Wellll...
Not going too well. With just the simple 'LED blink' code from Uze we have:
28.636MHz from crystal = no LED activity
28.636MHz from Crystal with CKDIV8 set = LED Blinking
8.0MHz from internal RC = LED Blinking
So, that tells us that the oscillator is starting up and running from the external 28.636MHz crystal-- so that's not the problem. (As verified by the code executing just fine when the CKDIV8 is set... That's just dividing down the oscillator with the system clock prescaler-- so the Core, SRAM, Flash, etc. get a lower clock rate.)
The only thing left that I can think of is to try to tweak the operating voltage and try to get it going that way. Not too encouraging though at this point... The DIP parts will be the same die, but maybe through some bizarre circumstance they'll behave better. (one can dream!)
-Clay
P.S. There's always the XMega192. 192K flash, 16K SRAM, two wire PDI with real hardware breakpoints/etc., seven 16 bit timers, event system, DMA, two channel 12bit DAC, up to 32MHz from 3.3V... And the 64-QFN fits inside a 40 pin DIP module outline. Same price as the 1284P...
Any news on the ATmega1284?
Re: Any news on the ATmega1284?
Ah shoot, that sucks then. I guess they are using some sort of new manufacturing process since the 644?Not going too well. With just the simple 'LED blink' code from Uze we have:
28.636MHz from crystal = no LED activity
28.636MHz from Crystal with CKDIV8 set = LED Blinking
8.0MHz from internal RC = LED Blinking
Oh well, seems that, unfortunately, any upgrade will mean the end of the Uzebox's DIY aspect. I'm unsure about what to pursue at this point. Stick with 644 for the foreseeable future or go ahead with a "Super Uzebox" driven by a Xmega. That sure looks exiting though. Any thoughts by the community??P.S. There's always the XMega192. 192K flash, 16K SRAM, two wire PDI with real hardware breakpoints/etc., seven 16 bit timers, event system, DMA, two channel 12bit DAC, up to 32MHz from 3.3V... And the 64-QFN fits inside a 40 pin DIP module outline. Same price as the 1284P...
Btw, any cheap QFN-64 to DIP adapters you know of? How hard would it be to make a Xmega AVCore that is pin and voltage compatible with the current one? In the mean time I just noticed Digikey has a nice xmega evaluation board using the xmega128A1. At a mere ~29$ with 8M onboard SDRAM, speaker, buttons, etc it's almost a steal!
-Uze
Re: Any news on the ATmega1284?
Yeah, all indications seem to be that they're optimizing for power consumption over their rated speed range. (ie, they're not interested in having it work faster than spec'd at the expense of getting beat up by other MCU manufacturers when it comes to power consumption which is the new 'in' thing.)uze6666 wrote: Ah shoot, that sucks then. I guess they are using some sort of new manufacturing process since the 644?
Xmega makes good sense for a logical step up from the Mega644-- familar CPU, good specs, enough new hardware toys to be exciting. On the other side there's something like a little ARM MCU which would make for a 32 bit CPU and likely more flash at a cheaper price... But software video generation is trickier.Oh well, seems that, unfortunately, any upgrade will mean the end of the Uzebox's DIY aspect. I'm unsure about what to pursue at this point. Stick with 644 for the foreseeable future or go ahead with a "Super Uzebox" driven by a Xmega. That sure looks exiting though. Any thoughts by the community??
The Event System and DMA in the XMega just *seems* like it could be pressed in to service for video generation. At least to ease the CPU load somewhat anyway. Be neat if it could blast out a scanline buffer without CPU overhead. Then we could likely support QVGA at standard monitor sync rates. If nothing else, it could definitely handle audio transfers to the built-in DAC...
I haven't seen any QFN's (but I haven't really looked hard-- they're trickier to hand solder though too. I use solderpaste and hot air.) There's tons of TQFP-64 adapters up on eBay for a few bucks each (those *are* hand solderable, although it does take a little bit more skill than DIP). I'd go for the QFN simply because it fits inside the 0.6" spacing on a standard 40 pin DIP socket. An Xmega version of the AVCore would actually be easier than the Mega644 version-- the XMega is spec'd for full speed at 3.3V, so there wouldn't need to be any voltage dividers on the SPI interface, etc.Btw, any cheap QFN-64 to DIP adapters you know of? How hard would it be to make a Xmega AVCore that is pin and voltage compatible with the current one?
