More package choices in 32 bit Micros - DIP28, TSSOP20 etc
jmg
Posts: 15,185
I see NXP and Microchip are playing new product snap : More sensible (non BGA) small packages
PIC32 are now in 28-pin SPDIP, SSOP, SOIC and QFN packages
http://www.microchip.com/pagehandler/en-us/press-release/microchip-introduces-smallest-lowest-cost-pic32-mi.html
and LPC111x are also now in SO20, TSSOP20, TSSOP28 and DIP28
http://www.nxp.com/products/microcontrollers/cortex_m0/lpc1100l/#products
The PIC32 models include USB option, so that gives another choice for a really smart Download and Assist/Measure/Debug? to a Prop ?
The PIC QFN is a more usable 0.65mm than the 0.5mm of some, and SSOP28 is easy enough to handle, but compact.
Prices are lower than FTDI alternatives, tho the PIC info is vague on bus-powering.
PIC32 are now in 28-pin SPDIP, SSOP, SOIC and QFN packages
http://www.microchip.com/pagehandler/en-us/press-release/microchip-introduces-smallest-lowest-cost-pic32-mi.html
and LPC111x are also now in SO20, TSSOP20, TSSOP28 and DIP28
http://www.nxp.com/products/microcontrollers/cortex_m0/lpc1100l/#products
The PIC32 models include USB option, so that gives another choice for a really smart Download and Assist/Measure/Debug? to a Prop ?
The PIC QFN is a more usable 0.65mm than the 0.5mm of some, and SSOP28 is easy enough to handle, but compact.
Prices are lower than FTDI alternatives, tho the PIC info is vague on bus-powering.
Comments
Many, many things can be simpler in 32-bit. If only they have DMA inside...
Ever since we started down the road of emulating a complete 8080/Z80 based CP/M computer on the Prop I have been keeping half an eye out for other MCU's that could do this in a single chip (No external RAM). Perhaps just now there are some. Otherwise I'm still looking forward to do this on the Prop II which will have plenty enough space for the complete emulation.
Edit: Ah, right, you said DIP.. didn't notice that until Leon's post. The one I indirectly linked to is one-chip, but not DIP.
-Tor
64k RAM seems to be the maximum available.
The DIP ones aren't available yet, I've just ordered some of the SOIC ones (PIC32MX110F016B-I/SO ). I'll get them on 2 November. I'll design a PCB for them. I started designing one some time ago for a pin-compatible 16-bit dsPIC, so it won't take long.
Here are the chip details:
http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en555989
128K RAM is a must. 64K for CP/M and some room to move.
I like one comment over on the vintage forum "looks like emulating a CP/M computer is the new 'Hello world' for MCUs".
Well, the Prop still makes the smallest CP/M machine so far.
Nothing I have ever seen, has that mix of Flash and RAM. RAM is much larger on die, so flash size tends to be bigger.
DIP these days will also be SkinnyDip, and that limits the die size.
You could look at something like the new Atmel SAM4S - comes in TQFP48 and up ?
["The Atmel SAM4S16, the first member of our SAM4 family, operates at a maximum speed of 120MHz and features 1MB of Flash, 128KB of SRAM"]
Plans for "Up to 2MB of Flash and 192KB of SRAM"
- not sure if this family supports QuadSPI ?
However, I do see Atmel missed the chance to move to 32 bit timers
Or you could look at the LPC43xx, which has a M0 copro, and that may allow a nice peripheral memory map emulation ?
Also available in RAM only, which is cheaper :
LPC4320 200 KB SRAM 150 MHz LQFP-100 $5.36
and I think this family can boot/run from QuadSPI.
However, most micros are not aimed at the hobbyist, hence no DIP packages. Commercially, for any volume, t/hole is dead - its just a fact of life.
that is an AVR32, nothing to do with the ATMegas or ATTinys. The AVR32 assembler is extremely nice and compact. and supports unaligned access too !, Sadly they discontinued the AP7000 line, it had MMU and full linux support
BTW, who would have thought that a MIPS processor would fit such a package some 20 years ago on the days of DEC...
Just an observation... Laying out boards seems to be an odd avocation for a computer expert. At one company where I worked, laying out boards was what the new guy did. At another, board design was what the guy who sucked at programming did.
I'd expect someone with vast computer experience to be creating killer apps or a new operating system. Seems like demo boards aren't too hard to come by.
I actually like designing PCBs and it only takes a few minutes to design a board like that. Here is the schematic:
I'm envious! Enjoy!!
http://en.wikipedia.org/wiki/LEO_%28computer%29
Where I used to work, it was a smallish company with about 30 employees. Engineers were responsible for hardware design, including the PCB, and software development. I designed their main product, and had another engineer design the PCB for me, as I needed to get the software done ASAP. It was a six-layer board, and took about a month. That was hardly the sort of thing that would be given to a new employee. It worked first time.
