PIC32MM, a Cortex-M0/M0+ contender?

[benSTmax]

I got a newsletter today and one of the titles got my attention. So, it seems like Microchip has launched the PIC32MM series which seems to address the same market as the Cortex M0/M0+. 

Lately, I have been using MCUs from Microchip (8 and 32-bit parts) but also some CM0+ and CM4 (from the former Freescale and ST), so I'm not bashing the Microchip nor the ARM camps.
Looking at the high-level specs of the PIC32MM it seems it has some nice features (dual CLC and PPS) but in the same time some others seems outdated(e.g. the 300ksps 10-bit ADC and 5-bit DAC).
Some of the peripherals seems taken straight from some 8-bit parts (e.g. 5-bit DAC) while others from the 32-bit families.

I will definitely get a couple of these parts to try them out. I think Microchip has a nice offering right now with both MIPS and ARM core parts(from former Atmel).
Even though the low-power figures look good, I would take them with a grain of salt(especially when they came from Microchip  ) and declare myself satisfied only after I'll try out the chip and measure the low-power currents for myself.

Some links:
http://www.microchip.com/promo/pic32mm/
http://www.microchip.com/stellent/groups/pic32_sg/documents/devicedoc/en586396.pdf
http://ww1.microchip.com/downloads/en/DeviceDoc/40001851A.pdf
http://ww1.microchip.com/downloads/en/DeviceDoc/00002191a.pdf

[JPortici]

500 nA sleep current seem nice
soic,sdip package available, nice.
single cycle multiplication for 16x32 bit, two cycle for 32x32, meh
hardware divide. that's really nice
the errata is very short, nice (but it's been out for so little time so there might be a lot of nasty stuff happening)
but the peripherals on the chip are quite poor.

But i have no idea of what to use one for, instead of a dspic for example.
However, if microchip would integrate the core with a radio..

[hans]

Low power is still lacking behind the ARM competition though.

7mA at 25MHz
2.5mA at 8MHz

That's still in the order of 300uA/MHz. Most energy friendly cortex m4 ARM chips have much lower figures, but it's slightly better than other PIC32s I believe (they were in the order of 500uA/MHz).
0.44uA sleep with 0.4-1.0uA RTC and <0.1uA WDT seems very typical these days, but a major step compared to other PIC32s. So that's nice, especially when your board is in sleep for very long times. Additionally, in things like IoT, it's likely the radio will dominate the power budget anyway.

Seems like they have used the PPS peripheral from the PIC24 family. I didn't like the one used in other PIC32s with arbitrary patterns in the PPS choices. It was a bit of a hassle to figure out which PPS you need to pick for a particular SPI port to map correctly.

The peripherals seem quite typical of microchip; to the point and simple. Not excessively complex like some ARM chips are.

All in all I think it's a nice chip, but it's entering a field of some heavy competition with many cortex m0 chips.

[andersm]

The part support has been in Microchip's compilers for I think well over a year, you could get a pretty good high-level overview by datamining the headers. The current compiler has support for unreleased PIC32MK, PIC32WK and PIC32MZ DA parts (true to form, Microchip China have placed the MZ DA datasheet online).

EDIT: The PIC32 FRM chapter on the control DAC boasts 12-bit resolution, while the PIC32MM datasheet only gives 5 bits. I guess there's a reason this thing launches at silicon revision B0?

[benSTmax]

The part support has been in Microchip's compilers for I think well over a year

I remember seeing those PIC32MM files in the compiler installation directory, however this is the 1st time I see the full datasheet of this part.

[ebclr]

If you are looking for low power you must try  the gecko

http://www.silabs.com/products/mcu/32-bit/Pages/32-bit-microcontroller-technology.aspx#low-power

[JPortici]

PIC32WK

http://www.microchip.com/DevelopmentTools/ProductDetails.aspx?PartNO=DM320102-1

[dannyf]

Sounds like a msp432, a 32 bit core plus old peripherals.

I think it is the right aoproach. From a programmers perspective, the coore is uninteresting. What matters is the peripherals.

Personally a M0 or M0+ core plus old peripherals would have been preferred as a replacement for 8 or 16 bit chips.

[nctnico]

Personally a M0 or M0+ core plus old peripherals would have been preferred as a replacement for 8 or 16 bit chips.

Why would you want to get stuck with old 8 bit peripherals? Things like a 32 bit timer make life a lot easier. Also newer peripherals -in general- tend to have stuff like DMA, deeper FIFO's and other useful stuff. Tie that together with basically being locked in by buying the compiler from Microchip I see no advantage at all versus using an ARM based microcontroller.

[asgard20032]

Sounds like a msp432, a 32 bit core plus old peripherals.

I think it is the right aoproach. From a programmers perspective, the coore is uninteresting. What matters is the peripherals.

Personally a M0 or M0+ core plus old peripherals would have been preferred as a replacement for 8 or 16 bit chips.

MSP432, a 32 bit core with old peripheral? What are you talking about? The peripheral that msp432 got is  very similar to msp430 (and most of the time, it is the same one with no modification). But the msp430 is all but old peripheral. It is a nice modern von neuman architecture with many modern peripheral that are quite capable and powerful. The peripheral that msp430 got is comparable with baseline ARM (if we just forgot the dma part). Look at msp430f5529 datasheet and msp430f5 family user guide. The ADC, the timer, everything is powerful and modern. Of course, there is old msp430 with outdated peripheral, but msp432 is not based on those first generation, but on more recent family like f5 family. The only downside to msp430 is the lack of CAN and ethernet, but if we forgot those two, a msp430 got more than enough in peripheral than most people would need.

[JPortici]

Tie that together with basically being locked in by buying the compiler from Microchip

The compiler have improven, a lot. I am finishing an age old weekend project with a dspic33e, i don't know exactly when it happened but now optimization 1 (basic optimization, works in free version) reduce the code size a lot and you can still use the debugger with all the functions

i remember years ago i couldn't step with the disassembly or use the call stack, the debugger would return "can't access to xxx. are you using optimization > 0?"

then i took a look at the disassembly and i left it there, good code.

and PIC32 should be using GCC for the compiler, i don't know if they can actively put garbage on it, or anyway you should be able to recompile it without the licencing (i read it somewhere here that it should not be difficult as in theory the licensing only checks for the serial number, then enable opt > 0, it doesn't alter the compiler)

[Kjelt]

IMHO too late, M0 is already taking over that specific market segment in price/performance wise.
Unless these chips have something unique to serve some niche marketsegment.