Yeah, that's the "XPlain" board they give away at the seminars. It's impressive for the money, if a bit odd. It has a little 'helper' AVR on there with a USB interface, so it basically has a built in ISP Mk2. Can't beat it for the price, that's for sure. Not all the signals are pinned out (so it's not a full on 'devboard' for any random application), but still plenty to play with. I didn't look at the port assignments-- might just be some resistors stuck on the available ports and it's the Super Uzebox. I could make a baseboard with the A/V and controller connectors on it.In the mean time I just noticed Digikey has a nice xmega evaluation board using the xmega128A1. At a mere ~29$ with 8M onboard SDRAM, speaker, buttons, etc it's almost a steal!
-Clay
- DaveyPocket
- Posts: 378
- Joined: Sun Sep 14, 2008 8:33 pm
- Contact:
Re: Any news on the ATmega1284?
So, I'm guessing this means the Atmega644PA cannot be over clocked since it shares a similar low power design?uze6666 wrote: Ah shoot, that sucks then. I guess they are using some sort of new manufacturing process since the 644?
Yeah, all indications seem to be that they're optimizing for power consumption over their rated speed range. (ie, they're not interested in having it work faster than spec'd at the expense of getting beat up by other MCU manufacturers when it comes to power consumption which is the new 'in' thing.)
Re: Any news on the ATmega1284?
I'd say so, Davey. There's an atmel app note for migrating from 164p/324p/644p to pa, so I guess the 1284p just skipped straight to the pa manufacturing process.
I've seen the '1284 (non-picoPower design) mentioned in some datasheet comparisons, but I don't see it available (I guess it was cancelled?). Also, I've seen Uze claim instability on some '644's - is the point of the LED test to remove the possibility that the chip Clay is testing has this instability, or could Clay simply have a bad chip?
As far as the Super Uzebox goes, ultimately it really depends on Uze and whomever assists Uze in the design of the supporting hardware and kernel. However, if you consider the current Uzebox capability to be around that of the NES, it'd be nice to see the Super Uzebox make some serious inroads into the SNES for it to justify the extra time and effort. It's a careful balance not to go too far and encroach on Pandora territory unless you want to compete with a much bigger and arguably more experienced community. There could be an argument made that emulator possibilities could kill off some of the homebrew motivation, but would likely massively increase the uzer base. I don't think we're talking about those degrees of upgrade, though.
You also have to consider that the Super Uzebox pretty much kills off any future development for the Uzebox imo. Even its announcement will likely lessen motivation to create new games for the Uzebox until the Super Uzebox is complete (who knows how long that will be with jobs and other commitments). LadyAda would likely have difficulty moving existing stock and new owners may feel a little peeved (justifiably or not).
So I think that if the '1284 has been ruled out, I'd make the decision privately (maybe through PM's with Clay and those experienced enough to assist), and should you decide to try a Super Uzebox, you develop it in private and announce it a few months before it's ready. I'd definitely cease thinking up ideas for the Uzebox and move on to the Super Uzebox if it was announced, but I'm also not dissatisfied with the current Uzebox.
I've seen the '1284 (non-picoPower design) mentioned in some datasheet comparisons, but I don't see it available (I guess it was cancelled?). Also, I've seen Uze claim instability on some '644's - is the point of the LED test to remove the possibility that the chip Clay is testing has this instability, or could Clay simply have a bad chip?
As far as the Super Uzebox goes, ultimately it really depends on Uze and whomever assists Uze in the design of the supporting hardware and kernel. However, if you consider the current Uzebox capability to be around that of the NES, it'd be nice to see the Super Uzebox make some serious inroads into the SNES for it to justify the extra time and effort. It's a careful balance not to go too far and encroach on Pandora territory unless you want to compete with a much bigger and arguably more experienced community. There could be an argument made that emulator possibilities could kill off some of the homebrew motivation, but would likely massively increase the uzer base. I don't think we're talking about those degrees of upgrade, though.
You also have to consider that the Super Uzebox pretty much kills off any future development for the Uzebox imo. Even its announcement will likely lessen motivation to create new games for the Uzebox until the Super Uzebox is complete (who knows how long that will be with jobs and other commitments). LadyAda would likely have difficulty moving existing stock and new owners may feel a little peeved (justifiably or not).
So I think that if the '1284 has been ruled out, I'd make the decision privately (maybe through PM's with Clay and those experienced enough to assist), and should you decide to try a Super Uzebox, you develop it in private and announce it a few months before it's ready. I'd definitely cease thinking up ideas for the Uzebox and move on to the Super Uzebox if it was announced, but I'm also not dissatisfied with the current Uzebox.
Re: Any news on the ATmega1284?
Yeah, the super-simple firmware just does an absolutely minimal task (blink an LED) so we're sure that the 'black screen' isn't a result of some other hardware feature on the 1284P causing the Uzebox kernel to die. Basically if we can't blink the LED, the part isn't executing code.paul wrote:I've seen the '1284 (non-picoPower design) mentioned in some datasheet comparisons, but I don't see it available (I guess it was cancelled?). Also, I've seen Uze claim instability on some '644's - is the point of the LED test to remove the possibility that the chip Clay is testing has this instability, or could Clay simply have a bad chip?
I've not tried two parts with the same results, so while it may be possible for some chips/lots to work under the same conditions it's probably not economical to try to screen for any 'good' ones assuming they exist. The fact that the difference between 'blinking LED' and not is just a matter of the clock divisor to the core/peripherals pretty much says that it's an on-chip problem.
As a datapoint (and why I wasn't holding out a whole lot of hope for the 1284P), I had problems trying to get the 644P to work reliably as well-- hence why I stuck with the 644 on the AVCore module.
One last ditch effort on the firmware side we could try. Start up and execute code with CKDIV8, then switch on the fly and disable the divisor, or progressively reduce the divisor... Maybe 'sneaking up on' full speed would kludge around something if it's a matter of the first fetch failing at high speed or something... A long shot for sure...
Yeah, that's the problem. The Uzebox is definitely a certain niche. It's novel in the way it works. If we just wanted more power it'd be easier to just buy one of the SunPlus based units out there and play with it instead. (ARM9, SDRAM, 2.4" color TFT LCD, ~$34 http://www.dealextreme.com/details.dx/sku.21968 ) Even going down the "use an ARM7/9/CortexM3" gets to be an exercise in "why not just spend another $5 and have more CPU and hardware graphics, etc..." The Uzebox is neat because it's minimal, IMHO.It's a careful balance not to go too far and encroach on Pandora territory unless you want to compete with a much bigger and arguably more experienced community. There could be an argument made that emulator possibilities could kill off some of the homebrew motivation, but would likely massively increase the uzer base. I don't think we're talking about those degrees of upgrade, though.
There are ways to minimize some of the impact there-- if a 'Super Uzebox' was largely *source code* compatible with the old Uzebox that'd be kind neat. (ie, a new Kernel would support the old modes and legacy API's and act as the hardware abstraction layer-- so people could just 'rebuild' their games and they'd generally "just work". Then it'd be easy to make augmented versions that take advantage of the extra space, etc.) Similarly, something like the AVcore could just be swapped out for a new version and reuse the existing baseboard/power/controller setup,etc.You also have to consider that the Super Uzebox pretty much kills off any future development for the Uzebox imo. Even its announcement will likely lessen motivation to create new games for the Uzebox until the Super Uzebox is complete (who knows how long that will be with jobs and other commitments). LadyAda would likely have difficulty moving existing stock and new owners may feel a little peeved (justifiably or not).
Orphaning old stock is certainly a problem, although maybe not for anyone except Limor and I. (I've still yet to sell enough AVCore's to recoup development/manufacturing costs-- but that's just a risk you take when supporting a niche product like this!)
At some point too though you can be pretty sure that Atmel will EOL the Mega644 in favor of the -P/PA parts, so eventually we'll need a migration path to something else anyway.
-Clay
Re: Any news on the ATmega1284?
So it looks like the ATmega1284 isn't going to provide an easy upgrade path as we first thought. What are peoples opinions/thoughts?paul wrote:.. So I think that if the '1284 has been ruled out, ...
Do we look at the xmegas or an ARM ?
Strangely, the Xplain (ATAVRXPLAIN-ND) isn't listed on Digikey's UK or Canadian sitesuze6666 wrote:...Digikey has a nice xmega evaluation board using the xmega128A1. At a mere ~29$ with 8M onboard SDRAM, speaker, buttons, etc it's almost a steal!
