ATmega1284P schedule

[havok1919]

Ok, got some info back from Atmel. The '1284P is in general sampling now, but only in TQFP and that's as far as it'll go for now. They have to do a mask change to fix something or other, so the B rev silicon is scheduled to tape-out and sample in Q1'09 with production to follow in the same quarter.

So, it's a good 5-7 months out.

I admit that one thing makes me nervous about it for this application... They only refer to it in the "P" (picopower) family. If the '644P's have issues with overclocking the '1284P might as well-- but if that's the *only* SKU available for the 1284 architecture... Well, I guess at that point UZEBOX2 is just Xmega.

-Clay

[uze6666]

Geez, they announced this devices ages ago. Is it always that long? Let's hope the xmegas don't take that much time.

Regarding overclocking problems, as mentionned in another thread it *seems* to affect only the UART and causes glitches on the data bus during RX. But again, that could be a bug in my software. It's just fun it stops when I plug back a regular 644. Or maybe the driving of the pin from my MIDI circuit? Note that I didn't test other external peripherals like SPI, but programming works ok, so this should be ok. It could be interesting to test it alone to see if I get the same glitches.

Uze

[havok1919]

Geez, they announced this devices ages ago. Is it always that long? Let's hope the xmegas don't take that much time.

Yeah, I assume they're running a slightly new/modified process for these so they're likely tuning that for yield in addition to getting the masks right.

Xmega 128/64A1's are supposed to be ramping full production now (Q3'08). The A3's and A4's are Q4'08 samples and pre production and Q1'09 volume production. XMega 384/256/192A1's are Q2'09 sampling, Q3'09 poduction. The Xmega128A4 is also Q2'09 general sampling and Q3'09 volume production.

-Clay

[uze6666]

Many thanks for the tips. While the first generation will only have 8k of RAM,it is still very interesting since, at worst we could have external memory. Talking of this, at 9mm for the 100 pins BGA package, would it fit on the stamp?

Also, since the xmegas are clocked at 32Mhz, and there no available crystals multiple of the NTSC color burst crystal past 28 MHZ, what do you know about custom or programmable crystals? Could they do the job and how much would they cost?

Uze

[havok1919]

Many thanks for the tips. While the first generation will only have 8k of RAM,it is still very interesting since, at worst we could have external memory. Talking of this, at 9mm for the 100 pins BGA package, would it fit on the stamp?

Yeah, that'd fit no problem. The CBGA could be a little annoying. In general that will likely require 5mil trace/space (and more importantly) 8mil drills to do the escape routing. Technically it's no big deal, but it makes the prototypes much more expensive since Advanced Circuits and the like consider 8mil drills to be 'production spec' so you're looking at ~$2K USD for prototypes. And then assembly is a little more expensive. Still doable though. The part fits inside a 0.6" DIP spacing easily. Here's a little board with a pair of 64 pins CBGA's I did:

Also, since the xmegas are clocked at 32Mhz, and there no available crystals multiple of the NTSC color burst crystal past 28 MHZ, what do you know about custom or programmable crystals? Could they do the job and how much would they cost?

I do believe that Epson Toyocom have programmable oscillators from Digikey and the like. You can also get custom cut crystals, but the minimum would probably be in the 1000's of pieces. I'd seriously consider just sticking with the 28.6MHz clock-- we wouldn't need to overclock and no painfully expensive SMT programmable oscillators... (they're in the $4-8/ea range in 100's depending on size/vendor/etc.)

-Clay

[uze6666]

I see, very expensive to use CBGA. I mentionned it because the QFP seemed much bigger at 14 mmx14mm. I guess we will have to wait for 64 pins versions...

As for the custom cut or crystal, that was also my understanding: expensive! Boosting the speed by another 3.5Mhz would have given about 200 more cycles per scanlines, interesting, but not much worth the cost I guess. And it would make generating the 14.3Mhz clock for the AD723 much more complicated. Thats too bad it does not have a clock multiplier like the AD724. We could have just fed it 3.56Mhz...

Uze

[uze6666]

And there I just see this: http://www.atmel.com/dyn/products/tools_card.asp?family_id=676&family_name=MCU+Wireless&tool_id=4291
It's about the AVRRAVEN which is an Atmel radio dev kit selling at digikey and which contains an...Atmega1284p! Why the hell don't they sell it if its production ready???

Hey Clay, why not do like the multitaps and rip the 1284 off those boards?

Just joking, hehehe!

Uze

[havok1919]

It's about the AVRRAVEN which is an Atmel radio dev kit selling at digikey and which contains an...Atmega1284p! Why the hell don't they sell it if its production ready???

Interesting! I wonder if they had some glitch in it that didn't affect any of the Raven functions but such that they didn't want to widely release it. (Or it could just be yield issues on a new process or something too-- not enough to go around.)

Would be interesting to see one up close... See if it's ES marked or something.

-Clay

[gabossy]

What sources are good for some one who likes to gather open source intelligence research as a hobby? I am the kinda of guy who likes to know about what is going on in the news, before it becomes news. Seeing how I am in the military, I can see such a hobby being an asset to my current job.
___________________________
yahoo keyword tool ~ overture ~ traffic estimator ~ adwords traffic estimator

[uze6666]

Good question, I'm not too much into open source apart the Uzebox project. Check out Adafruit forums, since they are a big open source hardware proponent.

-Uze