I do not understand their strategy, they bought Atmel who sells M0 chips and now they offer something similar but not compatible? Betting on multiple horses or are these the last painfull convulsions of the dying old microchip uC division?

[mikeselectricstuff]

I see no advantage at all versus using an ARM based microcontroller.

Choice of packages, including DIP
Same devtools for 8,16 and 32 bit parts
Manufacturer with a long track record of good availability and not obsoleting parts
Availability of low-cost programming/marking service from manufacturer

....And that's before looking at details of the actual parts

[dannyf]

"I do not understand their strategy, they bought Atmel who sells M0 chips and now they offer something similar but not compatible"

What else do you expect from microchip? The only thing consistent about microchips strategy is that they are always late to the game. Their focus sseems always to protect their 8bit business.

In the end, their success there is killing them.

[ebclr]

"Choice of packages, including DIP"

Very few when compared to all arm producers

"Same devtools for 8,16 and 32 bit parts"

You will need to pay for each one and is not the same, devtool only the IDE is the  same all, compilers is a different product

"Manufacturer with a long track record of good availability and not obsoleting parts"

Only ONE manufactures, arm space have dozen with a muck long track record


"Availability of low-cost programming/marking service from manufacturer"

For sure microchip dev tool isn't low cost, maybe low quality but not low cost, arm programming tool is cheaper  and better, in fact you can program arm with more than 10 different programmers from different vendors

....And that's before looking at details of the actual parts

[mikeselectricstuff]

"Choice of packages, including DIP"

Very few when compared to all arm producers

How many ARM makers do multiple packages for the same part ?  Or evan make any low pin-count packages at all?
Not looked recently but I only recall seeing NXP doing any DIPs

"Same devtools for 8,16 and 32 bit parts"

You will need to pay for each one and is not the same, devtool only the IDE is the  same all, compilers is a different product

My angle here was familiarity & re-learining new tools. Unless you're doing high volumes, using a familiar devtool can easily work out cheaper overall than using a cheaper part that means you need to learn a new toolchain.
The IDE, programmers and debuggers are the same. The compilers are similar enough that you can switch easily.
Purchase cost is a non-issue when doing commercial work, and the free versions of the 16 and 32 bit compilers are entirely useable for a large proportion of applications.

"Manufacturer with a long track record of good availability and not obsoleting parts"
Only ONE manufactures, arm space have dozen with a muck long track record

Which ones exactly ? NXP have been bought so uncertain, and they had a track record of up-reving chips and obsoleting the old ones, Atmel have a poor track record of obsoleting, and have been bought. Several other players haven't been around long enough to  establish a track record.

"Availability of low-cost programming/marking service from manufacturer"

For sure microchip dev tool isn't low cost, maybe low quality but not low cost, arm programming tool is cheaper  and better, in fact you can program arm with more than 10 different programmers from different vendors

What does that have to do with my point about being able to order pre-programmed parts?
As regards programmers, there are plenty of third-party PIC programmers, probably at least as many as ARM. And they fully document their ISP protocol. (Have ARM SWD docs ever been formally released by any manufacturer?)
And Microchip is one of the few makers that produce programmers, which means you're guaranteed that new parts will be supported as soon as they come out, which is rarely the case for third-party programmers.

[andersm]

I do not understand their strategy, they bought Atmel who sells M0 chips and now they offer something similar but not compatible? Betting on multiple horses or are these the last painfull convulsions of the dying old microchip uC division?

The PIC32MM would have been started long before the Atmel acquisition even came up as a possibility. Besides, Microchip will want to continue catering for their customers who have bought into the ecosystem.

[andersm]

and PIC32 should be using GCC for the compiler, i don't know if they can actively put garbage on it, or anyway you should be able to recompile it without the licencing

Both XC32 and XC16 are based on GCC, with source code available as per the GPL. The libraries and support files are not.

[andersm]

How many ARM makers do multiple packages for the same part ?  Or evan make any low pin-count packages at all?
Not looked recently but I only recall seeing NXP doing any DIPs

Several manufacturers have low pin count versions of M0/M0+ parts, I don't know anyone else bothering with DIP (Nuvoton or some other white-goods supplier may have some).

As regards programmers, there are plenty of third-party PIC programmers, probably at least as many as ARM. And they fully document their ISP protocol. (Have ARM SWD docs ever been formally released by any manufacturer?)

The debug port is defined by ARM, and the docs are available from ARM's website if you register. Microchip have (AFAIK) never publicly documented their two-wire debug interface used in the PIC32s, or their debugger protocols, so if you don't want to use their tools you have to use JTAG. Not a huge problem, but an annoyance nevertheless.

[JPortici]

"Choice of packages, including DIP"

Very few when compared to all arm producers

not really, there is only one M0 in dip package as far as i know, there are several pic32 in soid/ssop/spdip

You will need to pay for each one and is not the same, devtool only the IDE is the  same all, compilers is a different product

when is the last time you used them? you don't need to pay nothing, optimization 1 has become very good recently.

For sure microchip dev tool isn't low cost, maybe low quality but not low cost, arm programming tool is cheaper  and better, in fact you can program arm with more than 10 different programmers from different vendors

pic32 can be programmed/debugged using jtag. all new dspics (and i think a couple of the new 8 bitters) can be debugged with jtag.
oh, sure a j-link cost less than an ICD3.. we also built our own production programmer based on the available schematic anyway

[nctnico]

My angle here was familiarity & re-learining new tools. Unless you're doing high volumes, using a familiar devtool can easily work out cheaper overall than using a cheaper part that means you need to learn a new toolchain.

That is the fun part: I have been using GCC + Eclipse for 99.9% my microcontroller development(from various vendors) for the past 10 years and Eclipse for all (including PIC).

Which ones exactly ? NXP have been bought so uncertain,

No, it is the other way around: NXP bought Freescale and I strongy doubt NXP is flushing over a decade worth of business building in favour of the Kinetis parts which have been around for 2 or 3 years.

