my older (and much cruder) work using AVRs to generate VGA...
http://lucidscience.com/pro-vga%20video ... tor-1.aspx
But a plan is in the works for 512x384 using XMega and only an external 512k SRAM.
I am also working on a project to bitbang HDMI at 640x480 without anything but passive components and an XMega. I hope to be the first to do it!
Brad |
_________________
I Like to Build Bikes : http://www.AtomicZombie.com
I Like to Hack Stuff : http://www.LucidScience.com
|
| |
|
|
|
|
|
|
Posted: Apr 28, 2012 - 04:13 PM |
|
Joined: Sep 04, 2007
Posts: 356
Location: Oxford (England)
|
|
| Whoa! HDMI?! Doesn't that need crazy speeds? e.g. 10 bits per pixel, and a pixel clock of 25.175Mhz? |
|
|
| |
|
|
|
|
|
|
Posted: Apr 29, 2012 - 03:13 PM |
|
Joined: Feb 13, 2007
Posts: 1025
Location: Gillies, Ontario
|
|
indeed! But since 50MHz is possible on XMega, that may be enough. I live for "impossible AVR tricks"! When I read on several sites that synthesizing full NTSC color in software would be either impossible or yield only a few colors and massive pixels, I just had to tackle the task. Now HDMI is on the radar!
Next rainy weekend I will try to get the 256 color XMega game system documented and online. A local engineering firm also offered to make boards for this thing, so I may go for it. Drop on a few switches and make it into a hand held XMega game system / learning kit.
Brad |
_________________
I Like to Build Bikes : http://www.AtomicZombie.com
I Like to Hack Stuff : http://www.LucidScience.com
|
| |
|
|
|
|
|
FAQ
Search
Register
Log in to check your private messages
Log in
Maximize
RSS
