What is the resistance of a big FPGA for VCCINT (unpowered)

Do you still have a Stratix-II device unassembled? If you've broken the vacuum seal pack, you'd have to re-bake anyway. You could check the resistance on the unassembled parts you have. I'd tend to worry only if the lead resistance showed up as 1.2 ohms; you're concerned about solder shorts under the BGA. Solder shorts between power and ground (or between powers) is a problem. If you can bring up the board without configuring the FPGA, a boundary scan could tell you if there are any signals stuck at VCC or Ground. The point where you might damage the chip is if the signal was shorted to a rail and driven hard for an extended length of time.

How would you gain confidence in any board that has an expensive part on it? If you aren't set up for boundary-scan or manufacturing defect analysis, you flick the power on and off and check for excessive warmth. You flick the power on... and off and check for excessive warmth. If you can check the current draw of the board during the power-supply ramp, you only need milliseconds to capture the trace on the digital scope and make sure it's around what you expect.

I believe this is normal for some of the devices when you measure with a DMM. There was a thread on this topic in the past:

However, I have never encountered this behaviour in the devices I've used so far...

/Mikhail

Marc ,

My temptation is to tell you it is bad, but I will resist the temptation!

Seriously, a ohmmeter check may be dangerous: many use a 9 volt battery.

Testing this way immediately violates the warranty (look at the absolute maximum ratings in the data sheet).

9 volts will kill a 1 or 1.2 volt Vcc part!

I suggest that the very low core voltages, combined with the very large static leakage at 90 nanometers, means you may no longer use an ohmmeter to tell you anything.

Instead, a 1.2 or 1.0 volt current limited power supply is required.

Austin

My ohmmeter walked away at work, besides - I'd need an ohmmeter to check my ohmmeter. Don't they have high impedance outputs even when the resistance measurement goes to the 0-20 ohm range? As long as the measurement - through it's drive impedance - provided less than 1.2V through the 1.2 ohms (1 amp?!), measuring the 1.2V rail should be fine. Negative 1.2V could be a different matter.

- John_H

"austin" wrote

Ouch! Hopefully it was not the case this time :)

Good idea, I will make one for the next boards.

Thanks to all.

BTW I powered the boards. They did not blowup and the power supplies are at their normal levels. :)

Marc

|> Seriously, a ohmmeter check may be dangerous: many use a 9 volt battery.

...but only as a supply voltage, not for testing.

Typically, the test voltage is below 2V on modern handheld instruments. My

4.5-digit Metex has 1.25V in all ranges, maximum current is about 0.5mA. The diode tester runs with 3V, the max. current is about 1.3mA. I doubt that these currents and voltages will in any way affect the chip, especially when the operating (and leakage) currents are larger by a few orders.

--
         Georg Acher, acher@in.tum.de
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias

Georg.

Good to know!

Austin

Marc,

1.2 Ohms is fairly normal for this device. The EP2S90 is about 2 Ohms. You should get worried if it goes below 1 Ohm. Had one customer see 0.2 Ohm - turned out to be a short beween VCCint and PLL ground.

Best regards,

Ben

"Ben Twijnstra" wrote

Thanks Ben,

In fact I did found shorts on another board (cf link below). They were all between pads of the LLM21 decoupling caps and around 0.2-0.3 ohms. BTW I'm somewhat upset with this, considering that I got the boards inspected visually and by x-rays at the assembly shop. :(

Marc

Hi Marc,

Be thankful that you can actually see the short, and it's not under a part :(

Because semi parts are active, their "effective" resistance is a function of the applied "supply" voltage. To look for fab shorts with semi parts loaded you need a very low voltage ohm meter, sub 0.2v so that none of the forward diode junctions will actively conduct, and even then leakage currents can be a problem.

You will probably have to build this from something that does a very slow sweep from 0-0.5v (to allow caps to charge and equalize) and plots the on board voltage sweep (sensed by a high impedance amplifier) vs a known good gold board sweep.

current limited to a few hundred ma, and plot voltage/current during sweep

I once found a shorted part when I observed that the voltage rail in question was about 0.3R in the general area but 0.2R across the power pins of one particular part (QS3384, I think it was). I pulled it off and the problem went away.

I was surprised at the time that I was able to distinguish proximity to the short with a plain old Fluke 10. With a milliohmmeter it might be even easier.

--
Ben Jackson AD7GD

http://www.ben.com/

No.

So every time I check for a connection (sub-ohm) you think I'm shorting the battery?

I wrote:

John_H wrote:

No. There's a lot of room between "shorting the battery" and "high impedance".

I suppose it comes down to "high impedance" not having a clear quantitative definition.