1. Home
  2. Field-Programmable Gate Arrays
  3. Cyclone device misteriously overheats

Cyclone device misteriously overheats

Hello,

This problem has been getting on my nerves for quite some time now. First of all I'll let you know that I'm an inexperienced engineer - it's my first year in the field, so please bear with my inherent foolishness where applicable. Now to the point: I'm working on a design which uses the Altera Cyclone as a comm dispatcher/bridge/multiplexer between 3 Microchip PIC18F452, an FTDI 245 USB_to_parallel IC and an Cypress SL811HST USB host controller. The problem is the Cyclone overheats constantly (to about

85 C) from power-up. I have excluded the possibility of outputs colliding with other outputs by only soldering the FPGA to an empty PCB, suppling it with power and downloading the configuration to it and the problem is the same - it overheats whether configured or not. There is no difference in temperature between the configured and unconfigured state. I have tried all the obvious tests including short-circuit-to-ground-or-power testing of each IO pin and all are well. Also all the power pins are properly conected. One thing to know is that the PCB is designed in house by us so there may be errors there as well as this is the first version of the design. Another important observation is that the design was initialy developed on a breadboard using a Cyclone mini dev board purchased from
formatting link
and connecting the rest by strap-wire. It worked fine without overheating with the same configuration data for the Cyclone.

Please suggest any test path I should try to figure out this mess as deadlines ar pressing me quite strongly.

Thanks a lot!

Reply to
Alex Somesan
Loading thread data ...

Have you checked the power supplies with a scope? What frequency does it run at? Does it work despite its temperature?

Jeroen

Reply to
Jeroen

Yes, I've checked the powe supply (3.3 V from a LM317). Look OK on the scope. Even tried powering it from de old breadboard power supply which proved reliable.

It runs at 20 MHz from an oscillator device.

Yes, it works despite the heat but ocasionaly freezes up (I suspect it messes up configuration RAM contents because a reconfiguration, even without cutting power, gets it working again).

Reply to
Alex Somesan

Have you also checked the core supply of 1V5? How much ripple? Is the device properly bypassed? Can the supply deliver enough current? Especially on power up the FPGA needs more current. I don't which Cyclone you're using, but an 1C20 needs up to 1.2A at power up.

Are there no floating inputs? At certain input voltages the logic can't decide on a level and it will oscillate heavily at several hunderds of MHz. Are the inputs in your logic properly synchronised? Inputs that go into a metastable state may oscillate, and the oscillation will propagate through the device.

Jeroen

Reply to
Jeroen

"Alex Somesan" wrote in news: snipped-for-privacy@c13g2000cwb.googlegroups.com:

There is a 1.5V supply as well?

--
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com
Reply to
Al Clark

the

which

it

even

device

on

using,

can't

of MHz.

into a

through

The power supply is the same as the one used on the breadboard prototype and that one worked fine so I presume it does deliver enough current. Measured current for the whole board with all the devices mounted and the FPGA configured is around 300mA. The Cyclone is a EP1C3T144C8 which in the datasheet is speced at around 500mA at powerup. The power supply can deliver 1.5A considering the datasheet of the LM317. I scoped the power line and found around 20mV of ripple both over and under the 3V average wich sometimes fades to zero ripple. I can't recall of any 1V5 power for the core on the device pinout. The only 1V5 requirement is for the analogue part of the PLL. So as far as I know there is no separate 1V5 supply for the core. The PLL 1V5 supply is OK.

I don't realy understand what you mean by floating imputs?. Almost all of the device IO pins are used on the design so there are around 6 pins left unconected on the Cyclone. What sould I do with these? Shoud they be tristated from the design software?

As far as input oscillation I don't know if it is directly related to the problems as the heating occurs even when the device is the only chip soldered to the board and both in unconfigured or configured state.

Reply to
Alex Somesan

So you have all pins called VCCint tied to 3V3??? Then I guess it's no wonder the device gets hot, you're blowing it up!

Jeroen

Reply to
Jeroen

Stupid me! You were right! I cut the PCB traces of the VCCINT pins off the 3V3 rail and straped them to 1V5 - works perfectly! No heat, no config mess-up. Thanks for the advice. I guess it was obvious to the experinces guys out there. I slaped myself a few times after carefully looking over the datasheet. I guess now I have to clean up my reputation :). Thanks for helping!

Alex.

Reply to
Alex Somesan

Stupid me! You were right! I cut the PCB traces of the VCCINT pins off the 3V3 rail and straped them to 1V5 - works perfectly! No heat, no config mess-up. Thanks for the advice. I guess it was obvious to the experinces guys out there. I slaped myself a few times after carefully looking over the datasheet. I guess now I have to clean up my reputation :). Thanks for helping!

Alex.

Reply to
Alex Somesan

Alex Somesan wrote: (snip)

And now we all know how tough Cyclone is with excess VccINT.

-- glen

Reply to
glen herrmannsfeldt

Stupid me! You were right! I cut the PCB traces of the VCCINT pins off the 3V3 rail and straped them to 1V5 - works perfectly! No heat, no config mess-up. Thanks for the advice. I guess it was obvious to the experinces guys out there. I slaped myself a few times after carefully looking over the datasheet. I guess now I have to clean up my reputation :). Thanks for helping!

Alex.

Reply to
Alex Somesan

Stupid me! You were right! I cut the PCB traces of the VCCINT pins off the 3V3 rail and straped them to 1V5 - works perfectly! No heat, no config mess-up. Thanks for the advice. I guess it was obvious to the experinces guys out there. I slaped myself a few times after carefully looking over the datasheet. I guess now I have to clean up my reputation :). Thanks for helping!

Alex.

Reply to
Alex Somesan

Stupid me! You were right! I cut the PCB traces of the VCCINT pins off the 3V3 rail and straped them to 1V5 - works perfectly! No heat, no config mess-up. Thanks for the advice. I guess it was obvious to the experinces guys out there. I slaped myself a few times after carefully looking over the datasheet. I guess now I have to clean up my reputation :). Thanks for helping!

Alex.

Reply to
Alex Somesan

Stupid me! You were right! Tied VCCINT to 1V5 and all works well. No heat at all. I slaped myself a few times after going over the datasheet again, more carefuly. Anyway, stupid rookie mistake. I must clean my reputation now :). Thanks for helping.

Alex.

Reply to
Alex Somesan

glen herrmannsfeldt wrote in news:cpnrgj$qv$1 @gnus01.u.washington.edu:

Alex,

Don't let us "old timers" give you too much crap. We've all made this mistake in one form or another. It called experience.

--
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com
Reply to
Al Clark

Good to hear it's working now. Luckily it's only a cheap easily replacable Cyclone and not a Stratix housed in a 1508 pin FBGA package that's worth several thousand $ :)

Jeroen

Reply to
Jeroen

Spank yourself with a soldering iron while reciting the complete Cyclone manual

5 times (including appendix) and you'll receive an absolution from the Almighty FPGA...
Reply to
Jan De Ceuster

Don't worry about it too much-- AVNET shipped us a rev. of their Xilinx V21000 eval boards w/ a similar error. (turns out a 0 ohm resistor was mistaken for a DNP in their 1.5V regulator network) They've since fixed the problem, and have been very nice about replacing the boards, but from what I can tell, they're a major distributor, so it happens to EVERYONE.

--Josh

Reply to
Josh Model

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.