Improved performance 5V Arduino Nano/Mega compatible

[DC1MC]

HEAVY EDIT:

This project initially wanted to produce an f Arduino NANO compatible board with an Nuvoton M0 core, that after a good investigation by the thread participants has been replaced with cheaper and more capable Cortex-M4F core.
After important and useful feedback from the forum members, I've decided to design two boards as described:

Nanoton:
Arduino Nano compatible micro-USB connector, 30 pins, if everything goes well, is should be a performance/features increased drop-in replacement.

Nanonton+:
New design, software compatible with Nanoton (same MCU), USB-C connector, with more pins exported (for the EIB bus), an RGB led, 2 or 3 push-buttons besides the RST.

Because of consistency and price (currently 1.66USD) on BOTH boards the M471SE6AE (LQFP64: Body Size 7 mm x 7 mm, Pin Pitch 0.4 mm)

Full data sheet: https://www.nuvoton.com/export/resource-files/DS_M471M_M471R1_M471S_Series_EN_Rev1.00.pdf
(attached here as well for convenience)

Development kit: https://direct.nuvoton.com/en/numaker-m471r1 for whom lives in Germany and want to contribute to the SW development, I can order extra, to spare transport costs.
The manufacturer offers Eclipse/GCC based development IDE and packages for all the big contenders (Keil, IAR, ...). Also there is a pin configurator and a programming program, both graphical and command line.

On both boards, there will be only one MCU, no USB to serial converter, as the MCU is capable of ICP/IAP programming, host and device USB.

For "blanco" devices, the 6pin header will contain the SWD interface, used both for programming and debugging, I will attempt to produce an Arduino IDE compatible bootloader, using the IAP feature to program it over the USB port.

I initially wanted to use 48pin package with 0.8 pitch for the Nanonton, but is actually physically bigger and who can't solder 0.4, will not be able to manually solder 0.8 as well, we'll leave this job to machines  .

Deliverables will be the zip archives to be given to JLCPB or whoever offers services, I may even try the assembly service of JLCPCB.

Please let me know what you think about it and contribute with advice for the BOM and extra add-on like RTC battery/supercap and so on.

 Cheers,
 DC1MC

EDIT: The development kit and 10 samples have arrived, pictures attached.

[fchk]

For me the need for 5V operation is shrinking. Most things are 3V3 anyway now.

For the few times where 5V operation made sense I used PIC24FV and dsPIC33EV. The dsPIC33EV runs at 140 MHz internal oscillator clock/70 MHz instruction clock, has up to 256k flash, up to 16k RAM, CAN, LIN, Motor control PWM, SENT, is Automotive qualified and is qualified to run even at 150°C (with only 40MHz instruction clock above 125°C). It's meant for power electronics where 5V outputs can drive larger power MOSFETs without needing extra gate drivers. Internally these chips run at 1.8 or 2.5V, only the IO ring uses the 5V supply directly. This is an example where 5V still makes sense.

I'd drop the 5V requirement and use level shifters if absolutely necessary.

And no, I'm not an Arduino user. I usually use the vendor tools directly.

fchk

[DC1MC]

@fchk - well, for me the need for 5V is there and growing  , there are tons of Arduinos compatibles at 3V3 with STM32 and other chips, mostly nobody cares of them.
 On the other hand, the dsPIC33EV series, while a seemingly capable chip, has a bit of disadvantages: expensive, no real free toolchain for everybody and, oh God, the strange architecture, and most of all: Mouser when asked to show "Auf Lager" says "Es tut uns leid, für "dsPIC33EV" gibt es keine Ergebnisse." And that's a 52 weeks lead time from a main manufacturer and distributor for possible candidates from the series.

I don't contest that for some industrial or targeted professional projects they may be ideal, but Arduino is the proletariat option  .

But you did give me an idea, I will place a 3V3 regulator + a jumper or solder bridge, the chip seem to work at both voltages, so that should broaden the appeal of it  .

