Mixed Signal Microcontrollers

[Hiemal]

Is ST going to come out with any new mixed signal microcontrollers anytime soon... They're getting a lot of competition from other companies.

Gigadevices has the GD32G553.https://www.gigadevice.com/product/mcu/high-performance-mcus/gd32g5xx-series
ARM Core M33, clocked at 216 MHz, has a SLEW of incredibly useful DSP related peripherals (there's an FFT coprocessor?!) configurable logic, and fast ADCs. HRTimer @ 145 ps resolution.

Microchip with the new dsPIC33AK series.
Not ARM based, Microchip uses their own proprietary core clocked at 200 MHz, double and single precison FPU, 5 ADCs that can (supposedly) churn out 40 MSPS each. HRTimer @ 86 ps resolution.

Nuvoton with KM1M7xx series.https://www.nuvoton.com/products/microcontrollers/arm-cortex-m7-mcus/
Core M7 clocked in at 160 MHz. Double and single precision FPU. 5 MSPS ADCs. HRTimer @ 208 ps. This is probably the most comparable to the STM32G474.

Infineon with their new PSOC Control C3M series https://www.infineon.com/products/microcontroller/32-bit-psoc-arm-cortex/32-bit-psoc-control-arm-cortex-m33-mcu/psoc-control-c3m
Core M33 clocked at 180 MHz. Single 12 MSPS ADC. HRTimer @ 80 ps resolution.

Geehy, with their GD32R501 series. https://geehy.com/product/fifth/g32r501
DUAL Core M52 @ 250 MHz, Arm HELIUM, 3 3.45 MSPS ADCs, flexible logic blocks, and HRTimer @ 150 ps resolution.

And of course the various offerings from TI in their C2000 series.

The G474 came out in 2019; I'd really love for them to show that there's SOMETHING in the works that's updated...

[PCB.Wiz]

There is also AT32F435/7 from ARTERY, claims 288MHz on the timers.
12b DAC, 5.33Msps ADC, but I'm not sure it has high resolution PWM?
What would be nice is high resolution capture

[uer166]

What would be nice is high resolution capture

Yeah the base clock is more important than the "HR" resolution, capture or not. I use the ST HRTimers because of flexibility, not because of the analog delay line crap.

If they made a Cortex-M55 with timer/ADC bus running at 400MHz that would be unstoppable!

[NorthGuy]

Microchip with the new dsPIC33AK series.
MIPS core, clocked at 200 MHz, double and single precison FPU, 5 ADCs that can (supposedly) churn out 40 MSPS each. HRTimer @ 86 ps resolution.

Not a MIPS core. Microchip use their own core.

[Hiemal]

Ah, thank you. I'm not sure where I thought I read it was MIPS based.  I updated the original post.

[SiliconWizard]

The MIPS cores are only for the PIC32 line.

[Doctorandus_P]

During the virus thing a few years ago I was looking at the "STSPIN" uC's from ST. These things had a lot of integration specifically for motor controllers. Included in the STSPIN32F0 is:

• SMPS circuit for the uC voltage, max input was 45V or so, just add inductor and chicken fodder.
• Opamps for amplifying shunt resistor voltages.
• FET drivers (600mA Source / Sink).
• Bootstrap diodes
• 12V LDO for the FET drivers.

[uer166]

During the virus thing a few years ago I was looking at the "STSPIN" uC's from ST. These things had a lot of integration specifically for motor controllers. Included in the STSPIN32F0 is:

• SMPS circuit for the uC voltage, max input was 45V or so, just add inductor and chicken fodder.
• Opamps for amplifying shunt resistor voltages.
• FET drivers (600mA Source / Sink).
• Bootstrap diodes
• 12V LDO for the FET drivers.

Too bad they're tailored for BLDC where the frequencies and control loops are much slower than even modest DSMPS by 10x.. I would love something like that but with a higher base clock.

[hans]

There is also the STSPIN32G4. 170MHz M4 + FPU, 2x 12-bit ADC 4MSPS, 4x 12-bit DAC,  Unfortunately it integrates the G431 chip, not the one with HR timers.

I designed a three phase low-voltage motor driver board a few months back, and I opted for seperate MCU and driver instead. I figured that I'd rather have that flexibility (yet have higher driver integration) than optimize BOM count, cost or size..
I ended up picking the STM32H725 which drives 3 motors (which is still more CPU horsepower than a single G4).

[Doctorandus_P]

Too bad they're tailored for BLDC

Yes, but I've also seen some designed for multi phase power supplies, and those tend to have the high resolution timers.

I'd guess that the market for uC's is pretty much saturated. How much combinations of timers, uarts, SPI channels and flash / RAM size can you make? ST allone claims to have over 3000 variants, which seems quite silly to me. Apparently there are quite a lot of dies that are used for multiple IC's. Then you get IC's with twice as much flash as stated in the datasheet, and when enough of them are being sold, they create a new set of masks to make more chips, but with only the specified amount of FLASH.

But for the rest. In the last few years all sorts of extra's are being added to uC's. From small programmable logic blocks to bigger FPGA blocks, state machines such as with the RP2040, analog circuits such as opamps and PGA's. Dual core uC's are also becoming more common. I don't really keep up to date with what's happening in the market, but I'm still sort of curious what other manufacturers are doing too.

[Hiemal]

https://www.infineon.com/products/microcontroller/32-bit-psoc-arm-cortex/32-bit-psoc-control-arm-cortex-m33-mcu/psoc-control-c3-performance-line?ftab=04#Documents


New upcoming release from Infineon, they're releasing a much, much beefier boy in September!



Yes, TRIPLE M33's. Main runs at 180 MHz, and then two integrated into their "PPCA", programmable power control accelerator, at 200 MHz.

HRTimer running at some undisclosed speed.
4x 12 bit ADCs, up to 25 MSPS,
9 comparators,
2 DACs,
CORDIC accelerator, and more.

Very exciting, I hope it's not stupid expensive. 

[gspeed]

There is also AT32F435/7 from ARTERY, claims 288MHz on the timers.
12b DAC, 5.33Msps ADC, but I'm not sure it has high resolution PWM?
What would be nice is high resolution capture

Some C2000 DSPs from TI have it, for example TMS320F280049

[Xiamatsu]

small mixed MCU XHSC


HC32F334  -  Cortex-M4F, 120MHz, 128K/36K
  - 3xADC12 (2.5Msps), 3xDAC12, 3xCOMP
  - HRPWM 22bit - 130ps  (6x2 ch)
  - 16 PLA (programmable logic block)

HC32F336  -  Cortex-M4F, 150MHz, 128K/36K  (new)
  - 3xADC12 (2.5Msps), 3xDAC12, 3xCOMP
  - HRPWM 22bit - 104ps  (6x2 ch)
  - 16 PLA (programmable logic block)

HC32F558  -  Cortex-M4F, 240MHz, 512K/128K  (new / not on the site)
  - 4xADC12 (5Msps), 8xDAC12, 8xCOMP
  - HRPWM 22bit - 130ps  (up to 8x2 ch)
  - 16 PLA (programmable logic block)
  - CORDIC, FMAC, PID

[Shalomie42]

The G474 came out in 2019; I'd really love for them to show that there's SOMETHING in the works that's updated...

Oh. There is something coming. Don't worry. However like most things ST they like to tell you something is coming 3 years before it comes out and drag their feet... but it is coming. Maybe a little birdie perhaps told them it was still not up to analog and digital parity with the C2000 series from 15 years ago. However, that didn't stop them from steam rolling ahead ignoring such feedback.

[uer166]

The G474 came out in 2019; I'd really love for them to show that there's SOMETHING in the works that's updated...

Oh. There is something coming. Don't worry. However like most things ST they like to tell you something is coming 3 years before it comes out and drag their feet... but it is coming. Maybe a little birdie perhaps told them it was still not up to parity with the TMS320F28375S series PWM, analog & interconnect goodness from 15 years ago. However, that didn't stop them from steam rolling ahead ignoring such feedback.

It doesn't need parity since I wouldn't use a TMS320 anyway, however some newer analog and cores would sure be welcome...

[Shalomie42]

Let me tell you this. I have been telling TI for years that there is nothing even close to the C2000 analog & digital peripherals on the market currently. However, I hate their C2000 core. If they would pair their peripherals with a Cortex M core, it would be a match made in heaven. Many customers have been saying the same. TI finally got the message. Queue: The new AM13 series parts. There is a lot more where that came from. Trust me.

ST and the other vendors were playing catch up to the C2000 analog & digital stuff and now for most intents and purposes, the gate is closing very quickly for them. Once TI proliferates these lines, its game over for motor control & analog from any other vendor.

Look here. TI has given you 3-4 advanced ADCs, 22+ digital comparators each with its own DAC and digital output filter, 30+? "HR" PWM modules, and a crazy internal interconnect all in a 65nm process starting from 15 years ago. Modern chips from vendors like ST, Renesas, NXP, etc. have totally missed the plot for a very long time. These vendors give you 2, 3, 4 comparators and a wimpy interconnect, call it "rich analog", and then retire for the day. They are complete jokes in terms of analog goodness for motor control. They keep making the same 'mistake' with newer chips they release on even more advanced nodes better than 65nm. TI did it all in 65nm 15 years ago.

By the way, nobody cares about your "HR" PWM timers. Anybody and their dog can make one since its digital. TI already did it and has done it way way better decades ago.

The drawback for the C2000 has always been the C28 core, which will now be fixed. Game over

[Hiemal]

I mean... TI also has quite possibly the worst software and development package out of any microcontroller manufacturer, so ... there's also that... It's the primary reason I went with the G474 in the first place, rather than a TI part.

And a more complex part isn't always better. 22 digital comparators for what exactly??

The newest crop of mixed sig micros from the major companies seem to be quite nice. I'm especially excited for the infineon one because it ticks so many boxes for my company's applications, mainly the fast ADCs. I hope ST's next part, whenever the hell it does get released, has a faster ADC.

[uer166]

And a more complex part isn't always better. 22 digital comparators for what exactly??

