Forum Rule: Always post complete source code & details to reproduce any issue!
Page 1 of 2 1 2 LastLast
Results 1 to 25 of 30

Thread: Boot on Power-Up Problem with three T3.6s.

  1. #1

    Boot on Power-Up Problem with three T3.6s.

    After using T3.2's for a couple of years without a failure, I have been having all manner of problems migrating my (several) projects to the T3.6.

    I've now purchased 6 T3.6's, and over the last 8 months have not yet managed to get a single project ported over, and have spent countless hours chasing silly problems, including ILI9341 displays failing while connected to the 3.6. I'm starting to question whether the hassle is too much.

    The latest is that I have 3 new T3.6's that will not boot-up on power on - even running blink.ino. It's driving me crazy! Here's the situation:
    All are mounted on Tindie 64-pin breakout boards with headers so they can be swapped in and out. All 3 work well when powered via USB. In my lab prototypes, however, they are powered through an on-board 7805 regulator from a HP lab supply set at +9.0v. The voltage measured at the T3.6 Vin pin is 5.06v. There is an on-board on-off switch and a voltage reversal protection diode (1N4005) before the regulator. All measured voltages make sense.

    If I power up (with the on-board switch on) by turning on the HP power supply none of the Teensies will boot up. If I turn the PS on first, and then turn on the on-board switch they will "usually" - but not always - boot. It's independent of the code loaded - as I said above it happens with the humble "blink" sketch as well as my large app codes.

    If they fail to start I can kick them into action by grounding the RESET pad after which they perform fine until the next power on.

    What gives? The only common factor is the HP power supply. I'm wondering if the T3.6 boot cycle is edge triggered and the HP rise-time is just too slow? I have yet to try a RC delay network to hold RESET low briefly on power up, but I shouldn't have to go that far. BTW, I have played with the values of the input/output electrolytics around the 7805 regulator with no effect.
    Last edited by DerekR; 06-09-2017 at 03:20 AM.

  2. #2
    Here is a photo of a breadboard showing the minimal wiring to reproduce the power-on booting problem:

    Click image for larger version. 

Name:	BootProblemWiring (2).png 
Views:	276 
Size:	604.6 KB 
ID:	10768

    The red and black wires on the right are connected to the Hewlett-Packard power supply set at 9v. The 7805 regulator is on the left with a small slide switch.
    The Teensy 3.6 is mounted on the Tindie (loglow) 64-pin breakout board (the 3.6 has female headers attached to accommodate the Teensy audio board.) Note that the power connection on the breakout board are DIP pins 5 (GND)and 59 (Vin).

    It doesn't get much simpler - yet all three of my current working T3.6's fail to start when power is applied by turning on the HP power supply! I need to use the slide switch to interrupt the power, or ground the reset pad on the Teensy.

    I am supplying power this way because I need +9v for analog circuitry in some of my projects.

    I just measured the 10-90% rise-time of the HP power supply - it is ~70ms and nice and clean.
    Last edited by DerekR; 06-09-2017 at 02:39 PM.

  3. #3
    Senior Member+ KurtE's Avatar
    Join Date
    Jan 2014
    Posts
    3,638
    There have been a few other power supplies with similar issues with the T3.6... Most of them work fine on T3.5. You might do a quick search, but I think some of these have been resolved by putting a small cap on the reset line.

  4. #4
    Senior Member+ Frank B's Avatar
    Join Date
    Apr 2014
    Location
    Germany NRW
    Posts
    4,451
    yes, the supply is simply too slow at power-on. Use a secondary switch between the supply and teensy or add a small cap (~10nF ? try different values) to the reset-line.

  5. #5
    Yep, thanks, a 0.01uF cap works BUT, as others have pointed out, you cannot download code via USB with the cap in place.

    Seems like this is a design flaw. The T3.6 should work with any PS, particularly a high quality lab supply. @Paul -any chance of a revision or fix in the the near future?

  6. #6
    Senior Member+ Frank B's Avatar
    Join Date
    Apr 2014
    Location
    Germany NRW
    Posts
    4,451
    Quote Originally Posted by DerekR View Post
    Yep, thanks, a 0.01uF cap works BUT, as others have pointed out, you cannot download code via USB with the cap in place.


    hm, funny. i can.
    maybe a smaller cap helps ?
    Quote Originally Posted by DerekR View Post
    Seems like this is a design flaw. The T3.6 should work with any PS, particularly a high quality lab supply.
    yes, a design flaw of the power-supply.
    Quote Originally Posted by DerekR View Post
    The T3.6 should work with any PS, particularly a high quality lab supply.
    I bet, much more hardware will not work with your supply. In particular, every hw that needs faster power-on timing.
    It may be high-quality, but the startup is not so high-quality.

    Just use a better one. All smartphone-chargers I own work well.
    Last edited by Frank B; 06-09-2017 at 07:14 PM.

