Think big with the world's smallest microcontroller

[JoeRoy]

[mikerj]

Six years ago Maxim packed an M4 core, 256k flash and 96k SRAM into a 1.6x1.6mm package; the MAX32660.  The TI part gets you an M0+, 16k flash and 1k SRAM in a little over half the area, but the significant advantage I see in the TI part is that is has a built in ADC.

[Whales]

So it's an 0603 (1.6x0.8mm) with 8 pads.  Gorgeous.

[mikeselectricstuff]

There's a summary of other tiny MCUs in this vid from Cedric Honnet, who was researching chips for some ridiculously thin LED flex circuits.
https://youtu.be/fIyRMidTzso?t=397

[SiliconWizard]

Pretty cool.

[jnk0le]

Six years ago Maxim packed an M4 core, 256k flash and 96k SRAM into a 1.6x1.6mm package; the MAX32660.  The TI part gets you an M0+, 16k flash and 1k SRAM in a little over half the area, but the significant advantage I see in the TI part is that is has a built in ADC.

For 8 pad uC you would rarely need more than that.

[westfw]

NXP lpc1102 in a 16pin 2.17x2.32mm package, back in 2013 or earlier.

[SiliconWizard]

Maxim also has had smaller than 2x2 for years IIRC.

But here, 1.6x0.85 looks indeed like the smallest yet.

[RoGeorge]

The chip is for those that until now couldn't justify buying a wire bonding machine for their hobby. 

[PCB.Wiz]

It's cute, but you have to really need that size.

Their SOT26-8(0.65mm) package is relatively new in MCUs and perhaps more usable for projects.

I have seen a TSOT26-8(0.8mm) aka FSOP8 which looks very nice, but not yet seen that in MCUs, only in LED drivers like WS2811F tho those prove it must be a low-cost package.

[cunningfellow]

When ever I see something like this I feel like I should make a video game (tetris or tempest) built into a TRRS jack.

But then I realise I am too old and too slow.

[voltsandjolts]

To save you searching, the device in the OP is here;

https://www.ti.com/product/MSPM0C1104

[Warhawk]

Yeah, this is cool but I rather prefer the SOT23 version that MSPM0L1306 comes with. I posted it here before. I made a DIP-16 size adapter with a JTAG, buck-boost DC/DC and VREF that fits it all.
Most importantly, the SysConfig and libraries work surprisingly well.

<end of promotion of my own employer  >

[SiliconWizard]

Well of course the smallest of the series is for some particular, very dense applications, but the whole series is interesting in that it brings cheap western MCUs that are getting close to the chinese offering in terms of pricing, are well documented and come in various small packages that are not usual from other vendors/series. So that makes it a good move from TI.

[osmax_br]

It is as big as the required decoupling capacitor , isn't it?
I feel like the support circuit required to run this mcu is way bigger than the mcu itself thus removing some of the benefits of it smaller size (maybe?)
I'm wondering what new things (that weren't available before) that can be done thanks to this new tiny mcu
Maybe an mcu inside the type c connector/cable , but you still have to put a pcb and complementary parts

[Siwastaja]

It is as big as the required decoupling capacitor , isn't it?

01005 capacitor is much smaller. Random example: https://www.digikey.com/en/products/detail/murata-electronics/GRM022R60J224KE14L/5877378

With internal oscillator, there is not much ancillary circuitry needed.

[mikeselectricstuff]

It is as big as the required decoupling capacitor , isn't it?

01005 capacitor is much smaller. Random example: https://www.digikey.com/en/products/detail/murata-electronics/GRM022R60J224KE14L/5877378

With internal oscillator, there is not much ancillary circuitry needed.

annoying that the supply voltage doesn't extend high enough for direct operation for a lithium rechargeable.
AFAICS the IO mux only supports certain permutations for each pin - on a chip this small it would be good to be able to map to any peripheral to any pin.
And what's with that stupid beeper peripheral that only supports 1/2/4/8 kHz - don't they know that most piezo and magnetic beepers need to be driven at resonance for maximum output? And it doesn't support differential drive.

[SiliconWizard]

It is as big as the required decoupling capacitor , isn't it?

01005 capacitor is much smaller. Random example: https://www.digikey.com/en/products/detail/murata-electronics/GRM022R60J224KE14L/5877378

With internal oscillator, there is not much ancillary circuitry needed.

Even a 0201 capacitor (0.6x0.3), which is more common, is significantly smaller than this, and you can always mount it on the opposite side. As Whales mentioned, this MCU is the exact size of a 0603.

[Warhawk]

It is as big as the required decoupling capacitor , isn't it?

01005 capacitor is much smaller. Random example: https://www.digikey.com/en/products/detail/murata-electronics/GRM022R60J224KE14L/5877378

With internal oscillator, there is not much ancillary circuitry needed.

annoying that the supply voltage doesn't extend high enough for direct operation for a lithium rechargeable.
AFAICS the IO mux only supports certain permutations for each pin - on a chip this small it would be good to be able to map to any peripheral to any pin.
And what's with that stupid beeper peripheral that only supports 1/2/4/8 kHz - don't they know that most piezo and magnetic beepers need to be driven at resonance for maximum output? And it doesn't support differential drive.

Are you sure that you are reasonable with your criticism? Have you seen the 1ku price? From a western company with "OKish" support?

The beeper is limited but can be easily replaced by any PWM channel.
The 5V operation? ...wait...
MUX every pin to every pin? How much would you pay for this feature? Because most customers don't need this and this is not an MCU that you may etch the PCB at home for. What on earth do you want to mux on an 8 pin device?
Piezo differential drive - how many times have you really seen this implementation and in which product?

