Well duh, the Teensy 3.6 is NOT 5v tolerant

Very disappointing to have the most expensive chip die so easily by feeding it its own USB power pin to a data pin :(

There is a misunderstanding. Looking at the Teensy 3.6 schematic, you see that the USB power is not "the chip's own". It goes to an external voltage regulator (LP38691) which then feeds 3.3V into the MK66FX1M0 for the user's and developer's convenience, allowing VUSB to power everything during the development and prototyping phase. There is not a single direct connection between the microprocessor and +5V.

Look into your TV set. It has a 90 to 264V power supply input but this voltage is not applied directly anywhere in the TV circuit, it just goes to the internal SMPS which then provides 5V, 12V, 24V or whatever to the different circuit parts. And you would not blame the TV manufacturer after you applying "accidentally" 110 or 230V for example to the TV's audio board which runs on 12V.

You should also not willingly misinterpret Paul's statement about level shifters. These are neither expensive or slow by themselves if the circuit designer choses the optimal component for his specific use case. Level shifting happens all the time and everywhere, in PCs, industrial machine control, TV & Hifi, and mobile devices. One of the tasks of a circuit developer is to have an eye for what could happen even if it should not, and anticipating that by a thoughtful selection of components and peripherals. That is called "engineering"...
 
As for Paul's comment that level shifters are expensive and slow. Sure, I understand that, but 5v tolerance, is not level shifting, it's just not having the expensive chip destroy all of itself just because 5V appeared on one of its IO pins.
I'd be fine with the chip shutting down in protection mode (some polyfuse), or even hard fuses that just kill the pins you overvolt without killing the whole chip.
But having the whole chip die, is just crappy.
Along with the serial port having never worked reliably for me (which I understand may be a more localized problem), this still steers me back towards ESP32.
If there is to be a new teensy, I hope it will address those downsides (and yes, I could have bought a teensy 3.5, but I got the faster one with more RAM because I needed that).

"could have bought a teensy 3.5, but I got the faster one with more RAM because I needed that"

That compromise is the very reason there are TWO versions - the T_3.5 and the 3.6!

In order to have the speed and added features for the T_3.6 there are more parts of a more sensitive nature in closer proximity internally - T_3.6 can run much faster - have better cache memory and architecture - larger flash storage (as offered) - provide a USB Host port - Touch sensitivity on pins - etc. All in the exact same chip footprint on the exact same PCB - with generally no lost features in the T_3.6 … except 5V tolerance. Even the T_3.5 is only 5V tolerant on the Digital GPIO capable pins - and it turns out they both have the same 256KB of RAM.

There is one polyfuse on the board to protect the board/voltage regulator from consuming too much power - and clear documentation that the T_3.6 pins are not 5V tolerant so it must be handled with care in that regard.

The next Teensy product from PJRC is a 600 MHz processor generally keeping T_3.6 capabilities and extending them {1MB RAM, 1.5+GB Flash, better cache and processor architecture speed, CANFD, 7 UARTs, ...} - but only 1.4" long like the T_3.2 - and that MCU likewise cannot tolerate 5V either and oddly enough the package with more GOOD stuff is in an even smaller space - but requires many more PCB support components to function including off chip flash because the on chip flash version is probably not yet shipping in quantity. Those added components nearly fill the top and required a like number on the bottom side between the pads that bring out another ~18 data pins on the bottom to add to the 22 on the edge pins - many fewer than T_3.6 - but at great effort to expose most of the chip's feature in a Teensy format.
 
thanks both for your answers.
Yes, I was aware that 5V was only there long enough to be stepped down to 3.3V, it's just easy-ish to mistakenly have it touch a pin that's close, and fry the whole chip.
Thanks for correcting me that teensy 3.5 and 3.6 have the same amount of RAM, I forgot that teensy has a faster CPU instead.

For the rest, yes, I understand it's all about compromises. I think I'm still bitter here because I bought the expensive
https://www.tindie.com/products/loglow/teensy-3536-breakout-revision-a-standard/
and now that my chip is fried, the entire thing is useless. Sadly, it uses a lot of pins soldered on the pack of the chip, instead of just pogo pins, so I can't easily replace the chip with a new one.
Waiting to see if the seller is still around and willing to sell me a new chip with all the soldered pins.
But more generally, I'll still stick with chips that are 5V tolerant if at all possible, there are just too many ways to have a 5V connection happen for one reason or another.
 
"and it turns out they both have the same 256KB of RAM."
Just to be clear on that tangent: I've always read that teensy 3.5 has 192KB or RAM while 3.6 has 256KB.
I have a 320x240 24bpp framebuffer, which ends up being 225KB, so that only works with teensy 3.6, and not 3.5.
Honestly all the extra IO is not something I need, I just needed more RAM :)
Technically ESP32 has more RAM while being 5V tolerant, but that RAM is not contiguous, so I can't use it for my framebuffer.
 