  7. #7
    Senior Member
    Join Date
    Jan 2013
    Posts
    843
    Quote Originally Posted by DerekR View Post
    Seems like this is a design flaw. The T3.6 should work with any PS, particularly a high quality lab supply. @Paul -any chance of a revision or fix in the the near future?
    +1

    This is something, I would like to see addressed in future Teensy revisions as well. I've had issues with Teensy starting up on battery power.

  8. #8
    Senior Member
    Join Date
    Mar 2013
    Posts
    643
    2 things.

    1. Most bench supplies are slow to turn on because they have CC limiting and many I have run into have large output caps. (I have a BK Precision 5A that struggles to start up unloaded power supply's on one of our MCU boards.)

    2. That Large capacitor on the output of the 7805 is causing ramp issues at startup. LM7805's are typically very stable with only a small amount of capacitance on the Output. Try using a 1 or 10uF instead of that monster.

    Some of the TI chips we use will go stupid on you if you don't bring the power supplies up on them fast enough. This is probably a similar issue.

  9. #9
    Senior Member+ Frank B's Avatar
    Join Date
    Apr 2014
    Location
    Germany NRW
    Posts
    4,451
    the only way to fix it, is to add a reset-circuit. i'm not sure that there is room on the 3.5/3.6 board for a more enhanced power-on reset (?)
    Last edited by Frank B; 06-09-2017 at 07:29 PM.

  10. #10
    Senior Member+ KurtE's Avatar
    Join Date
    Jan 2014
    Posts
    3,638
    The interesting thing I saw earlier with the one BEC I had that failed to work with the T3.6 boards, was I believe it worked on the first beta boards (the ones with the bigger chips...).

    But did not work on the actual boards. So not sure if there was subtle differences in the reset and/or power section?

    Also noted then, this same BEC worked with T3.2, LC, 3.5, also RPI2/3, also Intel Edison (not sure if I tried Galileo ), Odroid (C1, C2, XU3 lite, XU4)

    Hopefully with my new board (RPI hat with T3.6) the Pololu 5amp 5v board will work with it... I know it does with T3.2

  11. #11
    Senior Member
    Join Date
    Dec 2016
    Location
    Montreal, Canada
    Posts
    2,710
    i use those 5V 30A dc-dc epoxy sealed waterproof psu for T3.5, it's input is 8-40V and output is a solid 5V. it powers basically everything, including lcds and servos, and survives cranking for cars

  12. #12

    Some success

    I couldn't get both USB program loading and power on reboot using a single capacitor from RST to ground, so I designed my own POR (power-on-reset) circuit using a bunch of old discrete components in my junk-box:

    Click image for larger version. 

Name:	Untitled-1.png 
Views:	195 
Size:	50.0 KB 
ID:	10774

    R3 and C1 define the time-out, D1 is a low voltage zener to define the threshold voltage (you could equally well use a resistor, a couple of diodes, or an LED to define the threshold). There is nothing special about the NPN transistors, something like 2N3904 would work just as well. Immediately after power-on, the second 2N4401 is turned on, holding RST low. As C1 charges up, the diode D1 starts to conduct, and the first 2N4401 turns on, turning off the second one, thus releasing the T3.6 from reset mode.

    CAUTION: Do not let 5v get anywhere near the T3.6 pins, otherwise you will be sending Paul another $30 order.

    I'm not suggesting that this circuit is optimum in any way. But it has worked well for the past three hours with many, many power on and off cycles and many many compile/load cycles without a single failure.

    There is however a better way. Tomorrow I am going to order some TI LP3470 POR IC's from Digikey. They are very small and look to be ideal for the job in hand.
    Last edited by DerekR; 06-10-2017 at 01:48 AM.

  13. #13
    Quote Originally Posted by DerekR View Post
    Tomorrow I am going to order some TI LP3470 POR IC's from Digikey. They are very small and look to be ideal for the job in hand.
    please keep us posted how well that chip works, i have trouble with a project sometimes not starting as well, and that looks like a nice solution, if you can still program with it.

  14. #14
    They are ordered, scheduled for delivery Thursday (6/15). Stay tuned....
    In the meantime, my kludgy homebrew design continues to work well :-)

  15. #15
    Senior Member+ Theremingenieur's Avatar
    Join Date
    Feb 2014
    Location
    Colmar, France
    Posts
    1,769
    It's perhaps time to replace the 1980s style 7805 electric heater by a more modern and more efficient voltage regulator with an additional "power good" output pin which could be directly connected to the reset pin.

  16. #16
    Senior Member
    Join Date
    Mar 2013
    Posts
    643
    @Theremingenieur. I disagree, it comes down to proper circuit design.
    The current issue appears to be excessive power draw at startup (ie large electrolytics). Powering the Teensy off an old school proven supply is fine for the majority of users. I believe it would be far better for users to spec their own power supply needs for additional circuits they add. The rise time for supplies like the one used in the Teensy is limited to the rise time of the supply its attached to. There are plenty of linear supplies with resets built in but I do not believe that adding more parts and increasing the cost of the Teensy would be worth it in the long run. And adding a switching supply would require a large amount of board space that does not exist (I personally have no issues with Paul making a Mega type board with every pin of the MK66/MK64 broken out but I doubt there is enough demand for it besides eccentrics like myself who prefer MCU GPIO's over expansion IC's).