I am asking serious questions and can forward your (constructive) feedback to right people.

[MadTux]

Best way to piss anyone off, who has to repair your trash
80s and 90s were the best, nice architectures, parts you could still grab with your fingers and good music and games

[mikeselectricstuff]

It is as big as the required decoupling capacitor , isn't it?

01005 capacitor is much smaller. Random example: https://www.digikey.com/en/products/detail/murata-electronics/GRM022R60J224KE14L/5877378

With internal oscillator, there is not much ancillary circuitry needed.

annoying that the supply voltage doesn't extend high enough for direct operation for a lithium rechargeable.
AFAICS the IO mux only supports certain permutations for each pin - on a chip this small it would be good to be able to map to any peripheral to any pin.
And what's with that stupid beeper peripheral that only supports 1/2/4/8 kHz - don't they know that most piezo and magnetic beepers need to be driven at resonance for maximum output? And it doesn't support differential drive.

Are you sure that you are reasonable with your criticism? Have you seen the 1ku price? From a western company with "OKish" support?

The beeper is limited but can be easily replaced by any PWM channel.
The 5V operation? ...wait...

WCH manage it just fine.

MUX every pin to every pin? How much would you pay for this feature? Because most customers don't need this and this is not an MCU that you may etch the PCB at home for. What on earth do you want to mux on an 8 pin device?

It is not an expensive thing to do on a low-speed device like this. In the past I've used pin muxes for all sorts of things on a PIC, both statically and dynamically.
 For a device like this aimed at very small things, ability to optimise routing by having free selection of pins is a useful feature and the limited number of pins  means you want maximum flexibility - I've not looked in detail to see what the constraints are on this device, but it looks like there are only certain permutations possible, so needs careful choice of pins if you need to use a particular combination of peripherals.
 

Piezo differential drive - how many times have you really seen this implementation and in which product?

If you're driving a piezo, for maximum output you generally need to either drive it differentially or use an inductor to boost the voltage.
Given the  choice to have a dedicate piezo peripheral, it seems odd that it's done in such a limited way that makes it not very useful for many real-world applications

[Siwastaja]

MUX every pin to every pin?

Muxing nearly every peripheral into nearly every pin costs almost nothing. The mux network is like 0.1% of the area (wiring, transistors) spent in everything else. Some microcontrollers have nearly full muxing and some others do not and it does not correlate with the price at all, some of the cheapest Chinese controllers do that and some expensive flagship chips do not.

I'm sure that for example ST's AF system, which offers partial and random muxing, including hours spent on design sessions where mappings are decided by throwing darts and serious amounts of alcohol/drugs consumed, then documentation having to be maintained, is in fact more expensive than e.g. Nordic Semi's simple choice of consistently and logically mapping each peripheral to any IO pins with a simple-to-use "use this pin" configuration register in every peripheral instance.

There is value in simplicity, too.

I consider full or near-full arbitrary IO mapping a fundamentally useful feature in nearly every project. Microcontrollers, even the cheapest series, include much more rarely needed stuff. How often do you use configurable pull-down resistors for example? I have exactly once.

Free IO mapping in unbelievably handy especially when space is constrained - routing and vias on PCB do take very real area. Also I enjoy it in my extension module connectors. No need to buy a larger MCU with more excess peripherals and bring out 20 pins to extension connector just because I want to reserve the possibility of I2C, SPI and UART. Can do with just a few pins, and decide later what to map to them.

[Warhawk]

MUX every pin to every pin?

Muxing nearly every peripheral into nearly every pin costs almost nothing. The mux network is like 0.1% of the area (wiring, transistors) spent in everything else. Some microcontrollers have nearly full muxing and some others do not and it does not correlate with the price at all, some of the cheapest Chinese controllers do that and some expensive flagship chips do not.

I'm sure that for example ST's AF system, which offers partial and random muxing, including hours spent on design sessions where mappings are decided by throwing darts and serious amounts of alcohol/drugs consumed, then documentation having to be maintained, is in fact more expensive than e.g. Nordic Semi's simple choice of consistently and logically mapping each peripheral to any IO pins with a simple-to-use "use this pin" configuration register in every peripheral instance.

There is value in simplicity, too.

I consider full or near-full arbitrary IO mapping a fundamentally useful feature in nearly every project. Microcontrollers, even the cheapest series, include much more rarely needed stuff. How often do you use configurable pull-down resistors for example? I have exactly once.

Free IO mapping in unbelievably handy especially when space is constrained - routing and vias on PCB do take very real area. Also I enjoy it in my extension module connectors. No need to buy a larger MCU with more excess peripherals and bring out 20 pins to extension connector just because I want to reserve the possibility of I2C, SPI and UART. Can do with just a few pins, and decide later what to map to them.

You both talk about how much it costs but I wonder if you have any semiconductor development experience. In the price, the die area is just one piece of the pie. There are additional costs such as testing or validation. In certain cases, adding just "one mux" doubles the testing effort.
And especially testing in production is something where you can cut corners and come with hard-to-compete prices. Also, multiplexing analog signals  (and pin protection) is also different than multiplexing digital. However, in this case, I admit, it is probably not a problem.

...No need to buy a larger MCU with more excess peripherals and bring out 20 pins to extension connector just because I want to reserve the possibility of I2C, SPI and UART. Can do with just a few pins, and decide later what to map to them.

This is a rock solid argument. I agree 101%

In any case, I respect your comments and will ask eventually the MSPM0 product line. I am interested in the feedback myself. MSP

[YTusername]

Here is maybe the first recorded attempt to use this new MCU