Not that it matters, but I don't think ESP32 is 5v tolerant either.

For example if you look at the different places on net, like: https://learn.sparkfun.com/tutorials/esp32-thing-hookup-guide/hardware-overview

It states:
The ESP32's operating voltage range is 2.2 to 3.6V. Under normal operation the ESP32 Thing will power the chip at 3.3V. The I/O pins are not 5V-tolerant! If you interface the board with 5V (or higher) components, you'll need to do some level shifting.

As for Frame buffer usage on it... At one point, I hacked up a version of the ILI9341_t3n like library where I think I was trying it on ESP32, and if the malloc failed to allocate the full size (note I was using 16 bits per pixel), it would try to malloc it in two pieces and that worked, so I then just had to play with the frame buffer code to know when I am going to certain locations it used first memory allocation and for others to second...
 
Sorry for getting off topic on the ESP32. It's actually the ESP8266 that is 5V tolerant despite its datasheet saying 3.3V max.
The ESP32 can take 5V but only with a resistor to limit the current. Straight 5V from USB is supposed to also damage it.
As for malloc on ESP32, indeed if you split your mallocs and join stuff in software, it can work, I've done it. I just didn't want to bother for this code.
If anyone is curious, this is what I'm working on:
http://marc.merlins.org/perso/ardui...GFX-for-SPI-TFTs-like-SSD1331-or-ILI9341.html
https://github.com/marcmerlin/FastLED_SPITFT_GFX
https://github.com/marcmerlin/Framebuffer_GFX

I've gotten this to work with 320x240 ILI9341 but at 24bpp, that only works on teensy 3.6, and only barely.

I've already had to do enough hacks to replace a bunch of arrays with mallocs on ESP32, and that was annoying to say the least (some memory can't be had as global heap, but can be malloced once the chip has initialized). This is definitely something you don't have to worry about on teensy.
But teensy has been such a terrible experience for me due to the serial port hangs I've described in
https://forum.pjrc.com/threads/56456-Teensy-serial-debugging-am-I-doing-this-wrong
that I've otherwise been avoiding it for development. I need serial support to just work. Never mind the hangs, I can't have teensy stop outputting serial debug after a little while, like it does "reliably" for me, sadly.

The other thing that really causes me to prefer ESP32, are the debugging tracebacks when things crash. Printf debugging is fun and all (although it doesn't even work reliably on teensy as per above), but it's sad that teensy prevents the in chip debugging capability without hacking it, and doesn't provide better debugging info.
 
i reckon PJRC should sell a small screw sheild adaptor that the Teensy plugs into that automatically protects the inputs and outputs - i would order one with every chip - there may be more profit in the shield adaptor than the teensy !
 
i would be happy to buy at $40 - much cheaper than making them myself

if you want to make some prototypes, I would be happy to commit to buying some

I kind of see it like a ruggedTeesnsy shield
 
I think it would make more sense for Paul & the PJRC staff to focus their limited development resources into the core products, the Teensies (is it ok to pluralize them?)

Given the relatively open nature of their products, nothing prevents a willing 3rd party to create & sell such a protection board.

Personally, I have only destroyed a single T3.5 out of sheer stupidity (that would make a great story by itself, given that it continued to work for almost a year in an over voltage condition!). That's out of 2 T3.2, 2 T3.5 and about 10 LCs, not counting those in my prototyping bin (1 or 2 of each).
I fared much worse with ESP8266 boards (2 out of 4 have died for no reason) although I would still used them when simple WiFi connectivity is needed; I also killed various STM32 flight boards (ok I reckon 30ft freefalls may have played a role here).

tl;dr: I use Teensies when they are a good fit with the project at hand

Marc
 
Given that an ESP32 can be had for less than $5, is more capable for some things (although less capable for some others too), it's indeed disappointing to lose a $30 teensy to a single VUSB application to the wrong pin.

In my case, I'm doubly sad, because I bought this:
https://www.tindie.com/products/loglow/teensy-3536-breakout-revision-a-standard/
It was $89 assembled with the teensy 3.6.
The main reason for it is to have all the pins on the bottom expanded to that carrier board, so that you can use them easily, and it also comes with a USB-A port.

Sadly, because that design does not use pogo pins, losing my teensy 3.6 means I can't just put another one, I have to get another one with the pins soldered at the bottom.
Obviously not Paul's fault, but also explains my sadness about losing the chip.
 