As regards programmers, there are plenty of third-party PIC programmers, probably at least as many as ARM. And they fully document their ISP protocol. (Have ARM SWD docs ever been formally released by any manufacturer?)

NXP ARM chips don't need a programmer. A USB-UART converter does the job due to their bootloader (the protocol is fully documented). As a bonus the programming port is also a UART so great for debugging and monitoring.

[mikeselectricstuff]

NXP ARM chips don't need a programmer. A USB-UART converter does the job due to their bootloader (the protocol is fully documented). As a bonus the programming port is also a UART so great for debugging and monitoring.

From what I recall when I used them a while ago  it was very slow compared to JTAG/SWD due to using low baudrates and inefficient ASCII transfer - is this stil the case ?

[mikeselectricstuff]

The debug port is defined by ARM, and the docs are available

Does that cover device programming though?
Last time I looked ( a while ago), device flash programming was typically done by manufacturer-specific undocumented protocols over JTAG/SWD.

[Kjelt]

Choice of packages, including DIP

who except some hobbieists will use this anymore? And from the hobbieists market no silicon manufacturer could ever profit.
If needed perse you can always use an adapter pcb.

How many ARM makers do multiple packages for the same part ? 

ST!

Or evan make any low pin-count packages at all?

ST!

[nctnico]

NXP ARM chips don't need a programmer. A USB-UART converter does the job due to their bootloader (the protocol is fully documented). As a bonus the programming port is also a UART so great for debugging and monitoring.

From what I recall when I used them a while ago  it was very slow compared to JTAG/SWD due to using low baudrates and inefficient ASCII transfer - is this stil the case ?

Define slow. It only takes a few seconds and IIRC the data is transferred in binary form with a CRC check (and retries if the data is corrupted). All in all it is a very safe way to program a microcontroller.

[nctnico]

The debug port is defined by ARM, and the docs are available

Does that cover device programming though?
Last time I looked ( a while ago), device flash programming was typically done by manufacturer-specific undocumented protocols over JTAG/SWD.

OpenOCD supports many common ARM chips so I guess the protocols are not that undocumented. The method is simple: use the information from the user manual for the flash controller. On NXP devices this would typically mean transferring the data into RAM and calling a flash programming routine in the bootloader.

[bktemp]

Choice of packages, including DIP

who except some hobbieists will use this anymore? And from the hobbieists market no silicon manufacturer could ever profit.

I once asked Microchip: They don't do this for hobbiests, but to allow simple and quick prototyping.

In most projects I do, I end up using 64 or 80 pin QFN or TQFP packages, but I often start developement using the small DIP28 version to get the most important peripherals working before I finish the layout.

[Yansi]

The debug port is defined by ARM, and the docs are available

Does that cover device programming though?
Last time I looked ( a while ago), device flash programming was typically done by manufacturer-specific undocumented protocols over JTAG/SWD.

FLASH programming in ARMs is done by arbitrary protocols, because every vendor uses different FLASH technology, different FLASH interface with different FLASH control registers. Therefore the SWD/JTAG interface for FLASH cannot be standardized. The protocols must be always open and documented, because the flash programing on ARMs is done mostly using a "flashloader" and usually the user of the device is allowed to program the flash using his own code. So the interfaces are documented.

The flashloader is a small piece of program loadedd into SRAM of the MCU and includes vendor specific algorithms for FLASH programing. Then the programmer transfers data to be programmed to the flashloader. For example for STM32 micros there is usually a separate PDF "Programming manual" covering the FLASH interface.

//EDIT: For example this PDF explains it quite well: https://www.silabs.com/Support%20Documents/TechnicalDocs/AN0062.pdf

[coppice]

who except some hobbieists will use this anymore? And from the hobbieists market no silicon manufacturer could ever profit.

There are a number of people making things on simple SRBP boards, especially white goods makers, who would rather have through hole parts. The number of recent MCUs available in DIP form may not be great, but pretty much every one is a response to specific customer demand.

[ebclr]

In modern days people can use breadboard with lqfp

[Kjelt]

I once asked Microchip: They don't do this for hobbiests, but to allow simple and quick prototyping.
In most projects I do, I end up using 64 or 80 pin QFN or TQFP packages, but I often start developement using the small DIP28 version to get the most important peripherals working before I finish the layout. 

Ok well we use ST chips and just get/buy a load of prototyping boards (the well known discovery boards) that costs €10 max to do our first software development.
DIP is a very expensive package and if the numbers are low it is a guarantee to throw money out the window but if Microchip does not have cheap prototyping boards........

There are a number of people making things on simple SRBP boards, especially white goods makers, who would rather have through hole parts. The number of recent MCUs available in DIP form may not be great, but pretty much every one is a response to specific customer demand.

What country is that? The white goods that are now failing in my and families household (so 10 yrs or older) already have 100% smt in them 

[coppice]

There are a number of people making things on simple SRBP boards, especially white goods makers, who would rather have through hole parts. The number of recent MCUs available in DIP form may not be great, but pretty much every one is a response to specific customer demand.

What country is that? The white goods that are now failing in my and families household (so 10 yrs or older) already have 100% smt in them 

No particular country. There are companies all over the world who still prefer DIL packages for SRBP boards in some applications.

[ebclr]

DIP is a museum piece, the volume production is automated, and is all based on SMD,  also DIP force to have wires longer than necessary adding noise.

I'm sorry Valves , Germanium transistor and DIP are all RIP

[coppice]

DIP is a museum piece, the volume production is automated, and is all based on SMD,  also DIP force to have wires longer than necessary adding noise.

I'm sorry Valves , Germanium transistor and DIP are all RIP

What does the use of DIP have to do with the use of automation?

[Kjelt]

No particular country. There are companies all over the world who still prefer DIL packages for SRBP boards in some applications.

If that is true I am amazed since it costs much more to build in TH technology, the parts are also more expensive and choices become limited.
I can imagine that one of a kind products or superextreme stuff like space tech still wants to use it for reliability or old proven designs but any commercial company that produces quantities >10000 pieces is plain dumb to stick to TH technology. Maybe they are also using airwires for their products as well?

[Yansi]

Space technology is also within SMT, for a decade at least.

[ebclr]

"What does the use of DIP have to do with the use of automation?"

Here you answer

[Kjelt]

Space technology is also within SMT, for a decade at least.

