How much does getting high-end PCBs custom made cost?

[Mark R]

A dude I met at work showed me some circuits he had been designing and fabricating. They looked quite high-tech, and I should have asked who he had found to make them, because they were like nothing I had ever seen, and they're certainly not something that express PCB would do.

Apparently, the placement and routing was so difficult, that off-the-shelf software couldn't do it, and they had to develop in-house software to do the footprint placement.

Brief specs:
160x100 mm
8 layers - Kapton substrate
Blind & buried vias
Min track width/spacing: 50um
Min pad spacing: 30um

The final assembly was interesting, as the board was to be populated with unpackaged ASIC dies, and they had to do the wire-bonding of the dies to the substrate in house - something like 8000 wire bonds per board.

So, anyone in the manufacturing business have any idea how much this sort of construction costs? I imagine that it's quite pricey.

 

[William Gaatjes]

That is indeed quite the price.
But the fact that they had to inhouse do wirebonding and develop their own design/placement rules for the software is interesting.
Especially the software. Only the design rule check rules describe the limitations of the manufacturing process : The creation of the printed circuit board, the size of the components and the placement of the components. You can even use it to impose limitations on component placement because of the air flow limitations during SMT soldering. Footprints that do not exists must be made by specification. That takes time.

 

[Jeff7]

Not sure of pricing, but maybe check with the Sales department at Sierra Circuits. That might at least be a point in a helpful direction. 🙂
They don't specifically list Kapton as a material, but they do work with polyimide film, and they look to have quite a lot of capability as far as more advanced boards go.

I've only used them at work for 2-layer FR-4 boards, not anything as high-end as what you're working on, but their quality there looks to be top-notch. And they actually read the construction guidelines sent along with the Gerber files, which is more than I can say for a several other board manufacturers I've worked with.

 

[polarmystery]

A lot. Thousands.

Not in manufacturing, but boards we send out for quote can cost upwards of 50k or more (our boards are typically >12 layers though)

 

[Modelworks]

If altium couldn't design it then that is one hell of a design as I haven't seen anything it can't handle.

Manufacturers are plentiful and cost will depend on how many boards are being done for the final run. Amitron can do 14 layer boards and their prices are reasonable for the amount of work involved.
http://www.amitroncorp.com/

A 10 layer board 6" x 4" with a 14 day return time cost $70 per board if you get 10 boards, so $700 minimum order. Order 100 of the same board and cost drops to $22 a board.

 

[Mark R]

I think the issue wasn't so much the routing, but placement of the components - which had to be precisely aligned and rotated to fit the pads on a special silicon wafer. (The ASICs were mounted on the PCB, half of the ASIC wirebonds were to the PCB, but the PCB was then glued to a silicon wafer, and the ASICs then wire bonded direct to the wafer). The angles between the ASICs and the wafer mountings had to be precisely controlled during placement, which off-the-shelf software could not do.

 

[soccerballtux]

what sort of technologies are you guys working on that use boards like this?

 

[Mark R]

High energy physics.

I found a pic of the boards I saw.

Check out the wire bonds from the ASICs to the silicon wafers. Yes, there are several hundred individual wire bonds to each ASIC.

 

[William Gaatjes]

High energy physics.

I found a pic of the boards I saw.

Check out the wire bonds from the ASICs to the silicon wafers. Yes, there are several hundred individual wire bonds to each ASIC.

Amazing ! Nice pcb. :thumbsup:
I assume the center is where the particles are detected ?

Why do they not use some kind of flip chip approach ?
Is that not available at that size or possible ?
I assume it is of no use to be using flip chip technique because those detection centers also are wire bonded ?
And cannot be soldered directly on the pcb to be connected to the asic with copper traces and solderpads ?