If they're easy to implement why not? Say you have a 3-phase PSU design with dual LLC for the last stage: that's 6 switches for PFC on primary + 12 on secondary assuming totem pole. I can't do it on a G474 because it's actually not enough HRTIM instances, comparators that feed HRTIM crossbar, dedicated ADC inputs, etc.

The infineon seems to tick all the boxes because you can offload a lot of stuff into dedicated CPUs and timers, but I dread to see the firmware development pipeline :/

[radiolistener]

Out of curiosity, what do you find missing in the STM32G431? It may be an older device, but it still offers a pretty solid feature set for mixed-signal applications: Cortex-M4 @ 170 MHz, hardware FPU, CORDIC and FMAC accelerators, dual 12-bit ADCs (up to 4 MSPS), 4 x 12-bit DAC's etc.

[Xiamatsu]

The small MCUs from Linko Semiconductor
LKS32MC45x series are very interesting.

Cortex-M4F; 192MHz;  256K Flash;  40K RAM
MCPWM (4Pair) ×2
ADC 14bit(2Msps) ×3;  DAC 12bit ×2
COMP ×6; OPA up to 6; HALL (3Phase) ×2;  QEP ×4
CORDIC, FMAC

The LKS32MC452 model contains also 6N  GateDriver


The very small MCUs
LKS32MC08x series (last from  MC0xx)

Cortex-M0+; 96MHz;  up to 64K Flash;  8K RAM
MCPWM (4Pair)
ADC 12bit(3Msps) ×1;  DAC 12bit ×1
COMP ×2; OPA up to 4; HALL (3Phase) ×1;  QEP (opt.)
CORDIC, DIV, SQRT

The many models contains also  6N or 3P3N  GateDriver

[uer166]

For G474: not enough HRTIM timing units, it has 6.

If you have a generic application where you have half bridges with fixed complimentary behavior (with dead time) that's 12 PWM outputs.

If you need some special behaviors like active clamping or diode emulation, that essentially drops it to literally 6 PWM channels which is the equivalent of one 3-phase bridge.
The reason is that the crossbar (the one that is fed by comparators, other timing units, GPIOs, etc) acts on the whole unit's output, not the individual complementary outputs post dead-time. I'm not actually sure you can even devise a system where you add events post-dead-time generator. In any case the easy solution is to add more instances like in one the STM32H7** parts without the shittiness of dealing with Cortex-M7.

[uer166]

The small MCUs from Linko Semiconductor
LKS32MC45x series are very interesting.

Cortex-M4F; 192MHz;  256K Flash;  40K RAM
MCPWM (4Pair) ×2
ADC 14bit(2Msps) ×3;  DAC 12bit ×2
COMP ×6; OPA up to 6; HALL (3Phase) ×2;  QEP ×4
CORDIC, FMAC

The LKS32MC452 model contains also 6N  GateDriver

Sorry but I'm not interested in non-tier-1 suppliers for MCUs. They don't save any material amount of money yet introduce a crazy amount of risk if you're able to get TI/ST/Microchip/etc.

[Shalomie42]

Believe you me, I am forced to work with the C2000 at my day job and I hate the software ecosystem as much as the next guy. I have complained to TI on many occasions. However, I still can't deny their analog / motor control peripheral superiority.

Why does somebody need a billion comparators & PWMs in a single chip? I work for a company that designs high power motor controllers. They need to be cost effective because we sell them in relatively high volume. We can't just put say 5 STM32G4 MCUs on there to cover all our needs because each individual one has 4 comparators or whatever. Its astronomically more cost effective for us to use a single chip that has everything in terms of cost, PCB board space, and complexity.

One of our motor controllers requires 3xADC, 12 PWMs, and 6 comparators (for example). However, again for the aforementioned reasons, we implement more than one instance of a "motor controller" on our boards. Imagine we want to implement 2 motor controllers on the same board. Multiply the required peripherals by 2. Instead of two STM32G4, we can still use the same single chip from TI. Now you should be able to see the scaling and why you would ever want that many peripherals in a single chip - When you are designing actual cost effective product you don't have the luxury of slapping multiple MCUs in a single design for multiple feature instances.

I always wondered myself why TI (for example) gives you more than one quadrature encoder instance. Don't I only ever need 1 for a single motor? Well, believe it or not many of our designs include a redundant encoder for safety reasons. That brings up the count to needing at least 2. Now if we do 2 controllers on a single board, 4... You get the idea.

Also, another thing is also often overlooked is that having multiple MCUs on a single board can increase complexity due to internal messaging, redundancy, fault tolerance, and safety. If I put 2 STM32G4's for example, now I need a whole messaging scheme to synchronize them externally. What if one fails ? Our high power controllers can have a ton of noise on the communication lines. With a single chip having everything we need, I don't have to worry about that. Its just one software load, single point of failure.

[dmendesf]

dsPIC33AK has 5 adcs capable of 40msps each but very little ram and no meaningful way to add more. This puzzles me a lot.

[Darkover]

LPC4370   32-bit ARM Cortex-M4 + 2 x M0 MCU; 282 kB SRAM; Ethernet; two HS USBs; 80 Msps 12-bit ADC;

Can someone beat this? 

Olaf