PCB fabrication costs for OMAP3530 designs

GoogleGroups
1

Hi-

I have recently discovered the OMAP3530 and Beagle board and would like to
develop my own board which will be based in large part on the Beagle board.
However I was concerned as to what the small geometries and need for blind
via's etc. might do to the PCB fab costs... leaving aside for the moment
what it might cost to have said board assembled with its VIP (Via In Pad)
design etc.

So as an initial sanity check I simply submitted the unmodified Beagle board
gerbers to my board house (Sierra Proto Express- see:
http://www.protoexpress.com/) for a quote. I asked the price for quantity 3
boards in 5 days or 8 boards in 20 days and here is the quote:

3 Boards 5 Days turn $7,664.37
8 Boards 20 Days turn $9,074.88

For me this is a non-starter. Are these prices representative of what other
board houses might charge? Does anyone know of other board houses that
might quote better prices for boards like the Beagle board?

I could instead content myself with using a Beagle board in my design IF the
LCD signals were brought out to a header. I have heard this Rev. C might do
just that. Can anyone tell me when Rev. C is due? If it's soon it might
well be that the schematic might be set already. Does anyone know if the
Rev. C schematics are out so that I could see if this second approach might
work for me?

Thanks,
Dan

__________ Information from ESET NOD32 Antivirus, version of virus signature
database 3661 (20081203) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com

2

First off, you need to go to a high end shop for these types of boards. Any shop that has not built this type of board will be high initially. As the shop gets more comfortable with the board the price comes down. Your quantities are also extremely low. You need to the quantity up there to get a better price. In quantit of 10K we pay under $20 per board.

The pricing you received is high even for these quantities. You are better off going to MEI, DDI, or Streamline, all of whom have already built this board before.

The REV C board will have the LCD signals brought out to a pair of headers. Rev C will be out late Q1 for purchase. We are testing the prototypes now. The REV C schematics are set, but will not be released until testing is completed. All of the signals that connect to the TFP410 are the signals that are brought out on these headers including the I2C3, 5V, 3.3V, and 1.8V.

Gerald

3

I have used streamline and advanced circuits.. For large quontity
streamline or for a proto board @ $99 a price with 4 unit minimum and
$50 panel fee
once you have proven your design with a local on shore fab house you
can use online fab houses over seas like in south Korea has very
reputable company that are very reasonable..

another option is to look at cogents ompa3530 module board
CSB740 - TI OMAP3530 SODIMM System On a Module (SOM)
http://www.cogcomp.com/csb_csb740.htm

4

I’m involved with a project doing a Beagle derivative, and I just spoke about pricing to someone who is talking to various board makers. Based on what he’s found out so far, he thought those prices seemed a bit high, even for prototype quantities. We’re going to be getting more pricing info in the next week or two from different vendors, so let me know if you would like further info, and I can pass it along when we get it.

Brett

5

I'm also concerned about the cost of prototyping beagle derived
hardware. I'm new to learning about ARM hardware and I like the
board. It looks to be a good way to get started playing with ARM
software/hardware. However, I just don't see myself being able to use
it as a reference platform for hardware hacking into customized
boards. Is there any small chance that some of the technology could
carry over to a simplifed board with less layers based on TSSOP style
packages that I can still solder myself to keep prototype costs down?

Ty

6

Anyone interested in doing there own PCB based on the OMAP35xx should look at TI's website. In particular they have a couple of very good application notes on laying out boards based on this chip.

The package of the chip used on the Beagleboard in some ways is great because of the POP for the memory. However, it does force some rather more exotic constraints on the board side. The other package that the processor is available in can be utilized with easier to manufacture PCB constraints (cheaper), but with other trade-offs.

Keith

7

The OMAP3530 does not come in TSSOP, so this is not likely to happen. All packages are BGA. There is a .65mm package that is available for OMAP35xx, but it is still BGA, although it can result in a lower PCB cost and is a bit larger.

Gerald

8

We too would like to build a derivative of the beagleboard.

However we are struggling to find a PCB manufacturer and PCB assembler
in the UK who have experience with the 0.4mm BGA and that can make low
volumn prototypes for a reasoable price. Has anyone found such a
manufacturer in the UK?

