Author Topic: Are CPLD's dead?  (Read 7593 times)

0 Members and 1 Guest are viewing this topic.

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Are CPLD's dead?
« on: April 04, 2018, 08:55:14 am »
So I have a design with a Mach4000 CPLD that I wanted to upgrade so started looking at the CoolRunners but it seems I'm 6 years late for that party.

Anyone know of some modern device families - Xilinx have buried their CPLD parts on the web site. I've not touched FPGA's yet but if I did it would have to be a flash device with no external components needed if such a thing exists. Basically those two points are what attracted me to CPLD's in general... that and the fact I'm only doing very simple logic.

The Mach 4000 (LC4256ZE) fits my current project design but I'm guessing in a year or so they will be gone so need something to migrate towards.

If I'm going to be forced to the FPGA parts then I think perhaps the ICE40's are an ok fit (low cost low pint counts). Perhaps I need someone to tell me to stop being a baby and just get on board the FPGA bandwagon?

[edit] reading more about the ICE40 they do have a one time programmable memory (NVCM), and an internal 48MHz oscillator that might fit the bill. Are the tools free?
« Last Edit: April 04, 2018, 09:11:14 am by David Chamberlain »
 

Offline Daixiwen

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: no
Re: Are CPLD's dead?
« Reply #1 on: April 04, 2018, 09:07:10 am »
Altera/Intel has the Max series, which are now basically small FPGAs with integrated flash, so from the user perspective they can be seen as CPLDs.
They usually require multiple power supplies (at least one for core, and one for I/O) but they also have some variants with integrated regulator that can work on a single supply.

Max V
Max 10
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #2 on: April 04, 2018, 09:18:58 am »
Thanks I'll check them out. I think I'm getting lost in FPGA land trying to translate 256 'macrocells' from my old part to something comparable in an FPGA device. Of course I just need to synthesize my design on every manufactures devices, reducing the part size until it fails but that sounds painful.

The Mach4000 (while it still is available) goes for about $13USD each.
 

Online mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 14497
  • Country: gb
    • Mike's Electric Stuff
Re: Are CPLD's dead?
« Reply #3 on: April 04, 2018, 10:49:08 am »
The Mach 4000 (LC4256ZE) fits my current project design but I'm guessing in a year or so they will be gone so need something to migrate towards.

Why do you think that ? have Lattice made any staement ? 
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline up8051

  • Frequent Contributor
  • **
  • Posts: 338
  • Country: pl
Re: Are CPLD's dead?
« Reply #4 on: April 04, 2018, 11:02:52 am »
Give us your requirements:
- power supply voltage
- I/O voltge
- pin counts
- max frequency
- power consumption
...

Regards
up8051
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #5 on: April 04, 2018, 11:03:55 am »
The only reason if a true CPLD can be better then an internal memory FPGA is boot speed. CPLD doesn't have the booting process. Their logic states are written in flash or EEPROM structures right in the logic cells. They don't have to load boot stream from memory.

FPGAs with internal memory still have to boot from that memory to SRAM which actually controls logic gates.

The exception is Actel ProASIC3 and Igloo and their nano versions. Those are true flash FPGAs, flash cells are placed right next to logic gates just like CPLDs.

iCE40 can boot in a few ms, so unless you need absolutely zero boot time, iCE40 will work.

Actually you are correct, I don't really need very fast boot time. I suppose I'm just looking to avoid an external EEPROM. My application is in a bit of an unwinnable position of preferring very low pin-out / small footprint (SPI + a handful of other IO) but needing large internal registers for counting
 

Offline agehall

  • Frequent Contributor
  • **
  • Posts: 390
  • Country: se
Re: Are CPLD's dead?
« Reply #6 on: April 04, 2018, 11:06:01 am »
CoolRunners are still available and I for one like them a lot.
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #7 on: April 04, 2018, 11:07:06 am »
The Mach 4000 (LC4256ZE) fits my current project design but I'm guessing in a year or so they will be gone so need something to migrate towards.

Why do you think that ? have Lattice made any staement ?

1) Because of the VIBE.
2) No Lattice have not made any particular statement.
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #8 on: April 04, 2018, 11:11:50 am »
Give us your requirements:
- power supply voltage
- I/O voltge
- pin counts
- max frequency
- power consumption
...

Regards
up8051

- power supply voltage = 3.3, 1.8.. what ever floats a boat.
- I/O voltage = as above.
- pin counts = Low, < 50 pin QFP or BGA
- max frequency = I'm counting rising edge clocks up to 100MHz, it's a frequency counter.
- power consumption = Don't care.
 

Online SiliconWizard

  • Super Contributor
  • ***
  • Posts: 17221
  • Country: fr
Re: Are CPLD's dead?
« Reply #9 on: April 04, 2018, 11:12:05 am »
Just for the record, Actel was bought by Microsemi a few years back, and Microsemi is now in the process of being bought by Microchip. Product line: https://www.microsemi.com/product-directory/1636-fpga-soc
Nice products, but I've always found their dev tools pretty awful. YMMV.

Lattice MachXO2 FPGAs can boot from internal Flash in 0.6 ms to 3.8 ms depending on the device size. Not too shaby.
The smallest MachXO2-256 probably fits the bill for replacing your Mach4000 and is ready in 0.6 ms.
 

Offline NivagSwerdna

  • Super Contributor
  • ***
  • Posts: 2531
  • Country: gb
Re: Are CPLD's dead?
« Reply #10 on: April 04, 2018, 11:29:33 am »
I'm toying with the idea about getting into CPLDs.... the prime reason being they are 5V tolerant... contemporary FPGAs seem less and less 5V tolerant.

