FPGARelated.com
Forums

ZYNQ temperature

Started by John Larkin May 11, 2015
On 5/12/2015 9:42 PM, John Larkin wrote:
> > It's probably shutting down at 125C, without our specifically > programming any temperature. > > Extensive searching, by us and by Avnet, finds no fan that matches the > hole spacing on the MicroZed board. So we'll fab a little aluminum > adapter plate and use a standard fan. With a pin-fin heat sink glued > to the 7020 FPGA, and the fan blowing down on that, we can run at 100C > ambient.
Reminds me of an array processor I worked on in the early 80's. It had ECL gate arrays in ceramic PGA packages with a heat sink on each chip and a specially designed plenum which slid over each one to direct air across the heat sink. This machine was as fast as a CRAY-1 and only a few years later. -- Rick
On Tuesday, May 12, 2015 at 9:42:58 PM UTC-4, John Larkin wrote:
> On Tue, 12 May 2015 13:57:47 -0700 (PDT), > lasselangwadtchristensen@gmail.com wrote: > > >Den mandag den 11. maj 2015 kl. 19.59.43 UTC+2 skrev John Larkin: > >> Does anyone know if the ZYNQ chips have an internal high-temperature > >> shutdown? They are behaving like they do. > >> > > > >looks like you have to enable it (it may be default) and you have to load the PL > > > >30.3.6 Critical Over-temperature Alarm > >Note: This feature sends an interrupt status to the PS and causes an automatic shutdown feature for > >the PL side of the Zynq-7000 device if enabled. Th e PL shutdown is enabled via the bitstream and the > >PL will only come out of power-down if th e over-temperature alarm goes inactive or a > >reconfiguration occurs. > >The on-chip temperature measurement is used for critical temperature warnings. The default over > >temperature threshold is 125�C. This threshold is used when the contents of the OT Upper Alarm > >register (listed in UG480) have not been configured. When the die temperature exceeds the > >threshold set in the XADC's Control register, the ov er-temperature alarm (OT) becomes active. The OT > >signal resets when the die temperature has fallen below set threshold. > >The OT alarm can also be used to automatically power down the PL upon activation. The OT alarm can > >be disabled by writing a 1 to the OT bit in the XADC's Configuration register. > >Note: these registers are in the XADC and are accessible using the DRP. > > > >-Lasse > > It's probably shutting down at 125C, without our specifically > programming any temperature. > > Extensive searching, by us and by Avnet, finds no fan that matches the > hole spacing on the MicroZed board. So we'll fab a little aluminum > adapter plate and use a standard fan. With a pin-fin heat sink glued > to the 7020 FPGA, and the fan blowing down on that, we can run at 100C > ambient. > > > > -- > > John Larkin Highland Technology, Inc > picosecond timing laser drivers and controllers > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com
The MicroZed has a -I part on it, right? Those parts are spec'd at a max junction temp of 100 C. You need the Expanded temperature grade parts (Q) to get the 125 C junction temps.
On Wed, 13 May 2015 05:37:17 -0700 (PDT), kkoorndyk
<kris.koorndyk@gmail.com> wrote:

>On Tuesday, May 12, 2015 at 9:42:58 PM UTC-4, John Larkin wrote: >> On Tue, 12 May 2015 13:57:47 -0700 (PDT), >> lasselangwadtchristensen@gmail.com wrote: >> >> >Den mandag den 11. maj 2015 kl. 19.59.43 UTC+2 skrev John Larkin: >> >> Does anyone know if the ZYNQ chips have an internal high-temperature >> >> shutdown? They are behaving like they do. >> >> >> > >> >looks like you have to enable it (it may be default) and you have to load the PL >> > >> >30.3.6 Critical Over-temperature Alarm >> >Note: This feature sends an interrupt status to the PS and causes an automatic shutdown feature for >> >the PL side of the Zynq-7000 device if enabled. Th e PL shutdown is enabled via the bitstream and the >> >PL will only come out of power-down if th e over-temperature alarm goes inactive or a >> >reconfiguration occurs. >> >The on-chip temperature measurement is used for critical temperature warnings. The default over >> >temperature threshold is 125&#4294967295;C. This threshold is used when the contents of the OT Upper Alarm >> >register (listed in UG480) have not been configured. When the die temperature exceeds the >> >threshold set in the XADC's Control register, the ov er-temperature alarm (OT) becomes active. The OT >> >signal resets when the die temperature has fallen below set threshold. >> >The OT alarm can also be used to automatically power down the PL upon activation. The OT alarm can >> >be disabled by writing a 1 to the OT bit in the XADC's Configuration register. >> >Note: these registers are in the XADC and are accessible using the DRP. >> > >> >-Lasse >> >> It's probably shutting down at 125C, without our specifically >> programming any temperature. >> >> Extensive searching, by us and by Avnet, finds no fan that matches the >> hole spacing on the MicroZed board. So we'll fab a little aluminum >> adapter plate and use a standard fan. With a pin-fin heat sink glued >> to the 7020 FPGA, and the fan blowing down on that, we can run at 100C >> ambient. >> >> >> >> -- >> >> John Larkin Highland Technology, Inc >> picosecond timing laser drivers and controllers >> >> jlarkin att highlandtechnology dott com >> http://www.highlandtechnology.com > >The MicroZed has a -I part on it, right? Those parts are spec'd at a max junction temp of 100 C. You need the Expanded temperature grade parts (Q) to get the 125 C junction temps.
It's 99% likely that all the chips come off the same wafer. The faster ones may get binned as the high-temp versions. The real issue is timing margins, and our fastest clock is only 128 MHz. (We buy two different Altera parts, one with twice the logic cells and RAM and price and stuff. They are actually identical, run the same bitstreams compiled for either part.) Here's the fan, with its adapter plate. The box runs fine at 100C ambient. Next time we recompile the design, in a couple of months maybe, we'll bring out the chip temperature sensor to see how hot it actually is inside there. https://dl.dropboxusercontent.com/u/53724080/Thermal/Uzed_Fan_Top.JPG https://dl.dropboxusercontent.com/u/53724080/Thermal/Uzed_Fan_Side.JPG It's really insane that we should have to do this. I once built and calibrated a ring oscillator to measure FPGA chip temperature. That's another story. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Den onsdag den 13. maj 2015 kl. 22.18.29 UTC+2 skrev John Larkin:
> On Wed, 13 May 2015 05:37:17 -0700 (PDT), kkoorndyk > <kris.koorndyk@gmail.com> wrote: > > >On Tuesday, May 12, 2015 at 9:42:58 PM UTC-4, John Larkin wrote: > >> On Tue, 12 May 2015 13:57:47 -0700 (PDT), > >> lasselangwadtchristensen@gmail.com wrote: > >> > >> >Den mandag den 11. maj 2015 kl. 19.59.43 UTC+2 skrev John Larkin: > >> >> Does anyone know if the ZYNQ chips have an internal high-temperature > >> >> shutdown? They are behaving like they do. > >> >> > >> > > >> >looks like you have to enable it (it may be default) and you have to load the PL > >> > > >> >30.3.6 Critical Over-temperature Alarm > >> >Note: This feature sends an interrupt status to the PS and causes an automatic shutdown feature for > >> >the PL side of the Zynq-7000 device if enabled. Th e PL shutdown is enabled via the bitstream and the > >> >PL will only come out of power-down if th e over-temperature alarm goes inactive or a > >> >reconfiguration occurs. > >> >The on-chip temperature measurement is used for critical temperature warnings. The default over > >> >temperature threshold is 125&#4294967295;C. This threshold is used when the contents of the OT Upper Alarm > >> >register (listed in UG480) have not been configured. When the die temperature exceeds the > >> >threshold set in the XADC's Control register, the ov er-temperature alarm (OT) becomes active. The OT > >> >signal resets when the die temperature has fallen below set threshold. > >> >The OT alarm can also be used to automatically power down the PL upon activation. The OT alarm can > >> >be disabled by writing a 1 to the OT bit in the XADC's Configuration register. > >> >Note: these registers are in the XADC and are accessible using the DRP. > >> > > >> >-Lasse > >> > >> It's probably shutting down at 125C, without our specifically > >> programming any temperature. > >> > >> Extensive searching, by us and by Avnet, finds no fan that matches the > >> hole spacing on the MicroZed board. So we'll fab a little aluminum > >> adapter plate and use a standard fan. With a pin-fin heat sink glued > >> to the 7020 FPGA, and the fan blowing down on that, we can run at 100C > >> ambient. > >> > >> > >> > >> -- > >> > >> John Larkin Highland Technology, Inc > >> picosecond timing laser drivers and controllers > >> > >> jlarkin att highlandtechnology dott com > >> http://www.highlandtechnology.com > > > >The MicroZed has a -I part on it, right? Those parts are spec'd at a max junction temp of 100 C. You need the Expanded temperature grade parts (Q) to get the 125 C junction temps. > > It's 99% likely that all the chips come off the same wafer. The faster > ones may get binned as the high-temp versions. > > The real issue is timing margins, and our fastest clock is only 128 > MHz. > > (We buy two different Altera parts, one with twice the logic cells and > RAM and price and stuff. They are actually identical, run the same > bitstreams compiled for either part.) > > Here's the fan, with its adapter plate. The box runs fine at 100C > ambient. Next time we recompile the design, in a couple of months > maybe, we'll bring out the chip temperature sensor to see how hot it > actually is inside there. >
if you have a jtag cable you should able to read it out and I think the driver is installed by default in the xilinx linux image some where down in /sys/bus/iio/devices/iio:device0 -Lasse
On Wednesday, May 13, 2015 at 6:37:19 AM UTC-6, kkoorndyk wrote:
> The MicroZed has a -I part on it, right? Those parts are spec'd at a max junction temp of 100 C. You need the Expanded temperature grade parts (Q) to get the 125 C junction temps.
MicroZed can be purchased off the shelf with either a -C or -I. A -Q is possible through a custom build by contacting customize at avnet dot com Bryan
On Thu, 14 May 2015 09:45:11 -0700 (PDT), bryan.at.avnet@gmail.com
wrote:

>On Wednesday, May 13, 2015 at 6:37:19 AM UTC-6, kkoorndyk wrote: >> The MicroZed has a -I part on it, right? Those parts are spec'd at a max junction temp of 100 C. You need the Expanded temperature grade parts (Q) to get the 125 C junction temps. > >MicroZed can be purchased off the shelf with either a -C or -I. A -Q is possible through a custom build by contacting customize at avnet dot com > >Bryan
We are buying AES-Z7MB-7Z020-SOM-G. I'm not sure which version of the FPGA is on that. Seems fine at 100C ambient, with our fan. I couldn't persuade the engineer to crank the temperature any higher. I like to test things to destruction. Nobody at Avnet wants to discuss the fan-mount hole spacing, or name a fan that fits. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
> Nobody at Avnet wants to discuss the fan-mount hole spacing, or name a > fan that fits.
I don't have the measurements, but it seems like it's a very common cooler among motherboards: http://www.ixbt.com/mainboard/msi/k9n4-sli/cooler.jpg Regards Tomas D.
> I don't have the measurements, but it seems like it's a very common cooler > among motherboards: > http://www.ixbt.com/mainboard/msi/k9n4-sli/cooler.jpg
Here it is: http://www.aliexpress.com/item/Northbridge-motherboard-graphics-heatsink-6cm-pitch-40-40MM-ultra-quiet-fan/1714057574.html Maybe fits the size? Regards Tomas D.
Den torsdag den 14. maj 2015 kl. 23.53.37 UTC+2 skrev Tomas D.:
> > I don't have the measurements, but it seems like it's a very common cooler > > among motherboards: > > http://www.ixbt.com/mainboard/msi/k9n4-sli/cooler.jpg > > > Here it is: > http://www.aliexpress.com/item/Northbridge-motherboard-graphics-heatsink-6cm-pitch-40-40MM-ultra-quiet-fan/1714057574.html > > Maybe fits the size? > > Regards > Tomas D.
not even close the mounting fan holes on the Microzed is 2mm, and the diagonal spacing is ~31.5mm, that is some where between a 25x25mm and 30x30mm fan and the heatsink has to fit in a ~17x17mm footprint because there are caps that are taller than the Zynq right next to it -Lasse
On Thu, 14 May 2015 16:55:06 -0700 (PDT),
lasselangwadtchristensen@gmail.com wrote:

>Den torsdag den 14. maj 2015 kl. 23.53.37 UTC+2 skrev Tomas D.: >> > I don't have the measurements, but it seems like it's a very common cooler >> > among motherboards: >> > http://www.ixbt.com/mainboard/msi/k9n4-sli/cooler.jpg >> >> >> Here it is: >> http://www.aliexpress.com/item/Northbridge-motherboard-graphics-heatsink-6cm-pitch-40-40MM-ultra-quiet-fan/1714057574.html >> >> Maybe fits the size? >> >> Regards >> Tomas D. > >not even close > >the mounting fan holes on the Microzed is 2mm, and the diagonal spacing >is ~31.5mm, that is some where between a 25x25mm and 30x30mm fan > >and the heatsink has to fit in a ~17x17mm footprint because there are caps >that are taller than the Zynq right next to it
The hole spacing may be English units, namely 1.25" I did look at a lot of CPU cooler fans. There are many, many hole spacings, except that one. We'll just order a bucket of those adapter plates and get on with our lives. We're gluing a Cool Innovations pin-fin sink to the top of the FPGA and blowing air down on that. Without the forced air, the pin fins are useless. But the base of the heat sink spreads heat laterally out from the central hot-spot (a flat metal plate works the same) and cuts junction temp by 5C or so. -- John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com