XC3S400 and XC3S500E in PQ208

Avnet lists some parts as available :

XC3S400-4PQ208I XC3S400-4PQG208C XC3S400-4PQG208I XC3S400-5PQ208C XC3S400-5PQG208C

Real availability is another issue, which I happen to be interested too.

Josep Duran

Hi Uwe,

Thanks for bringing this to our attention. I've passed this on to our distribution manager and he's working to remedy the situation. Both the XC3S400 and the XC3S500E FPGAs are most definitely available. The working theory is that the PQ208 package option is not the highest running product combinations. Consequently, our distributors aren't stocking them. I fully understand the attraction of the QFP package, however.

-- Steve Knapp

"Steve Knapp (Xilinx Spartan-3 Generation FPGAs)" wrote: ...

As explained in the XC3SAN thread, the availability of big PQ208 parts may not lead itself to big demand on the "many-pin" PQ parts. But the ease of prototyping may lead to more design wins and so for demand for "not-so-many" PQ and for BGA parts...

I got a quote for a small quantity XC3S400-5PQ208C from Avnet a while ago and they said the devices could be delivered from stock...

Hi Uwe, I think this is one of those few times where I'm tending to disagree with you. IMHO, the SI properties of the PQ208 package are so terrible that any prototype made using a PQ208 containing a die capable of sub-ns rise times is not going to accurately reflect what would happen when you build it with a decent FBGA package. Sure the internal logic will work ok, but that's what simulators are good at. It might be time to bite the bullet and invest in a toaster oven! I know Philip posted not too long ago about his set up, perhaps it might be of interest to you. The other choice that I can think of is to use an interposer to connect the FPGA to the board. I read recently about magic elastomer materials to connect BGA to PCBs with excellent SI properties. Best regards, Syms.

Is that the Samtec stuff? Wasn't in their catalog last time I looked.

Hi Tim, I've looked for that after Teraspeed referenced it in an article and, like you, not found it yet. I was looking at something else the other day that had an elastomer with tiny wires in it like this:-

| | | | | | | | | O O O O O O O O O | | | | | | | | | O O O O O O O O O | | | | | | | | | O O O O O O O O O | | | | | | | | |

So, vertical lumpy wires. When you used it the wires deformed :-

/ / / / / / / / O O O O O O O O \ \ \ \ \ \ \ \ O O O O O O O O / / / / / / / / O O O O O O O O \ \ \ \ \ \ \ \

A bit more higgledy-piggley than that! Each PCB pad/ BGA ball used several of these wires to make the contact, like a zebra strip in LCD connectors. Sadly, I can't remember the link. Anyone else see it?

Cheers, Syms.

I'm pretty sure that general design has been around for 10+ years, and my memory is that Samtec had an unsuccessful crack at productizing it some time ago. It's about time they tried again!

-- Tim

Good to know !! I'll forward this to the purchasing deparment.

I'm afraid -5PQ208 is not ROHS compliant. (Not that it matters for in- house prototipes)

Regards.

Symon wrote: ...

Not all applications are bound by outgoing traffic. If you have many input signals, plenty available Pin are welcome, and SI issues are not that big issue...

SI covers not just the I/O but also the Vcc/Gnds, as the core switching currents need to find a low impedance.

Another wrinkle in the "drop the die into a PQ208" idea, is that many now use flip chip spread bonding (bond pads over the whole die)

- works very well in BGA, but rather kills using a perihperal bond package. Even MLF only gets low Z on the Gnd bondings, and FPGA's have many supplies.

Having said that, there is a market for easy to deploy FPGAs (low PCB layer count ), and not just lab prototypes. The PC and Hard Drive markets are good examples of just how hard-nosed some high volume customers are about layer counts.

If the fpga vendors keep driving up the PCB layer counts, and MFG machinery costs, then they will be doomed to grow at a rate that is below the fabless industry average. ( Oh, right, that's already happening....)

-jg

Jim Granville wrote: ...

The National Ethernet PHYS DP83847 has an interesting LLP package with "split" exposed pad. So some package creativity not going the obvious BGA way.

It seem however newer National PHYS use "normal" packaging

Here it is:-

'pariposer'

Cheers, Syms.