Cheers,
DC1MC
 

[Infraviolet]

I agree that 5V is useful, but for a lot of things digital lines can all just be put through a MOSFET based level shifter to convert between 3V3 and 5V (works for input and output purposes).

[DC1MC]

I agree that 5V is useful, but for a lot of things digital lines can all just be put through a MOSFET based level shifter to convert between 3V3 and 5V (works for input and output purposes).

Why should I use 20 extra mosfets when I can change a jumper to switch the MCU voltage in between 5V and 3V3, actually where I would put 20 mosfets on a Nano sized board ?
And I seriously want to see the schematic with this mythical one mosfet that works simultaneously for the input and output 

[Psi]

I agree that 5V is useful, but for a lot of things digital lines can all just be put through a MOSFET based level shifter to convert between 3V3 and 5V (works for input and output purposes).

Why should I use 20 extra mosfets when I can change a jumper to switch the MCU voltage in between 5V and 3V3, actually where I would put 20 mosfets on a Nano sized board ?
And I seriously want to see the schematic with this mythical one mosfet that works simultaneously for the input and output 

I would use a combined IO expander level converter chip.
Then you have lots of 5V IO and the space needed is tiny.

[DC1MC]

...
I would use a combined IO expander level converter chip.
Then you have lots of 5V IO and the space needed is tiny.

Still I don't see the advantage of replacing the native latency with I/O expander latency and increasing the code size and costs against switching a jumper, please help me here, for a "hat", or "shield", or whatever they call them, it may make sense, but for the main board?!?!

[Psi]

Larger pin count AVRs tend to be expensive and all tap out around 20mhz, that's why manufactures don't tend to make Arduino compatible boards for them. A STM32 is much more powerful and much cheaper but at the cost of no native 5V logic output and limited 5V tolerant inputs.

If you find this a problem then yeah, probably time to outgrow off-the-shelf Arduino and make your own board using a MCU that suits your needs. But usually 5V logic and high speed don't go together.

I'm in the process of trying to get away from 5V. Everything I want to interface with is 3.3V and 5V mcus are getting expensive or just less cost effective.

[DC1MC]

Larger pin count AVRs tend to be expensive and all tap out around 20mhz, that's why manufactures don't tend to make Arduino compatible boards for them. A STM32 is much more powerful and much cheaper but at the cost of no native 5V logic output and limited 5V tolerant inputs.

If you find this a problem then yeah, probably time to outgrow off-the-shelf Arduino and make your own board using a MCU that suits your needs. But usually 5V logic and high speed don't go together.

I'm in the process of trying to get away from 5V. Everything I want to interface with is 3.3V and 5V mcus are getting expensive or just less cost effective.

The whole purpose of this post was to outgrow the off the shelf offers for the 5V range and to offer a cost effective replacement with updated parameters. I have selected some cheap MCU with a reasonable Cortex M0 core as start point. I was hoping for guidance regarding all the question in the first post,  but, so far, just got (in summary):

- "But don't you heard that bird 3V3 is word, 3V3 is word, weeell everybody knows that 3V3 is a word...".
- "I'm just using for 5V some high-end unobtanium dsPIC, that should make a nice, cost effective, Nano replacement...".
- "Don't bother with 5V native, just put some mosfet on the 3V3 signals and they will auto-magically transform in 5V compatible, no, I won't come with a schematic of this magic circuit...."
- "You can use a GPIO expander, screw latency and everything else, you only need to control some LEDs, do you..."
- "You want to use a high(er) speed core, heh, high speed and 5V do not mix..."
- "Do your own if you didn't like the off the shelf options, doh..."

Aaaand we're back at the beginning... 

I don't want to mock the genuine attempts to help, but really is little related to the original project questions and scope, all these stuff doesn't really help or contribute with anything useful, so I guess I have to proceed and do a Fait Accompli thingie, I think I've got a good candidate for the name  .

But being a hard-core optimist I still hope that some flower will grow in the pile of mist.

