Future Teensy features & pinout
- Thread starter PaulStoffregen
- Start date
Mcu32
Well-known member
MichaelMeissner
Senior Member+
I've not looked into it in detail, but I thought Xtensa used ARM at its basis. Perhaps not.
RISC/V uses the old MIPS ISA.
Basically, it was a hope that since NTP bought out Marvell (which provides wifi & bluetooth) support, that someday the 11xx IMX line would include wife/bluetooth by default.
With the T3.x gone, I think this oct-2019 thread (10-02-2019) should get again more attention and maybe some observations from Paul.
susan-parker
Member
The chips are now available... but which version: i.MXRT1175 or i.MXRT1176, and Consumer (at 1GHz), Industrial or Automotive (both at 800MHz)?
https://www.mouser.co.uk/c/semiconductors/embedded-processors-controllers/microcontrollers-mcu/arm-microcontrollers-mcu/?core=ARM%20Cortex%20M4%2C%20ARM%20Cortex%20M7&m=NXP&maximum%20clock%20frequency=400%20MHz%2C%201%20GHz&series=i.MX-RT1170&instock=y
It's a lot of work to design, develop, and test a new processor of this complexity. Even a "simple" design like a Teensy if one is setting up to do large scale manufacturing (hacking out a one-off as a hobbiest isn't the same deal).
The Teensy 4.1 is stonking, a new Teensy needs to be significantly better in performance to make the development worthwhile as the sale price is going to be significantly higher given the current chip prices (Mouser is at c. $18 in 100 up quantities, compared to c. $11 for the iMXRT1062).
https://www.mouser.co.uk/c/semiconductors/embedded-processors-controllers/microcontrollers-mcu/arm-microcontrollers-mcu/?core=ARM%20Cortex%20M4%2C%20ARM%20Cortex%20M7&m=NXP&maximum%20clock%20frequency=400%20MHz%2C%201%20GHz&series=i.MX-RT1170&instock=y
It's a lot of work to design, develop, and test a new processor of this complexity. Even a "simple" design like a Teensy if one is setting up to do large scale manufacturing (hacking out a one-off as a hobbiest isn't the same deal).
The Teensy 4.1 is stonking, a new Teensy needs to be significantly better in performance to make the development worthwhile as the sale price is going to be significantly higher given the current chip prices (Mouser is at c. $18 in 100 up quantities, compared to c. $11 for the iMXRT1062).
PaulStoffregen
Well-known member
PJRC is not currently in a financial position to launch any new Teensy using an expensive chip like RT1176.
Teensy 4.0 and Teensy 4.1 have been continuously in stock since November 2022, and we had them in stock about 75% of the days in early-to-mid 2022 and late-2021 when chip shortages were the most severe. This didn't happen merely by luck. PJRC took a pretty serious financial risk to buy large numbers of all the chips, with scheduled delivery dates much earlier than needed. We started this path in early 2021 when we saw lead times changing, and it mostly paid off for Teensy 4.x (and we kept Teensy 3.x in stock until the end of 2021). Until just recently, NXP has delivered almost every order quite late, and on the chips for Teensy 3.x they delivered only a tiny fraction (which is the reason Teensy 3.x is now discontinued). Even in the first half of 2023, we came very close to running out of Teensy 4.0 or 4.1 many times. It's been a tremendously stressful 2 years (after the first pandemic year), running out of stock on the old Teensy models and almost running out many times on the new ones, while simultaneously taking financial risks even most larger businesses would probably consider to be quite insane.
Now that the world's semiconductor supply is finally recovering, and just now NXP is starting to deliver chips we has requested delivered months ago, we now have the opposite problem. If there's one thing Robin and I are pretty good at, it's planning ahead. We did properly budget for the scenario of chip supply eventually catching up, so we will survive this. But it does mean PJRC simply won't be in a position to buy a large enough number of any new expensive chip until at least well into 2024. It's far from an ideal situation, but had we not taken this path, most of 2023 would have had Teensy 4.0 & 4.1 going in and out of stock, like so many of the other dev boards on the market have done through most of 2022 and 2023.
When I started this thread in October 2019, I had a lot of enthusiasm for upcoming chips and felt RT1176 or maybe RT1070 (which never happened) would become future Teensy boards. But at the end of 2019 it became clear RT1176 was a long way and so were other chips, so Teensy 4.1 was made with the same RT1062 to give access to more pins and features, and it released during the first months of the pandemic.
3 stressful years have taken their toll, mostly on the software side. So many things I wanted to do in 2020 are still barely even started. Right now, with the old products discontinued and finally stable supply of chips for Teensy 4.0 & 4.1, I'm determined to catch up to years of falling behind on software and documentation. Launching new hardware would mean even worse neglect and backlog on the software side.
But even if software and documentation weren't an issue, because of risky purchasing decisions Robin and I made since early 2021 to deal with shortages, PJRC simply won't be in a financial position to make a new Teensy using an expensive chip like RT1176 until probably mid-2024. I know this probably isn't the fun and exciting news anyone wants to hear, but after I've written so many messages with such enthusiasm about RT1176 in 2019-2020, I wanted to give you this explanation of the reality PJRC is facing today. PJRC is only a tiny 4 person company (plus contract manufacturers) and I believe we manage to do quite a lot for our size, but the reality of today's situation is a product using RT1176 or any other similarly expensive chip just won't be financially possible for a while while we use up a large number RT1062 chips.
Teensy 4.0 and Teensy 4.1 have been continuously in stock since November 2022, and we had them in stock about 75% of the days in early-to-mid 2022 and late-2021 when chip shortages were the most severe. This didn't happen merely by luck. PJRC took a pretty serious financial risk to buy large numbers of all the chips, with scheduled delivery dates much earlier than needed. We started this path in early 2021 when we saw lead times changing, and it mostly paid off for Teensy 4.x (and we kept Teensy 3.x in stock until the end of 2021). Until just recently, NXP has delivered almost every order quite late, and on the chips for Teensy 3.x they delivered only a tiny fraction (which is the reason Teensy 3.x is now discontinued). Even in the first half of 2023, we came very close to running out of Teensy 4.0 or 4.1 many times. It's been a tremendously stressful 2 years (after the first pandemic year), running out of stock on the old Teensy models and almost running out many times on the new ones, while simultaneously taking financial risks even most larger businesses would probably consider to be quite insane.
Now that the world's semiconductor supply is finally recovering, and just now NXP is starting to deliver chips we has requested delivered months ago, we now have the opposite problem. If there's one thing Robin and I are pretty good at, it's planning ahead. We did properly budget for the scenario of chip supply eventually catching up, so we will survive this. But it does mean PJRC simply won't be in a position to buy a large enough number of any new expensive chip until at least well into 2024. It's far from an ideal situation, but had we not taken this path, most of 2023 would have had Teensy 4.0 & 4.1 going in and out of stock, like so many of the other dev boards on the market have done through most of 2022 and 2023.
When I started this thread in October 2019, I had a lot of enthusiasm for upcoming chips and felt RT1176 or maybe RT1070 (which never happened) would become future Teensy boards. But at the end of 2019 it became clear RT1176 was a long way and so were other chips, so Teensy 4.1 was made with the same RT1062 to give access to more pins and features, and it released during the first months of the pandemic.
3 stressful years have taken their toll, mostly on the software side. So many things I wanted to do in 2020 are still barely even started. Right now, with the old products discontinued and finally stable supply of chips for Teensy 4.0 & 4.1, I'm determined to catch up to years of falling behind on software and documentation. Launching new hardware would mean even worse neglect and backlog on the software side.
But even if software and documentation weren't an issue, because of risky purchasing decisions Robin and I made since early 2021 to deal with shortages, PJRC simply won't be in a financial position to make a new Teensy using an expensive chip like RT1176 until probably mid-2024. I know this probably isn't the fun and exciting news anyone wants to hear, but after I've written so many messages with such enthusiasm about RT1176 in 2019-2020, I wanted to give you this explanation of the reality PJRC is facing today. PJRC is only a tiny 4 person company (plus contract manufacturers) and I believe we manage to do quite a lot for our size, but the reality of today's situation is a product using RT1176 or any other similarly expensive chip just won't be financially possible for a while while we use up a large number RT1062 chips.
susan-parker
Member
Dear Paul,
Thank you for your detailed explanation, your candour with the details is much appreciated.
Best,
Susan.
Thank you for your detailed explanation, your candour with the details is much appreciated.
Best,
Susan.
Thanks Paul for clear and candid answer.
I'm confident, that when situation stabilzes, your enthusiasm for novel groundbreaking approaches will return.
The Teensy 4.0 has been great. The faster USB makes a big difference and, for the analog interfaces, and I am pretty happy using the SPI with external ADCs and/or DACs (whispered: they're better). So, the T4.0/4.1 are pretty much all I need. The only thing missing so far, is that I dream of using FlexIO for parallel transfers for the newer ADC's, but have not had the free time to work it out.
chensx2012
Active member
Paul, maybe the donations from Teensy fans can help with the development of new teensy, if you consider this way.
Blackaddr
Well-known member
I'd much rather see PJRC continue to stabilize as a company than risk sink or swim with a new design. I'm definitely in the camp of always wanting more memory, more compute power for audio DSP, but PJRC's greatest contribution to this community is creating new generations of Makers. The core driver of that is not clock speeds, it is beginner-friendly tools and software libraries.
I'm glad to hear you've made the decision to ensure we have PJRC around for years to come.
I'm glad to hear you've made the decision to ensure we have PJRC around for years to come.
Paul does not need me to say this, but I think his position is right, and I agree with Blackaddr as well. I am excited for developments in the software. The Teensy still has a lot of untapped potential.
In terms of on-board features, there are always trade-offs in size, power, and cost. Admittedly, for something like video, it can also be about how many pins you need to access and whether they are available. But, for simple things, there are a lot of peripherals being sold on ebay, amazon, sparkfun, etc, that can work with the Teensy.
So that said, if someone needs a new peripheral for the Teensy, a new capability or a higher performance "something", and wants to sponsor it, I am happy to design it and build a few. Please contact me.
In terms of on-board features, there are always trade-offs in size, power, and cost. Admittedly, for something like video, it can also be about how many pins you need to access and whether they are available. But, for simple things, there are a lot of peripherals being sold on ebay, amazon, sparkfun, etc, that can work with the Teensy.
So that said, if someone needs a new peripheral for the Teensy, a new capability or a higher performance "something", and wants to sponsor it, I am happy to design it and build a few. Please contact me.
PaulStoffregen
Well-known member
I recall seeing your message about this ADC some time ago. Can you remind me with a link to that thread?
Indeed, and FlexIO is one of the places where more work and a lot more documentation is needed. I recently played with FlexIO for a special project, only the 2nd time I've created my own custom FlexIO peripheral. It's currently a lot harder than it really should be. NXP's documentation is disjointed and confusing!
Long term, I'm hoping we can collect up all the open source FlexIO designs and merge into Kurt's FlexIO_t4 library.
I'm curious to take another look at that ADC and FlexIO, if hardware is available and documentation can be shared in public.
Paul, maybe the donations from Teensy fans can help with the development of new teensy, if you consider this way.
Just buy more Teensy 4.0 and 4.1.
Just buy more Teensy 4.0 and 4.1.
I'm a bootloader addict, I'm aiming to become your largest bootloader customer. Let's see how that goes!
PJRC is hardly alone in this cash-flow situation. I have some visibility into the contract manufacturing world and many/most companies are in the same situation. It's going to take a couple of years for this to sort itself out completely, especially if consumer spending continues to trend down.
I think Paul is being smart to focus on maximizing the use/sales of the current product (which is already a great product!) by continuing to improve software support and overall ease-of-use. I would suggest that should also include continuing to improve documentation, especially of some of the software to help people get started. Most buyers of Teensy are not at the level of many of the software contributors on this forum. Having a great library doesn't help people if they can't figure out how to use it.
I think Paul is being smart to focus on maximizing the use/sales of the current product (which is already a great product!) by continuing to improve software support and overall ease-of-use. I would suggest that should also include continuing to improve documentation, especially of some of the software to help people get started. Most buyers of Teensy are not at the level of many of the software contributors on this forum. Having a great library doesn't help people if they can't figure out how to use it.
@PaulStoffregen
Yes, absolutely! I think that was this https://forum.pjrc.com/threads/72874-dual-sdo-spi I am close to finished with the design. As usual, it is about money to make the first few. After that, I plan to post it. I am pretty excited about it, it should give a lot of performance as an instrumentation input, and still be inexpensive.
One of the nicer features for this part, besides low cost and dual SPI, is that it can do external clocking and triggering. There is an input that starts the sample and then it asserts a line to trigger the host for the transfer. So it needs single word triggered transfers on the host side.
Yes, absolutely! I think that was this https://forum.pjrc.com/threads/72874-dual-sdo-spi I am close to finished with the design. As usual, it is about money to make the first few. After that, I plan to post it. I am pretty excited about it, it should give a lot of performance as an instrumentation input, and still be inexpensive.
One of the nicer features for this part, besides low cost and dual SPI, is that it can do external clocking and triggering. There is an input that starts the sample and then it asserts a line to trigger the host for the transfer. So it needs single word triggered transfers on the host side.
I think the situation Paul explained is understandable and i think was the right path to go. Having a stable stock in nowadays market is really difficult and that's making developers and production's life easier, kudos for that. Like someone said before, Teensy 4.0/4.1 has a lot of unexploded potential and I prefer more software and documentation (the most difficult, annoying but helpful thing to do) is super exciting.
If I were to make a request for hardware the only thing besides access to more pins will be to modify the USB connector position. In order to make the board as tiny as possible (grateful for that) is hard to use the board in an enclosure and access the USB. Moving the connector a bit forward, away from the firsts connection pins will do a long way in solving this issue. Making accessible pins to solder a header and route the USB connector in the base PCB can be an awesome alternative.
I'm super happy about the Teensy, keep going forward!
If I were to make a request for hardware the only thing besides access to more pins will be to modify the USB connector position. In order to make the board as tiny as possible (grateful for that) is hard to use the board in an enclosure and access the USB. Moving the connector a bit forward, away from the firsts connection pins will do a long way in solving this issue. Making accessible pins to solder a header and route the USB connector in the base PCB can be an awesome alternative.
I'm super happy about the Teensy, keep going forward!
Paul, do not underestimate how much trust and goodwill being open with your customers can generate.
You're right too. I keep looking for a TeensyLC replacement, but really all of my projects that use it will use the 4.0 just fine. It is absurd overkill for a keyboard and joystick controller, but $24 is still a reasonable price. What I really valued from the LC is your quality work with the USB drivers. The 4.x has that.
You're right too. I keep looking for a TeensyLC replacement, but really all of my projects that use it will use the 4.0 just fine. It is absurd overkill for a keyboard and joystick controller, but $24 is still a reasonable price. What I really valued from the LC is your quality work with the USB drivers. The 4.x has that.
theboot900
Well-known member
I would love to see a teensy 4.2 with extra pinouts with the 24 bit display pins and software support for them (I believe the 1060 has some 2d graphics functions built in)
Add the pinouts for an sdram ic and can make full functioning 24 bit displays with buffers.
Add the pinouts for an sdram ic and can make full functioning 24 bit displays with buffers.
Make a custom board. There is a library using FlexIO that can be built out to support 24 bits. It's super fast!
Check it out: https://github.com/david-res
isaacjacobson
Well-known member
Completely understandable!
However, is a new form factor of Teensy 4 out of the question? I think many people find themselves somewhere between the Teensy 4x and bootloader based custom pcb:
The Teensy 4x is fantastic for single-use projects, experimentation, and tinkering. Yet, it's pretty restrictive for repeated use in medium-scale projects/products. On the flip side, creating a custom Teensy board using the bootloader, sourcing quality BGA manfacturing, and advanced engineering is pretty out of reach for many (unless you're large scale, but even to get to large scale production, you have to start with small/medium scale production). I suspect that there is a substantial market for a product that falls between the 'hobbyist' Teensy 4x and the 'advanced engineer' use of the bootloader chip in large scale production.
The MicroMod Teensy was the perfect answer to this problem... except there seems to be ongoing mysterious problems and qc issues.
Admin Edit: see discussion of recent MicroMod experience on this other thread (looks like Sparkfun has likely addressed early MicroMod issues)
A few examples of why the teensy 4x is a restrictive form factor:
- pins are the only way to attach to a custom pcb
- the usb +/- is only accessible via pogo pins, which make for pretty clumsy assemlies
- impossible to effectively breakout the SDIO pins to your own pcb
- etc
- summary of restrictions: it's very difficult to merge teensy to your own circuit and bring IO access (usb, sd, etc) to your own enclosure.
Even something with castellated holes would be a huge step in the right direction. Adafruit and Raspberry Pi have tons of products featuring castellated holes–– but I'd rather not move from arduino code / PJRC libraries to RP or settle for Adafruit's less powerful chips.
I would personally greatly benefit from a Teensy in a more integratabtle form factor. I would immediately 10x my order quantities if there was something like that. I've also seen several people voicing their desire for something like that over the years.
I feel like a form-factor that solves that issue would be a really good business decision. With that addition: PJRC would have better offerings for all 3 categories of customers: 1) Hobbyists, 2) small/medium scale production, and 3) large scale production.
In days of tighter budgets, hobbyists will stop buying Teensy a lot quicker than those who need it and use it in a product.
Last edited by a moderator:
PaulStoffregen
Well-known member
I have a lot of mixed feelings about any new form factor. It's not completely out of the question, to answer directly. But it's absolutely the sort of thing that's much easier said than done (well). MicroMod was (and still is) intended to be that new form factor for medium scale manufacturing.
If anything, the challenges Sparkfun has encountered should serve as a warning of how a new form factor which seems like it ought to be simple can turn out to be quite a difficult challenge. I really can't honestly predict that PJRC would do any better if we attempted such a form factor. We use a local contract manufacturer for SMT soldering, whereas Sparkfun has it all in house. At least in theory they should have quite an advantage in ability to quickly address manufacturing challenges.
A manufacturing issue that has come up many times over Teensy's long history is rigidity of the PCB panel during pick and place assembly. We've tried many different panel designs over the years. The current design is working very well. For Teensy 4.0 the panel has 30 pieces arranged as 2 groups of 15. The long side of Teensy uses V-score and the short sides are defined by routed slots. The center of the panel (between the groups of 15) is a solid bar of unused PCB material, which keeps the slots relatively short spanning only 3 pieces. Quite a bit of fine tuning has gone into getting the PCB vendor to v-score at a shallow depth but consistent on both sides, so the panel remains strong and vibrates very little during pick and place, but also works nicely with the panel cutting machine which separates the panel into individual Teensy boards.
Castellated holes requires a completely different panel design. You can't use v-score. Usually tab routing is needed around the entire edge with only very small tabs connecting each piece to the panel. From a panel design point of view, it's pretty much the complete opposite of the panel design we have for Teensy. Done the simply and obvious way, it's a worst case nightmare scenario for panel rigidity during pick and place assembly!
I've talked with some of the other companies who do products with castellated holes. When I ask about manufacturing and panel vibration during pick and place, they suddenly clam up and want to change the subject. Apparently the Particle folks came up with some special way to control this problem, but they consider it secret sauce and won't say a word about it.
My point is manufacturing a product like Teensy or Arduino with consistently good quality isn't easy. Castellated holes or M.2-like form factor seem like relatively simple ideas. But the reality of manufacturing isn't so easy. There a lot of very difficult trade-offs to be made, and the path to making all those decisions is a long and painful journey of low yields and often times customers suffering if the problems manifest after final testing.
So a new form factor isn't completely out the question, but it's not something to be taken lightly. Especially with Teensy's software so in need of more attention, and with Sparkfun making progress on the MicroMod form factor, I'm really reluctant to go down a path where my time and attention gets sucked into manufacturing rather than software. I can tell you from a lot of painful experiences, when those problems come up they tend to suck all the oxygen from the room when it comes to getting any other development work done. Right now, I'm really not feeling much appetite for taking on that kind of risk.
brtaylor
Well-known member
Paul, would you be ok with another manufacturer going that route? I really wanted MMOD to be successful, but I'm hesitant using them in our products since it's our name on the box and our customers would (correctly) blame us for unreliability if it fails in the field. So far, we've only used MMOD on products meant for R&D and have stuck to chip directly on PCB for commercial products.
I've toyed with the idea of a Teensy 4.x aimed at commercial use - probably closer to an Edison footprint with the wider temperature range chips. Planning on investigating more early next year, but the price would likely have to be several times higher than Teensy for it to make sense. Really just thinking about it since the higher volume could help us lower the costs to manufacture our products.
Last edited: