Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
10 years ago

CycloneIV-E: Heating issue

Hey Members,

it's my first post here and I'm also pretty new in FPGA-PCB-Design ;)

I designed a 10-Layer CycloneIV-PCB with 5 different switching regulators (LT3690). I use the CycloneIV EP4CE15F23C7N.

My Powersupply chain for the FPGA is as follows:

24V_in

=> LT3690 for 1.2V (VCCINT)

=> LT3690 for 3.3V (VCCIO+µC)

=> Out of 3.3V with LDOs:

=> 2.5V (VCCIO)

=> 1.8V (VCCIO)

The other 3 switching regulators (LT3690) are used for powering up a sensor board and for two TECs.

I use a passive heatsink with 24.5°C/W.

First I used the Early Power Estimator (excel sheet) from Altera and with that design+heatsink it calculates around 28°C chip temperature @ambient temperature=24°C

I have a strong heating issue with that FPGA board :/

I download a test design with only an ethernet core installed (25% FPGA usage) and my results are frustrating:

With ambient temperature of 24°C my FPGA is heating-up to 33°C.

So even with this small design my FPGA has a deltaT of +9°C :/

After this result I disabled all switching regulators and power up the FPGA with my lab power supply.

So 1.2V and 3.3V (=> 2.5V, 1.8V) are now comming from a clean power supply and the result is good:

With ambient temperature of 24°C my FPGA is heating-up to only 27.5°C. This fits very good to the estimation from the excel-sheet.

So disabling the switching regulators reduce the deltaT to +3.5°C.

Now my question:

I assume now that the switching noise of my 5 switching regulators are the reason for that strong heating. Could that be? Did you ever had the issue that with high switching/ac-noise your FPGA is getting too hot?

I missed to put some PI/LC-Filters after and before the switching regulators and it looks like that the combination of 5 equal switching regulators produce a lot of noise...

Thanks so far!

If you need further information - please tell...

Fizz

3 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    10 years ago

    I think it's very doubtful that "switching noise" is causing your FPGA to heat up. Isn't it more likely that your switching regulators dissipate some heat of their own, and this contributes to heating the FPGA?

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    10 years ago

    --- Quote Start ---

    I think it's very doubtful that "switching noise" is causing your FPGA to heat up. Isn't it more likely that your switching regulators dissipate some heat of their own, and this contributes to heating the FPGA?

    --- Quote End ---

    I thought the same in the beginning but the only switching regulator which has significant heat dissipation is the 3.3V-regulator. The 1.2V for example drives only 0.05mA for VCCINT. Even when I only have this enabled the FPGA is heating up more than with ALL voltages from the lab power supply.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    10 years ago

    Can I assume that, with the switching regulators, VCCINT & VCCIO look OK at the FPGA? Is there suitable decoupling locally such that the switching noise the FPGA is exposed to is minimised?

    I feel this is more likely to be conducted heat rather than heat generated by the FPGA. Do any of the LT3690 power circuits generate much/any heat of their own? How close are they to the FPGA? Are there any other circuits on the board dissipating heat?

    By bypassing them, and powering from the bench supply, your target is not going to generate as much heat - heat which could be conducted to your FPGA. The temperature of the device is dependent on both what it is doing and the heat generated around it.

    The Altera Power estimator is only going to consider power the FPGA generates and will assume no other heating sources around the device.

    +9 degrees doesn't seem particularly bad to me. What delta do you see at the FPGA if you leave the board powered but don't configure the FPGA? Do you get this same behaviour on multiple boards (assuming you have more than one)?

    Cheers,

    Alex