PSoC examples

[Harvs]

Every now and again someone pipes in with the suggestion that PSoC would be the answer to a problem, and recently jmole has a thread over here about his/her kickstarter project.

https://www.eevblog.com/forum/crowd-funded-projects/thinking-of-starting-your-own-crowdfunded-project-ask-me-anything!/

Rather than hijack that thread completely with PSoC discussions, I thought it would be wise to start one over here.

I've done some basic research on these in the past, and the inclusion of the cortex Mx into there range certainly pricks a lot of interest.  But I honestly still don't understand where they fit into the picture.

If anyone has come across some examples (or has some), of where using a PSoC is "better" (e.g. lower cost or could actually get the job done), than with the more regular parts I'd love to read it.

Seriously I'm not having a go, I just genuinely haven't seen a real world example where using a PSoC made for a better design, and their marketing videos are just crap.  I keep reading examples of where PSoC are better than arduino, but come on...

For example could a PSoC do the iPod nano display drive in hardware that Mike recently did with the small FPGA?

[mikeselectricstuff]

For example could a PSoC do the iPod nano display drive in hardware that Mike recently did with the small FPGA?

That's something I will be looking into. Enough logic to generate the DDR and LLVDS voltage levels and have a fast interface to the host could be useful - I'd rather use an MCU than an FPGA for this project (now looking at a couple of thousand screens BTW) as we're looking at doing fonts & overlays over the image data 

Last time I looked at Psoc a few years ago, ( the earlier ones), they were expensive unless you could make good use of the analogue stuff to save cost on external hardware

[jmole]

I worked on a project a while ago that measured conductance between electrodes placed on a waterfall to make music based on how much water was flowing between various segments.

Probably 10-12 electrodes if I remember correctly. I wanted to measure conductance between each pair, so something like 100 or 144 readings (yes, some are redundant), where each reading consisted of setting both electrodes to 0, setting the measured electrode to high-Z, setting the drive electrode to 1, then taking an ADC reading shortly thereafter. Essentially it was like an integrator, using the R of the water and the C of the input pin on the PSoC. The input pins were buffered with one of the internal op amps, so there was little charge leakage or current demand on the pins, and it worked pretty well.

Anyway, I don't have the project at hand, so I don't remember if that's exactly how I did it (might have used a current DAC, rather than the integrator I described), but being able to switch between analog/digital operation on those pins was pretty useful, as was being able to buffer/multiplex them internally to a SAR ADC.


Another example, with no analog at all, is the work I've done on ZeroTouch. If you take a look at the CHI 2012 paper, you can see the system architecture on page 4 of the PDF. Using the verilog capabilities in PSoC, I created some glue logic that lets us quickly collect data from the sensor and pump it to the EZUSB FX2 chip. So rather than wasting MCU cycles on data transfer, we can flip a pin and let the digital logic do the work automatically.

[amspire]

I went to the Cypress Website to get Creator and their download page died when I was trying to get Creator. I imagine they will fix this pretty quickly. It might actually be working as data sheets also went down and you can get them again.

Anyway, after a lot of digging, I worked out some direct download links for anyone who just wants to get Creator directly:

The PSoC Creator 2.2 Sp1 install (alternative locations)
http://dlm.cypress.com.edgesuite.net/akdlm/downloadmanager/software/PSoCCreatorSetup_2.2_sp1.exe   (614MBytes)
http://dlm.cypress.com.edgesuite.net/akdlm/downloadmanager/software/creator/PSoCCreatorSetup_2.2_sp1.exe  (614MBytes)

PSoC Creator 2.2 ISO of install disk.  Sp1 and Sp1 release candidate 11
http://dlm.cypress.com.edgesuite.net/akdlm/downloadmanager/software/creator/psoc_creator_2.2_sp1.iso  (1136 MBytes)
http://dlm.cypress.com.edgesuite.net/akdlm/downloadmanager/software/creator/psoc_creator_2.2_sp1_rc11_b572.zip  (1222MBytes) (Seems to be the latest, but I went with SP1)

The ISO to make an install DVD is probably not needed. I think it just has a lot of prerequisites like Microsoft NET packages.

[mikeselectricstuff]

I had a quick look at the Cypress website, and they are not doing themselves many favours in publicising the family - several broken links, and despite looking at the product page and watching a video I couldn't tell if their IDE actually included a C compiler, or you had to use it with Keil.....

