Author Topic: NXP Cortex-M chips, which ones are popular? (Read 8447 times)

richardman · « **on:** June 22, 2016, 08:41:07 pm »

Looks like there is a lot of love for STM32F chips on this forum. We love them too ;-)

How about NXP Cortex-M? Which ones are popular with people. Do their -M0 chips have similar peripheral subsystems as their -M3/M4/M7 chips?

Experience, gotchas? Pearls of wisdom?

dannyf · « **Reply #1 on:** June 22, 2016, 08:43:39 pm »

The hottest is 1114. Just the right compromise between features, packaging and price.

Like stm32f103.

nctnico · « **Reply #2 on:** June 22, 2016, 09:17:29 pm »

Quote from: richardman on June 22, 2016, 08:41:07 pm

How about NXP Cortex-M? Which ones are popular with people. Do their -M0 chips have similar peripheral subsystems as their -M3/M4/M7 chips?

I have been using NXP ARM controllers for almost a decade. The peripherals are basically the same across the entire ARM7TDMI, M0 and M3 line (perhaps also M4 and M7 but I never really looked at those). There are some small differences though. Some chips don't have specific set/clear bits for the GPIOs and some have a toggle function. On newer devices (like the 11U6x / 11E6x) NXP replaced the extra UARTs with non 16550 compliant ones. But overall there is much less variation between the devices than you probably got used to when dealing with ST. When choosing a device for a design I don't have to care whether I can use existing code because I know it will work or will need very minor changes.

mubes · « **Reply #3 on:** June 22, 2016, 09:25:02 pm »

EDIT: Its just struck me that by saying NXP you might mean 'old' NXP (LPCxxxx) or 'new' NXP (which incorporates the Freescale CORTII). You might need to clarify which you mean, but I can only comment usefully on the LPCs...below;

LPC1549 is the unsung hero of the M3s. For some reason they badge it as a motor controller but it's got a great mix of peripherals and the configurable switch matrix (which also exists on the LPC81x and LPC82x) is a killer feature. Probably my fave chip at the moment.

I am actively using LPC433x and LPC437x is various designs. Quite powerful but a few 'gotchyas' like specific pins not having GPIO functionality and the fact you can't use a couple of peripherals at the same time because of a decode bug in the silicon (one of the CAN busses, if memory serves). The serial flash effectively gives you virtually unlimited memory and the embedded M0s give you a couple of spare processors for drudge tasks when you've got other things to keep the M4 busy with.

The LPCExpresso platform (==Eclipse + GCC for ARM) is powerful and the JTAG/SWD well integrated with features like profiling.

The LPC40xx'es are newer kids on the block, but not too sure what they bring to the party. They have the secondary M0 but when I glanced at them they didn't look compelling....didn't look too deeply.

The future of a LPCs is a bit cloudy with the Freescale takeover though...hopefully they'll clear the air a bit soon.

DAVE

richardman · « **Reply #4 on:** June 22, 2016, 10:06:42 pm »

Yes, I do mean the "old"LPC lines. I know the Kinetis lines are very different.

0xdeadbeef · « **Reply #5 on:** June 22, 2016, 10:13:00 pm »

I think LPC1768/1769 were and still are very popular. They lack some features of the newer derivates though and LQFP100 is a bit on the big side.
LPC1347 was also pretty popular but has been superceded by the LPC1549 -which is also my current favorite.
The only thing I really miss compared to the LPC1768 is the more powerful "general purpose" DMA of the LPC17xx. And yeah, the maximum frequency is a bit lower.

dannyf · « **Reply #6 on:** June 22, 2016, 10:26:33 pm »

Specing lpc chips now can be a risky business, and in my view will likely put you on the losing end of the stick. I suspect that nxp is going To rationalize it's mcu offering around freescale parts and framework.

dannyf · « **Reply #7 on:** June 22, 2016, 10:39:13 pm »

On the freescale deal, it shows something people like new have been arguing for a long time. The mcu business is a software business for the most part.

Once you are specd in, it is very difficult to get kicked out and that customer is a cashhcow for you.

To get specd in, however, you have to win by software going forward - renaisas is an exception giving it's root in auto and it's roll up history. The single bigger hurdle in getting designed in is software, or ease of development. PE was absolutely best in class and for the most part got freescale where it iss today.

Soo I think for guys like St to be successful, they have tobright thee software side of theiir business.

That bring me to cypress. Of all the vendors, those guys are the only one who is doing absolutely innotive things on the software side (also hardware side as well).

Given their small size, they would be a perfect acquisition target for someone with a wide product offering.

nctnico · « **Reply #8 on:** June 22, 2016, 11:00:17 pm »

Quote from: dannyf on June 22, 2016, 10:26:33 pm

Specing lpc chips now can be a risky business, and in my view will likely put you on the losing end of the stick. I suspect that nxp is going To rationalize it's mcu offering around freescale parts and framework.

Why would they do that? That would basically kill the business around their LPC line-up. The LPC line is much older than the Kinetis line so it is likely they have more customers for the LPC line than the Kinetis line. Ofcourse it depends on the volume but I strongly doubt NXP sells more Kinetis devices than LPC. Besides that the LPC line looks way more polished with shorter errata sheets. A quick glance at the peripherals shows me Freescale has re-used many 8 or 16 bit peripherals where you'll find more real 32 bit peripherals in the LPC line. It is very easy to be able to load a timer with a 32 bit value instead of needing pre-scalers. Also for programming the Kinetis parts you'll need to use JTAG or SWD dongles which are a major pain in the ass (drivers, USB incompatibilities, susceptibility to pulses) compared to the (UART) serial bootloader in the LPC series.

TheDirty · « **Reply #9 on:** June 23, 2016, 01:41:49 am »

Quote from: dannyf on June 22, 2016, 08:43:39 pm

The hottest is 1114. Just the right compromise between features, packaging and price.

Like stm32f103.

LPC1114 is my goto, but timed sleep mode and power saving in deep sleep have been a pain in the ass and I still haven't got them perfect. This is one of their early chips and it shows.
LPC1769/1759 for larger work.
LPC812/824 chips are great and I'm using them whenever I can. The peripherals are different, but pretty simple. That SCTimer peripheral is amazing. I implemented an AC dimmer trigger just using the SCTimer and no other code or interrupt. Just need to change one reload value to change the dimmer. Super flexible, wish this was in every chip everywhere. Easy flexible peripheral to pin assignments as well.

I have not tried newer mid-range chips, but I might go try the LPC15xx chip mentioned above. Just don't like paying for features I'm not using.

Also find it difficult to believe NXP would standardize on the Kenetis line, which is not as well formed.

richardman · « **Reply #10 on:** June 23, 2016, 02:14:56 am »

Well, as a 3rd party vendor, we do have to make the right bet, so it's important to have a good crystal ball to know whether NXP will settle on LPC or Kinetis. Too bad my crystal ball isn't very clear

And of course similar worry abound for Atmel SAM stuff. Will Microchip kill it? Probably not, but...?

Ugh.

dannyf · « **Reply #11 on:** June 23, 2016, 10:01:38 am »

The good thing about the lpc line, from a programming perspective, is the similarity of peripherals from one chip to another. Surprisingly, gpio is the least similar.

Kjelt · « **Reply #12 on:** June 23, 2016, 01:32:15 pm »

The bad thing is that they don't have such a large portfolio as ST on their micro's so say you choose one type then ST has them with 128kiB,256kiB,512kiB,1024kiB and a lot of different packages, ideal to tweak and choose for your purpose.

nctnico · « **Reply #13 on:** June 23, 2016, 01:54:12 pm »

Quote from: Kjelt on June 23, 2016, 01:32:15 pm

The bad thing is that they don't have such a large portfolio as ST on their micro's so say you choose one type then ST has them with 128kiB,256kiB,512kiB,1024kiB and a lot of different packages, ideal to tweak and choose for your purpose.

I don't see it that way. Due to the peripherals being (very) similar on the NXP LPC devices you can easely move code from one to the other. I do this code mix&matching all the time.

cgroen · « **Reply #14 on:** June 23, 2016, 02:31:30 pm »

I have used a lot of NXP (LPC) chips, both professionally and for my own stuff.
On the top of my list right now is LPC1549, LPC824 and LPC11c24.
The 1549 is brilliantly with regards to its peripherals (SCT, SWM, CAN etc)

(I have also been using a lot of LPC2148, LPC2458 and LPC1778)

ebclr · « **Reply #15 on:** June 23, 2016, 02:45:51 pm »

http://www.cypress.com/products/32-bit-arm-cortex-m0-psoc-4

Sal Ammoniac · « **Reply #16 on:** June 23, 2016, 03:48:38 pm »

My preferences are the LPC1768 and the LPC4337 or 4357. The 43xx series is dual core--there's an M4 and an M0 on the same die--but tools support for the M0 is somewhat lacking.

NXP are late to the game with the M7. They don't have any in their lineup and as far as I know they haven't announced any either.

jnz · « **Reply #17 on:** June 23, 2016, 08:59:36 pm »

I haven't used it yet, but I like that some of the NXP 1100's have built in transceivers, like for CAN. I very well may use that.

Otherwise, I've been using STM32 and Nordic's nRF52.

doobedoobedo · « **Reply #18 on:** June 23, 2016, 10:52:57 pm »

I've just started with an LPC11U35. I really like it. Mind you I've only used AVRs before. Porting my arduino code to LPCOpen was easier than I thought it was going to be too.

I decided to go ARM and this particular chip for reasons of:
Flash space
USB
Performance

64k flash vs 30k usable flash (328p) - no contest, my ported code compiles smaller too

.
I tried using a 32U4 for USB but lost too much flash space and my code didn't fit

Performance is stellar in comparison, including the peripherals like ADC. Timers are much nicer to use. 32bit and higher clock rates , I wish I'd swapped over sooner.

ebclr · « **Reply #19 on:** June 24, 2016, 10:34:36 am »

You don't even need to port, Already have a Arduino compatible who compiles for arm from several difference projects

Jeroen3 · « **Reply #20 on:** June 24, 2016, 10:58:46 am »

There are also mbed compatible chips. Mbed library is made open source, so you can use them for free.

doobedoobedo · « **Reply #21 on:** June 24, 2016, 02:04:04 pm »

Quote from: ebclr on June 24, 2016, 10:34:36 am

You don't even need to port, Already have a Arduino compatible who compiles for arm from several difference projects

If that's a SAMD, I need real eeprom.

Sal Ammoniac · « **Reply #22 on:** June 24, 2016, 04:09:56 pm »

How about the robustness of the peripherals on NXP versus STM32 MCUs? Which is better under typical error conditions? I'm particularly interested in I2C, SPI, and CAN.

Sal Ammoniac · « **Reply #23 on:** June 27, 2016, 04:03:55 pm »

One of the things I find annoying about the NXP LPC user's manuals is how they depict bits in a register. Here's what an example looks like. It's hard to visualize the layout of the bits easily.

Here's how ST does it in the STM32 user's manuals. It's much easier to visualize the bits.

Jeroen3 · « **Reply #24 on:** June 27, 2016, 04:10:37 pm »

I don't agree with that. Nxp has a shorter overview. You'll be using shifts or defines anyway, so you don't need to visualize the bits. Or at least, I don't.
Or are you using magic numbers?


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: NXP Cortex-M chips, which ones are popular? (Read 8447 times)

Share me