MSP432P401R: Adding I2S to the MSP432

Look at the CC3200. It has I2S port, parallel camera port, multilayered data bus, 32-channel DMA, dedicated 2-pin SD-Card port, all data peripherals are buffered, 80 MHz CortexM4, 256 kB double access SRAM.

CC3200 will be discontinued in favor of CC3220. But this should not be discussed on forums. :)

I would suggest you to look at the TMS320C5545 DSP also. But it needs much more technical expertise to develop with it than for general purpose MCU in all areas.

Ilmars said:

If I am wrong, then please correct me ;)

AFAIK you are correct - at least concerning Cortex M. The TM4C familiy is missing this peripheral as well.

Ilmars said:

There's plenty of ARM microcontrollers from another manufacturers with i2s support.

Exactly.

DSPs, as suggested here, are a quite different class of animal. Albeit they might be appropriate for audio applications.

Alexey Bagaev said:

I know what is on all minds, - everyone wish all-in-one dedicated MCU which would be as low power and debuggable as MSP430, as powerful as TMS320C5545 DSP and MSP432, and as good in peripherals as C2000 and probably as cheap as MSP430G series.
TI's engineers read these posts. :)

Sort microcontrollers for your favorite distributor by price, and some Cortex-M device near MSP430G series will pop up for sure. There is no problem anymore, to find M0 / M0+ / M3 for 1$ (or less) in low quantity. Don't know what is situation with TI DSP's, but TI Cortex-M devices are too expensive compared to other producers.

Alexey Bagaev said:

In high quality products the MCU could be one of the least expensive element of the system. Dumping the price of MCU to beat competitors doesn't change BOM of the end device even to 10%.

True.

But, to be really honest, just a few products are  really "high quality" - except in the eyes of the project managers and sales people.

I have been working on projects that selected chips and MCUs purely by the price, and swapped vendors and toolchains for a few cents. Just "mass products" in the purest sense of the word. Just numbers counted ...

Alexey Bagaev said:

I know what is on all minds, - everyone wish all-in-one dedicated MCU which would be as low power and debuggable as MSP430, as powerful as TMS320C5545 DSP and MSP432, and as good in peripherals as C2000 and probably as cheap as MSP430G series.

Look at this one. Low power, lower price than msp432. Packed with peripherals. I2S is there too. Performance-wise 16-bit DSP SIMD instructions of Cortex-M4 are not that far behind tms320c5x. Notice that only one kind of msp432 is available in stock: MSP432P401RIPZR :) Disclaimer: I am mentioning competing chip only because TI does not have one (low power /w I2S).

Ilmars said:

Look at this one. Low power, lower price than msp432. Packed with peripherals.

The same vendor has a series of Cortex M0/M0+ devices (price-wise directly competing with the MSP430), which almost all feature I2S. Such a device would suffice if no complex operations on the audio data are required. It's a pity TI does not have M0 ...

I agree that TI have plenty of chips with i2s and adding such interface to msp series would be just waste.

Alexey Bagaev said:

This setup doesn't intended to be ultra-low-power

Why msp430 shall be limited to low power setups? Even TI reference designs with msp430 on-board often are far from low power. One to mention: solar MPPT charge controller. That device at max power dissipate more than 5W heat in losses alone, not to mention output power of that thing :)

Sure, msp430 you tend to shift focus to, does not match performance needed for i2s. Please note that we talk about msp432.

The limitation of power supplies cannot make MSP432 be same efficient at high peak power and in ultra-low power save modes in the same package. That's why it has integrated hybrid power management solution.

The idea of providing or getting consulting in a way of counterarguments and arbitrary doesn't look interesting to me. To be straight - I don't like it.

Alexey Bagaev said:

The idea of providing or getting consulting in a way of counterarguments and arbitrary doesn't look interesting to me. To be straight - I don't like it.

Right. You like that your arguments are not challenged.

Alexey Bagaev said:

Is it bad that MSP430FR5994 can handle DSP process 40 times faster than M0+?

Do you have any reference for that claim ?

What about the power consumption during this "40 x faster" DSP operations ?

Alexey Bagaev said:

Or is it bad that MSP430FR5994 was designed for ultra-low power applications, for peak power consumption less than 500 mW.

While this is surely true, audio (especially digital audio) applications using I2S are very likely not low-power applications.

Alexey Bagaev said:

Including I2S peripheral means claiming DSP capabilities (in real situation you wouldn't send/receive and/or store RAW audio data.

I don't think so.

A music player application might just channel audio data from permanent storage (SD card) to an audio DAC using I2S, with very few or no "DSP" operations involved. That would not make it a real DSP application IMHO. Otherwise, one can call almost anything "Digital Signal Processing" ...

f. m. said:

Do you have any reference for that claim ?

f. m. said:

What about the power consumption during this "40 x faster" DSP operations ?

Reference: slaa698a

The situation with CortexM4 is much better since it has full emulated DSP stack. But MSP430FR5994 is still approximately 3 times faster than CortexM4 in cFFT-512. (No reference, don't ask).

Alexey Bagaev said:

(No reference, don't ask).

That would be interesting, however.

I suspect this advantage implies equal clock frequencies as well, and is limited to data types directly supported. For 32-bit types, float and double, the results would be different. And the Cortex M4, albeit having "DSP-like" instructions, is not at all a DSP.

But I still wouldn't consider both (MSP430 and Cortex M) as predestined for "real" DSP applications.

For an audio-output application - like a music player - I would choose the smallest/cheapest MCU that could channel through the data fast enough. A missing I2S interface, implying bit-banging and the corresponding performance loss, is clearly a disadvantage here.

Alexey Bagaev said:

and this process is not for short period of time - this is usually for always-on operations.

And thus not really a low-power application. I strongly suspect that HiFi data (ADC resolution > 16 bit, 48kHz sample rate) would "crash" that performance. But TI has real DSPs.

f. m. said:

I suspect this advantage implies equal clock frequencies as well, and is limited to data types directly supported. For 32-bit types, float and double, the results would be different.

Of course, as I mentioned, in ticks, equal conditions - cFFT512.

Alexey Bagaev said:

But MSP430FR5994 is still approximately 3 times faster than CortexM4 in cFFT-512

Faster only if they run at the same clock frequency. Why run msp432 at 16MHz if it can do 48MHz? 3x clock-efficient, more than 3x energy-efficient, but not necessarily faster.

Reference here.

Alexey Bagaev said:

Now you start to support my side.

Sure - LEA accelerator is good, msp430 is good, many Cortex-M microcontrollers are good. I don't stand in any side or "chip worshipping religion" except - pick what is best for given task and conditions, don't fall in love with single kind of devices :)

p.s. Adding i2s to msp432 is waste anyway.

Alexey Bagaev said:

I have no intention to brake you position.

Neither do I have any intention to "fight" your position.

My contention is with the TI microcontroller portfolio. Does TI neglect Cortex M0 to avoid internal competition ? Many customers love continuous portfolios, covered by one toolchain. It allows them to scale up their product portfolio in a quick and easy way. While the MSP432 has many MSP430 peripherals implemented, it is a continuation break (their assembler code is not portable). Neither there are much similarities to the TM4C controllers. And no Cortex M7 ?

While TI seems to have an established hold in the DSP and application processor (Cortex A) market, I see other vendors better positioned in the MCU arena. Just my personal opinion.

BTW, has anyone noticed the ongoing merger frenzy in the MCU market ?