The very last Teensy 3.5 and Teensy 3.6 are available today, limit 2 per person.

Today is my first time in months checking PJRC.com as I had very wisely chosen your excellent products 4 years back when designing a very high performance requirement piece of embedded remote solar-power sensor sampling electronics, and have been very happy with the product ever since with no issues.

It is terribly unfortunate to hear.

The Teensy 4.0 had absolutely horrible power consumption at the same clock rate when compared to T3.5 when I tested it when it first came out a while back.

I run my continuous-execution-needed production code at 8Mhz on the T3.5 (low power modes do not apply to my real-time, 100% duty cycle usage of FreqCaptureMulti use case) and remembered quite clearly the T4 both had inexplicably worse power consumption, and FreqMeasureMulti would report very noisy high-frequency false pulses around real pulses in the same exact hardware and software as the T3.5 in the same place output only clean, valid timestamps of digital edges.

I would have to resolve both issues before porting my project to T4/T4.1. While the latter could probably be fixed or already has been, the power consumption difference was very striking and a deal-breaker for me.

With this discontinuation of an excellent product in which I invested much time, perhaps you will choose to focus efforts now on a suitable low-power replacement? I remember concluding that the absolute best thing for my project was the Ambiq Apollo 3 MCU SparkFun uses in their Artemis, but it didn't come with a library offering perfect, easy encapsulation of input capture hardware buffered pulse timings, nor a form factor containing an SD-card built in, so I ultimately stuck with the Teensy 3.5.

In the future I will have to use a different MCU due to the discontinuation. I love your business and would prefer to stick with PJRC, but I will be looking for an acceptable solution in terms of power consumption for logging timestamps of sub-2khz pulses to SD card continuously and outputting some statistics over serial. This should be far less than 10ma draw. The Sparkfun Artemis would have been able to do this around 2ma @ 3.3V if I remember correctly.

In summary, this unfortunate discontinuation increases the business case for PJRC to focus on a modern low-power-optimized solution which has an SD card slot and T3.5 form factor, to support legacy T3.5-based projects for which the T4/4.1 is not suitable.

EDIT: A cursory search does reveal multiple others corroborating unresolved high power consumption issues with T4.x, especially when compared to T3.x:
https://forum.pjrc.com/threads/69159-Teensy-4-0-at-24Mhz-still-drawing-100mA-Why
https://forum.pjrc.com/threads/70191-Teensy-4-1-using-Snooze-library-with-deepSleep
https://forum.pjrc.com/threads/71957-Putting-Teensy-4-0-to-sleep-does-not-reduce-power-consumption
 
Last edited:
Indeed it's sad these old products came to an end this way. I'm especially regretful for all the people who made special products like modular synths with custom PCBs.

Indeed the power scaling to very low clock speeds was another thing these old chips had going for them. Maybe in future software we'll make some improvements to Teensy 4 power consumption, but it's pretty unlikely to ever achieve very low power like other chips can. It's sad to lose that capability.

But on the other hand, the 4 old Teensy models with Kinetis chips have taken a huge amount of Robin's time and my time over the last 20 months. I can't share behind-the-scenes details, but believe me, it's been a huge distraction from everything else. Selling the very last of them is sort of a relief for us.
 
@Paul - from purely a business perspective it’s the right decision!! This way you can free up your time to work off that todo list. Still sad though.
 
Well it will be sad to officially have the last 3.5 and 3.6 boards sold and gone. And then as you mention on the Teensy 3.2 product page (https://www.pjrc.com/store/teensy32.html), you are expecting one last shipment of the chip used in the 3.2 in September and October, and presumably those well be retired also. While I don't recall an official announcement, the LC page says that you don't expect to make any new LC's. All understandable, given they need 90mm chips.

But as you close down the gear, some minor suggestions that I noticed in looking at the pages tonight. You might want to start updating the comparison pages, and such to identify that only the Teensy 4.x boards will be produced in the future.

One of the things I've been doing is pointing to some pages I've written in the past to highlight the differences. Maybe you want to incorporate some of the ideas in the official documentation, maybe not:

In the 3.2 page, you mention you might be able to use the MQS support (pins 10 and 12) to replace the DAC with a 'resistor-capacitor filter'. But the audio tool should have a link in the page of how to build the filter.

In both the 3.2 and 3.5/3.6 pages, you mention using the PT8211 as a replacement. I would suspect you may want to mention the full audio shield also (assuming the target is somebody who has an old project that just uses a DAC, and is looking for what to buy to upgrade). However, I recall from previous discussions, that some said the chip used in the PT8211 also might be headed towards being retired. If it is on its final legs, it may make sense to start recommending something else.

In terms of the boards, I may have been one of the few people that actively used the prop shield. But it was nice and convenient in that I didn't need a lot of other boards for making props with neopixels and plays some audio. I recall that you've said the motion sensor used in it is now impossible to get. I've personally never used motion sensing, but I believe a lot of people do. But perhaps people want a new version of the prop shield that is based on the audio shield, but includes amplifier for at least a mono speaker, neopixel/dotstar level shifting, and with/without other forms of motion sensing. I would ask if you do something like a prop shield mark II, please don't put the neopixel/dotstar pins on the spi pins. I did once do an uncanny eyes with neopixels, and it was a hassle to switch between neopixel mode (first turning off the CS pins) and SPI mode to run the displays.

As a slightly humorous note, I was checking my stash. I noticed in storing the shields, that I had a revision B audio shield stacked on top of a prop shield. I don't think it would be useful to use both at the same time.
 
Last edited:
Thanks for the 5V compliant Teensy 3.1, 3.2 and 3.5 and support for this long. Legacy systems can be a real time sink for continued support.

I've deployed many 3.5s on adapter boards to replace proprietary 40 pin PIC processors on legacy TTL circuit boards produced for model railroad signaling. They allowed me to eliminate some nasty bugs that caused data unreliability. I hope you may find and release another 5V tolerant processor in the future.

Thanks for the great products and support!
 
Will you still make the 3.x boot loader chip available (in larger quantity orders) if we're
still able to get our hands on the kinetis chips?
 
Personally I am gutted (to say the least) to hear that the Teensy 3.5 has just been discontinued.

I write neuroscience software controlling a wonderful scientific instrument that uses the Teensy 3.5. So not only was there a special circuit board and instrument made to use the Teensy 3.5, custom software has been written (by me) and used to control the instrument, and the instrument being a piece of scientific kit, had to be rigorously tested. Now all this has to be redone - and my understanding is that the Teensy 4.1 is not even completely compatible with the 3.5 for this instrument.

I think that the instrument manufacturers would continue to buy the Teensy 3.5 at double or triple the price. Heck of a lot cheaper than redoing the whole design, manufacture, software and testing process.

PJRC should look to great companies like National Instruments. NI doesn't demand software rewrites when they introduce a new data acquisition board, NI continues to make and sell older boards, and/or make new boards that are downward compatible with the older ones.
 
Personally I am gutted (to say the least) to hear that the Teensy 3.5 has just been discontinued.

I write neuroscience software controlling a wonderful scientific instrument that uses the Teensy 3.5. So not only was there a special circuit board and instrument made to use the Teensy 3.5, custom software has been written (by me) and used to control the instrument, and the instrument being a piece of scientific kit, had to be rigorously tested. Now all this has to be redone - and my understanding is that the Teensy 4.1 is not even completely compatible with the 3.5 for this instrument.

I think that the instrument manufacturers would continue to buy the Teensy 3.5 at double or triple the price. Heck of a lot cheaper than redoing the whole design, manufacture, software and testing process.

PJRC should look to great companies like National Instruments. NI doesn't demand software rewrites when they introduce a new data acquisition board, NI continues to make and sell older boards, and/or make new boards that are downward compatible with the older ones.

Do note, that the main reason for eliminating the Teensies is the chip manufacturer is discontinuing the chips used making the Teensy LC, 3.2, 3.5, and 3.6. If PJRC can't buy the chips, they can't make the Teensies. As I understand it, the whole maker market (PJRC, Adafruit, Arduino, etc.) consumption of chips, is probably too small to be noticed, compared to all of the embedded devices using them. I.e. we are not a big enough of a market to keep a multi-billion dollar fab running profitably churning out the older 90 nanometer chips. Chip manufacturers look at their bottom line, and they have decided to retire the 90nm fabs in favor of newer fabs that have denser chip packing.

As an example, I work at IBM, and while I can't comment on details that haven't been made public, during the design of the current power10 processors, the Global Foundries fab that IBM used would not be able to deliver the 10nm chips that IBM had planned to use. IBM had to hunt around, and eventually switched to Samsung:

 
We’ve ordered some National Instruments kit for a project at work. We need one PC compatible ADC with a bridge amplifier, though the selected card has two channels - always nice to have room for expansion, it’s often wanted by the client. Ordered a couple of weeks ago, hoping for delivery from Hungary in early October. The price? Just shy of £5k.

Different worlds…
 
Glad to know that the bootloaders will still be available. There seems to be a batch of MK20DX arriving in September and more coming in January and we're looking forward to receiving them.

Separately, can we be optimistic about looking forward to faster teensies (T5?) being designed by next year?
 
Just came back to an old project today (which uses 3.6), minor software update. This lead me to re-download teensyduino and re read my old source code.

Want to chime in and say Thank You! :D

Last time I was on this project was Arduino 1.6 or 1.8, now seeing the new Arduino 2+ documentation, and how easy you've made it to install teensyduino, and all the options you've given the community to work with teensy. Really glad to see, you've helped so many people get into this I'm sure.

Cheers from Canada!
 
For custom board developers: NXP longevity program

Our current product depends on 150 MHz Kinetis K60 and K66 parts, and we would be in deep trouble if these devices become unavailable. Seeing the announcement of PJRC's last-device availability caused us to look into the question. We are relieved to see that NXP has a product longevity program, and our Kinetis parts appear to be on that program. (We do not use the K64F so I am not sure of its availability; I think it may be). Our product is not very sensitive to microcontroller price or to lead time (given recent supply problems, we have plans in place to mitigate these issues).

NXP Product Longevity

This leads to a few results:
  • It is likely that one can buy parts but no promise of lead time
  • No promises about price
  • non-PJRC custom boards can probably continue

PJRC has a stellar reputation for manufacturing test of Teensy boards, and making a Teensy3.5 or 3.6 "clone" would be a major challenge. That said, it may be worth asking: Could a developer develop a process around Paul's design under protection of an NDA? PJRC can still sell the bootloader chip.
 
non-PJRC custom boards can probably continue

Yes, if you can get the chips. And if PJRC can continue getting the bootloader chip, though I can confirm we have a large number in stock today, so hopefully we'll be good for a while.

NXP's product longevity program sadly makes no assurances of lead times or pricing. After waiting more than 2 years for our 2021 orders and seeing at least 3 retroactive price increases (well beyond inflation and semiconductor market trends), and then finally getting only small quantity on those 2021 orders with the rest all pushed out to 2024, we just couldn't continue Teensy 3.5 and 3.6. The way PJRC operates ongoing shortages becomes a huge distraction and drag on development.


Separately, can we be optimistic about looking forward to faster teensies (T5?) being designed by next year?

Extremely unlikely in 2024. If we make anything new in 2024 (a very big "if"), odds are it would be a lower cost replacement for the Teensy 3.x and LC which (for all practical purposes) we lost in 2022.

Options for a faster chip are pretty limited. So far NXP only has RT1170 which is faster, but it's only made it packages too physically large to fit in Teensy's form factor.

They have a new line of microcontrollers coming "soon", and while some have pretty interesting new peripherals, overall their performance is on-par with Teensy 3.6. Predicting the future is always difficult, but as a general rule when the US or EU or China or world economy is in recession, or fear of coming recession, large corporations like NXP tend to scale back ambitious new product plans and focus on projects seen as lower risk and needing less investment. I would be very surprised if we see much new from NXP in 2024 other than their new 150 MHz parts.
 
Extremely unlikely in 2024. If we make anything new in 2024 (a very big "if"), odds are it would be a lower cost replacement for the Teensy 3.x and LC which (for all practical purposes) we lost in 2022.


Would love to see a drop-in replacement for the LC come out! including the DAC pin. Not sure how many other people are in the same situation, but I've designed a product that uses the LC and have ordered several hundred of them with only about 30 left in stock. The LC is a great product and has plenty of processing power for lots of embedded applications.
 
Would love to see a drop-in replacement for the LC come out!
Couldn't agree more. For a lot of my little projects, the LC processing power is enough. And I do like the 5V output on pin 17 for driving NeoPixel strips/rings.
Wish I had bought more LC's in advance...

Paul
 
Would love to see a drop-in replacement for the LC come out! including the DAC pin. Not sure how many other people are in the same situation, but I've designed a product that uses the LC and have ordered several hundred of them with only about 30 left in stock. The LC is a great product and has plenty of processing power for lots of embedded applications.

While it isn't a Teensy, and you would have to completely redo your pinout, you might want to look at the Adafruit ItstyBitsy line. It has one pin that has the level shifter built-in, and the M0/M4 versions have 2 DAC pins (the 32u4 and nRF52840 versions don't seem to have DAC). That being said, Adafruit also has problems getting M0 and M4 parts (the M4 ItsyBitsy is currently sold out). Who knows if they are going to be able to keep getting them.
 
They have a new line of microcontrollers coming "soon", and while some have pretty interesting new peripherals, overall their performance is on-par with Teensy 3.6.

Assuming you are talking about the MCX line here. They do look super interesting! Did you see the inference demo they used the AI co-processor for? Pretty cool stuff. A Teensy with that sort of capability sure would be interesting. I have a dev board, so I don't think it will be too long before they are in the market. We've been using their LPC55 line lately on a few products. It's starting to get some decent availability, and could be appropriate for an LC.

Personally I'd like to see a return of the cap sense stuff. A large footprint Teensy with ~32 capable touch inputs would enable some really cool MIDI controller type stuff.
 
Back
Top