The ompa35x is a wonderful device however due to its small size and
state of the art PCB requirements it will hamper its adoption in the
wider embedded market becuase its so expensive to manufacture. The
1.8V i/o pin is also a limitation. 3.3v would be a certain advantage
for us as we would not need level shifters which saves space and
lowers component cost.

The 0.4mm BGA is great if you are making very small mobile phones in
large volumns but is not so important for general embedded control
systems which are much larger and manufactured in lower qnatities and
by smaller organisations.

Its a shame that TI cannot offer the same 515 pin device in a much
larger gemortry package to keep the PCB manufacting and assemble costs
down. Even the 0.65mm package is far too small and has less I/O pins
than the 0.4mm 515pin BGA .

Andrew

9

The OMAP3530 comes in .4mm pitch, .5mm pitch, and in .65mm pitch packages. The .65mm can actually be routed using .8mm pitch rules. Actually, .4mm pitch can easily be done by those who have a desire to try it. Granted, .4mm is not a leading edge technology these day as .3mm and .2mm are becoming more prevalent, but it can be handled by anyone that has a desire to try it. Most people are just afraid of it. I think that you will find most of these SOC suppliers are at least .5mm and some even .4mm. It is where the industry is going.

On Beagle assembly we are getting 99% yields on .4mm pitch along with the POP device. TI has a lot of customers that are not cell phone suppliers that are using these devices, take Archos for example. The biggest hurdle one has to get over is the old days. You need to find a way to move forward. A lot of TI customers screamed about the same issues. Now they swear by the technology. You also need to reflect that the bigger the physical package, the higher the cost and the higher the price of the device.

The companies that make the Beagle PCB are DDI, MEI, and Streamline. When we started working with these companies, they said it was impossible as well. There are also companies in China and Korea that can make the board. One thing nice about Beagle, the information is there and the PCB house can easily to a trial run to see how they do. I would encourage you to encourage some of the UK sources that you have contacts with to give it a try. It is not impossible. Uncomfortable at first, yes, but not impossible. I would be happy to work with these companies to get them over the hump.

Gerald

10

We too would like to build a derivative of the beagleboard.

However we are struggling to find a PCB manufacturer and PCB assembler
in the UK who have experience with the 0.4mm BGA and that can make low
volumn prototypes for a reasoable price. Has anyone found such a
manufacturer in the UK?

The ompa35x is a wonderful device however due to its small size and
state of the art PCB requirements it will hamper its adoption in the
wider embedded market becuase its so expensive to manufacture. The
1.8V i/o pin is also a limitation. 3.3v would be a certain advantage
for us as we would not need level shifters which saves space and
lowers component cost.

Just so that you don't think I'm unwilling to acknowledge any issues, I'll agree that having more 3.3V I/Os would save many developers in the wider embedded market some money and board area. Now for the package...

The 0.4mm BGA is great if you are making very small mobile phones in
large volumns but is not so important for general embedded control
systems which are much larger and manufactured in lower qnatities and
by smaller organisations.

Its a shame that TI cannot offer the same 515 pin device in a much
larger gemortry package to keep the PCB manufacting and assemble costs
down. Even the 0.65mm package is far too small and has less I/O pins
than the 0.4mm 515pin BGA .

What extra I/O pins are you looking for on the 0.65mm ball-pitch package?

The 0.65mm ball-pitch package can be used with design rules typical for 0.8mm ball-pitch packages (actually a bit looser than is even required for that package). The entire device can be "escaped" in just 2 signal layers without any blind or burried vias thanks to the "via channel array" technology that reserves large areas for the vias. There is an example layout that shows how to escape the signals from the package.

I'd like to understand if you really need a larger geometry package with more pins or if you simply haven't explored the 0.65mm ball-pitch package sufficiently enough yet.

11

Here is an idea that may work. Why don’t we pool our resources and see if we can generate a board specification that will work for most of us. This way we get the benefit of more than one set of eyes to review the design and we can all share in the costs for manufacture prototypes. Our combine volume will also reduce the cost of manufacturing this board. Ultimately, we could have Digikey sell the board and recover our costs, or potentially make a small profit. The idea is that we collaborate in an open source manner and make the design available to everyone. BTW, TI said that they have people available who would review the design and provide input as well.

Regards,

12

Yes they do. BTW, the Beagle is also open to everyone. All of the files are freely available.

Gerald

13

True the Beagle is open and low cost but was not designed to have daughter cards and has a clumsy form factor. But the manufacturing costs I was quoted were out of the question for me to do a series of prototypes of my own OMAP3530 designs. The solution I’ve adopted is to simply buy a Gumstix Overo OMAP3503 board (they’ll have OMAP3530 versions soon) and design a card for that to plug into. The Overo is very small and has all the mechanically difficult parts on it and the cost for me to do a simple (no blind via’s etc.) board with only what I need to add is FAR FAR lower that tackling the 0.4mm blind via stuff.

I’m not at all reluctant to embrace the new wave of fine BGA’s, POP, VIP, etc. I’m not a Ludite- I’m just a cheap skate!

14

Gumstix, Cogent, and LogicPD are all viable solutions for the thrifty class!

Gerald

15

Gerald, Jason and people supporting BB,

First I would like to give my thank to you for such a great work you
are doing.

I like the idea of a beagleboard rev D or (how you wanna call it) with
CUS package and another interfaces.
I know the intention of beagleboard was to "bring your own
peripherials" or stuff like this, but that's not my case and I think
there are alot of people in the same situation.

What's my beagleboard dream?
1) 4 Layer PCB and not complicated packages. -> DIY board, Low Cost
and easy to assembly.
2) LCD TTL or LVDS, Touch Screen (SPI), AUDIO -> This is the standard
way to interact with a user.
3) Ethernet 10/100 -> Standard interface.
4) USB Host/OTG -> Standard interface.
5) Expansion port: UART, SPI, I2C, GPIO. -> Here, you interface the
circuit for your application.

I would love to participate/help in a project like this.

Taking about other suppliers:
I've used boards from LogicPD for example, but they don't have the
community that beagleboard has. So is not the same to just buy a board
from other supplier like LogicPD or Congent. For example last time I
bought a board from LogicPD was an IMX31, I had to wait like 8 month
to get the first BSP to test a linux kernel with some patches. Overo
from gumstix is supported for some ppl that is working with
beagleboard but:
- Has no SGX chip.
- I don't see any .SCH or gerber available in their site.
- Has the POP package
- I don't want/need a mini-mini-mini board like that.
- When you add the expansion board Is more expensive than the BB.
- Their website and documentation sux :slight_smile:

Leandro

16

Gerald, Jason, and people supporting BB,

First at all, I want to say thanks for such a great work you are
doing.

I know that the BB was born for the idea of: "bring your own
peripherials", but there're alot of people (included me) interested in
a beagleboard (new version) with CUS package and other peripherials.
This is a great opportunity to make a BIGGER BB community and bring
more people from the embedded world!!

What this new version should have?
- 4 Layers PCB -> cheap and easy assembly.
- LCD signals (TTL or LVDS), touch screen support (SPI) and AUDIO ->
This is what everybody (embedded) waits for a user interface.
- Ethernet and USB host/otg -> Standard communication interfaces.
- I/O Ports: UART, I2C, SPI, GPIO, external bus with some ADD and Data
(for parallel chips).
- SCH, BOM and Gerber available.

Why I don't buy Overo, Pandora or LogicPD 35xSOM?
- All use the POP package (I love to work with the last tech, but I
don't have the resources for this kind of thing).
This just kill the developer's dream of make their own PCB.
- Overo: Minimi board like stick, this is good for robotics maybe!?
(btw, their website sux).
In my app, size is not important, so why I will have to pay extra cost
to make for a mini-mi. (this extra cost in pcb/assembly/test are
higher than make a bigger 4-Layer-PCB).
- LogicPD linux support (I've used boards from them and you don't get
any support if you don't pay). If not take a look of how many post are
in their forum under OMAP35 or OMAP34.
- Pandora?: this is a cool device but I just want a board.....
- TI, Mistral, developments tools? nah too expensive I better hack a
Rev C when come out.

An one of the most important thing is that BB is the only one with
excelent documentation + an IRC channel with 85 developers/users where
to learn or ask for help.

Leandro

17

Some responses below...

Gerald, Jason and people supporting BB,

First I would like to give my thank to you for such a great work you
are doing.

I like the idea of a beagleboard rev D or (how you wanna call it) with
CUS package and another interfaces.
I know the intention of beagleboard was to "bring your own
peripherials" or stuff like this, but that's not my case and I think
there are alot of people in the same situation.

Some people honestly need as little as an OMAP3530+TPS65960+RAM
connected to the USB port of a PC. With that, you could run VNC,
native compilers, gdb, etc. to do code development, performance
testing, and many more things. Given the cost pressures in some
environments, I think we must be very careful in every dollar that we
add that *everyone* must pay, just so that a few might have some extra
features.

What's my beagleboard dream?
1) 4 Layer PCB and not complicated packages. -> DIY board, Low Cost
and easy to assembly.

I agree this is critical to make it where more other people can
actually build their own derivative boards. Right now, I believe we
are committed to delivering on the Rev C boards with minimal changes
and the USB host enabled. We've been talking about doing that for a
while, so I think something drastic would need to change for us not to
complete that plan prior to moving on to a CUS-package board.

2) LCD TTL or LVDS, Touch Screen (SPI), AUDIO -> This is the standard
way to interact with a user.

SPI already comes to the expansion header. Gerald has seriously
looked into bringing out LCD signals, but adding TTL or LVDS would
more significantly add to the cost. Beagle seeks to make things
simple by sticking with standard buses. If you really want to extend
the hardware, building your own board with the CUS package or using
one of the system-on-module-style boards seems to me to be the way to
go.

3) Ethernet 10/100 -> Standard interface.

Everybody pays the cost for what a subset of people want, even if it
is probably the majority. What are people willing to pay for this
change? Each of the board changes cost us roughly $20,000, plus there
are the per-board cost increases. I believe we need to keep the price
down and continue to drive the volume up.

Given that there is a reasonably affordable and functional solution
for software developers, can you give a better justification?

4) USB Host/OTG -> Standard interface.

Agreed on the need for a standard-A USB host. This is planned.

5) Expansion port: UART, SPI, I2C, GPIO. -> Here, you interface the
circuit for your application.

This exists on Beagle.

I would love to participate/help in a project like this.

Great. I think we are struggling to find a good way to collaborate on
hardware. Personally, I like hardware description languages, like
VHDL, but I don't know if that is viable for everyone involved.

Taking about other suppliers:
I've used boards from LogicPD for example, but they don't have the
community that beagleboard has. So is not the same to just buy a board
from other supplier like LogicPD or Congent. For example last time I
bought a board from LogicPD was an IMX31, I had to wait like 8 month
to get the first BSP to test a linux kernel with some patches.

Not sure what that was about. The BSP for LogicPD SOM should be
there, though I haven't tried it myself.

Overo
from gumstix is supported for some ppl that is working with
beagleboard but:
- Has no SGX chip.

Hopefully there will be the demand and required software support for
them to provide a version with the SGX.

- I don't see any .SCH or gerber available in their site.

I don't think they share gerbers, but I thought they did share
schematics. Really shouldn't be an issue if you are just creating the
board that connects up to the motherboard/SOM.

- Has the POP package

Again, not a problem if you are just using that the processor board
with your own system board. Makes the system board really easy to do
and put what you want on it.

- I don't want/need a mini-mini-mini board like that.

???

- When you add the expansion board Is more expensive than the BB.

True, but it has more RAM and gives you more flexibility.

- Their website and documentation sux :slight_smile:

Perhaps with some good feedback it will get better? Would the Beagle
Board System Reference Manual be a good guideline for other board
makers?

18

Oops!

Sorry, I didn't know about the delay when you post in google groups.
And I wrote two times the same (almost) ...

Leandro

19

The moderation delay is only for non-subscribers. Subscriptions are
meant to keep out spammers.

20

Maybe is time to use marketing tools like a survey to know how many
people could be
interested in this kind of board I've described.