Hot!PIC32MZDA and Heat sink potential need

Author
flatbush
Senior Member
  • Total Posts : 123
  • Reward points : 0
  • Joined: 2009/01/30 07:51:14
  • Location: 0
  • Status: offline
2018/11/08 09:23:42 (permalink)
5 (1)

PIC32MZDA and Heat sink potential need

Hi everybody,
 
I am conducting severe tests with a PIC32MZ2064DAG176I2J.
 
First I noticed that the component has on its bottom part a metal plate which was, I thought, a "safety" heat sink.
There is not such a thing with the previous versions of the PIC32MZ. Curious...
In the design of my board I took that into account and I connected this plate with a copper area approximately the same size to the ground layer of the board. Finaly, when running, the MZ is hot but not so much, I would say 30/40°C.
 
Now I am using the 32MB internal DRAM very intensively i.e. 20% of the actual time is filled with Read and Writes operations.
Temperature of the MZ has rised but again not so much, anayway it is not burning.
But problems occurred with the DRAM: my application test detects errors between what should be the content of the memory cells and their actual value. Very strangely, the difference is very often a multiple of 4 (???).
To give numbers the probability of such an error is 1 every 1000 000 "operations". Conditions of the test are with a 28°C ambiant temperature.
This rate is reduced to 0, yes 0, when ambiant temperature decreases to 20°C or when I simply put a heat sink only on the microcontroller.
 
I insist that the application continues to work normaly there is no other bug, except this quantitative difference. It is as if the internal DRAM were very sensitive to external temperature. But the PIC32MZ Graphics is normaly able to sustain -40ºC to +85ºC. (???).
 
Did some of you experience any problems this type with the MZDA ?
#1

19 Replies Related Threads

    friesen
    Super Member
    • Total Posts : 2008
    • Reward points : 0
    • Joined: 2008/05/08 05:23:35
    • Location: Indiana, USA
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 09:44:56 (permalink)
    0
    The internal DRAM is errata'd to 0-70C, are you sure you aren't getting there?
     
    Also, are you using standard clock values like harmony puts in there?
     
     

    Erik Friesen
    #2
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 10:06:28 (permalink)
    5 (1)
    HI Eric
     
    Yes I am sure that the temp is below 70C. I don't know the exact value, but I think it is even below 40°.
     
    I am using Harmony and its standard values. For the DRAM I trusted what was given as an example in the demos.
     
    I tried 2 kinds of heat sink
     
    - first one is 1.3X2X4X CM and the error rate is 0.
    - the second is much smaller 2X0.6X0.4 and the error rate is divided by 50
     
    the 2 heat sinks are just in contact with the plastic top of the component, with thermal grease in between.
     
    For the anecdote, I noticed the errors when the air conditionning system broke down in my office. Very funny.
     
     
    #3
    friesen
    Super Member
    • Total Posts : 2008
    • Reward points : 0
    • Joined: 2008/05/08 05:23:35
    • Location: Indiana, USA
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 10:21:09 (permalink)
    0
    I have a 100+ of the qfp version in the field, I haven't noticed anything funny. (PIC32MZ2064DAH176)
     
    I am using dram much less than you though, maybe 2% wild estimate.  The chip runs comfortable warm to the touch, maybe 5-10c above ambient.
     
     

    Erik Friesen
    #4
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 10:31:54 (permalink)
    0
     
    The ambient temp is 26°C (!) in my lab. I just put a probe 2 minutes ago on the top of the MZDA. It shows 42°C. The contact of the probe is enough to decrease the rate of errors by 100. So something like one error every 10E8 operations.
    #5
    friesen
    Super Member
    • Total Posts : 2008
    • Reward points : 0
    • Joined: 2008/05/08 05:23:35
    • Location: Indiana, USA
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 10:56:59 (permalink)
    0
    How about your layout and schematic for all ddr related pins?

    Erik Friesen
    #6
    maxruben
    Super Member
    • Total Posts : 3292
    • Reward points : 0
    • Joined: 2011/02/22 03:35:11
    • Location: Sweden
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/08 15:24:23 (permalink)
    0
    See this thread about ambient temperature, thermal resistance and junction temperature.
     
    Check with the internal temperature diode how hot the die actually is.
     
    Is the exposed die pad at the bottom soldered to the board?
     
    /Ruben 
    post edited by maxruben - 2018/11/08 15:30:14
    #7
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/09 11:20:07 (permalink)
    0
    Thank you Erik and Ruben for your ideas.
     
    About layout and schematic:
     
    The proto board has 6 layers : 3 signal layers, one for GND, one for +3.3, one for 1.8 and +5V.
    the 1.8V plane area covers 15MM all around the CPU only. This area is not shared with the +5V. 1.8V is made with a REG1117A 1.8 which is able to supply 1A, from the +5V. A large copper area under the regulator acts as a heat sink. There is also a 100µF capacitor and a ferrite bead to clean the VDDR1.8V. Furthermore there are 6 100nF (805) capacitors very close to the VDDR1.8 pins.
     
    The die pad at the bottom is soldered to the board. In fact a top layer copper area larger than the pad acts also as a heat sink. This copper area is of course connected to the GND plane thu a large hole (1MM dia).
     
    The phenomenon is abrupt : the temp probe in contact with the  CPU shows: 
     
    50°  7600 Errors
    47° 1600
    45°C 444
    43.6° 38
    43° 14
    40° 0
     
    Of course one could think that other components around the CPU could cause the errors. But I have eliminated this possibility:
     
    - I have the option not to use the DRAM for the acquisition but instead the IRAM (512K) -> no errors
    - there is clearly a relationship between the writing percentage and the reading percentage: Strangely , it seems that the DRAM gets some improvement during the readings. So when the percentage of reading is higher, the rate of errors decreases...I must quantify this behavior
     
    #8
    maxruben
    Super Member
    • Total Posts : 3292
    • Reward points : 0
    • Joined: 2011/02/22 03:35:11
    • Location: Sweden
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/09 11:51:28 (permalink)
    0
    Did you check with the internal temperature diode what the die temperature really is?
     
    /Ruben
    #9
    Jim Nickerson
    User 452
    • Total Posts : 5578
    • Reward points : 0
    • Joined: 2003/11/07 12:35:10
    • Location: San Diego, CA
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/09 12:03:29 (permalink)
    0
    maxruben
    Did you check with the internal temperature diode what the die temperature really is?
     
    /Ruben




    I think maybe the errata has something to do with this.

    Attached Image(s)

    #10
    maxruben
    Super Member
    • Total Posts : 3292
    • Reward points : 0
    • Joined: 2011/02/22 03:35:11
    • Location: Sweden
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/09 12:29:17 (permalink)
    5 (1)
    Ok then. Can you measure the power consumption of the mcu and calculate the die temperature based on power and thermal resistance between die and package?
     
    Perhaps there are other erratas connected to your problem?
     
    I see that the link to the thread about temperature in post #7 above didn't stick. Here it is again https://www.microchip.com/forums/FindPost/1059462
     
    /Ruben
    #11
    maxruben
    Super Member
    • Total Posts : 3292
    • Reward points : 0
    • Joined: 2011/02/22 03:35:11
    • Location: Sweden
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/09 12:29:17 (permalink)
    0
    Ok then. Can you measure the power consumption of the mcu and calculate the die temperature based on power and thermal resistance between die and package?
     
    Perhaps there are other erratas connected to your problem?
     
    I see that the link to the thread about temperature in post #7 above didn't stick. Here it is again https://www.microchip.com/forums/FindPost/1059462
     
    /Ruben
    #12
    LostInSpace
    Super Member
    • Total Posts : 153
    • Reward points : 0
    • Joined: 2016/03/11 22:47:59
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/10 09:17:47 (permalink)
    0
    This doesn't 'smell' like a temperate problem to me per-say, I believe that you have a temperature dependence problem, I really doubt that it is the chip temperature itself that is the issue.
     
    If it was me troubleshooting this - I would be looking for a timing issue. I bet there is some timing constraint that is being violated on a clock and/or data lines from the PIC to your Memory. Time to get out the scope (or good logic analyzer).
     
    HTH
    #13
    friesen
    Super Member
    • Total Posts : 2008
    • Reward points : 0
    • Joined: 2008/05/08 05:23:35
    • Location: Indiana, USA
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/10 10:10:13 (permalink)
    5 (1)
    @lostinspace it is internal dram the op says.

    Erik Friesen
    #14
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/10 15:55:52 (permalink)
    0
    I am far away from my office now, but I have read all your posts with great attention. Next week, I continue the tests.
    I need to make more measurements and the idea to get the die temperature thanks to the internal temperature diode is certainely very interesting.
    I read that inside the component packages there are in fact 2 chips. One is the MCU and the other is the DRAM. I don't know if it is true or not. In this case may be the area occupied by the DRAM is small and can reach higher temperatures than the rest of the die. So may be, locally the limits are reached.wink: wink
    I must check if in Harmony there are some functions to read the temp diode. If you have already done that, I would certainly apreciate the information.
     
    Many Thanks
    #15
    maxruben
    Super Member
    • Total Posts : 3292
    • Reward points : 0
    • Joined: 2011/02/22 03:35:11
    • Location: Sweden
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/11 12:03:43 (permalink)
    5 (1)
    Jim Nickerson showed in post #10 that the internal temperature diode unfortunately doesn't work. Errata31.
     
    While quickly skimming through the datasheet there seemed to be a lot of registers connected to the internal dram. Perhaps it is a setting issue?
     
    /Ruben
     
    #16
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/14 09:44:55 (permalink)
    0
     
    After a lot of time observing the errors, I can say there is a problem above 38°C.  Errors are more and more numerous, increasing exponentialy with the Temp. I can avoid the phenomenom by simply gluing a heat sink 20X20MM on the MCU. This work around is a sad solution, because 38° is very low, and I can't accept such a limitation.
    It is very difficult to answer about the current consumption. Many components are involved in the process. But I can distinguish a 30/40ma surge when continously writing the DRAM.
     
    May be  in the internal DDR demo, they have set the ddr registers with too border values.
     
    When I take a look to this:
     
    void SYS_MEMORY_DDR_Initialize(void)
    {
     DDR_PMD_Init();
        DDR_PHY_Init();
        DDR_Init();
     DDR_PHY_Calib();
    }
     
    I find there are more than one hundred settings involved. So before exploring that, I need a map.
     
    /Flatbush
    #17
    flatbush
    Senior Member
    • Total Posts : 123
    • Reward points : 0
    • Joined: 2009/01/30 07:51:14
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/29 10:15:48 (permalink)
    0
    I have made some tests concerning the power comsumption  of my prototype:
     
    All other things being equal:
     
    When the DDRAM is continuously read the consumption indicates 700mA with the +5V
    When the DDRAM is at rest (no readings and no writings)  I note an INCREASE of 30 mA with the +5V -> 730 mA
    When the DDRAM is continuously written I note an increase of 150 mA with the +5V -> 850 mA
     
    I don't have an explanation why it seems that reading the DDRAM drains less current than leaving it without any access.
     
    Of course, with continuous writings, the temp of the component increases rapidely.
     
    A difference of 3/4 watt in the 2 configurations explains why they expect to dissipate the heat with the bottom metal plate.
    Being an absolute ignorant in the technology of the DDRAM, I am very curious to hear something from the Microchip guys.
     
    /Flatbush
    #18
    Antipodean
    Super Member
    • Total Posts : 1668
    • Reward points : 0
    • Joined: 2008/12/09 10:19:08
    • Location: Didcot, United Kingdom
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/29 10:25:32 (permalink)
    0
    Did you connect the heat sink pad on the PCB to anything? Normally this should be full of vias to make a connection through to the 0V plane to help take the heat away - but check that the pad on the chip is at 0V first.
     
     

    Do not use my alias in your message body when replying, your message will disappear ...

    Alan
    #19
    andersm
    Super Member
    • Total Posts : 2485
    • Reward points : 0
    • Joined: 2012/10/07 14:57:44
    • Location: 0
    • Status: offline
    Re: PIC32MZDA and Heat sink potential need 2018/11/29 12:13:44 (permalink)
    5 (1)
    The datasheet says
    The metal plane at the bottom of the device is internally tied to VSS1V8 and should be connected to 1.8V ground externally.

    Ie. it's ground for the DDR2 memory.
    post edited by andersm - 2018/11/29 12:15:52
    #20
    Jump to:
    © 2018 APG vNext Commercial Version 4.5