 Cheers,
 DC1MC

 

[fchk]

no real free toolchain for everybody

Many people get this wrong. Yes, it's not GNU, but is free enough to download and to use, and the Pro license is mainly intended for mass production where you can save $$$ by choosing the next smaller version.

and, oh God, the strange architecture

This is also a popular lie. The architecture is rather similar to AVR (just wider), and the CPU core has nothing in common with the old 8 bit architecture most people know as PIC. Same for PIC32, which is MIPS32 in fact.

Availability: yes, this is a problem, but it's the same everywhere.

fchk

[fchk]

I don't want to mock the genuine attempts to help, but really is little related to the original project questions and scope, all these stuff doesn't really help or contribute with anything useful, so I guess I have to proceed and do a Fait Accompli thingie, I think I've got a good candidate for the name  .

Maybe the main takeaway is that 5V support is a non-issue anywhere else. The world has changed since the mid-90s and there is no way back.

fchk

[DC1MC]

...
Maybe the main takeaway is that 5V support is a non-issue anywhere else. The world has changed since the mid-90s and there is no way back.
fchk

Good then, it will be a vintage Nano, everybody loves vintage around here, isn't that OK  ?

[pcprogrammer]

I somewhat like the idea. 

To the 3.3V versus 5V argument relating speed, why is it then that the 3.3V 328P version is slower then the 5V version?

But a problem I see with projects I do, is the need for a mix of 3.3V and 5V IO.

As an example I made a proto to do some development using a STM32F103 based bluepill and a NRF905 transceiver module plus for debug info an I2C module for driving a 1602 display.

The bluepill and the NRF905 module are 3.3V so direct connection, but the I2C module and the display are 5V. Even though the pins might be 5V tolerant I decided to use mosfet based level converters to interface between the two.

The two 2N7000 mosfets provide the single transistor bidirectional digital level converters others wrote about. Might not work with high frequencies but have used it for I2C in several projects without problems.

So having a faster 5V board sounds nice, but in this case would still need some level converters.

[fchk]

To the 3.3V versus 5V argument relating speed, why is it then that the 3.3V 328P version is slower then the 5V version?

Because the 328 is manufactured in a rather old semiconductor process. The whole chip runs at the external Vcc level, and the transistors used there switch faster with higher voltages.

Modern controllers like AVR128 or all PIC18F...J... and PIC18F...K... use a much lower voltage for the core and much smaller transistors (which are much faster because of less capacitance etc). Only the IO ring with the pin drivers runs at the external Vcc voltage. This is also much more energy efficient.

fchk

[djsb]

DC1MC,
I think what you propose is a good idea. Not having to use the level shifters would help me a lot at work. We use the Arduino Nano and Uno and a 5V I/O compatible board which would save us money would be great. Making it compatible with BOTH 3.3v AND 5V would be even better (but you say the MCU is ALREADY usable with both). It would have to be COMPLETELY software compatible with the Nano/Uno codebase. The built in USB 2.0 interface would be a plus also.


PS I'd go with a USB connector that is the most robust (so Type B as used on the UNO) or an option for this on the pcb.

[DC1MC]

DC1MC,
I think what you propose is a good idea. Not having to use the level shifters would help me a lot at work. We use the Arduino Nano and Uno and a 5V I/O compatible board which would save us money would be great. Making it compatible with BOTH 3.3v AND 5V would be even better (but you say the MCU is ALREADY usable with both). It would have to be COMPLETELY software compatible with the Nano/Uno codebase. The built in USB 2.0 interface would be a plus also.


PS I'd go with a USB connector that is the most robust (so Type B as used on the UNO) or an option for this on the pcb.

