Tiniest Teensy?

Status
Not open for further replies.

onehorse

Well-known member
I was challenged to make as narrow a Teensy as I could so as to fit inside a "standard" 14-mm-ID hoop. I easily got to 15.4 mm and decided to stop here for a test run.

Prodigy3.jpg

It is of course a stunted Teensy since it has only those ports exposed that are required for the application. Getting to 14 mm will be quite difficult; I will have to make more use of the back of the board and possibly replace the side button with a top-mounted button. I have a battery charger and motion sensor on the back so i am running out of room. The last 10% will require 90% of the effort. The good news is it might not be necessary, as this size might be small enough for the use case. If not the tiniest Teensy, it surely qualifies as one of the cutest Teensy's ever!
 
Cool! Those next few mm added back are really worth it when you have the space :)

onehorse / Paul : Suppose one had such a Teensy using the PJRC supplied programming CPU - what would the serial number be?

Now that I've written this I can mostly answer my question recalling the Serial# is pulled from the MCU flash - not on the programming chip CPU? So outside of PJRC control.

On the PJRC retail units it is resolved to a PJRC MAC family ID - but on a onehorse teensy unit I got the serial number from TYQT is reported as : The number 4,294,967,295 is an integer equal to 2³² − 1.

I've asked Koromix if this is something on his end - but expect there is something outside his purview.
 
@onehorse I noticed from Prodigy kickstarter that they are open source. Did you make the board for them or can you sell it?
 
Both I think. I mean I made it for them, it is open source, you can find it on OSHPark I believe, and I will sell you one if you want. If you want more than one, you might want to customize it for your own use. What would you use it for?
 
Thanks.

cartere, maybe you want the finished product, and not the cute Teensy board?

The kickstarter is going on for another week or so. I am not part of it other than making some electronics for the project, but I will likely get a bit more involved.
 
Really teeny...
My current project (a team) has the PCB being 0.8mm in height of highest component. It's 1mm after it's packaged. Quite a challenge to find all the many components that small, including bypass capacitors and an inductor. One of the microprocessors on this board is a Cortex M4 w/1MB flash, in die form and 16GBytes of NAND flash dies w/SDIO controller.
The PCB is flex.

That's maybe a nano-pico teeny. Lots of art in building so small.

I have to use a strong magnifying glass to even find the M4 where my code goes! Kind of makes me feel insignificant, even though a huge amount of work has gone into that code which fills over half of the 1MB flash.
 
Last edited:
Mine is smaller than yours!

Not really---the hardest board I have designed so far is 40-mm-long by 10.4 mm wide one-sided 4-layer board with a 6 mm x 6 mm nRF51822 QFN and a bunch of sensors, boost converters, and motor controllers. It has to fit inside the frame of eyeglasses, so we had some 0.5-mm-thick pcbs made to reduce the thickness as much as possible. I put two together yesterday. They are a beautiful sight.

I am tasked with reducing the width even further which is very difficult. The current board can and has been made at OSH Park (in the 1.6-mm thick version) but going to the WLCSP nRF51 package and 0201 passives, as well as a six-layer board in order to get the width down another 10% will require much more expensive via-in-pad methods, not to mention strain my design and hand assembly skills. I'm not looking forward to it.
 
I wonder if making a teensy board on flex would make sense for hoop purposes... if it could be slightly bent you could probably make it slightly larger and still fit inside the 14mm hoop
 
Flex might help with length but it is width that is the problem. I am tasked with reducing the width to 14 mm and I am able to double the length but this really is of no help. The K20 chip is almost 13 mm wide with the pins so 14 mm is very challenging and requires several pretty ugly compromises to good hardware design practices.
 
My guess would be this chip, the MK20DX128VMP5. Mouser has a brochure and the external pin/pad assignments are 100% compatible. I'd like to think that the internal stuff would be similar as well.

It's eqv to a Teensy 3.0, with 40 i/o, 128kB Flash, 64 pads, etc. If everything I/O, register related to the chip is compatible with the MK20DX128 then you have a chip that may be compatible with the bootloading chips that PJRC sells... as long as you use the Mini54. The MKL02 only lists compatibility with the MK20DX256 (Teensy 3.1 and Teensy 3.2) as well as the MKL26Z64 (Teensy LC) on the PJRC store page. Curiously, the PJRC MKL02 store page also mentions the MiniTan54 in its discussion of MKL02 features and has spelling errors. Very odd!

Thus, it's not clear if the MKL02 will support a MK20DX128-style CPU. I know Paul won't claim compatibility with anything but what he has tested (as he should) so your best bet might be to whip up a test board. Could be as simple as a BGA-LQFP converter that allows you to site a BGA board on an existing Teensy board (with the LQFP removed, of course). The cost would be moderate, the results interesting. For the purposes of the test, I'd only focus on seeing if I could get one ADC channel to work, one digital channel to work, and perhaps SPI (the LED) as well. Minimizes the routing complexity and may allow you to get away with a 2-layer board (for this test only!). Or, design a larger board (4 layer?) that incorporates everything needed to create a full-fledged Teensy 3.0 with a BGA MCU.

I didn't see a BGA eqv to the Teensy 3.1/3.2 MCU, i.e. a MK20DX256. Neither Mouser nor Digikey carried one.
 
Last edited:
Did you look at Freescale's package data sheet spec (not the reference manual or vendor sites)? The package spec will list all the package options for a given die.
 
My guess would be this chip, the MK20DX128VMP5

I believe this is correct. I can rather easily make a prototype since I have some spare MiniTan54's left over.

Did you look at Freescale's package data sheet spec

Yes, unfortunately there is no 64-pin equivalent to the MK20DX256VLH7. There is a 7 mm x 7 mm BGA but it has more/different pins and I think this would be a lot riskier.

The other alternative is the LC-chip but this doesn't really have the processing power required.

I'll let you know what I learn...
 
Here is page 7 off this product manual:
page 7.png

Seems like the 128kB version is the only one open to you. 5x5mm is pretty teeny weeny! Hope you get the alignment right. Would scare me! It also is only rated to 50MHz, whatever that means.

There appears to be a 256kB version, but that has fewer I/O, making me think it'll likely not work without some serious work re: the files that define the chip. Way above my head!
 
Last edited:
The 5 mm x 5 mm is 0.5 mm pitch, which should be fine for me. I worry a bit about the trace routing with OSH Park 5 mil trace minimum but I don't need more that I2C, UART and SPI plus a couple of GPIOs so I should manage. Yes, with this if I can get it to work I can make the 13 mm desirement with some ease. But the design will take a bit of time...

I'll report back in a month or so!
 
Good luck! I presume 4-layer, if not 6, and the pleasures associated with double-sided construction. I hope it works out!

Would be really neat to add another, unsupported yet working, MCU to the teensy lineup. The funny thing about that board will be that the bootloader chip will be almost as big as the actual MCU.
 
Thanks, yes there are quite a few applications that require an appallingly small approach...

Right, I will be using the smallest 4 mm x 4 mm MiniTan54.
 
Well I made a lovely package for the MK20DX128VMP5 and designed the symbol so it was nearly a one-for-one replacement for the LQFP part. It was easy to replace the K20 in the schematic with the new part but the trouble came after I started trying to layout the traces on the board. Even at 5 mil trace width I cannot route traces into the interior of the K20 without triggering errors in EAGLE CAD. Several of the GND and VDD pins are in the interior so it's not just a matter of using the outer pins either, unless somehow I can connect several GPIO pins with a GND or VDD trace and hope this doesn't cause trouble.

How does one usually make use of such parts then? 3 mil traces???
 
Status
Not open for further replies.
Back
Top