    @DerekR, I would like to point out the MIC803 series, they are very similar to the TI part but much less expensive.

  17. #17
    @Donziboy2 - Thanks, they look good. I wish I'd seen them before placing the order! Too late - just paying the shipping would make it too expensive.

  18. #18
    Senior Member PaulStoffregen's Avatar
    Join Date
    Nov 2012
    Posts
    18,153
    I am considering a hardware rev and/or bootloader tweak to improve compatibility with these types of power supplies.

    More specific info about the problematic power supply would really help. I see mention of a 7805, a "HP power supply", and 70ms startup time. But I don't see a scope screenshot. I don't know exactly which HP power supply, or what settings (know positions?) is it using. I also see a big capacitor in the photo, but I don't know its value. What I'd *really* like to see is a very specific description of every part and setting to precisely reproduce the troublesome power supply here. A scope screenshot of the startup behavior would also be really nice, so I could compare any attempt I make against what you've actually got.

  19. #19
    Senior Member+ KurtE's Avatar
    Join Date
    Jan 2014
    Posts
    3,638
    Hi Paul, for what it is worth, this is the UBEC that I had issues with: https://www.amazon.com/Castle-Creati...&keywords=ubec

    Don't have full scope but could hook up again to newer saleae LA with analog if that would help.

  20. #20
    Senior Member+ defragster's Avatar
    Join Date
    Feb 2015
    Posts
    6,755
    Quote Originally Posted by KurtE View Post
    Hi Paul, for what it is worth, ... if that would help.
    Or on loan ... for $7 USPS (LiPo not included) would hand a complete repro unit over for Paul to plug in

  21. #21
    what startup behavior do you want to see, the 3.3v teensy voltage and?

  22. #22
    @Paul: Thanks for getting back on this issue. I seriously think that addressing the start up problem would be a wise and welcome decision. I know that it has caused me a lot of frustration, and I may have consigned two T36's to the trash can because they were catatonic. Here is the info you asked for:
    1) The power supply is a HP 6236B triple output 0-6v 2.5 amp, 0 - +/-20v 0.5 amp). It's an oldie but goodie, and built like a battleship.
    2) Here are a couple of screenshots of the startup transient voltages using the power supply. Note that these are not taken using the breadboard I showed previously because I had torn it down. They are taken using one of my prototype systems and has a much smaller capacitor (47uF) on the 7805 regulator output. The dc current load on the 7805 is 150mA. It also suffers from the power-on problem.
    Photo 1: The evolution of the power supply output voltage (red), and 5v 7805 output voltage (yellow).
    Click image for larger version. 

Name:	PS&5Vpower_on.jpg 
Views:	155 
Size:	107.4 KB 
ID:	10812
    Voltage scales are 1v/division, time scale is 10ms/div.

    Photo 2: The evolution of the 5v 7805 output voltage (red), and the Teensy 3.6 3.3v voltage supply (yellow).
    Click image for larger version. 

Name:	5V&3.3Vpower_on.jpg 
Views:	153 
Size:	90.6 KB 
ID:	10813
    Please let me know if you need anything else.

    The LP3740s just arrived as I was typing this. I'll wire one up and let know how it works out...

  23. #23
    The good news is that the LP3470 power-on-reset works with very few connections to the Teensy.
    Click image for larger version. 

Name:	POR.png 
Views:	171 
Size:	44.5 KB 
ID:	10816
    I soldered it to an Adafruit SOT-23 breakout board, and built up the breadboard system running blink.ino again.
    Click image for larger version. 

Name:	lp3470breadboard.jpg 
Views:	185 
Size:	120.5 KB 
ID:	10818
    The lp3470 is in the lower right corner. It seems to work fine without the pullup that the datasheet says is necessary, and with or without the delay capacitor. The following scope traces show the reset line startup with a 0.01uF delay cap (without the cap reset simply goes high 20ms earlier). Yellow is the power supply input, red is the reset line. Voltage scale is 1v/div, time is 10 ms/div.
    Click image for larger version. 

Name:	lp3470.jpg 
Views:	165 
Size:	80.0 KB 
ID:	10817
    BTW - I'm using the 2.93 v threshold version of the LP3470...

    The not so good news is that I built a second one and installed it permanently on one of my prototype SDR boards. It starts up but is acting weird - with it connected something is generating a reset every 500 ms - I'm looking into it...

  24. #24
    I have just installed a MIC803 (2.63 threshold and >1 second delay but im pretty sure the faster 140ms or 20ms ones will work)
    I have a eurorack module with t3.6 which gets used with various different power supplies. and before would not start on some of them sometimes. now it will start every time on every supply.

    i used the recommended 100k pullup resistor from MIC803 data sheet.

  25. #25
    here is a "hack" where you can add a MIC803 under the teensy.
    Click image for larger version. 

Name:	19402668_10155452882628545_1752656905_o.jpg 
Views:	193 
Size:	159.7 KB 
ID:	10864

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •