Future Teensy features & pinout

DavidWH1968 said:
As far as their libraries go they are what most any other manufacturer provided libraries are: Turned out quickly and not optimized. I've run into this with virtually all supplier related libraries.
Click to expand...
I agree. In all fairness to Adafruit, their libraries are not in any way intended for production applications. They have always been clear their libraries are for rapid deployment of Adafruit products by hobbyists, many of whom are beginners. For instance, Adafruit created the CircuitPython fork or MicroPython specifically to make it easier for beginners to use. They did not implement features that engineers would likely require, but that hobbyists are unlikely to understand or want. For instance, no threading and no interrupts.

I can't say this all that much different from my experience with much larger hardware companies like Atmel, ST, or Microchip. I always need to modify, and sometimes rewrite, manufacturer-supplied libraries to get production-ready functionality, reliability, and performance. I think this is OK; the intent of these libraries is not to be deployable solutions, rather to demonstrate the underlying hardware or technology in "application note" style.

Sorry for going off topic.
 
mlewus said:
I agree. In all fairness to Adafruit, their libraries are not in any way intended for production applications. They have always been clear their libraries are for rapid deployment of Adafruit products by hobbyists, many of whom are beginners. For instance, Adafruit created the CircuitPython fork or MicroPython specifically to make it easier for beginners to use. They did not implement features that engineers would likely require, but that hobbyists are unlikely to understand or want. For instance, no threading and no interrupts.

I can't say this all that much different from my experience with much larger hardware companies like Atmel, ST, or Microchip.
Click to expand...
This is true, and many manufacturers have followed suit since it speeds the development of ever more complex libraries. These also help to get those of us that have more experience up and running more quickly. So definitely not disappointed. (Just look at an API from a company like Intel and you'll know what I mean!)

This being off topic is really brought on by the drooling (on this end, too) over the prospect of an RT1176 board from PJRC! I like the development board from NXP for all its features, but the size of that board is just too much for my project. And it's nature of being ultra complete is what defines the size. But also increases the complexity of getting it up and running reliably, or adds time desoldering many parts on a brand new $200 development board.
Hopefully the final product from Paul S. will include the FFC LCD header and somewhere to solder one or two SDRAM's. A board about half the size of an Uno would yield a nice form factor and plenty of room to be very expandable with SDRAM, Flash, and the LCD.
 
DavidWH1968 said:
Hopefully the final product from Paul S. will include the FFC LCD header and somewhere to solder one or two SDRAM's. A board about half the size of an Uno would yield a nice form factor and plenty of room to be very expandable with SDRAM, Flash, and the LCD.
Click to expand...
I agree. IMO the Teensy 3.x and 4.x boards were successful because they were very small and orders of magnitude faster than an Atmega Arduino. The recent release of the Pi pico and its “M0+ on steroids” (100+ mhz, dual core, PIO for high speed IO processing) makes that board a drop in fit for many small projects previously well suited to the Teensy, as long as it does not require fast floating point.

If Paul were to make a slightly larger board that was more expandable and easier to debug I think it would open up a market segment for him. At the very least it would make it possible to build projects that more readily absorb all of the horsepower that the RT1170 brings to the party.
 
Regarding the teensy 0.1" header which is so fantastic for breadboarding stuff, but to go faster maybe denser is needed. Maybe make a teensy with 0.50" posts, and sell adapter board for 0.1" header to make bread boarding soldering easier. Also for higher speed I think at some point more grounding is needed G-S-G to keep things under control.
 
kito said:
Maybe make a teensy with 0.50" posts
Click to expand...
please not.

Also for higher speed I think at some point more grounding is needed G-S-G to keep things under control.
Click to expand...

A optional extra shielding, housing would make sense to sell... the faster Teensies more and more become a transmitter.. a lot of EMI :)
Try to use a DAB radio in the same room! (a little desaster...works..works not...works..works not..) it began with the 3.6 but was easy to handle there.
T4x is a little worse. Especially, if you try to do something which involves fast pins.. oversampled I2S, very fast SPI etc.
The 1170 has the potential to send on GSM frequencies!

Don't know wether the 1170 has spread spectrum PLLs - if yes, we should enable it by default, perhaps...
T4 has spread spectrum on PLL2 only.. tried that..does not help that much to enable it (did not try much to run the Arm-Core on PLL2, too - this might help more - tried only once, and it became unstable - might be I used wrong or suboptimal settings)
 
Last edited:
What about 0.050 pins with a pin out where every other pin gives you a Teensy 4.0 pin out and the pins between bring out extra signals? Cut off the even pins and the 0.050 device goes in a breadboard. OK, throw stuff now.
 
kito said:
Regarding the teensy 0.1" header which is so fantastic for breadboarding stuff, but to go faster maybe denser is needed. Maybe make a teensy with 0.50" posts, and sell adapter board for 0.1" header to make bread boarding soldering easier. Also for higher speed I think at some point more grounding is needed G-S-G to keep things under control.
Click to expand...
I much prefer standard 2.54mm/0.1" mounting. 0.050" headers, especially female headers, are more difficult to come by and more expensive. And the only way you can mount them is with a printed circuit board. Other manufacturers have developed boards with standard headers, and where they need to break out more pins they use a high-density connector: Rasp. Pi Compute Module, PyBoard type "D" etc. I think this is more practical as it allows low-density applications to use breadboards or hand wiring. High-density applications will need a printed circuit board anyway so a high-density connector is not a problem.
 
Frank B said:
please not.
Click to expand...

Agreed! The idea of 50mil headers makes me cringe! It also defeats the idea behind these boards: A convenient and rapid way to prototype using breadboards.

I'm also in agreement about the RF noise these will begin generating. The pin switching speed on the RT1062 is at most 200MHz with edge rates that echo into the gigahertz range already!. Some extra ground pins scattered throughout the headers would be helpful, maybe every 4th or 5th pin. It would mean sacrificing some GPIO's, but would enable high speed switching with less hassle in addition to electromagnetically quieting the board somewhat. The omission of a periodic ground pin, however, does not kill off the high speed switching. It would just mean that during development we have to continue doing what we've already been doing: Adding our own noise suppression as needed and appropriate. I'm not actually having any noise issues with my current project which is continually switching pins at 30MHz to 125+ MHz. I would only see a problem if adding something like a 200MHz SDRAM has to be done through the pin header, and hopefully that would happen using a footprint on the new board.

We'll just have to wait and see.
 
Spectrum from my SSA3021X using small B-field probe 1cm above T4.0:
t4_emi.png

background level from same probe:
emi_background.png

These are 0 -- 2.1GHz range, with T4 running at 600MHz clock.
 

Attachments

  • t4_emi.png
    t4_emi.png
    15.5 KB · Views: 99
@Mark: Great! Thank you.
Can you repeat this with FPU usage, and I2S enabled? (don't know if the if makes a difference wether there is a load on the pins or not.. maybe try with audioshield added?)

My DAB Radio 2 Meters away denies receiption when doing things with Teensy (ok,overall receiption is weak anyway in my room)

SDR# shows many peaks (almost all frequency-ranges) with Antenna ~ 1.5 Meters away from Teensy - without USB traffic.
 
Place in a metal box with ferrite toroid on the USB cable?

A different T4 with audioshield and HX8357D TFT attached:
t4_audio_emi.png

I think that one was programmed for 150MHz
 
This noise is not from the frequency of the CPU, but is more from the edge rate when switching. The RT1062 does have adjustable edge rates, so if you are using a lower switching frequency you can slow the switching speed down to reduce those high frequencies somewhat. If you are switching the pins at more than 150MHz the noise is unavoidable. But for switching speeds below 50 MHz the edge rate can be slowed considerably to reduce the high frequency RF noise.
 
In fact, it's both <- that was with a Teensy 3.2, many years ago (turn on Audio).
All pins disabled.

Single pin: With T4 and using 5cm wire, I can receive it with WFM in best quality HIFI stereo 10 Meters through two rooms (walls reinforced concrete) with a cheap tuner.
You're right you can modify the PAD settings. It was done for this test...:)

Don't get me wrong. I'm not complaining. But it is an issue in some cases. Theres nothing Paul can do about it (or, maybe by enabling the spread spectrum features for the 1170)

Now lets not misuse this thread further .)
 
Last edited:
Still, the noise levels are fairly minimal. Mostly less than -70db above 1GHz, and nothing higher than about -55db. This type of noise is easily suppressible with grounds added between transmission lines.
I'm using a harness with 10 grounds in it, isolating data lines and control lines alike, and the grounds come in very close to the pins before diverting to a common ground, and I'm getting a transmission rate of up to 50MHz on DUPONT WIRE with 17 of the 21 lines active (fastest edge rate, highest transmission strength). Haven't noticed any interference with anything in the 2.4GHz range, either. (SSD1963 based 7" display, and the display itself is making FAR more noise than the Teensy 4.1.)

One note about these NXP iMXRT chips: I'm having much better luck with them than I've ever had with signal output than the SAMD51 ever gave me in terms of consistency and signal integrity. The SAMD51's outputs seem to get unstable past about 20MHz switching. The iMXRT has allowed me to get signal rates up to 125.5MHz on 6" DuPont wire while testing. Nice to use a part that actually meets its I/O specs. :)

Can't wait to get a hold of an RT1176 or RT1172!
 
Last edited:
Yes, the signal quality is good.

Edit: In Europe te least, GSM can be at 900MHz (parts).
Air traffic is in the >100MHz Range. Transmitting an ILS signal would be not a good idea ;)
 
Good idea. After all, this is supposed to be a discussion about the next Teensy, not signal integrity.
Sorry for the off-topic discussion.
 
Hi, I am concerned about that glitches that would occur on adjacent pins when toggling at high speed rather than EMI emissions which is another issue maybe have to be handled externally by the designer. Also the impedance mismatch on the shorter pitch header would be better using tighter pitch connector assuming we are trying to get closer to 50-75 ohms range rather than the 100ohm range the 0.1" pitch would see. I guessing slew rate can be controlled to slow the IO pins so most users would can run slower i2c/ serial stuff in this G-S-S-S.. pin out.
 
The "absolute maximum" switching frequency of the RT117x series is listed as 208MHz (i.MX RT1170 Crossover Processors Data Sheet for Consumer Products, Rev. 0, 12/2020, page 50, table 38). Since the maximum configurable switching frequency for the RT1062 is nearly identical at 200MHz, I don't see that there would be any real problem compared to the T41 with on exception: There are about 100 more I/O lines on the RT1176 versus the RT1062 and that could bring more possibility of glitching due to cross-talk since the traces will, obviously, be thinner and closer to each unless the layer count of the board is increased. I am still hoping for the double row of pins at either side of the board, mostly because I am an I/O hog :), but giving up 1/3 of the I/O pins in exchange for a ground pin using the format G-S-S would be fine. And the pins that are most likely to be used for high speed I/O are really the only ones that would need it, so we would still end up with about 80 GPIO pins if there are four rows of 24 pin headers to work with.
 
I like the double rows, too.
The "problem" with the double row is, that soon there will be users who solder both and place the Teensy into a breadboard and wonder why there is smoke or something does not work.
Perhaps with 2nd row in a non standard distance to the first, so it does not fit into a bradboard? But then you can't use normal connectors too, and need a custom PCB for everything.
 
The solution is to solder one row on either side so they will insert into the breadboard, the remaining rows get soldered facing up when inserted into the breadboard. all pins available and still breadboard friendly.
When producing an expansion board for this type of setup, all pins are soldered to the Teensy for socketing on the expansion board. I don't see there being any problem. If someone puts their pin headers on wrong then that is their fault, just like it would be their fault today.
 
thinking of the 2nd row.. i had the idea, it could be a (oh my english..searching for the right vocable.. ) mouser says it is a "Card-Edge-Connector" (German "Leiterplattenverbinder")
First row - normal, pinheader - second "row" a Card-Edge-Connector - this would allow even more pins, as it has two sides! And the old ones (2.54mm ) are very robust, too.
Even inexperienced can easily solder a cable or something else if the connector is not desired.

Yes, a very unusual solution...
 
You must log in or register to reply here.
Back
Top