Well, I've decided in advance if I get ONE positive feedback, I will proceed, I'll try to make the BSP as compatible as possible with the exiting SW, some small changes may be needed, especially for projects that exploits the original 328P features, but this is life, blinky will work anyways .
Now if you want to have an influence on the design, please do have a look at the data sheet (I'm thinking to do a Nano compatible with NUC029LEE and better but not-compatible variant with NUC029SEE, that will have a "classical" 8/16bit 6802 and friends bus, is a feature of SEE).
Of course if you find on Nuvoton's offer a MCU that you consider to be more suitable, suggestions are welcome, it seems that no one likes USB-C connector, and I consider the Type B too bulky for Nano, maybe 4 the other one, so we meet in the middle with mini-USB like the original Nano .

 Cheers,
 DC1MC

[tooki]

@fchk - well, for me the need for 5V is there and growing  , there are tons of Arduinos compatibles at 3V3 with STM32 and other chips, mostly nobody cares of them.

Huh? There are plenty of popular Arduino framework compatible 3.3V boards, including STM32 and ESP32 boards. Some are also physically compatible with Arduino (like the STM32 Nucleo series).

5V is sometimes handy — and when I absolutely need that I will often grab an Arduino Uno — but most of the time, I just use 3.3V and call it a day.

As for the port: first choice USB-C, second choice either micro-B or standard B. Just not mini-USB, which needs to shrivel up and die.

[Psi]

IO Expanders don't have to be annoying to use. If you pick one that can do SPI at the speed of your MCU, or near it with a couple of NOP()s, then you can simply clock data in or out in real-time as you want to read write it the pin.
And you can setup macros to do it all for you so your code reads normal.

IO Expanders can get annoying when you don't have the spare cycles to wait for the SPI/I2c data to clock out.
So you have to use a function/macro that only schedules the SPI/i2c write which happens later via interrupt.
Or when you need to put I2C/SPI/UART data on the IO Expander pins.

[DC1MC]

@tooki - no micro-USB will be in a design of mine  , the Nano like board with LEE will have micro-USB like the original Nano, the "advanced" board, with SEE and 40 pins, will have USB C connector. The 3V3 variants that nobody cares, is not that they aren't significant or super widely used, nobody cares to make a new one because most of the sweet spots for that platform have been already reached.

@psi - I have selected an MCU that works at 5V as well as 3V3, I'm interested in using it at 5V, you prefer 3V3, no problem, so for the moment this issue is solved. There is no need for I/O expanders to be used on base board, they can be added externaly if needed.

Generally speaking, I'm planning two boards, one as much as possible compatible with Nano, with NUC029LEE (7x7mm LQFP44) and one, an extended version with 40pins and the EBI bus out with NUC029SEE ( 7x7mm LQFP64) and USB-C connector to make it a bit future proof. because the belly of the boards will be free from the USB-to-Serial programming chip, both boards will get a RTC battery holder (or soldered battery for really low budget), I'm open for suggestions of what to place on the rest of the belly area that will be uesfull but not increase the costs too much.

Now that the 3V3 vs. 5V topic has been  until it become a hamburger patty, maybe we can let it rest and the knowledgeable people have a look at chosen MCU, eventually suggest another one from Nuvoton, that has a similar packaging, and better performance or peripherals (multimedia excluded).
Why Nuvoton, one because is not Mainland Chinese or mainstream, two because I can get reasonably cheap and easy chips from their portfolio and three, and the most important, YOU can get reasonably cheap and easy chips. I tried to avoid either mainstream manufacturers, or the extreme obscure ones and I think that Nuvoton fits well the middle ground.
Also when suggesting better alternatives IYNSHO, kindly please remember the scope of the project, that is practically a MCU on a PCB, attempting to be cheap, but with a bit of performance, avoid astronautic, automotive grade, mainstream stuff, that is plundered by everybody and unobtainable since 2020, also lets further restrict the architecture to ARM Cortex-M compatibles capable to run at 5V.

Any useful advice or suggestion will be thanked and appreciated  , any contrarian, "muddling the water" and "poisoning the well" shite will be ridiculed and/or ignored  .