-Uze
Re: Any news on the ATmega1284?
Speaking of XMegas, I see Sparkfun have just listed a nice looking ATxmega128A1 breakout board for $24.95
Re: Any news on the ATmega1284?
That's a very valid point. Considering the astonishing amount of work done on the current Uzebox (done by me and many others) I'm not too fond on the idea of starting over with "Super" Uzebox. For one, I think there's still juice to extract (IHMO) in the current version. We are starting to have a nice game library and the unit in my living room get played on more and more by friends & family. Additionally, I personally won't have time to design, research or develop on such a venture for the foreseeable future.Paul wrote:You also have to consider that the Super Uzebox pretty much kills off any future development for the Uzebox imo. Even its announcement will likely lessen motivation to create new games for the Uzebox until the Super Uzebox is complete (who knows how long that will be with jobs and other commitments). LadyAda would likely have difficulty moving existing stock and new owners may feel a little peeved (justifiably or not).
So I think that if the '1284 has been ruled out, I'd make the decision privately (maybe through PM's with Clay and those experienced enough to assist), and should you decide to try a Super Uzebox, you develop it in private and announce it a few months before it's ready. I'd definitely cease thinking up ideas for the Uzebox and move on to the Super Uzebox if it was announced, but I'm also not dissatisfied with the current Uzebox.
A super Uzebox would be great but we have to recall it would have its cons (as Paul mentioned). The whole DIY-friendly aspect goes away if we go with a non-DIP available part like the xMegas. By then, its easy to add external DRAM and a bunch of other stuff and we slowly go away from the minimalist aspect. Add more power with an ARM and it's so powerful we could do 3D graphics...so long for the retro aspect. I made careful trade-offs when I originally designed the Uzebox. I made them so the final thing would be novel, cheap, easy to assemble and the whole concept geeky enough to appeal to other guys like me.
I'm probably repeating myself (again), but I have nothing against a super "Super Uzebox" project. Quite on the contrary. But unless we come up with a design that respects the original vision, it wouldn't be appropriate to call it Uzebox. With all this said I kinda like the AtxMega192A1 Clay mentioned since it would still be an 8-bit MCU with relatively limited resources and be backward compatible. Plus, it would have the 16K RAM we were looking for with 1284p plus plenty of neat functions. If Sparkfun can make a breakout board like Tony mentioned (*very* nice btw) but with that XMega192 I may give it a second tought. I *may* even consider external SRAM!
Yes indeed. I guess we'll see then. In the mean time..buy a lot?!havok1919 wrote:At some point too though you can be pretty sure that Atmel will EOL the Mega644 in favor of the -P/PA parts, so eventually we'll need a migration path to something else anyway.
-Uze
-
- Posts: 30
- Joined: Sat Nov 14, 2009 10:07 am
Re: Any news on the ATmega1284?
Hey, I know that I haven't been involved as long as pretty much anyone here, I just wanted to add something to the discussion because I love the Uzebox and want to see it continue to grow and be awesome!
I just wanted to point out that with personal fabrication being "the next big thing" in DIY, it is not out of the question to have a next generation Uzebox utilize a surface-mounted part like the xmega and still make its production possible for the hobbyist. I have a couple of friends who are making home-built pcb mills, and have used them to prototype circuits using xmegas!
For example, here's one that costs a couple hundred in parts to build - and can wield a 10 mil endmill:
http://mtm.cba.mit.edu/machines/mtm_az/index.html
(being part of the MTM project: mtm.cba.mit.edu)
Anyway, just my $0.02. I totally agree with the points about it being a bummer abandoning the existing install base and redoing the whole kernel, as well.
I just wanted to point out that with personal fabrication being "the next big thing" in DIY, it is not out of the question to have a next generation Uzebox utilize a surface-mounted part like the xmega and still make its production possible for the hobbyist. I have a couple of friends who are making home-built pcb mills, and have used them to prototype circuits using xmegas!
For example, here's one that costs a couple hundred in parts to build - and can wield a 10 mil endmill:
http://mtm.cba.mit.edu/machines/mtm_az/index.html
(being part of the MTM project: mtm.cba.mit.edu)
Anyway, just my $0.02. I totally agree with the points about it being a bummer abandoning the existing install base and redoing the whole kernel, as well.