I agree that a soldering problem doesn't seem likely, but you could try pulling off the FPGA and analyze the damage to see how they were soldered.

Xilinx sells a Spartan 3A/3AN Starter Kit using a 484-pin XC3S700 and they have a schematic on their website.

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

I may have to resort to this. I have a couple of 200ANs off the boards from previous attempts, having to do with solderpaste application.

ElectricEyE- Thanks for the schematic pointer. I will refer to it and check the wiring. These boards usually have way too many devices connected, but should be helpful anyway.

Possible explanation

Because my circuit is a 2-layer board, I routed some power and ground buses directly across unused IO pins deep inside the grid, thinking that it should not matter. At some later point I think I grounded the PUDC_B pin, enabling internal pullup resistors during configuration (I can't remember why I did this). This, of course, turnes the grounded pins and their respective pullups into little heaters.

If this is the case, the boards are actually usable since the heating is not terminal and the issue should go away after configuration, which should not take more than a few seconds.

Is the 0.8A current consumption consistent with 10-20 shorted IO pins? It seems a little high, but well within the range of possible. I am away from the circuit, will check this later today.

Edit: looking at the layout I counted 49 inner pins that are grounded by the power distribution bus. With a meter, 3.3V power grid draws approximately 730mA. That's about 15mA per pin, perfectly explaining the problem.

Next task: create a simple test circuit and configure the FPGA, which should fix the problem...

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

I created a simple test circuit that turns an LED on basted on another pin. That works fine.

Power consumption remains .8A! WTF...

I modified the UCF file with names for all the grounded pins, set them up as inputs in the 'top.v' file. Still drawing .8A ... Checked with FPGA explorer to make sure the IOBs are not optimised out - they are all there. Out of desperation, listed all pins shorted to the 3.3V bus as inputs as well, even though it makes less sense. Nope, still 0.8A, running hot.

So as far as I can tell, it works but runs in 'ungreen' mode, sucking power somewhere...

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut

It's a short between VCC and VINT. Bare board shows the 2 buses as isolated, fine. The soldered version with 1.2V only shows 1.2V on the 3.3V bus, and vice versa when connected to one supply only. With both connected, VINT reads 2.9V. How did I miss it before? Will go looking for the problem.

Interestingly, connecting just VCC (that is driving VINT at 3.3V) works fine! Probably reducing the lifespan of the chip substantially, but surprisingly resilitent.

PROBLEM FOUND

Stupid every day! Pin P11 was erroneously connected to VINT instead of VAUX/VCC. It sits in the VINT checkerboard pattern, so I see why I missed it. 3 XC3S700AN chips added to the future 'reballing' box.

Thank you for your help.

_________________
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut