May want to revisit Teensy pricing

[metalgimp]

There's a new kid on the block that may severely impact the Teensy product line.

Adafruit's "Espressif ESP32 Development Board"

It has a load of features along with built in wifi. It currently is listed for $15 and is labeled as a "developer board." I do not know if that will be the final price.

All of my development thus far (four fab'ed boards) has been around Teensy 3.2. I cannot afford the latest (3.6). No need to flame me; I like the Teensy product line, and I don't want to go through a whole new board layout, fab work, and testing.

If interested in what I have accomplished (my chance to show off):
1) A robotic arm controller with FTDI.
2) A touch sense swipe reader. (Allows more than a binary result from touch sense.)
3) Second generation robotic arm controller with recalibration LEDs and improved layout.
4) Third generation robotic arm controller with independent power coupling, breadboard, improved LED panel, power regulator, two headers for the "Color 320x240 TFT Touchscreen" plus a breakout for a third.

For pics https://www.facebook.com/sean.walton.395

BTW, the purple test board didn't work for me. Probably my error. A lot of money wasted.

 

[BJB]

To be clear, respect Limor Fried for her principles and goals - she seems to have built an ethical and sustainable business. But much adafruit stuff is not industrial grade in either code or hardware. Use mostly T3.2s at work, and have used two T3.5s for home stuff. But neither myself nor my employer can afford the risk of using some of the stuff from sparkfun or adafruit, and certainly not banggood...

As for Tensilica stuff, have had some rather interesting problems with their stuff - so approach with caution; and do not want to use their tool chain as it is 'based' on eclipse.

But am looking at ESP32 for nephew's plant robot school project as the shortest development distance available, as the typical 8 to 10 bit ENOB for ESP stuff will not matter where only doing digital I/O. Also am hoping his project will not be affected by the poor I2C/SPI (three DMA) and buffering issues.

 

[MichaelMeissner]

Given they only had limited quantities of the chip and it is now sold out, it may not be the wisest choice to base your project on those chips. And note, the toolchain/libraries are DYI and not a finished product. So, if you are comfortable with doing everything yourself, and writing library code, etc. go for it.

Perhaps somebody will help you, perhaps they won't. Companies like PJRC and Adafruit fund their development of boards and libraries by the sale of boards. If that is not important to you, go search for a cheaper board, but remember you are on your own in terms of adapting stuff to the board.

 

[metalgimp]

Your points are good. Still, I am going to try them out once available, only with caution and low expectations. If I had the wherewithal in terms of electronics background I might consider the WVOOM32 module by itself. But judging the amount of electrical on their breakout board, I would be very hard pressed to get very far. In fact I would still have to make a breakout board *anyway*, so that might be a fool's journey.

I just wish that the T3.6 were cheaper. When I buy things like that I usual by 2+ with the absolute certainty that I will fry one. So, $40-60 for 2-3 T3.2s is far more reachable than $70-105. I am after all a mere hobbyist, an embedded systems programmer by trade. As an aside, I am very excited that I can actually order board fabs, a technology completely unavailable to me as a kid 40 years ago. T3.2 has paved the way for my making my own boards. Arduino was tempting at first, but when they jumped rails and went with PIC and other low MIP processors, I was glad I went with Teensy. It seems that much of the Arduino world is targeted for "Jughead" hobbyists, not serious personal scientific research.

I have always wanted to dabble in the junction between hardware and software (I am on the team that made the CaptionCall phone and am the go to guy for such). When I was young I bought a 6502-based desktop. It naturally had a whole lot of limitations, one being memory. I came up with banked memory on my own, but had no way to do it. Now G-jobs are reachable and relatively cheap.

 

[Epyon]

I don't think a focus on the price is that important. I am a researcher and I run a spin-off company where we (among other things) develop and sell products based on Teensy, and we integrate the Teensy as a whole. Sure, I could easily save €15/per product by integrating the MK20 on my PCB directly and omitting the bootloader chip, but I'd rather focus on other things that make me more money and save me more time, like device functionality or communicating with the users of our application about their experiences and wishes.

Teensy is a semi-professional product with some very efficient device specific libraries and actually quite good user support. So I think the current price point is just good. Especially if you compare it to the bonkers price of an Arduino Uno and what you get in return for that money.