Yes I know, however often there are still systems in use from more than 20 years ago such as the spaceshuttle in the time that kept on using the same tech because of reliability and it was tested and proven. A redesign would cause new testing and an incredible amount of (paper)work.
BTW, it would not surpise me one bit if the russian Soyuz spacecraft still use TH technology, but it looks like they are upgraded every few years also. So perhaps one of our russian "friends" on the forum can comment on this.

[bktemp]

The current compiler has support for unreleased PIC32MK, PIC32WK and PIC32MZ DA parts (true to form, Microchip China have placed the MZ DA datasheet online).

If I understand the datasheet correctly that chip has an embedded 32MB DDR2 SDRAM in the small package variants!
Combined with the TFT controller and GPU it makes a nice little HMI controller with a greatly reduces BOM (basically single chip). For DDR2 you typically use a 6 layer PCB, but for the QFP variant a 2 layer board should work.

[coppice]

Space technology is also within SMT, for a decade at least.

Some of the first applications of surface mount electronics were in space craft.

[coppice]

"What does the use of DIP have to do with the use of automation?"

Here you answer

What point is that video supposed to make? Surface mount component placers are no more automated than through hole ones.

[Yansi]

The current compiler has support for unreleased PIC32MK, PIC32WK and PIC32MZ DA parts (true to form, Microchip China have placed the MZ DA datasheet online).

If I understand the datasheet correctly that chip has an embedded 32MB DDR2 SDRAM in the small package variants!
Combined with the TFT controller and GPU it makes a nice little HMI controller with a greatly reduces BOM (basically single chip). For DDR2 you typically use a 6 layer PCB, but for the QFP variant a 2 layer board should work.

I doubt the onchip DDR memory. The datasheet says "interface", when translated from whatever hell that language is. It is indeed nice to have a DDR2 controller, but purpose of 533MHz DDR2 controller on a 200MHz CPU... I smell some microchip bullshit there. It's 32MByte only capability is also little bit funny.
I would be much more pleased, if vendors such as the damn ST would implement a DDR controller into their Cortex M4 and M7 (STM32F4 and STM32F7) processors. And also make the damn STM32F7 go fast. 400MHz. Cortex M7 at only 216 is a "marketlooser".

[mikeselectricstuff]

[bktemp]

The current compiler has support for unreleased PIC32MK, PIC32WK and PIC32MZ DA parts (true to form, Microchip China have placed the MZ DA datasheet online).

If I understand the datasheet correctly that chip has an embedded 32MB DDR2 SDRAM in the small package variants!
Combined with the TFT controller and GPU it makes a nice little HMI controller with a greatly reduces BOM (basically single chip). For DDR2 you typically use a 6 layer PCB, but for the QFP variant a 2 layer board should work.

I doubt the onchip DDR memory. The datasheet says "interface", when translated from whatever hell that language is. It is indeed nice to have a DDR2 controller, but purpose of 533MHz DDR2 controller on a 200MHz CPU... I smell some microchip bullshit there. It's 32MByte only capability is also little bit funny.
I would be much more pleased, if vendors such as the damn ST would implement a DDR controller into their Cortex M4 and M7 (STM32F4 and STM32F7) processors. And also make the damn STM32F7 go fast. 400MHz. Cortex M7 at only 216 is a "marketlooser".

For the smaller packages google translator says "internal" and it says 32MB in the next column. For the BGA288 package it translates to external. It looks like the smaller ones have embedded DDR2 SDRAM and the larger package has an DDR2 interface allowing up to 128MB. Further down it reads 200MHz DDR clock, this translates to 800MByte/s maximum data rate, eg. 32bit per cycle.
The SDRAM interface on STM32F4/7 is quite slow, becaue it runs only at half the CPU clock and the memory controller/caching (at least on the F4) is rather inefficient.

[coppice]

The machines which form and insert leaded components are more fun to watch. Especially the brutal way most of them trim the leads after the wave machine.

[Yansi]

The current compiler has support for unreleased PIC32MK, PIC32WK and PIC32MZ DA parts (true to form, Microchip China have placed the MZ DA datasheet online).

If I understand the datasheet correctly that chip has an embedded 32MB DDR2 SDRAM in the small package variants!
Combined with the TFT controller and GPU it makes a nice little HMI controller with a greatly reduces BOM (basically single chip). For DDR2 you typically use a 6 layer PCB, but for the QFP variant a 2 layer board should work.

I doubt the onchip DDR memory. The datasheet says "interface", when translated from whatever hell that language is. It is indeed nice to have a DDR2 controller, but purpose of 533MHz DDR2 controller on a 200MHz CPU... I smell some microchip bullshit there. It's 32MByte only capability is also little bit funny.
I would be much more pleased, if vendors such as the damn ST would implement a DDR controller into their Cortex M4 and M7 (STM32F4 and STM32F7) processors. And also make the damn STM32F7 go fast. 400MHz. Cortex M7 at only 216 is a "marketlooser".

For the smaller packages google translator says "internal" and it says 32MB in the next column. For the BGA288 package it translates to external. It looks like the smaller ones have embedded DDR2 SDRAM and the larger package has an DDR2 interface allowing up to 128MB. Further down it reads 200MHz DDR clock, this translates to 800MByte/s maximum data rate, eg. 32bit per cycle.
The SDRAM interface on STM32F4/7 is quite slow, becaue it runs only at half the CPU clock and the memory controller/caching (at least on the F4) is rather inefficient.

Yes, i know very well how slow and buggy (and sometimes plain stupid) the external memory controller is on these STM32 devices. The synchronous interface indeed runs at HCLK/2 or HCLK/3, but that mostly may (or may not) be due to the fact, that SDR SDRAM memory devices over 166MHz are unobtanium (and those 166M ones mostly too), so doubling the speed will only bring problems finding a memory fast enough. Actually - i think there really are no SDR SDRAMs over 166MHz - never seen one. So the right solution here would be to implement a DDR controller (for a  400MHz  Cortex M7 it should be a MUST, otherwise its bummer...). But let me tell ya, if ST will ever add a DDR controller to their MCUs, I think it will be same or worse than the one used so far: buggy as hell and with stupid pinout configuration. (For exampel allowing only one chipselect on a 100pin package, because the two chipselects are both on a single GPIO pin, wtf? Pure idiocy...)

