FPGA temperature measurement

We're experimenting with heat sinking an Altera Cyclone 3 FPGA. To measure actual die temperature, we built a 19-stage ring oscillator, followed by a divide-by-16 ripple counter, and brought that out.

The heat source is the FPGA itself: we just clocked every available flop on the chip at 250 MHz. We stuck a thinfilm thermocouple on the top of the BGA package, and here's what we got:

formatting link

We can now use that curve (line, actually!) to evaluate various heat sinking options, for both this chip and the entire board.

The equivalent prop delay per CLB seems to be about 350 ps. The prop delay slope is about 0.1% per degree C.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

Cute graph.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I had a minion photograph my whiteboard data and type it into Excel. I don't do Excel.

Being now calibrated, I stuck a short pin-fin heat sink on top of the FPGA with some grease, and the chip temp dropped 4C. A tall pin-fin dropped it 4C. A 0.7" square of 0.062 thick aluminum, greasy-stuck to the top, dropped the chip temp 4C.

Neat!

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

Suggesting that the main mechanism is helping transport heat to the leads (or the more distant solder balls), rather than to the air.

Pin fin heatsinks are a crock. You don't get any more surface area than a parallel-fin design, and all the discontinuities interfere with the airflow very badly.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Right. It spreads the heat laterally from the hot spot in the center of the package. So, why didn't Altera do that for me?

But I can buy one with a thick flat base and peel-off acrylic sticky on the bottom. The pin-fins are just for show.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

It would be amusing to calculate how thick the aluminum has to be before the sticky stuff dominates the thermal conduction. My guess is about 10 mils.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Neato!

Say, how's the voltage coefficient on that, with respect to Vcore I suppose? Would be good for calculating how much ripple it can tolerate.

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://seventransistorlabs.com 
 Click to see the full signature
Reply to
Tim Williams

Vcore comes from an LM1117 with the ADJ pin grounded, so it's not really easy to twiddle. I might if I'm ever feeling energetic.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

Generally diode characterization is how you measure die temperature. [I make it a point not to look at stuff on drop box. I don't trust it.]

I can document the procedure, but I'm sure it is on the internet somewhere.

Reply to
miso

Not me! The ring oscillator lets us just scope-probe a connector pin to get FPGA temperature, and we can/might build a frequency counter into the design so we can remotely read chip temp on all 16 boxes in the system.

[I

Do you refuse to view any jpeg image that's hosted by Amazon?

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

Did you try different kinds of heat sink grease? Apparently toothpaste and Vegemite work pretty well:

formatting link

Matt Roberds

Reply to
mroberds

I believe it about vegemite--it's clearly not a foodstuff.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Clearly you are not a convert! :)

Reply to
Scromlette

Won't this frequency vary due to different lot characteristics, so it can only be used for a pre-calibrated device, calibrated with temperature sweep or twopoint regresssion?

Cheers

Klaus

Reply to
Klaus Kragelund

I haven't tried different FPGAs, and I'd expect some variation. My interest here was in evaluating heat sinking, delta-t, and a single chip curve does that. I doubt that my calibration is absolute, but I think the slope is pretty close.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

Here are comparable experiments done with Xilinx Virtex 5:

formatting link

regards, Bart Fox

Reply to
Bart Fox

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.