If you really want to cut pennies in half, you can always turn to sites as Aliexpress to buy clone boards for a ridiculous low price.

 

[tonton81]

why are people comparing apples to oranges? if you like to use another product no one is stopping you, don't start going public about why a completely different product should better the one here. Should we start comparing banggood or aliexpress products prices vs teensies here as well? Indeed not. There is a time and place for this kind of discussion, and this, is neither.

 

[metalgimp]

why are people comparing apples to oranges? if you like to use another product no one is stopping you, don't start going public about why a completely different product should better the one here. Should we start comparing banggood or aliexpress products prices vs teensies here as well? Indeed not. There is a time and place for this kind of discussion, and this, is neither.

Really? This is hardly what I would call a contributive argument. And this is "General Discussion," btw.

Consider this. If there really is a product that can do everything that a Teensy does, wouldn't that need addressing? If the claims of WROOM32 capabilities are true, Teensy will have up its game or lose out on new customers. Return customers are great, but growth comes from new customers. Every successful, long lasting business heeds this rule. Another inviolate rule is: a company must reinvent itself every XX years depending on the industry. Today, that period is shortening substantially.

 

[metalgimp]

...Teensy is a semi-professional product with some very efficient device specific libraries and actually quite good user support. So I think the current price point is just good. Especially if you compare it to the bonkers price of an Arduino Uno and what you get in return for that money.

Arduino Uno (or any other modern ilk) is a professional misalignment--I wholeheartedly agree. Teensy is superior. I look forward to writing a goal-sync OS for it. Honestly it would be silly to productize the OS, so it will be for pleasure only. I am working on a windowing subsystem. These could only be accomplished with enough MIPS. Teensy's MK choice is okay but limited. I have looked into making my own MK breakout, but like I said, I do not have the EE experience to do it correctly.

If you really want to cut pennies in half, you can always turn to sites as Aliexpress to buy clone boards for a ridiculous low price.

Okay, let's stay on the boat of reality. Those parts are "Cheap Chin" (excuse me for the nationalist invective) and have a level of quality that everyone today spurns.

Consider reading my other reply which covers business practices.

 

[tonton81]

This General Discussion as you put it looks more into an advertisement for another product that your so deeply interested in rather than just running your mouth about how teensy is not capable of doing or having what you want, you say you want to extort pjrc to save pennies yet frown upon it's missing features and praise everything the board you are advertising as the best out there, go buy it and enjoy it, theres nothing benefitting this discussion you raised other than you belittleing the teensy product line and advertising other units here. who the hell are you to want extra features, saying teensie is lacking all of this this that and that etc but you want a better price? if you want all that, whats the point of lowering the price if you still dont get all that in a teensy? just go buy your product and shutup, we already heard enough of your "this board is better" advertisements. enough.

 

[Gibbedy]

Gotta say this, just because I was thinking about it a few days ago while looking at some adafruit products.

I'm very happy with teensy pricing for hardware/software/support.

 

[Epyon]

The Adafruit product page clearly states There are no refunds, exchanges, store credit or support in any way for this product. Color of PCB may vary. I don't think PJRC has to worry as long as this is the case. The reason for the success of Teensy is it support and code portability between Arduino and Teensy. The ESP32 is still nowhere there. Yes it has a lot of power under the hood, but little to no means for the general public to access it.

Remember that Teensy is not just the hardware, it's also the Teensyduino software and libraries you're 'buying', or at least funding development of.

 

[defragster]

I'd agree Apples to Oranges. ESP8266 had some value - and some limitations - after a year on the market it became useful for what it was and stabilized - super cheap PCB's pumped out in large numbers as was repackaged many times. ESP8266 boards are $16 at SparkFun - with a limiting flash size.

Just $4 more for the ESP32 version is there - reading its reviews it may be a great build but it is not yet generally useful - maybe in 2017 will have usable support for what it has, depending on your use case that might be what you want. It has a dedicated core for radio work and one for applications. Complete radio functionality isn't usable yet and time may allow it to surpass the ESP8266 and it has some more I/O points - but like apples and oranges there are pros and cons depending . . . price can be critical - but there are lots of other things to consider - and those things change . . .

There will always be cheaper items out there ... Or get a Redboard clone Uno for only $20 - or a genuine ZERO for only $50 ... you don't always get what you pay for - so far with PJRC Teeny I think I have.

 

[PaulStoffregen]

I understand you want a lower price. Everyone always does. But the K66 chip we're using on Teensy 3.6 isn't cheap. You can look it up at various distributors who carry NXP/Freescale to see for yourself. PJRC is also a small & independent shop, and one of the few in this "marker market" that actually invests heavily in new software development. There's no way we can offer subsidized or loss-leader pricing, at least not any way that involves staying in business and keeping up ongoing long-term development & support.

It's my personal hope we'll see the entire semiconductor industry become more competitive as they all move towards 28nm process in the coming years. I have a long history of "learn new tricks" as more powerful hardware become available, and I hope to continue. But such trends and strategic decisions of the semiconductor giants are far outside of my influence. It's entirely possible we'll see things move in other directions, especially with the business trend of mergers, acquisitions and consolidation.

One thing I'm absolutely not going to do is make products with woefully undocumented parts, especially the ESP chips. Most of the peripherals in those chips are secret. Their registers aren't documented. Binary blobs are published by Espressif. Even the ones which are documented have scant info, for this new part only long list of registers with little or no info about what they actually do. Compare Freescale's rather terse reference manual over 2000 pages to Espressif's under-200 page manual, which only lists a small subset of the chip's features, and contains almost no useful info. I can understand how you probably don't care about such thing. Most people who use Teensy probably don't look at the core library code either. But in terms of developing a platform, such closed hardware means PJRC or other 3rd parties can't really develop much platform-level code. With secret undocumented hardware, only Espressif can have that role. PJRC and others are locked out, even if we wanted to develop for these chips.

I know this isn't the answer you want, but it's honest and it's the only one I have for you.

 

[Epyon]

I have always wondered what exactly the business model of Espressif is. How can you release a product with such limited documentation and support, especially knowing that your chips are hugely popular with the maker crowd? Or do they also have an application development team for customers wanting to integrate their chips in their products? Then why bother selling the chips loose anyway?

 

[metalgimp]

I have always wondered what exactly the business model of Espressif is. How can you release a product with such limited documentation and support, especially knowing that your chips are hugely popular with the maker crowd? Or do they also have an application development team for customers wanting to integrate their chips in their products? Then why bother selling the chips loose anyway?

Look at Paul's point on acquisitions.

 

[metalgimp]

The Adafruit product page clearly states There are no refunds, exchanges, store credit or support in any way for this product. Color of PCB may vary. I don't think PJRC has to worry as long as this is the case. The reason for the success of Teensy is it support and code portability between Arduino and Teensy. The ESP32 is still nowhere there. Yes it has a lot of power under the hood, but little to no means for the general public to access it.

Remember that Teensy is not just the hardware, it's also the Teensyduino software and libraries you're 'buying', or at least funding development of.

Careful, I did not find return policies on PJRC. However, the fact there is a warehouse of docs is notable.

 

[metalgimp]

*{DANG BLASTED TABLET deleted my reply.}*

At last a reasoned, non-invective, non-"I'll take my marbles and go" reply. For those who think I am a shill: I devoured MSP, Atmel, Freescale, etc., back when they first introduced while doing my PhD research on sensory net wireless networks. I hold no alliances.

I understand you want a lower price. Everyone always does. But the K66 chip we're using on Teensy 3.6 isn't cheap...

Confirmed. DigiKey:
MK66FN2M0VLQ18-ND - $16.50/ea
MK20DX256VLH7-ND - $6.50/ea
WROOM32 - $9/ea

...trends and strategic decisions of the semiconductor giants are far outside of my influence. It's entirely possible we'll see things move in other directions, especially with the business trend of mergers, acquisitions and consolidation.

This is an excellent point. I find it interesting that Espressif eagerly approached Adafruit for partnership. Adafruit is not a big fish which means that Espessif is looking for a buyout by legitimizing themselves with *anyone* out there.

One thing I'm absolutely not going to do is make products with woefully undocumented parts, especially the ESP chips...

DING! This is the real nail-biter. I spent several days trying to assemble docs with insufficient results. I am used to researching and cobbling together a packet of docs--look at NesC and the associated hardware. The fact that there are few little essential documents corroborated by someone else means a lot.

I know this isn't the answer you want, but it's honest and it's the only one I have for you.

On the contrary, I was convinced with docs.

 

[MichaelMeissner]

I have always wondered what exactly the business model of Espressif is. How can you release a product with such limited documentation and support, especially knowing that your chips are hugely popular with the maker crowd? Or do they also have an application development team for customers wanting to integrate their chips in their products? Then why bother selling the chips loose anyway?

For most chip makers, the DIY/maker community is such a small niche market that it would not add appreciatively to the number of chips sold if they did completely open up the design. As to why bother selling the chips loose, if it doesn't cost much to sell the leftovers, then it adds a little spare change to the bottom line, but doing the full documentation does cost more time and effort.

Another factor is often the add-ons are the intellectual property of another company that is handled with a cross licensing agreement. You would have to get agreement from all of the owners of the IP to release it.

 

[metalgimp]

I'd agree Apples to Oranges. ESP8266 had some value - and some limitations - after a year on the market it became useful for what it was and stabilized - super cheap PCB's pumped out in large numbers as was repackaged many times. ESP8266 boards are $16 at SparkFun - with a limiting flash size.

I won't compare head to head the processors, because I will be accused of being a shill again (in my domain such accusations are taken seriously). Saying "apples to oranges" is an inconsistent comparison in and of itself. Aren't SoC processors comparable?

...what you want. It has a dedicated core for radio work and one for applications.

This is a patch for bad programming. I had no problem, because writing a proper asynchronous I/O is simple. So, in the end I get an extra core for free.

Complete radio functionality isn't usable yet...

Now THAT is something I did not know about and is the crucial piece I was relying on. In other words, I was really asking: "Why is Teensy 3.6 more expensive and yet does not have wireless?" If its true that wireless is not yet solid, I'm no longer interested in it.

 

[Jay1011]

PJRC is also a small & independent shop, and one of the few in this "maker market" that actually invests heavily in new software development. There's no way we can offer subsidized or loss-leader pricing, at least not any way that involves staying in business and keeping up ongoing long-term development & support.

This. ESP32 may eventually be a great part (ESP8266 was a PITA in my opinion), but it will never be backed by the dedication and contributions to the community that Paul/PJRC have made and continue to make over the years. Even if you never buy a Teensy, you're benefiting from these bugfixes and other contributions, so ante up and pay the extra few dollars. You're getting your money's worth whether you realize it or not.

 

[CollinK]

This. ESP32 may eventually be a great part (ESP8266 was a PITA in my opinion), but it will never be backed by the dedication and contributions to the community that Paul/PJRC have made and continue to make over the years. Even if you never buy a Teensy, you're benefiting from these bugfixes and other contributions, so ante up and pay the extra few dollars. You're getting your money's worth whether you realize it or not.

I have a few Teensy 3.6 units and I have a few ESP32 WROOM32 boards. The documentation surrounding the Teensy platform is about 1000x better than the documentation for the ESP32. The ESP32 libraries are like 15-20% done. Large swathes of functionality are non-existent. Documentation for much of the chip is being written "as we speak" so to speak. Every week or so the documentation is updated with more stuff albeit very slowly. I don't see any real binary blobs like Paul was talking about but the code to support the various chip features is very much under construction. There is an Arduino environment for the ESP32 but it too is incomplete. I had hoped to use an ESP32 as a daughterboard type arrangement where my Due or Teensy based projects could use the ESP32 for the radio. Yes, it's kind of silly to use a 240MHz ESP32 as the add-on for a Due or Teensy but the WROOM32 is really cheap so it could kind of work out. However, due to existing pinouts for my boards, I'm trying to use SPI to talk between my stuff and the ESP32. Well, I would need the ESP32 to be in slave mode on the SPI for this. As luck wouldn't have it, the library doesn't support this yet and the documentation basically doesn't exist. However, quite a few of the pieces are already there for wifi and BLE. I think that vanilla bluetooth is lagging a bit behind. I may end up having to switch to the use of a UART (somehow...) in order to talk between the WROOM32 and my other stuff even though I'm trying to match a layout that has no UART run to where the daugherboard would go... It's a mess. And, that should be the ESP32 slogan "It's a mess, but it's cheap" Quite the catchy slogan.

All of this is to say: you get what you pay for. The ESP32 is kind of a fly by night, seat of your pants sort of chip. It's cheap, it's fast, it has a lot of features, good luck making it work. I believe that the audience for the ESP32 is actually larger than that for the Due or the Teensy boards. Go look at the ESP32 forum - it is very active. Look at their GITHub repo for the ESP32-IDK. It is very, VERY active. But, actual, real support is quite lacking. The answer to most inquiries will be "we'll write that part some day" I think that mid to late 2017 the ESP32 will be a nice chip to work with but I'd still be much more comfortable doing 90% of my work on something else and just using the ESP32 as a glorified radio. Just ignore all the other stuff it can do. It's a radio that can do wifi, BT, and BLE. That's where the development is going to concentrate and that's what people are mostly going to care about. I don't know that I'd ever feel comfortable replacing a Teensy 3.6 with that turdball of a processor / module.

 

[metalgimp]

I have a few Teensy 3.6 units and I have a few ESP32 WROOM32 boards. The documentation surrounding the Teensy platform is about 1000x better than the documentation for the ESP32. The ESP32 libraries are like 15-20% done.... I don't know that I'd ever feel comfortable replacing a Teensy 3.6 with that turdball of a processor / module.

Thank you for that fantastic analysis.

 

[defragster]

... should be the ESP32 slogan "It's a mess, but it's cheap" Quite the catchy slogan....

The same was true of the ESP8266. All the progress on that platform seems to have come from reverse engineering a terse Chinese language doc and copious time with hardware. If they have a 200 page doc for the ESP32 that may be an improvement. There is enough info (under NDA?) for Adafruit or SparkFun or onehorse to put an ESP 8266 or 32 on a PCB - but all software support is at the hands of some mysterious overlords.

I think the 'binary blobs' Paul refers to are the bulk of the ~250KB included in the minimum Arduino upload. To get a web serving sketch the upload gets to and over 300KB quickly.

As ColinK notes "vanilla bluetooth is lagging a bit behind" goes along with what I read when I noted 'Complete radio functionality isn't usable yet...'

Like cars - SOC's - are not a simple comparison - Thus Apples to Oranges. Ford, GM, Porsche, etc make a variety of 4 wheeled vehicles - but their interchangeability for 'your use' is based on more than sticker price. Read a 'car depreciation' article - that showed KIA loses a large share in the first year. If you get a good one you like and keep it you don't care - but when trading it in that bargain price correlates to a deferred penalty.

 

[metalgimp]

Like cars - SOC's - are not a simple comparison - Thus Apples to Oranges. Ford, GM, Porsche, etc

Please forgive me for being persistent and dense. A car analogy is specious.

Can you give me 1 (one) feature, operation, wiring, or logic at any level of abstraction which would solidly prove (ad demonstrandum) incomparability? I look at the two with my limited knowledge of the two and have to say that the only incomparable feature is wifi, but that's a given, differentiating feature, and since it isn't fully functional (according to this discussion thread) its advantage is moot.

Some examples from other sets:
. Cores are native vs p-code
. Ram/rom/flash is Harvard vs. von Neumann
. No exposed ports, interrupts, buses
. Clockless pipeline or no pipeline at all
. Stack vs register device
. Opcode vs inference vs. Petri net

The incomparability has to be core to say "apples to oranges," but I'll give you the benefit of experience. I believe that I'm reasonable to ask for 1 incomparable thing.

 

[defragster]

- end use on the grand scale - not mechanical similarities. My analogy was about this deep ...

Porsche makes AWD 911's and AWD SUV's - depending on the course there would be a better choice between the two. And on the wrong course A Cadillac or Mustang would beat the Porsche SUV and a Jeep would beat the 911.

... where the ESP32 - when it matures - will be a different class of device in large part - that may or may not mature to do what a Teensy can do today. Not as wildly different as an R_Pi and an UNO - but at neither of those are in the 'emerging from vapor' state of an ESP.


... assume that many have been aware of much of what has been made evident on this thread :: difference in documentation, and maturity/completeness of the product and support. Those are significant features as now somewhat detailed that make Teensy worth more than many offerings that costs much more for much less - without all those attributes - when the Teensy is the right product.

For examining minute details make up a spreadsheet and enjoy - they only factor into the big picture of what choice can be bought.