That's not very nice. It's a serious put down to all the worlds PCB designers. It implies that software engineers are some how smarter than board designers. In my experience there is no evidence to suggest that. As it stands the software guys are dependent on PCB designers, not to mention logic designers, chip designers, analog electronics experts, mechanical designers etc etc. A whole world of stuff most softies know nothing about. Else they have no where for their oh so brilliant software to run.
Me, I used to work for Racal on software for a schematic capture/PCB layout package, I soon discovered that laying out boards is not necessarily as easy as it looks when you have a huge list of electrical, thermal, mechanical even aesthetic constraints to work within.
On another level, laying out boards can be a nice puzzle to relax the mind after a hard day churning out code. Like some people tackle crossword puzzles or sudoku for a good change.
That wouldn't be me by any chance? http://www.vintage-computer.com/vcforum/showthread.php?27703-Single-Chip-CP-M-Board post #10
In a way I think CP/M can be a useful test because even on 64k CP/M you ran out of memory after a few days of coding, and then we had so called "progress" to pic chips with only a few hundred bytes of ram. Fortunately in the last few years microcontrollers are getting a decent amount of ram.
I very much agree that we need more micros in proper packages that you can put on a breadboard and experiment with. The DIP package was what got me into the Propeller in the first place and I still think it is a huge selling point. You can get things working on a breadboard without having to solder, and IMHO that is a great way to get a younger generation interested in electronics.
@Leon I like that. Good on ya, mate!
Ray
I have layed out a number of professional pcbs in my time. At times I was able to use professional pcb design houses and instead devote my time to the hardware and firmware (software burnt in eprom). Anyone can lay a pcb out, but it takes a proper understanding to get it right the first time, both in hardware design and correct layout which means proper decoupling, ground and power planes, correct part locations, sizes of tracks, hole and pad sizes, etc, etc. Almost all my designs remain at the first pcb revision which is the prototype. TriBlade and RamBlade are first pcbs, as are all my latest pcbs with 1 exception, and I overclock all my props to 104MHz.
At Kidsgrove, I worked in the main development lab where the prototype KDF9 was being developed. I spent a lot of time working on a LACE analogue computer. Next to it was a little (wardrobe-sized) KDN2 computer which my boss had programmed to generate football pool permutations for a syndicate that had been organised. They never won anything.
However, that'll be another story in NXP's Cortex-M0 (also in DIP form factor) offerings. Either someone else in DangerousPrototypes or Adafruit will make a simple programmer for this one once it gets popular. Also, the compilers for ARM are very difficult to use.
What can I do on these little things? It would be a dsPIC replacement, but not a 100% one due to the absence of DSP-specific instructions in PIC32 and the M0.
The PIC32 has some DSP instructions, such as MADD, which performs a MAC function.
Leon, a friend of mine iirc worked on the Leos. I worked on Frieden/Singer/ICL System Ten and 25 including teaching both hardware and software and I designed interfaces for them including a multiprocessor pcb that plugged into the bus of the 25. This was a most enjoyable time and spanned 1974-2000. The Ten was released in 1969 and a large number of 25s were still being used right up to the end of the 90s. It only stored the date as YYMMDD.
In so many ways, the Ten and 25 are like the prop... risc with src and dest in the instruction, calls and jumps similar, hub and cog memory, etc. One of these days I will get around to emulating them on a prop - I did it on a pc writing it in 486 assembler on a 30MHz pc. It was validated at about 3x performance.
Other instructions had the index as part of the instruction itself. You could write self-modifying code to update the index if you wished to, but as the CPU could also switch to separate code and data banks (which would give you another 64kilowords of memory to work with), it was better to use another feature: An instruction which would execute whatever was in a given register as an instruction. So just keep updating that register, and execute its content.
That disappeared quite a while ago, but I recently wrote a user-level emulator and had to dig up all of that again. A fun architecture in lots of ways.
-Tor
The System 25 use used 10 bytes of 8 bits for the native instructions. Up to 32 instructions were now possible (think about 25 were implemented??), and the memory size was increased by more bits.
Just one problem I found with a keyboard PCB was long traces next to each other going to the keys the furthest away from the top where the cable was attached.
Because these were next to each other for quite a distance, they had a bit of capacitance between them.
And you clock these things fast, then the microcontroller thinks the previous key scanned is a key press because it still is high (due to the capacitance). But you need to get the traces down there somehow!
Then you get multilayer boards, etc. I can't imagine being able to design this stuff and take into consideration problems such as the above.
Even worse is something which involves capacitive sensing. All sorts of things you must do or can't do. Here is paper on that...
http://ww1.microchip.com/downloads/en/AppNotes/01102a.pdf
Mind boggling to me!