[Howardlong]

DIP is a museum piece, the volume production is automated, and is all based on SMD,  also DIP force to have wires longer than necessary adding noise.

I'm sorry Valves , Germanium transistor and DIP are all RIP

Not really RIP: travelling wave tubes, magnetrons, klystrons, SiGe transistors in high speed applications, DIP in power electronics, for example.

For super quick prototyping, nothing beats DIP, and it's certainly one reason I still use PICs extensively. I can easily get a breadboard running in five minutes with DIP. It'll take twice that long to get an SMD part mounted onto a breakout board.

While I understand where you're coming from, and indeed most of what I do is SMT, your statement is far from absolute.

I did buy a small stock of DIP ARM chips a few months ago, LPC810M021FN8 (8 pin) and LPC1114FN28 (28 pin). I haven't used the 28 pin yet (it's not a skinny DIP either) but I did use the 8 pin. It was a fight to use because you need a lot of those pins for program and debug. In the PIC world you have a choice of using a debug header or a larger chip (also in DIP) for test and debug.

[Howardlong]

PIC32WK

http://www.microchip.com/DevelopmentTools/ProductDetails.aspx?PartNO=DM320102-1

This is quite interesting. One of the big problems I have had over the past few years with Microchip's WiFi offerings is the throughput, or lack of it, the bottleneck being the SPI interface.

I am wondering if the PIC32WK will be available in two formats like the TI CC3200, as both a chip and a module, the module having had some precertification.

(The TI CC3200 chip-only solution I gave up on, trying to get it to work on your own board was a nightmare, even following their own reference design. Just getting the oscillators and on chip SMPSs to work was a chore.)

[Kjelt]

For super quick prototyping, nothing beats DIP, and it's certainly one reason I still use PICs extensively. I can easily get a breadboard running in five minutes with DIP. It'll take twice that long to get an SMD part mounted onto a breakout board.

For hobbieist use and small company small batches stuff, sure. But the point was this is a small niche market. The big numbers and quantities are smd components.
For product development in larger quantities we are talking about at least 3 prototype board releases before a final design is ready.
EMC, groundplanes, oscillations other stuff you want to tackle asap, you can not do that with a different chip package or a carrierboard then the final product has.

[Howardlong]

For super quick prototyping, nothing beats DIP, and it's certainly one reason I still use PICs extensively. I can easily get a breadboard running in five minutes with DIP. It'll take twice that long to get an SMD part mounted onto a breakout board.

For hobbieist use and small company small batches stuff, sure. But the point was this is a small niche market. The big numbers and quantities are smd components.
For product development in larger quantities we are talking about at least 3 prototype board releases before a final design is ready.
EMC, groundplanes, oscillations other stuff you want to tackle asap, you can not do that with a different chip package or a carrierboard then the final product has.

I'd agree with that, but, like I say, if I have the choice to prototype with a part in DIP and an SMD part, for a functional proof of concept in the MCU world, I'll almost always go for the DIP part. Production, yes, of course you are right, most of the time it will end up as surface mount, and in fact I much prefer dealing with surface mount to through hole when soldering.

Keep in mind that at the PoC stage, for MCUs, EMC and grounding issues more often than not are not of the greatest concern, even well into the 10s of MHz: the 70MIPS PIC24E & dsPIC33E  parts for example work fine on solderless breadboards, even with 140MHz internal clocks. The same is not, of course, true for high speed analogue or high speed digital interconnects, but in the MCU world that is frequently not a problem at PoC or unit testing stage unless there's some analogue stuff involved.

Simply put, I don't want to lose hours designing and fabbing a unit test board when I could have a functional PoC working in minutes.

[ebclr]

Because they didn't have the 10 buck development board like those one

http://www.mouser.com/ProductDetail/Cypress-Semiconductor/CY8CKIT-049-42XX/?qs=sGAEpiMZZMtw0nEwywcFgDSEQBlaQVlqGniHMHCOrgJzeKXcAmR6Hw%3d%3d

http://www.mouser.com/ProductDetail/Infineon-Technologies/KIT_XMC_2GO_XMC1100_V1/?qs=sGAEpiMZZMtw0nEwywcFgL8I6Kimd0toFxqtm0%2fEJ5FHa4SHTgwyeg%3d%3d

http://www.mouser.com/ProductDetail/STMicroelectronics/STM32L100C-DISCO/?qs=sGAEpiMZZMtw0nEwywcFgAGAFIfFhSOz5M%252bHJPc6DDk%3d

http://www.mouser.com/ProductDetail/Cypress-Semiconductor/CY8CKIT-059/?qs=sGAEpiMZZMtw0nEwywcFgAVWPmrmL%252bbOWTKlhe6oaTU20idmQOzazQ%3d%3d

and many others

They have no other option than cry for a DIP version since they are lazy enought and don't want to use 2 minutes to solder a LQFP DIP adapter

[Howardlong]

They have no other option than cry for a DIP version since they are lazy enought and don't want to use 2 minutes to solder a LQFP DIP adapter

You're making quite an assumption there. I, and I am sure many others do exactly that _very_ frequently, in fact it's pretty much an everyday occurrence for me. I have a drawer full of assorted breakout boards for exactly that reason.

I absolutely guarantee you that it is impossible to hand solder an LQFP to a DIP adapter in 2 minutes from start to finish, and I have a section in my lab already set up ready to do exactly that. All I have to do is turn on the irons, find the right breakout board from the stash, and take the dust cover off the microscope and I can get started. But with LQFP, after soldering the chip and the header pins, and cleaning the flux off, not only do you need to do a visual inspection, you also need to do an adjacent pin continuity test too. I'm sure with your experience you've had two adjacent pins shorted on an LQFP that didn't show up on your visual inspection. Miss that step at your peril! Realistically I'd have a finished LQFP in ten minutes start to finish from scratch.

With a DIP package, in the same time, I'd already be well on the way to having the answer before you'd even put your breakout board on the breadboard. If you want to do things your way, sure, go right ahead, I'm not stopping you, but I'd rather take the path of least resistance (see what I did there?)

[ebclr]

If fact I do this in less than a minute on a daily basys. If you have the right tools everything is very easy

Simple 2 lines on solder paste pasted with a descent dispensere and a hot air station and everything is ready in less than a minute

[Howardlong]

If fact I do this in less than a minute on a daily basys. If you have the right tools everything is very easy

Simple 2 lines on solder paste pasted with a descent dispensere and a hot air station and everything is ready in less than a minute

Video or it didn't happen, end to end, from deciding you need to do it, to having a finished and tested board. ;-)

[Rerouter]

If he used a bit less solder paste or used longer pads on his design, he would have been done in under 35 seconds, with a hand iron, hot air would more or less match this, if you had a small station you would do a lap around the pins, and with a bigger one could do the whole chip at once,

Arguably as someone why regularly hand solders 0201's i may lie outside the norm...

[ebclr]

[benSTmax]

They have no other option than cry for a DIP version since they are lazy enought and don't want to use 2 minutes to solder a LQFP DIP adapter

...
Realistically I'd have a finished LQFP in ten minutes start to finish from scratch.

With a DIP package, in the same time, I'd already be well on the way to having the answer before you'd even put your breakout board on the breadboard. If you want to do things your way, sure, go right ahead, I'm not stopping you, but I'd rather take the path of least resistance (see what I did there?)

I think the DIP package is very convenient for people with poor soldering skills. In fact I know a couple of very good firmware guys(with poor soldering skills) and the last time one of them tried to solder a LQFP package on an adapter, he spent a day hunting non-existent bugs because of some bad soldering joints.
On the other hand, Howardlong was spot on when he said the DIP & breadboard approach might save a lot of time and headaches for many folks. I fully agree with him

[dannyf]

"I absolutely guarantee you that it is impossible to hand solder an LQFP to a DIP adapter in 2 minutes from start to finish, and I have a section in my lab already set up ready to do exactly that"

Put some of your money behind it and your guarantee may sound more credible.

No, I'm not good at soldering. Yes, I can do it well within 2 minutes.

[bktemp]

After soldering a SMD package to breakout board you are not done yet:
First you need a breakout board for the particular package. You maybe have a selection of generic ones, otherwsie you need to make and order a pcb or etch one, but this takes hours or days.
If you use a generic breakout board, you still need to plug it into a breadboard or solder it to some other board, because you need to add some capacitors to all the Vcc/Vss pins.

Regardless how fast you solder, using DIP will always be faster.
If you use a generic breadboard, the traces from the package to the capacitors will be much longer than when using the DIP variant.
So for a first prototype, DIP has some advantages compared to SMD package unless you buy a ready made evaluation board.

Soldering a QFP IC in 2 minutes is doable, but if you include all the time needed to set up the work bench for soldering and clean up afterwards and inspect the PCB, I doubt it is possible for a single IC (maybe if you have a dedicated work bench for soldering SMD stuff and do not need to clean up your tools afterwards).

[dannyf]

"After soldering a SMD package to breakout board you are not done yet:"

Don't you think you could do the soldering before hand?

Sounds like you are more interested in making an argument rather than making a fair comparison.

If we want to go down that path, how long does it take you to write firmware on a pdip snapdragon 820? Or Xeon e3-1230? ...

[bktemp]

"After soldering a SMD package to breakout board you are not done yet:"

Don't you think you could do the soldering before hand?

I only compare the time from getting the chip until I can load software into the microcontroller.
If I do the soldering beforehand, I only need to plug the DIP ic into the socket because everything else is done, while I still have to solder the SMD ic onto the breakout board (it is unlikely to have a SMD socket, because they are expensive).
DIP will always be faster unless you have a ready made board where you can reflow solder the SMD part.

[dannyf]

"DIP will always be faster unless you have a ready made board where you can reflow solder the SMD part."

In the same spirit, i woukd say that dip will always be slower unless you have a ready made dip part.

[dannyf]

I think the point is that there are advantages and disadvantages to any packaging. The fact that dip doesn't work for you doesn't it it will not work for others, and vice versa.

To me, the pains and sufferings assodiciated with prototyping a smt part are dwarfed by the production convenience as well as wide selection of smt parts.

Obviously, that equation will change if you don't go production or you do extensive prototyping on breadboards, etc.

[Howardlong]

"I absolutely guarantee you that it is impossible to hand solder an LQFP to a DIP adapter in 2 minutes from start to finish, and I have a section in my lab already set up ready to do exactly that"

Put some of your money behind it and your guarantee may sound more credible.

No, I'm not good at soldering. Yes, I can do it well within 2 minutes.

Go on then, video or it didn't happen, start to finish.

[Howardlong]

If he used a bit less solder paste or used longer pads on his design, he would have been done in under 35 seconds, with a hand iron, hot air would more or less match this, if you had a small station you would do a lap around the pins, and with a bigger one could do the whole chip at once,

Arguably as someone why regularly hand solders 0201's i may lie outside the norm...

1) How long to find the breakout board?
2) Clean the board?
3) Visual inspection, hardly good enough.
4) Adjacent pin continuity test?
5) Header pins?

[Howardlong]

1) That's not an LQFP.
2) How long to get the breakout board?
3) Where are the header pins?
4) Where is the visual inspection and adjacent pin continuity test (not really needed for the SOIC demonstrated, but would be needed for LGFP).
5) At least the board was cleaned!

That's demonstrating just one part of the whole end to end job.

[ebclr]

OK, Let's use DIP if you like to be outdated is up to you, SMD is a reality not an option

[JPortici]

it might be me but i keep finding that tqfp adapter are a mix of
-cheap ass shit (pads that comes away with drag soldering in some cases. in other cases the "universal adapters" have traces going up to the middle of the board and they are useless as there are ALWAYS solder shorts between traces)
-very expensive but good quality pads
-useless shit anyway, like having the pins spaced 1.27 mm (WHY? just... WHY?)

and i have to use protoboards anyway so i end up with a LOT of wasted space, so one of three things happen:
- i have to use a pic16/18 for cost/legacy/whatever purpose -> literally no point in using a smd part
- i get a demo board/starter kit and build an expander board around it (but only when absolutely necessary as i want to have freedom in deciding what pins do what, but for first prototypes is okay i guess)
- i have to use a dspic that has a DIP variant. i may be able to use the same pic16/18 board from another job as the dspic is probably pin-to-pin compatible. how cool is that?

so, for what i have to do DIP is far from dead or outdated

[Howardlong]

OK, Let's use DIP if you like to be outdated is up to you, SMD is a reality not an option

Er, that is not what I said, now was it? 

[ebclr]

Believe or not I Like PIC, i simple no more use that, I used a lot 10 years ago

[ebclr]

Mamy I drink to much and I can't solder a lqfp without destroying everything


Here a solution


https://world.taobao.com/item/524702904631.htm?fromSite=main&spm=a312a.7700824.w4002-9391622304.92.sqvXrw

[andersm]

1) That's not an LQFP.

If you're OK with low pin count versions anyway, prototyping with SOIC or TSSOP versions shouldn't be an issue. Adapter boards cost pennies, and I just consider them an essential part of the toolbox.

[Howardlong]

For the avoidance of doubt, here are few from the prototyping pile I prepared earlier, so as you can see I'm hardly averse to making up boards like these, including SOIC, SSOP, QFP, QFN, BGA etc etc, it's just that if I don't need to, I don't.

[nctnico]

"DIP will always be faster unless you have a ready made board where you can reflow solder the SMD part."

In the same spirit, i woukd say that dip will always be slower unless you have a ready made dip part.

I think I have not used my breadboards in over 2 decades. I got fed up rather quickly with wires going out or making poor contact. I rather just make a PCB (cheap nowadays) and have a good base to start from.

[Howardlong]

1) That's not an LQFP.

If you're OK with low pin count versions anyway, prototyping with SOIC or TSSOP versions shouldn't be an issue. Adapter boards cost pennies, and I just consider them an essential part of the toolbox.

Indeed, I have a drawer full of them, from SC70 to 256 ball BGA, but while SOIC, SSOP, QFP and QFN etc are reasonable enough, even down to 0.4mm QFP, BGA is an exception. I certainly don't mount BGAs on breakouts every day, so I only buy in boards as needed with a small buffer. Unlike other technologies, fabbing your own boards for BGA is nigh on impossible not least because you really do need solder mask: without solder mask it's not pretty!

[Sal Ammoniac]

Unlike other technologies, fabbing your own boards for BGA is nigh on impossible not least because you really do need solder mask: without solder mask it's not pretty!

But is that really necessary? Getting a PCB fabbed professionally is dirt cheap these days, especially if you're willing to wait a few weeks, that I just don't see the need to fab them myself any more.

[Howardlong]

Unlike other technologies, fabbing your own boards for BGA is nigh on impossible not least because you really do need solder mask: without solder mask it's not pretty!

But is that really necessary? Getting a PCB fabbed professionally is dirt cheap these days, especially if you're willing to wait a few weeks, that I just don't see the need to fab them myself any more.

It's still not dirt cheap to prototype small volume runs if you want BGA: realistically doing a BGA fanout will need 4mil track and gap, not typically found in pool services.

But having to wait a few weeks is a deal killer for me, by then I'll have moved on.

If you're very patient (more so than me!) and don't mind paying a premium for the tolerances, then sure. Which in some respects brings us back the earlier discussion about how quick it is the prototype in surface mount compared to DIP.

[TheDirty]

Not a professional, just a hobbyist, but I still make my own boards.  Shipping from China to Canada has gotten to be at a minimum 4 weeks so Chinese proto boards are taking a ridiculous amount of time.  As far as time, I find free wiring stuff like the picture above takes too much time and is prone to error.  Making up a single sided board, etching, and drilling takes 20 minutes (timed myself).  No BGA of course, but everything else I've tried, DFN, QFN...

No comment on the speed of dip prototyping, but I haven't used a DIP package in a few years.  EDIT:  Not true, ordered a couple PIC12f1840's in DIP8 package.  I can't think of anything else.

[Howardlong]

Not a professional, just a hobbyist, but I still make my own boards.  Shipping from China to Canada has gotten to be at a minimum 4 weeks so Chinese proto boards are taking a ridiculous amount of time.  As far as time, I find free wiring stuff like the picture above takes too much time and is prone to error.  Making up a single sided board, etching, and drilling takes 20 minutes (timed myself).  No BGA of course, but everything else I've tried, DFN, QFN...

No comment on the speed of dip prototyping, but I haven't used a DIP package in a few years.  EDIT:  Not true, ordered a couple PIC12f1840's in DIP8 package.  I can't think of anything else.

My own PCB fab process takes 30 minutes beginning to end (assuming I have the board already designed) including getting the chemicals out to putting them away again. Unfortunately living in an apartment I don't have a dedicated space for this, it happens in the kitchen    I almost uniquely prefer surface mount for PCBs because only vias need drilling.

The typical design process I take for prototyping any non-trivial design is incremental. Rather than put everything together in one go, I unit test each section. This works reasonably well with bread- and proto boards on the low to medium speed digital side, but for RF I fab each section on a PCB for unit test and glue those together like Lego.

Once I have everything unit tested it's time to put them all together onto a single board design, and go to the board house. If I'm lucky, only one or two spins later I have a production board.

[Sal Ammoniac]

If you're very patient (more so than me!) and don't mind paying a premium for the tolerances, then sure. Which in some respects brings us back the earlier discussion about how quick it is the prototype in surface mount compared to DIP.

I always have several projects in the pipeline, so when I send a board out to be fabbed I just switch to another project and write firmware, or do board layout, or whatever, until the PCB arrives and then I switch back to that project.

[nctnico]

I do the pipelining too but looking at Howard's boards having those made will take too long. I'd probably etch such boards myself to have a prototype to work with the same day.

[boz]

Nice chip. Good to see Microchip making the investment for us poor deluded retards who quite like the pic32 (DIP and SMT Versions).  I also like Pepsi over Coke and prefer Fords over GM for those that like a real argument.

[asgard20032]

The problem of surface mount vs dip, is not soldering, but getting those breakout board. They also can get quite expansive and even cost the same as the chip. MCU : 3$, breakout : 4$... Its not rare to see such price.

[andersm]

They also can get quite expansive and even cost the same as the chip.

For simple prototyping supplies like that, I have no problem ordering from eBay where you can get fifty of the smaller boards for three bucks, delivered. They're probably not going to be of as high quality as your five-dollar DigiKey board, but they're good enough.

[Howardlong]

They also can get quite expansive and even cost the same as the chip.

For simple prototyping supplies like that, I have no problem ordering from eBay where you can get fifty of the smaller boards for three bucks, delivered. They're probably not going to be of as high quality as your five-dollar DigiKey board, but they're good enough.

I'd whole heartedly agree with that, investing in a stash of various boards is worthwhile, the biggest problem being that there are so many varieties.

I also keep a stock of Schmartboards many of which have the benefit of a ground plane. While it's still no substitute for a properly designed PCB you can do at least some mixed signal on them. They are expensive though. I have in the past made my own breakout boards, and still do if necessary, but having a stock is one less step to your next prototype.

These are the contents of my breakout drawer...

[westfw]

Speaking in isolation of political factors, I think that it is very good that there is at least one microcontroller vendor that is implementing something other than the most recent ARM core, and the MIPS core is a natural alternative.  It ought to be technically viable as well, but it's a little surprising how being the only implementer works against microchip...

[Howardlong]

While I've been using PICs or their forerunners for decades, I do also work on hardware and firmware for ARM from M0 to M4F (TI & NXP) quite extensively, and quite recently I was working on an Atmel AT32UC3C for the first time.

What concerns me is that there's a school of thought that seems to suggest that once you have picked an ARM then you can use that experience across every ARM. Practically speaking, there's so little common ground across the mutitude of ARM platforms, including both the chips implementations themselves and the development platforms, there isn't much benefit to stipulating ARM as a fundamental requirement.

More important is going to be if a device has the right amount of memory, performance, peripherals and price, plus of you already have a dev team with skills in a given vendor's devices, there's a significant cost in time and money to switch.

Having awareness of what's out there is certainly a benefit, but I think that to simplify it down to ARM vs everything else doesn't bear much scrutiny.

[bktemp]

I'd whole heartedly agree with that, investing in a stash of various boards is worthwhile, the biggest problem being that there are so many varieties.

It can be really annoying. Almost every manufacturer has its own package variations. I recently had to build a prototype using a small 8 pin IC. I had 0.5mm breakout boards but the package width was completely different, so I had to cut the breakout board to adjust the distance of both pin rows. It was some extra wide DFN package I had never seen before.
0.4mm, 0.5mm, 0.635mm, 0.65mm. So many variants, some with pins others DFN/QFN. For QFP/SOP many different pin counts can share the same board, but that does not work for QFN/DFN because of the thermal pad shorting any exposed pads. So for QFN/DFN every package needs its own breakout board.

[mikeselectricstuff]

. For QFP/SOP many different pin counts can share the same board, but that does not work for QFN/DFN because of the thermal pad shorting any exposed pads. So for QFN/DFN every package needs its own breakout board.

As long as you don't need the thermal pad, which is often the case, you can usually get away with covering it with kapton tape

[bktemp]

. For QFP/SOP many different pin counts can share the same board, but that does not work for QFN/DFN because of the thermal pad shorting any exposed pads. So for QFN/DFN every package needs its own breakout board.

As long as you don't need the thermal pad, which is often the case, you can usually get away with covering it with kapton tape

That's the quick&dirty solution I often use if I don't have a perfectly matching breakout board.
Some parts require the pad for electrical reasons. One example is Altera Cyclone IV in TQFP. It has a pad that needs to be connected to GND, otherwise it does not work at all. My TQFP144 breakout board needed some modifications to make it work.
I also had some QFN/DFN ics that didn't require any connection to the pad, but if left unconnected the chip behaved different. That's why I always try to connect the pad to GND (or the most negative voltage for some ics), even when using breakout boards. If possible, I avoid QFN/DFN because of the centre pad for prototypes.

[Howardlong]

The ground pad is one of the problems of QFN prototyping boards. Some chips have it connected to ground, others it's floating, but still needs connecting. QFN boards with proper ground lands and vias aren't a problem, but some don't have both of these facets. Some have the plated through via but no land. It's about a 1mm via, so bigger than normal, but you can't get an iron down it. I have no idea what the point of that is, it's very difficult to make the ground pad connection as the solder is wicked down the via, and the device slightly floats above the pad. Reflowing is equally hot and miss.

One trick I do if the pad is connected internally to ground is to use it as a reference, and switch the meter to diode test. You can then quickly test every land has connection. If the pad's not connected to anything on the chip you're out of luck.

[nctnico]

With QFN packages I put a couple of vias in the the pad for the ground pad. When soldering I fix the outer pins first and then I let the solder flow into one via of the ground pad and wait until the solder comes up into the other vias. Sometimes I use hot air in case of a multilayer board but this trick works very well either way (also for TSSOP and other packages with a ground pad).

[dannyf]

The logic for a capacitor tester seems to be simple, measure the voltage and the current, and update the energy discharged. done.

[benSTmax]

It is funny to read all these intricate methods for just soldering a package like (QFP, QFN, DFN, ...) when in the end the whole 2nd part of this discussion on the packages started between people defending the DIP and other blaming it. Yes, DIP is old but it still gets the job done for quite a few of us.

The neat thing about Microchip is they still offer this old packages which is easy to prototype with.  Once, I was in contact with a Microchip support guy and he told me the DIP variants are being specifically offered, so students, hobbyists and others can play with their chips. They weren't selling high volumes of this package, however the DIP helped in the long run to sell more of the QFP, QFN, ... (much smaller than DIP) packages.
I guess he was right. I followed the same path with one of their chips. I put it on the prototype board and once the module I built around it worked as expected, I integrated with the rest of the modules. Afterwards, I switched to a small PCB and a QFN variant of the same chip. Indeed, the DIP package I played with helped the sale of its QFN variant.

[Howardlong]

Anecdotally I'd agree with you in my experience, I prototyped two designs recently with low pin count PICs in DIP. In system testing and production I switched to QFN, we're now at 20ku as I write. While it's not 7 figures, it's not chicken feed either.

Back to QFN prototyping, the boards I was referring to with the rubbish ground lands are Schmartboards. I generally rave over these because they have a solid ground plane, just a shame they make if so hard to solder the ground pad.

[andersm]

With DIP you're seriously limiting your choices of parts, and not just when it comes to MCUs. And while they may be appreciated by beginners, I think a modern hobbyist or student is more likely to use a header board or a complete module like an Arduino.