@marcmerlin - Many of us have been there, done that... (more or less) - I have seen magic smoke more than once... Example try plugging in 12v wall wart into an UP board who want's 5v... (Same connectors on both wall warts...) Been there: https://forum.up-community.org/discussion/2069/now-that-i-fried-my-only-up-board

As for ESP32... - Many I see are at least $15 like: https://www.adafruit.com/product/3269 ... Also it is a 3.3v device and that device does have 5v output pin... Not sure what happens if you connect that to one of their IO pins...

Just about everything is a trade off... Example Pogo pins. Pain to solder in. Get too many of them, and if your chip is socketed, the springs in the pogo pins, can cause the chip to slowly lift out of socket, making bad contacts... Seen that with first T4 Beta adapter.

Anyway good luck. Have you contacted Tall Dog? Maybe he will sell you a T3.6 (or 3.5), that has the pins soldered in.... Probably a long shot, but who knows!
 
@marcmerlin - Many of us have been there, done that... (more or less) - I have seen magic smoke more than once... Example try plugging in 12v wall wart into an UP board who want's 5v... (Same connectors on both wall warts...) Been there: https://forum.up-community.org/discussion/2069/now-that-i-fried-my-only-up-board

As for ESP32... - Many I see are at least $15 like: https://www.adafruit.com/product/3269 ... Also it is a 3.3v device and that device does have 5v output pin... Not sure what happens if you connect that to one of their IO pins...

Just about everything is a trade off... Example Pogo pins. Pain to solder in. Get too many of them, and if your chip is socketed, the springs in the pogo pins, can cause the chip to slowly lift out of socket, making bad contacts... Seen that with first T4 Beta adapter.

Anyway good luck. Have you contacted Tall Dog? Maybe he will sell you a T3.6 (or 3.5), that has the pins soldered in.... Probably a long shot, but who knows!
No offense to adafruit, but they are far from the cheapest source of ESP32s. If you look online, you'll find some in the $3.xx range.

ESP8266 is supposed to resist to 5V (although not in spec) while ESP32 does not, but again, they're less than $5 and I don't have to solder to the back or buy a new pre-soldered chip if I lose one (but to be fair, they also don't have nearly as many IO pins as a teensy 3.6 does).

As for pogo pins, point taken, I hadn't realized that it could push the chip out of the pins it's in if you have too many.

I did contact tall dog, waiting for an answer, but yeah, it won't be cheap (and it's already been 2 weeks).

But now that I know that teensy 3.5 has as much RAM as 3.6, I'll stick to 3.5 for it's 5V resistance in the future if I need something that an ESP32 can't do (it's now my to go default chip due to the built in stack traces and working serial support for debugging, 2 huge pluses for me).
 
When you are talking $5 or less disposable devices - that isn't UNO or Teensy, Adafruit or Sparkfun other supported quality devices. There are $5 or $10 raspberry Pi Zero's too - in limited quantities. All with alternate design ideas and limitations. I just ordered some ESP Pico's that were $18 to $25 each - the $18 dev kit actually has hardware added that added to the $25 base PICO board costs $45 in a kit including simple display, 3DOF, AMP, Speaker - but like cheap rPi Zero it is subsidized or bulk made to expand the user base.

I've collected 4 years worth of evolving Teeny ARM units and each one I got and used it still working - and my workspace and handling skills are the opposite of static safe and electronics friendly - but each one still works - because I've managed to not put 5V where it shouldn't go. Any added protection or cost ruining the size or design speed/sensitivity or cost factor would be a waste of effort and harm the general utility it provides.
 
Yeah, it's all a matter of handling / care.

Look at it this way:

If you want to blame hardware for your mistakes, or think it would be a problem with the hardware if you expose it to conditions it's not built for, you have a problem - not the hardware.

I know some people professionally who never make mistakes (of course) - always others are responsible. That's what I call bad education. In the end, they lose all the more. That's my life experience.

Just my two cents.

Mistakes are made to be learned from. That applies to everything, every hobby, and even more so to electronics.
Take a red piece of paper, write "NOT 5V tolerant" to it, and attach it to the teensy. Done. Costs less than 1 cent (not 40 or 80$)

I killed two teensys (3.1/3.2) and two audio shields (+countless other hardware) - my own mistakes. I know now what have to watch out for. I will continue to make mistakes in the future ;) - But...not the same :)
 
Last edited:
Yeah, it's all a matter of handling.

Look at it this way:

If you want to blame hardware for your bugs, or think it would be a problem with the hardware if you expose it to conditions it's not built for, you have a problem - not the hardware.

I know some people professionally who never make mistakes - always others are responsible. That's what I call bad education.

Just my two cents.

Mistakes are made to be learned from.
True that.
My first mistake was using an expensive $90+ shipped fancy teensy 3.6 that was difficult to swap/replace, and in an application where I didn't even need it because 3.5 had just as much RAM.
 