Pricing on the M3 parts is pretty high - significantly more than a CM3 MCU+CPLD, but the M0 parts look good, when they become available.
 

[andyturk]

Pricing on the M3 parts is pretty high - significantly more than a CM3 MCU+CPLD, but the M0 parts look good, when they become available.

It gets slightly more reasonable if you can use some of the other stuff to replace external components. I just ordered a board (a shield for jmole's freeSoC) that uses the internal op amps instead of outboard components. The "special" I/O pins also mean I don't have to mess with level translators. The chip even has an on-board SMPS controller (not sure if I'll use that).

[tinhead]

I couldn't tell if their IDE actually included a C compiler, or you had to use it with Keil.....

Cypress IDE installation always contains C compiler, PSoC Creator 2.0 is using Keil and/or Gcc (for ARM and 8051).
So all you need is to run installer, with Cypress products it is the best to leave paths to what the installation
is proposing (not sure if that fixed today, but in the past it was always a small fight to get things running when
path was customized).

PSoC Designer 5.x is using Hitech C compiler, it is as well included in the installer.

It make sense to not touch nor play with the inluded gcc/hitech/keil, they works even if you have tons of other
compilers on your PC.

[jmole]

...and despite looking at the product page and watching a video I couldn't tell if their IDE actually included a C compiler, or you had to use it with Keil.....

Definitely comes with a free compiler, GCC. Works very well for me.

Pricing on the M3 parts is pretty high - significantly more than a CM3 MCU+CPLD

The upshot is, you can do verilog development without another IDE. And it's trivial to integrate with the MCU, since you can reconfigure it in software. If you've got two chips, like an NXP M3 and a Lattice CPLD, changing the interface between the two could be frustrating at the design/proto stage, and impossible in the field.

Plus, have you ever used an FPGA/CPLD tool that you've liked? That was easy to use out of the box? That you could easily version control with your other bits of firmware? Probably not. Not in my experience anyway.

[blewisjr]

Ok so I hear and read all these benefits about these PSoC chips.  From jmole's freeSoc site even the cypress site.  But what real advantage do these things give to the hobbyist / beginner.  For instance right now I spend a lot of my time with 8 bit avr's.  I have been thinking about getting involved in some arm development as well.  Besides speed and elimination of some discreet components what do these things give me that and AVR or ARM chip can't from a hobbyist bang / buck perspective?  As far as I can tell it is really not saving any money except from a time perspective.

[jmole]

Ok so I hear and read all these benefits about these PSoC chips.  From jmole's freeSoc site even the cypress site.  But what real advantage do these things give to the hobbyist / beginner.  For instance right now I spend a lot of my time with 8 bit avr's.  I have been thinking about getting involved in some arm development as well.  Besides speed and elimination of some discreet components what do these things give me that and AVR or ARM chip can't from a hobbyist bang / buck perspective?  As far as I can tell it is really not saving any money except from a time perspective.

It all depends on the application in question. If you have a need for analog hardware and it's not too exotic, odds are you can do an implementation with just a PSoC. Ditto if you need some basic high-speed glue logic.

If you're just doing SPI, I2C, and some general processing, an ARM will serve you just as well as an Ardunio.

[amspire]

Ok so I hear and read all these benefits about these PSoC chips.  From jmole's freeSoc site even the cypress site.  But what real advantage do these things give to the hobbyist / beginner.  For instance right now I spend a lot of my time with 8 bit avr's.  I have been thinking about getting involved in some arm development as well.  Besides speed and elimination of some discreet components what do these things give me that and AVR or ARM chip can't from a hobbyist bang / buck perspective?  As far as I can tell it is really not saving any money except from a time perspective.

In the case I was looking at, I was considering using 3 Atmega chips and an extra FPGA as well. The reason I had to look at using 3 micro's is that two of them wire running a fast timer loop with regular timer interrupts and they didn't have the capacity of doing any other tasks without the risk of missing a timer interrupt. I was basically running the two Atmega's at full speed doing one task, so they could not do anything else.

I am hoping that by going to the Cypress chip, I can offload the time-critical tasks to hardware logic, and also add the logic from my external FPGA as well. If it works, I can replace the 3 Atmega's and one FPGA with one Cypress chip. That would be fantastic. Just one device to program and a big simplification of the circuit.

Plus I can make the circuits behave exactly as I want. I need to have some 8 bit PWM's but I need them to behave properly when set to 0 (a constant 0V out from the PWM) and when set to 255  (a constant 1 out from the PWM). Many micro PWM's do not behave correctly at one or both of these limits. With the Cypress chip, I can force it to work exactly the way I want, even if it means adding extra logic after the PWM to force the output to 0 or 1 in the specific case the PWM is set to 0 or 255.

With other micros, you have to live with the way the designers decided the internal features should work. With the Cypress, you can overwrite the way these embedded features work by adding extra logic.

[amspire]

Also, the feature set of these Cypress chips is just incredible. It is easy to get a micro that include LCD drivers, Capacitive touch sensors inputs, comparators, opamps, 0.1% voltage references, 20 bit delta sigma A/D's, D/As and lots more. You can chose what output goes to what pin, so board layout can become very, very easy.

[BravoV]

Also, the feature set of these Cypress chips is just incredible. It is easy to get a micro that include LCD drivers, Capacitive touch sensors inputs, comparators, opamps, 0.1% voltage references, 20 bit delta sigma A/D's, D/As and lots more. You can chose what output goes to what pin, so board layout can become very, very easy.

Wait till you played with the on fly dynamic "re-assignment" at those highly re-configurable analog/digital blocks, it feels like having a mutant mcu that can morph into other beast, a dynamically change configurations repeatedly while the device is running. 

[blewisjr]

Ok I see now.  They are a great choice for very complex applications as they can simplify much of it.  This is probably something I would not need for quite some time but could be very fun to play around with at some point.

[Harvs]

Also, the feature set of these Cypress chips is just incredible. It is easy to get a micro that include LCD drivers, Capacitive touch sensors inputs, comparators, opamps, 0.1% voltage references, 20 bit delta sigma A/D's, D/As and lots more. You can chose what output goes to what pin, so board layout can become very, very easy.

And there's my problem with understanding where these devices fit into the grand scheme of things.

The only feature you've mentioned there that a current model ARM chip wont have is opamps, and the 20bit ADC, DAC you'd need to get one that specifically has those in it (which there are a number of parts that do.)  From the spec sheet the op-amps are only fairly basic garden variety with 2-3mV input offset (PSoC5), and an unspecified input bias current (though I'm assuming they're a FET input.)

You also mentioned in the previous post that you replaced 3 ATmega's.  It is not my intention to pass any judgement at you're design, but I can't help but wonder what the result would have been with a Cortex M3 or even M4 thrown at it?

I guess it comes down to the configurability, and building you're own components.  Which is where I wonder about how much fabric they have compared to say a medium sized CPLD.  It's a shame there's no VHDL support, but I guess it wont take much to pick up Verilog knowing VHDL.  I'll just have to buy a dev board and try out some things.

[amspire]

You also mentioned in the previous post that you replaced 3 ATmega's.  It is not my intention to pass any judgement at you're design, but I can't help but wonder what the result would have been with a Cortex M3 or even M4 thrown at it?

Given the same clock speed, probably worse. An 8 bit micro can often change a pin in one clock cycle. ARM processors usually have pipelining so it will probably take several clock cycles.

If you are talking about replacing a 20MHz 8 bit Atmega with a 400MHz ARM, then yes - it will be much faster.  Not as fast as a logic circuit though. If you are going to choose a M3 ARM core, why not choose the Cypress PSoC Cortex M3 Arm chip and get the extra features of the Cypress ICs thrown in?

[Harvs]

You also mentioned in the previous post that you replaced 3 ATmega's.  It is not my intention to pass any judgement at you're design, but I can't help but wonder what the result would have been with a Cortex M3 or even M4 thrown at it?

Given the same clock speed, probably worse. An 8 bit micro can often change a pin in one clock cycle. ARM processors usually have pipelining so it will probably take several clock cycles.

If you are talking about replacing a 20MHz 8 bit Atmega with a 400MHz ARM, then yes - it will be much faster.  Not as fast as a logic circuit though. If you are going to choose a M3 ARM core, why not choose the Cypress PSoC Cortex M3 Arm chip and get the extra features of the Cypress ICs thrown in?

But they're not the same clock speed, STMF0 -> 48Mhz, F3 -> 72Mhz, F4->168Mhz.

From what I've seen, cost.  An STM32F3 is in the 5-10$ range.  An F4 ranges between $11-20.  A PSoC F3 @ 67MHz, is in the $14-30.

But as Mike said earlier in the thread, the M0's look a lot more promising.  With chips in the $3-4 range, which is similar to the STM32F0, except with half the RAM and Flash, so I guess they've traded memory for their configuration on the silicon.  However, you can't buy them yet...

[amspire]

If you do not need any of the Cypress PSoC special features, you can get a cheaper processor from other suppliers. If I can use a $1 micro, I am not going to use a $14 Cypress M3 PSocC instead. In my case with the 3 Atmegas, I could probably make the two dedicated ones ATTiny processors instead at about $1 each - or something else.

If you need some of the extra features, then even at $20, it may be a costs saving. How much does it cost when you have an non-Cypress ARM chip + an ASIC chip to attempt to partially get the same capability? How much time are you going to waste trying to get the two different ICs to work together? How much time will it take to learn two different development environments?

It is all about choosing the best device for the job. In some cases, nothing else will come close to a Cypress PSoC.

I haven't used a Cypress micro myself, but it really does look a huge time saver. I could end up using a 8051 based PSoC instead of the Arm version as it is much cheaper and it may do the job. The micro speed for me is not a big issue if I have programmable hardware. The main issue with the 8051 core is you have to request a free license for the 8051 compiler, and you have to reapply for the license every year. It is possible that the free version of the compiler could disappear in the future.

[Corporate666]

Ok so I hear and read all these benefits about these PSoC chips.  From jmole's freeSoc site even the cypress site.  But what real advantage do these things give to the hobbyist / beginner.  For instance right now I spend a lot of my time with 8 bit avr's.  I have been thinking about getting involved in some arm development as well.  Besides speed and elimination of some discreet components what do these things give me that and AVR or ARM chip can't from a hobbyist bang / buck perspective?  As far as I can tell it is really not saving any money except from a time perspective.

We used to use exclusively Atmel chips for our designs (makes it easier having one hardware/software platform).  With Atmel, the chip you choose depends on the features you need.  One chip may support CANbus, but will have 3 timers and one capture pin.   The part with 2 capture pins may have 6 times and 8 PWM channels, but is a much "bigger" chip than you need - you're just buying it for the extra capture pin.

With the PSoC, it's different.  You have blocks that can be anything - timers, counters, PWM's or whatever you like.  So you do not spec your chips based on individual functions, but rather based on how many blocks you need.  And there will be several versions of the same chip family in the same package with differing numbers of blocks, flash, SRAM, etc.  And I think the chips are competitively priced.   

As an example, I went to Digikey and typed in ATMEGA, and selected one with 32k of flash and 25 or more IO's.  The cheapest one is the ATMEGA329 that is $3.89 in 1,000 unit quantities, or $6.97/ea.   That ATMEGA runs at 16Mhz and has 3 timers, 1 input capture, and 4 PWM channels.  It has no USB support or DAC and but has SPI, I2C and a UART.

I went to Cypress (direct is cheaper than Digikey) and the cheapest PSoC3 with 32k of RAM is $3.44 in 1,000 quantities, or $4.52 for one.  It runs at 50Mhz  and whatever features you need you can just drag-and-drop those components into the digital blocks.  So if you need 10 PWM's, fine.  If you need 4 capture pins, fine.  If you need 8 timers, fine.  If you need less or more memory, fine.  If you need I2C, SPI, etc - you just add the necessary components.  Your consideration is just how many blocks the chips gives you to work with, not what pre-determined functions the chip supports.

The other big thing about PSoC is the analog and processing you can do in the chip.  You can build logic and do filtering/conditioning with hardware and reconfigure it any time you like.  If you need to invert a signal, or buffer signals, or multiplex signals, you just add those blocks and do whatever you need.  And the blocks have API's to make them easy to use.  You're not dealing with registers directly unless you want to.

Another thing that is nice is being able to abstract the function from the pin.  So you can route your PWM to pin 1, or pin 10, or whatever pin you like.  If you make a layout error in your board, you can usually just reconfigure your pins to fix it.  You can also pass signals right through or put some logic or components like op-amps or comparators inline for fast processing without actually having to do any code at all.

Considering the power and pricing, I am not sure why most people wouldn't use PSoC's for most stuff. 

[andyturk]

Also, the feature set of these Cypress chips is just incredible. It is easy to get a micro that include LCD drivers, Capacitive touch sensors inputs, comparators, opamps, 0.1% voltage references, 20 bit delta sigma A/D's, D/As and lots more. You can chose what output goes to what pin, so board layout can become very, very easy.

