Forums

MicroZED

Started by John Larkin October 19, 2013
On Sunday, October 20, 2013 1:04:32 PM UTC-7, Lasse Langwadt Christensen wr=
ote:
> Den s=F8ndag den 20. oktober 2013 21.25.25 UTC+2 skrev edward....@gmail.c=
om:
>=20 > > > > The $89 LX9 evaluation kit also claim to run Linux. Sure, a z80 Li=
nux, if=20
>=20
> > > > you can still call it Linux. To run a true 32 bit machine, you nee=
d at least=20
>=20
=20
> > > > the LX16. Unfortunately, the LX16 kit is loaded with junks for $28=
9.
>=20
> > > but once you get a microblaze, memory, memory controller and the usua=
l peripherals implemented how many fpga resources are left?
>=20
> > I am not using the microBlaze, but a subset of BM32, which is around 20=
0K Q=20
> > in full size. I can probably strip some instructions and addressing mo=
des. =20
> and that includes how much memory and what peripherals?=20 >=20
None. But it includes some complex instructions that can be stripped: MULT= IPLE, DIVIDE, STR/BLK, etc. as well as instruction cache and pipeline. The= resources are better used for I/Os. The result is pretty close to MIPS, but with 8 bit opcodes vs. 6, 16 regist= ers vs. 32.
Den s=F8ndag den 20. oktober 2013 22.26.43 UTC+2 skrev edward....@gmail.com=
:
> On Sunday, October 20, 2013 1:04:32 PM UTC-7, Lasse Langwadt Christensen =
wrote:
>=20 > > Den s=F8ndag den 20. oktober 2013 21.25.25 UTC+2 skrev edward....@gmail=
.com:
>=20 > >=20 >=20 > > > > > The $89 LX9 evaluation kit also claim to run Linux. Sure, a z80 =
Linux, if=20
>=20 > >=20 >=20 >=20 >=20 > > > > > you can still call it Linux. To run a true 32 bit machine, you n=
eed at least=20
>=20 > >=20 >=20 > =20 >=20 > > > > > the LX16. Unfortunately, the LX16 kit is loaded with junks for $=
289.
>=20 > >=20 >=20 >=20 >=20 > > > > but once you get a microblaze, memory, memory controller and the us=
ual peripherals implemented how many fpga resources are left?
>=20 > >=20 >=20 >=20 >=20 > > > I am not using the microBlaze, but a subset of BM32, which is around =
200K Q=20
>=20 > > > in full size. I can probably strip some instructions and addressing =
modes.
>=20 > =20 >=20 > > and that includes how much memory and what peripherals?=20 >=20 > >=20 >=20 >=20 >=20 > None. But it includes some complex instructions that can be stripped: MU=
LTIPLE, DIVIDE, STR/BLK, etc. as well as instruction cache and pipeline. T= he resources are better used for I/Os.
>=20 >=20 >=20 > The result is pretty close to MIPS, but with 8 bit opcodes vs.=20 > 6, 16 registers vs. 32.
so you have an engine,=20 all that is needed is a gastank, a gearbox, four wheels, some seats, and a = chassis and we can drive some where -Lasse
On Sun, 20 Oct 2013 10:46:34 -0700 (PDT), Lasse Langwadt Christensen
<langwadt@fonz.dk> wrote:

>Den s&#2013266168;ndag den 20. oktober 2013 07.28.08 UTC+2 skrev edward....@gmail.com: >> On Saturday, October 19, 2013 9:34:41 PM UTC-7, John Larkin wrote: >> >> > On Sat, 19 Oct 2013 19:27:49 -0700 (PDT), edward.ming.lee@gmail.com wrote: >> >> > >> >> > >> >> > >> >> > > >> >> > >> >> > >> The time and effort that I'm trying to avoid is buying a BGA FPGA and an Arm >> >> > >> CPU, and designing and laying out a board with DRAM, flash, ethernet, switching >> >> > >> regulators, jtag, and all the things you need to make a complex signal >> >> > >> processor, and then starting in on the software. >> >> > >> >> > >How about a soft CPU core? There are IP cores for 8051, 6502, z80, AVR, MIPS and ARM, as well as non-standard one like ZPU and single customer core like BellMac32. We were just discussing this a few days ago in C.A.E. You can even make something in-between. >> >> > >> >> >> >> > >Spartian 6 ($20) based evaluation board sells for $400+ >> >> > >I don't believe Zynq ($40) boards will last too long for less than $200. >> >> > >> >> > The soft cores eat a lot of FPGA resources and tend to be slow. Zynq has two >> >> > hard ARM cores, 800 MHz, pretty spiffy. And an on-chip ADC. >> >> > >> >> >> >> That's why the chip retails for more than $50. I worry about putting a $10 >>chip in a product. $20 FPGA is at the top end. $50 is not my cup of tea. >> > >it of course depends on what the product is going to sell for, but if >that 50$ part does the job of five 10$ parts with less hassle and in >the case of fpga+cpu on a single chip potentially more performance >I'd say it is a winner
No, what the product will sell for is largely irrelevant (only to the point that cost + desired profit > price). The only metric that matters is what does the alternative cost. Whether it has a higher performance is likewise irrelevant. If there is no cheaper way to do it, fine (as long as it meets cost + profit > price).
>> >> If you really need the speed, then go for it. >> >> >> >> > Avnet is selling and supporting ZED, so it's not entirely an amateur thing. >> >> >> >> Yes, but their goal is to sell chips, taking losses in boards. As long as >> you realize the real cost and not counting on the intro pricing on these >> boards for long. > >indeed, but as long as you take the real price in to account you'll be >saving until you might have to do you own production
You'd better take that cost into account before you spend the money on development.
>The end user license for the STM32F4Discovery development board actually >says you cannot use it in a product, whether that is enforceable or not >is probably debatable
Everything is debatable (except Obamacare within the government) but it is enforceable. They'll get you on the software license. As long as you can program it without using any of their software IP (either directly or indirectly), fine. You could probably sell them as doorstops without worry. ;-)
On 10/19/2013 7:11 PM, John Larkin wrote:
> On Sat, 19 Oct 2013 18:15:35 -0400, "Tom Del Rosso" > <td_03@verizon.net.invalid> wrote: > >> >> Joerg wrote: >>> >>> Now I really don't want to belittle community-based projects, they are >>> excellent tools to get youngsters into tinkering again and our society >>> really needs that. But I never use one of those products in designs >>> that have to run for years or decades. There usually isn't much of a >>> long term supply chain guarantee. >> >> If it's open-source can't you just make your own if they stop selling them? > > All the schematics and PCB layouts are published, open-source. Ditto > lots of code. We could make our own if we had to... if the parts were > still available. > >
Of course Avnet is showing zero stock... Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 USA +1 845 480 2058 hobbs at electrooptical dot net http://electrooptical.net
Den l=F8rdag den 19. oktober 2013 19.52.25 UTC+2 skrev John Larkin:
> Has anybody used the MicroZED? It looks like a good subassembly to use >=20 > on a controller board. The documentation, especially on mechanicals, >=20 > is spotty. >=20 >=20 >=20 > We have an upcoming project that needs a uP, FPGA, Ethernet, DRAM, all >=20 > that stuff. It's tempting to buy a little board that has all that done >=20 > and working, running Linux out of the box. >=20
I just stumbled over this: https://wiki.trenz-electronic.de/display/TE0720/= Overview http://shop.trenz-electronic.de/catalog/default.php?cPath=3D1_65_260 it's more expensive and it doesn't have any connectors, though that might = be an advantage if you make a carrier board and want to put it in a box looks like on of the guys behind it hangs out in comp.arch.fpga and they se= em=20 to offer 10+ years life time -Lasse
John Larkin wrote:


> We prefer BGAs to leaded parts.
OK, well, someday, I'm sure, I will be led kicking and screaming down that path. The lack of ability to examine the solder balls worries me. I do all my own assembly using an aging P&P machine and a modified toaster oven. Jon
Jon Elson wrote:
> John Larkin wrote: > > >> We prefer BGAs to leaded parts. > OK, well, someday, I'm sure, I will be led kicking and screaming down > that path. The lack of ability to examine the solder balls worries me.
That's by far not the only thing that concerns me. http://www.cartoonstock.com/lowres/jfa2492l.jpg [...] -- Regards, Joerg http://www.analogconsultants.com/
On Mon, 21 Oct 2013 12:23:13 -0700, Joerg <invalid@invalid.invalid>
wrote:

>Jon Elson wrote: >> John Larkin wrote: >> >> >>> We prefer BGAs to leaded parts. >> OK, well, someday, I'm sure, I will be led kicking and screaming down >> that path. The lack of ability to examine the solder balls worries me. > > >That's by far not the only thing that concerns me. > >http://www.cartoonstock.com/lowres/jfa2492l.jpg
In all the places I've worked, we had the least trouble with BGAs. At the PPoE the big problems were QFNs that didn't wet on the sides. BGAs never had a problem, until they have to be replaced. That never goes well locally. It's something that one has to do constantly to get right. We send it all out at the CPoE.
On Mon, 21 Oct 2013 14:25:56 -0500, Jon Elson <jmelson@wustl.edu>
wrote:

>John Larkin wrote: > > >> We prefer BGAs to leaded parts. >OK, well, someday, I'm sure, I will be led kicking and screaming down >that path. The lack of ability to examine the solder balls worries me.
The great thing about inspecting all those BGA balls is that you can't inspect them. Actually, there are some optical gadgets that let you peek under the chip, from the side, and see the first 3 or 4 rows of balls fairly well. And you can see through the entire array, for debris and such. We just bought a good one for around $30K.
>I do all my own assembly using an aging P&P machine and a modified >toaster oven.
People do BGAs at home, too. These are cool: http://www.stencilsunlimited.com/bga_rework_stencils_stencilquik.php We use them for rework, but they would be good for hand assembly, too. I have overcome my fears of 1206s, 0805s, 0603s, SC70s and US8 packages (almost), TSSOPs, and BGAs. SC79s and 0402s still terrify me. -- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
On Mon, 21 Oct 2013 19:11:53 -0400, krw@attt.bizz wrote:

>On Mon, 21 Oct 2013 12:23:13 -0700, Joerg <invalid@invalid.invalid> >wrote: > >>Jon Elson wrote: >>> John Larkin wrote: >>> >>> >>>> We prefer BGAs to leaded parts. >>> OK, well, someday, I'm sure, I will be led kicking and screaming down >>> that path. The lack of ability to examine the solder balls worries me. >> >> >>That's by far not the only thing that concerns me. >> >>http://www.cartoonstock.com/lowres/jfa2492l.jpg > >In all the places I've worked, we had the least trouble with BGAs. At >the PPoE the big problems were QFNs that didn't wet on the sides. BGAs >never had a problem, until they have to be replaced. That never goes >well locally. It's something that one has to do constantly to get >right. We send it all out at the CPoE.
We rework BGAs in-house. We had one bunch of products that kept getting expanded functionally so we had to pull the 780-ball FPGAs and put in compatible ones with twice the CLBs. We did about 20 boards, no problems. BGAs are great. Much more reliable than fine-pitch leaded parts. -- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation