| Author |
Message |
Guest
|
 Posted:
Thu Dec 15, 2005 5:15 pm Post subject:
Xilinx DCM Shuts down at 75degree centigrade |
|
|
Hi,
I am using a Xilinx Virtex Pro P20 FPGA.The system clock (155M) is
comes out of a DCM FX which is phase locked to input 19 M. Everything
works fine till 75 degree temperature. But as the temperature crosses
75 degree I see my logic behaving badly. I clearly suspect the DCM
because at 80 degree centrigrade the system clock (155M) stops coming
out of DCM. It is always low.
The virtex pro data sheet says the maximum temperature range to be
85-90 degree centrigrade.Is there is solution to this. There is good
heat sink on the FPGA. Has anyone seen this problem before ? Logic
utilization is around 65 %.
--Debashish |
|
| Back to top |
|
 |
PeteS
Guest
|
| Posted:
Thu Dec 15, 2005 5:15 pm Post subject:
Re: Xilinx DCM Shuts down at 75degree centigrade |
|
|
debashish.hota@gmail.com wrote:
| Quote: | Hi,
I am using a Xilinx Virtex Pro P20 FPGA.The system clock (155M) is
comes out of a DCM FX which is phase locked to input 19 M. Everything
works fine till 75 degree temperature. But as the temperature crosses
75 degree I see my logic behaving badly. I clearly suspect the DCM
because at 80 degree centrigrade the system clock (155M) stops coming
out of DCM. It is always low.
The virtex pro data sheet says the maximum temperature range to be
85-90 degree centrigrade.Is there is solution to this. There is good
heat sink on the FPGA. Has anyone seen this problem before ? Logic
utilization is around 65 %.
--Debashish
|
The temperature range is for die temperature usually. How and where are
you measuring the temperature?
With internal logic running at 155MHz (at least the clock section) a
10-15C die temperature rise is not unusual (for the internal sections,
power dissipation is generally directly proportional to toggle rates).
Alternaltively (and I admit to not having looked closely at the data
sheet), the temperature may be for ambient, but one has to be very
careful - the power dissipation ratings get derated above a certain
temperature, which may be 70C, but is more commonly device dependent.
When a manufacturer states their device will work in ambient
temperatures of (for instance) 85C, they mean 'work at ambient
temperatures of 85C with this [you have to work it out] maximum power
dissipation.
The ultimate issue is always die temperature.
Cheers
PeteS |
|
| Back to top |
|
|
Austin Lesea
Guest
|
| Posted:
Thu Dec 15, 2005 11:49 pm Post subject:
Re: Xilinx DCM Shuts down at 75degree centigrade |
|
|
Debashish,
All Xilinx specifications are based on die temperature (junction
temperature).
As long as the junction is below the specified temperature (85C for
commercial, 100C for industrial), all specifications are met.
It sounds like your junction temperature is much higher than what you
think it is.
How much power is your part dissipating?
What sort of heatsink is on it?
What is the airflow over the part (heatsink)?
Have you predicted the junction temperature from the above information
(and what is the predicted value)?
Austin |
|
| Back to top |
|
|
Symon
Guest
|
| Posted:
Fri Dec 16, 2005 12:14 am Post subject:
Re: Xilinx DCM Shuts down at 75degree centigrade |
|
|
A quick addition to the replies you've already received:-
Always route out on the PCB the connections to the on-chip temperature
diode. DXN, DXP. That way you know what's going on, directly on the die. I
suspect that the DCM isn't your problem, I've run parts near the temperature
specs and occasionally beyond (when the fan broke!) and didn't have any
problems with DCMs. Mind you, I do take great care of my Vccaux supply.
HTH, Syms. |
|
| Back to top |
|
|
Peter Alfke
Guest
|
| Posted:
Fri Dec 16, 2005 12:43 am Post subject:
Re: Xilinx DCM Shuts down at 75degree centigrade |
|
|
Ages ago, IC manufacturers used to specify a max ambient temperature
(70 degr.C for commercial parts). We soon found out that this makes no
sense for programmable parts. The FPGA manufacturer does not know what
the user implements inside the chip (what clock frequency, what
utilization) and how the chip is cooled. The only reasonable
specification that we can guarantee is the max junction temperature. So
we pioneered the Tj max 85 degr.C specification for all commercial
parts.
The user is the only person who then can evaluate whether the chip is
operating within these guaranteed limits, and can modify the design or
the cooling to bring it within specification.
It would be nice if the FPGA-manufacturer could guarantee operation at
a given ambient temperature, but that is inherently impossible,
especially for parts where the user can program the logic and choose
the clock speed.
As a rule of thumb, no package without a heatsink has a thermal
resistance (junction-to-ambient) below 10 degr.C per Watt. ( the very
biggest get down to 8 degr./W), and a heatsink is needed when power
exceeds a few watts.
Peter Alfke, Xilinx Applications |
|
| Back to top |
|
|
PeteS
Guest
|
| Posted:
Fri Dec 16, 2005 3:24 pm Post subject:
Re: Xilinx DCM Shuts down at 75degree centigrade |
|
|
Peter Alfke wrote:
| Quote: | Ages ago, IC manufacturers used to specify a max ambient temperature
(70 degr.C for commercial parts). We soon found out that this makes no
sense for programmable parts. The FPGA manufacturer does not know what
the user implements inside the chip (what clock frequency, what
utilization) and how the chip is cooled. The only reasonable
specification that we can guarantee is the max junction temperature. So
we pioneered the Tj max 85 degr.C specification for all commercial
parts.
The user is the only person who then can evaluate whether the chip is
operating within these guaranteed limits, and can modify the design or
the cooling to bring it within specification.
It would be nice if the FPGA-manufacturer could guarantee operation at
a given ambient temperature, but that is inherently impossible,
especially for parts where the user can program the logic and choose
the clock speed.
As a rule of thumb, no package without a heatsink has a thermal
resistance (junction-to-ambient) below 10 degr.C per Watt. ( the very
biggest get down to 8 degr./W), and a heatsink is needed when power
exceeds a few watts.
Peter Alfke, Xilinx Applications
|
For all the VLSI devices I have used in the last few years, the maximum
temperature ratings were die temperature, not ambient for the reasons
you specify. As VLSI power dissipation is very dependent on operating
mode, ambient ratings are pretty meaningless not only for programmable
parts, but for the majority of VLSI. Things become quite difficult when
the mfr has neglected to put a thermal sensing diode (really a
transistor) on the die somewhere.
I suggest the OP grab the rather useful power dissipation calculator
from the Xilinx website and then apply it using the Theta J-A for the
device he is using.
Cheers
PeteS |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in