Thanks and cheers,
DC1MC

[Psi]

@psi - I have selected an MCU that works at 5V as well as 3V3, I'm interested in using it at 5V, you prefer 3V3, no problem,

I prefer to run it at whatever voltage makes sense for the circuit.
But i don't really interact with many MCUs that can run at 5V and are also high speed, It's usually a decision between sub 20mhz and up to 5V or over 20mhz and under 3.4V.   There are exceptions obviously.

[DC1MC]

@psi - I have selected an MCU that works at 5V as well as 3V3, I'm interested in using it at 5V, you prefer 3V3, no problem,

I prefer to run it at whatever voltage makes sense for the circuit.
But i don't really interact with many MCUs that can run at 5V and are also high speed, It's usually a decision between sub 20mhz and up to 5V or over 20mhz and under 3.4V.   There are exceptions obviously.

Please, please, pretty please, with sugar on top, let the voltage shite rest 

Also, it is worth mentioning, for whom didn't realize yet, that the fact the core runs at a higher-speed, it doesn't mean that it HAS to run at that speed, but offers better latency at processing, lower jitter for RT tasks and the possibility to actually sustain comfortably the USB transfer rate. Yes, a 5V signal, given the same slew rate, will reach its logical 1 or 0 levels slower than a 3V3 logic, and even slower than a 2V5, 1V2 or 0V9, that's not the point, one could use LVDS if they need speed, we are talking here about improved Arduino-like stuff.

[pcprogrammer]

You could also consider the M471 or the M4521 series of the Nuvoton MCU's. These are Cortex M4 which brings a bit more power to the table, which can make it interesting for people that need a bit of DSP. Did not look into availability and pricing, but they do run from 2.5V to 5.5V according to the website and there are versions with USB 2.0 FS host/device interfaces.

https://www.nuvoton.com/products/microcontrollers/arm-cortex-m4-mcus/m471-series/
https://www.nuvoton.com/products/microcontrollers/arm-cortex-m4-mcus/m4521-usb-series/

Might even be that they are pin compatible with the ones you mention, which would be better, because the PCB can be the same for both versions.

[DC1MC]

You could also consider the M471 or the M4521 series of the Nuvoton MCU's. These are Cortex M4 which brings a bit more power to the table, which can make it interesting for people that need a bit of DSP. Did not look into availability and pricing, but they do run from 2.5V to 5.5V according to the website and there are versions with USB 2.0 FS host/device interfaces.

https://www.nuvoton.com/products/microcontrollers/arm-cortex-m4-mcus/m471-series/
https://www.nuvoton.com/products/microcontrollers/arm-cortex-m4-mcus/m4521-usb-series/

Might even be that they are pin compatible with the ones you mention, which would be better, because the PCB can be the same for both versions.

THIS ^^^, exactly this is the type of reply I was expecting, the M4 and M4F cores looks nice, and the price difference from the original candidates for LQFP44 and LQFP64 packages is not big, I will make a list with "worthy candidates" and let discussion participants vote, eventually a qualified vote (with arguments, but sane .

 Thanks and cheers,
 DC1MC

[DC1MC]

Dear thread participants and potential interested, I have shortlisted some device in an Excel file attached to the first post, unfortunately I could only attach two of the three datasheets, I'm attaching it the last one here for the 471 series.
The prices are interesting and a bit unexpected  , so please have a look, a good overview is in the Excel file, I think I was able to catch all the features, if is not there, it will probably doesn't have it.

Many thanks,
DC1MC

[coromonadalix]

I too would need a more powerful "uino" variant   5vdc i/o tolerant,  yeah   3.3v interfaces / levels shifter exists   .... chice is made   thks  op

but i need simpler "direct" interfacing

pricing   i would go as far at 15-20$ cad  / single board if the design is solid and practical

I have a mega, never played lot with it compared to 32u4 variants, or uno  .....  even one with an real db9 serial port was fun (yeah they still exists loll)


About time the use of an usb-c  plug, and a rock solid one will be good, others are too easy to rip off the pcb
a good reset circutry  once programmed,   and the famous pin 13 led ....... if you get the idea,  just put a way to recover the pin if we need it   lolll    say a smt jumper