Building a board with Spartan 3 FPGA.

Hi, I am trying to build a very simple Spartan3 - based FPGA board. The idea is to have FPGA + Flash PROM & Jtag header and I/Os. Nothing else.

I am running into a problem: Using my DMM, VCCAUX & VCCO seem to be shorted to GND. At this point I am absolutely sure that they are shorted through the Xilinx chip.

Does Xilinx chip need to be programmed with some kind of default configuration, before it can even be powered on? In other words, should there be some kind of *safety configuration* programmed into the Flash PROM, prior to applying power to Xilinx FPGA?

If you have any references, links, suggestions, guidelines for putting a single FPGA chip on a board with nothing other than the basics for programming it, I would be grateful if you point me to them.

I copied my design from Digilent Spartan3 Starter Kit, getting rid of all the *starter kit* peripherals there and just leaving the essentials: JTAG header, Flash PROM, FPGA, expansion I/Os

Thanks!

Reply to
Telenochek
Loading thread data ...

no. DMM is not good to measure but before config the VCCxx pins should take some normal current, eg not short circuit.

doing a s3 proto is rather simple, just connect the GND, VCCxx, JTAG and pull PROG_B thats it

Antti

Reply to
Antti

If VccAux appears to be grounded in a powered board, your problems will never be with the device.

If VccAux never comes up, there is no configuration to do anything negative to your board.

A newly-powered Spartan3 comes up in an innocuous state until the device is programmed at which point it does what it's programmed to do.

Check your soldering. Check your gerbers. I've seen problems with both despite the best attempts of engineers and design software.

Reply to
John_H

----------------------------------- doing a s3 proto is rather simple, just connect the GND, VCCxx, JTAG and pull PROG_B thats it

------------------------------------ That's exactly whats on my board + Flash PROM.

On a fully working Spartan3 board from Digilent, when the board is not powered up: the VCCO and VCCAUX appear to be shorted to GND (using DMM). Of course, this could be due to low impedance between supplies and GND, due to high Iq through the FPGA. But the board is not powered at all.

How do I debug the shorts if I can't even test for them?

Reply to
Telenochek
1) Before you assemble, check for supply and ground shorts. Too late? 2) Do a visual inspection of soldered TQFP leads with a stereo microscope If you're doing a BGA instead, it's a much different issue. Xray inspection of assembled boards would be needed to troubleshoot BGA shorts; this inspection is available from most professional assemblers.
Reply to
John_H

formatting link

EASY - the board as above DOES work, I used it for SPI indirect programming testing

Antti

Reply to
Antti

Its not too late, I have multiple boards for this. Without FPGA (TQFP144) chip on board, there aren't any shorts. I have checked connections with stereo microscope, no solder bridges.

I think the next thing I will try is to assemble everything, but the FPGA and verify that everything without FPGA is fine. Trouble is that FPGA will be harder to solder with all the stuff on the board already. But doable.

Thanks!

Reply to
Telenochek

EASY - the board as above DOES work, I used it for SPI indirect programming testing

----------------------------- Very nice but where is the schematic for it?

Thanks!

Reply to
Telenochek

Have you verified where pin 1 is? I got tripped up a while ago with Xilinx using a center-of-side pin 1 on CPLD packages but a corner pin 1 on FPGAs. Really dumb mistake, but pretty easy to do.

Jon

Reply to
Jon Elson

When you say "shorted", how many ohms (without power applied) are you seeing between VCC* and GND? Is it a dead short (i.e. a couple of ohms max), or is it in the ~15 ohms or more category? If the latter, it is probably not shorted, and may just be the unpowered resistance of the power system (load and/or regulator). If it is a dead short, and it does not exist without the devices on the board, then there are a few options. Solder bridges or faulty components, or you may have a mistake in the pinout of one of the devices, and have a device ground pin tied to the power plane, or vise-versa.

Andy

Reply to
Andy

I second this as a necessary check. The file I looked to for Xilinx CPLD orientation guidance (last time I specified orientation for a tech) had superfluous information. The text was shown along with a circle (or two). There was a specific cirle locating the pin 1 location that wasn't pointed out as *the* unique identifying feature and was not readily visible on the part without the right shadowing.

Reply to
John_H

No, they really are shorted, 1.2ohms.

I will try verifying the pins again. Then soldering everything except the FPGA & Flash PROM to see what the state of the pins is.

It would be very nice to compare to a really simple board/schematic however.

Pavel

Reply to
Telenochek

It's probably nothing to worry about. Here are a couple of excerpts from a similar discussion on the SI mailing list:

The reason is because of leakage current in transistors. As process technology progresses (180nm -> 130nm -> 90nm ...), the leakage current in a transistor gets bigger. That is, a transistor, even when it is "off", will have some current flowing through it. Also, you can pack more transistors onto a chip. Therefore, you end up with tens to hundreds of millions of leaky transistors, which leads to a low resistance when measured with an Ohm-meter.

Because IC components are nonlinear, how much resistance you measure strongly depends on the ohmmeter's applied voltage. (Most ohmmeters don't apply a voltage, so don't expect a fixed value from your meter.)

When measuring semiconductor junctions with an ohmmeter, the voltage developed (i.e., resistance) across a DUT will vary with the applied current, therefore, a different reading will be indicated depending on the scale (current source value) used.

/Mikhail

Reply to
MM

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.