Re: [linux-audio-dev] Re: [Alsa-devel] Firewire Audio Card Support

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: Re: [linux-audio-dev] Re: [Alsa-devel] Firewire Audio Card Support
From: Simon Jenkins (sjenkins_AT_blueyonder.co.uk)
Date: Sat Nov 20 2004 - 02:19:44 EET


Jussi Laako wrote:

>On Fri, 2004-11-19 at 21:03 +0000, Simon Jenkins wrote:
>
>
>
>>Pessimistic. I had to specify a single 6-layer board, 5 days delivery
>>and 100 component placements before I got an online quote approaching
>> 500Euro. And the price drops *fast* when you want a few more boards,
>>a bit slower. Its down to 162Euro per board if you want 5 boards in 10
>>days. Its down to 55Euro per board if you want 10 4-layer boards in 15
>>days. That's manufactured and *assembled* (but not including the
>>component cost). These are actual quotes from an actual company who I've
>>actually used.
>>
>>
>
>Then it's way more cheaper at UK than it's here in Finland. What I've
>been ordering the starting cost (for making the films, etc) is around
>250¤ and the board itself from 100 to 250¤ for series of 1 board
>depending on the size. Minimum line width and isolation 4-6 mils*
>(because of high density chips). For two boards the cost of single board
>about halves. This for 4-layer board, chemical tin, two silkscreens. For
>hidden/buried vias the price goes higher because of laser drilling.
>Proto boards may have to be assembled somewhere, as soldering BGA chips
>by hand is a bit difficult.. ;)
>
My quotes included assembly. (But not BGA, for which they charge extra).

Min line width & isolation is 0.15mm, ie 5.91mils, ie just meets your
4-6 mil
requirement.

How it seems to work is: They paste everybody's boards together into a
super-
board, manufacture it, and then cut it up afterwards and send you your
bit. So
you can't have any weird requirements. The "5 days", "10 days" is counted
from the scheduled day that they are starting their next batch, not from the
actual day of placing the order. (They start a 4-layer batch twice per week,
for example).

>>>would be possible for about 10000¤ if not counting price for the work.
>>>
>>>
>
>
>
>>Either you typed an extra zero here, or you're factoring in 9000+Euros for
>>development software and hardware, or you're expecting 10 generations of
>>prototype before it works, or something.
>>
>>
>
>It contains the development software, IP blocks and cost of prototype
>boards.
>You can count 2000¤ for license of some decent PCB CAD, 500-1000¤ for
>the FPGA development environment (incl. VHDL compiler and JTAG
>programmer), about 2000¤ for PCI-Express IP block. About 3 rounds of
>prototype PCBs (3 * 500) and sack of components for the board (I think
>Virtex or Stratix is around á 50¤ for <10). You may also need USB,
>Firewire or PCI-Ex bus analyzer and a logic analyzer, but after that the
>cost goes over my figure. So I think 10k is a bit high, would be a safe
>bet without too much risk for going over that in reality.
>
>OK, the cost is around IP blocks + prototypes ~ 3700¤ if your employer
>allows you to use company's tools for the project. And in this case it
>would be around 1700 if you are going to use 32-bit PCI and use some
>ready bridge ASIC. But I think it would be stupid to start developing a
>32-bit PCI card at this point when the 32-bit PCI bus is going away and
>PCI-Express rather quickly replacing it.
>
>Because something can go wrong you should have at least 20-25% margin. I
>think there should be a reasonable margin for possible problems if that
>kind of project is started, because what really pisses off is when you
>run out of money just before hitting the goal...
>
>Are we talking in the same context here?
>
You were talking about a slightly more difficult and expensive thing than I
was. I was thinking about:

1. A DSP implementation, not FPGA
2. A firewire interface, not PCI
3. Creating a "reference implementation" not a finished product.

Compiler cost would be a factor in choice of DSP.

The prototypes would be a platform for developing an open protocol, open
firmware (by which I meant the s/w in the DSP, not FPGA IP) and open
drivers.

Provided you stick to 4 layers a slightly misrouted first prototype can
usually be hand-modified to work. Ugly, but useable as a development
platform. So you could go straight to 5 boards (or however many
developers there are) on the first run, get them working and get on
with developing the software.

Once there was functioning h/w and s/w... well, things could go in any
number of directions.

Simon Jenkins


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Sat Nov 20 2004 - 01:16:31 EET