Yes, I've spent much more money for useless things ... such is life. The 3.5 additional RAM is a bonus, for free. It was found by a user of this forum. Don't you like it?
The 3.6 is faster, so, it's still worth its price.

Edit: Soon we will see controllers with less than 3v - and it might be they are not even 3v tolerant. This whole 5V thing is outdated, since years...
Keep in mind that these controllers are not made for hobbyists - they are made for industy use. Producing them for lower voltage is much better in all ways. If you apply 3V to the CPU in your PC, it will die in nanoseconds..
 
Last edited:
LOL,

the faster the chip.....The quicker it can die?

Yes mistakes are generally not cheap, either in money or in time. Been there and done that for sure and as Frank B said... I will probably keep making mistakes and learn from them.

I love the Teensy's small, fast and yes, not the cheapest of controllers but so much fun to use. I never bought an arduino again after my first T3.1
Regards all, Otto
 
the faster the chip.....The quicker it can die?

Yes, or well, sort of...

Ultimately this depends on many details of how the chip is designed, and how it's fabricated, which are usually a closely guarded trade secret. But for a general high-level picture of chip fabrication, it's common to talk about the chip's transistor channel length, in nanometers. As a general rule, small numbers mean faster transistors. Sometimes this number is called "process node", which is really just jargon for talking about different fabrication of chips. Sometimes the 1 quoted number is called "feature size".

At the extreme high end, most of Intel's CPUs today are 14nm, and 10nm & 7nm are on the horizon. It's important to remember this number isn't the entire picture. For example, some chips are using 12nm or 10nm transistors, but the achieve only slight improvement over Intel's 14nm, because this one number only captures 1 spec about the chip fabrication.

When Freescale announced the Kinetis microcontrollers, 90mn with a particular type of flash memory process was advertised. Since then, they've been reluctant to give any info about the fabrication process used. As far as I know, Teensy 3.2 and 3.5 are using 90nm. What size transistors are used in the chip on Teensy LC and 3.6, I do not know. I suspect the K66 chip on Teensy 3.6 may be using something like 65nm, but that's really just guesswork. We also don't know for sure what size is used in the new chip for Teensy 4.0, but there's reason to believe it's 40nm, mostly because so much of it is so similar to their iMX6 chips which were previously advertised as using 40nm.

As the transistor size (or really, the channel length which is distance between drain and source) shrinks, generally the transistor is able to have lower on resistance when the same voltage is applied to the gate. With digital logic, usually propagation delay is mostly determined by the transitor's on resistance and the amount of capacitance on the wires and gates of other transistors it drives. Generally the rest of the stuff in the chip also grows smaller, so capacitance also goes down, again giving faster performance for small sizes.

But there are tough trade-offs to be made. As the transistors get smaller, there's a lot less surface area for the gate to electrostatically couple to the silicon below. So smaller & faster transistors require incredibly thin insulation between the gate and drain+source. The thickness of the chip's oxide layer puts an upper limit on the voltage the transistors can withstand. There are many other trade-offs, especially regarding power consumption and heat, which also drive these modern chips to require lower voltages.

The finer details are a huge and complex subject. Sadly, much of this stuff is shrouded in secrecy these days. Searching for info is also quite difficult, because the internet is filled with articles written by business analysts (or I supposed you could call them by other less charitable terms) who throw around phrases like "process node" without the faintest idea of how a mosfet transistor really works.

But as a general trend, faster chips tend to be made with smaller transistors, and the reduction in size tends to require lower voltage and gives less resilience to damage if accidentally exposed to higher voltages.
 
Hi Paul, I generally get what you're saying, but the naive person in me is still wondering what it would cost to put some schottky diodes in front of each IO pin to act as voltage sinks if you exceed 3.5V, or whatever other protection, polyfuse, whatever. Or at the very worst, it would kill that specific IO pin without killing the whole chip (a simple non resettable fuse on each pin).
I know it adds costs, but then again I can buy (probably used) pentium 4s for less than $30, so at that pricepoint for a much lower/slower chip that's meant for tinkering (or I guess that's the problem, likely it is not meant for tinkering unlike an arduino chip), I was kind of hoping for some level of protection.
 
Think of the many mice, rats, apes, and pigs which are killed during the development of new médicaments or perfumes...

And think of what you will earn when you’ll have sold the first 5000 units of your project. Then, the loss of a single Teensy will be neglectable and you’ll even use this expense for your income tax reduction.

No reason to be sad. Buy another one, move on, and don’t make the same mistake twice. I/O protection either adds cost or sacrifices speed, both can be avoided by professional handling of your equipment.
 
If I happened to roast a Teensy accidentally, I even wouldn’t post that here. No need to advertise my silliness in public since whatever happened, it was my fault. The correct procedure is trashing the dead part shamefully and silently, before ordering a new one. ;)
 
Back
Top