It's like getting a breadboard full of goodies inside your mcu.

I'm looking to convert a prototype that's based on a STM32F103. It has four buffered 5V analog sensors, some voltage dividers to bring the signals down to 3.3V, buttons, a TFT display and a few LEDs for debugging. I'd never considered a PSoC before, but I read a few of jmole's posts and got intrigued.

I think I can ditch most of the external analog stuff in previous prototype (op amps and dividers). That'll make assembly easier and (hopefully) reduce the errors introduced by all the discrete components. The analog sensors are 5V and another point for the PSoC is that it can run the ADCs with the same VREF that the sensors are using (instead of having to divide the outputs down to 3V).

Another benefit for prototyping is the flexible pin mapping. The STM32 devices are pretty flexible, but you still have to be careful about which peripherals you use because the pin assignments overlap. It requires some fairly serious datasheet study to make sure you're not painting yourself into a corner by needing to run SPI_X and UART_Y simultaneously. Even worse, the the pinouts might work, but there can still be conflicts with overlapping DMA channels (DAMHIK  ).

The PSoC approach seems much more flexible. I've got one of jmole's freeSoC boards with an Arduino I/O pinout for "shields." It took a few hours to slice off the mcu and power supplies from the design I'd been working with and convert it to a freeSoC shield. I didn't pay much attention to the Cypress datasheets, since all the pins can be remapped in firmware.

Neat stuff!

[kripton2035]

I love the Psoc too. the only drawback I found is that you can at best have 4 opamps inside
and for some projects it can be too small.
for the other digital and analog parts it's always enought and you can choose the chip for the amount of block you need
but it would be great to have a psoc with 8 opamps and a few digital block.

[TerminalJack505]

These Cypress PSoc chips sound awesome.  Which one would you recommend for a hobbyist? 

That is, one that comes in a hobbyist-friendly package.  0.65mm pin pitch package or bigger, for example.

And free (as in beer), unencumbered development software.  I don't want to have to pay to use the vendor's hardware or do yearly re-licensing nonsense.

[blewisjr]

These Cypress PSoc chips sound awesome.  Which one would you recommend for a hobbyist? 

That is, one that comes in a hobbyist-friendly package.  0.65mm pin pitch package or bigger, for example.

And free (as in beer), unencumbered development software.  I don't want to have to pay to use the vendor's hardware or do yearly re-licensing nonsense.

I agree these things actually sound really awesome and might be really helpful in teaching me how circuits work without having to go out and buy crap loads of components and breadboard them.  Is PSoC Creator 100% free to use with no little commercial gripes hidden below?  I might grab on of jmole's freeSoC just to learn about various circuits and interfacing type stuff.  Then if by some chance I need more op amps or whatever just some perfboard and pins to make a shield for the top.  As a circuit noob this thing might teach me a bucket load.

[Hypernova]

I love the Psoc too. the only drawback I found is that you can at best have 4 opamps inside
and for some projects it can be too small.
for the other digital and analog parts it's always enought and you can choose the chip for the amount of block you need
but it would be great to have a psoc with 8 opamps and a few digital block.

Ha! once you get get some you will always ask for more. Next thing you know people would start demanding a million gates and multi amp outputs. Personally I think it's a shame they didn't release a M4 version with FPU, given all the analog stuff the core looks a bit gimped for DSP work. I find 67MHz quite low as their top speed considering most other M3's I've seen reach 100+.

As to the comment that it should be more popular, I think it suffered to due to the 1st PSoC5's (the non-LP version) having long erratas, promised CAN bus removed, looser analog specs etc, the years of development time didn't help. These were only fixed in the new LP series. Atmel Xmegas were in a similar boat, their ADC's had shortcomings that were only fixed in the later -U series with USB ports.

[TerminalJack505]

These Cypress PSoc chips sound awesome.  Which one would you recommend for a hobbyist? 

That is, one that comes in a hobbyist-friendly package.  0.65mm pin pitch package or bigger, for example.

And free (as in beer), unencumbered development software.  I don't want to have to pay to use the vendor's hardware or do yearly re-licensing nonsense.

After some research I think I have determined the answer to my question is: PSoC 4.  These are Cortex M0-based and (soon to be) available in TQFP and SSOP packages.

It looks like they haven't hit the streets just yet, though. 

It looks like they are going to be very nicely priced.  I'll have to check them out when they are available.