[Either ATF1504AS-10AU44 or M4A5-64/32-10VNC48]
 

Offline up8051

  • Frequent Contributor
  • **
  • Posts: 338
  • Country: pl
Re: Are CPLD's dead?
« Reply #11 on: April 04, 2018, 11:36:49 am »
Lattice MACHXO are  much cheaper then MACH4000.
For example LCMXO256C costs about 3$.

Regards,
up8051
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3508
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Are CPLD's dead?
« Reply #12 on: April 04, 2018, 05:41:33 pm »
AFAIK most devices sold as CPLD now are really low-end FPGA with built-in Flash and maybe in-cell configuration. I have some experience with Intel MAX II series, which is for all intent and purpose that, with in-cell configuration Flash for the purpose of eliminating boot times.
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #13 on: April 09, 2018, 03:30:18 am »
Just an update.. After a lot more reading I've decided to play around with the MachXO2/3 parts.

Interestingly my Mach4000 design did not fit the MachXO2-256 so I'm going for the 640 at least, I was really working hard to get the most out of that 4000.

Then wanting a little more room I settled on the LCMXO3LF-1300E in a 36-WLCSP is going for under $4 in singles and the non-flash L even cheaper (when if ever it's in stock at Digi,Mouser). I've ordered the breakout board to mess about with.

So It's going to be fun to see if A) Get the PCB manufactured well enough B) reflow in my own oven. The chips are   0.4mm pitch with 0.25mm recommended landing pads leaving 0.15mm for clearance, just on the nose for my current preferred PCB manufacture.

Thanks for the suggestions.
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: Are CPLD's dead?
« Reply #14 on: April 09, 2018, 04:48:03 am »
I'm toying with the idea about getting into CPLDs.... the prime reason being they are 5V tolerant... contemporary FPGAs seem less and less 5V tolerant.

[Either ATF1504AS-10AU44 or M4A5-64/32-10VNC48]

I wouldn't make that assumption, not all CPLDs are 5V tolerant. I'm not aware of any current generation FPGAs that are, although in many cases you can get away with just using a resistor in line with the IO pin to protect it from a 5V signal.
« Last Edit: April 09, 2018, 06:02:02 pm by james_s »
 

Offline David ChamberlainTopic starter

  • Frequent Contributor
  • **
  • Posts: 251
Re: Are CPLD's dead?
« Reply #15 on: April 09, 2018, 06:49:25 am »
I'm toying with the idea about getting into CPLDs.... the prime reason being they are 5V tolerant... contemporary FPGAs seem less and less 5V tolerant.

[Either ATF1504AS-10AU44 or M4A5-64/32-10VNC48]

I wouldn't make that assumption, not all CPLDs are 5V tolerant. I'm not aware of any current generation FPGAs that are, although in many cases you can get away with just using a resistor in line with the IO pin to protect it from a 5V signal.

This is mentioned at least in the MachXO SysIO Usage Guide.
http://www.latticesemi.com/~/media/LatticeSemi/Documents/ApplicationNotes/MO/MachXOsysIOUsageGuide.PDF

Quote
All the I/Os on the top side of the MachXO1200 and MachXO2280 devices (Banks 0 and 1) have a clamp diode
that is used to clamp the voltage at the input to VCCIO. This clamp diode can be used along with an external resistor
to make an input 5V tolerant.

Also this thread from a couple years back might have some more info
https://www.eevblog.com/forum/microcontrollers/small-ish-fpga-or-cpld-5v-tolerant/

« Last Edit: April 09, 2018, 07:22:27 am by David Chamberlain »
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2946
  • Country: us
Re: Are CPLD's dead?
« Reply #16 on: April 09, 2018, 09:31:46 pm »

Anyone know of some modern device families - Xilinx have buried their CPLD parts on the web site. I've not touched FPGA's yet but if I did it would have to be a flash device with no external components needed if such a thing exists. Basically those two points are what attracted me to CPLD's in general... that and the fact I'm only doing very simple logic.
Xilinx is still selling their XC9500XL series, though not actively promoting it.  They are quite affordable.  if you need wide gates, like to recognize an address, they are great.  Classic CPLD internal architecture.

Xilinx also has the CoolRunner 2 family, these are much closer to FPGA fabric architecture with non-volatile configuration.  But, available in relatively small packages and also inexpensive.  Very low power, but they DO require TWO power supply voltages.

Jon
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2946
  • Country: us
Re: Are CPLD's dead?
« Reply #17 on: April 09, 2018, 09:40:12 pm »
Thanks I'll check them out. I think I'm getting lost in FPGA land trying to translate 256 'macrocells' from my old part to something comparable in an FPGA device. Of course I just need to synthesize my design on every manufactures devices, reducing the part size until it fails but that sounds painful.

The Mach4000 (while it still is available) goes for about $13USD each.
The Xilinx XC9536XL (smallest device, 36 FF and 36 36-input gates) goes for just over $1 in small quantity.  The XC2CR64 (I think that's the right part # for the smallest one) is about $2.

Jon
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2946
  • Country: us
Re: Are CPLD's dead?
« Reply #18 on: April 09, 2018, 09:45:46 pm »
The only reason if a true CPLD can be better then an internal memory FPGA is boot speed. CPLD doesn't have the booting process. Their logic states are written in flash or EEPROM structures right in the logic cells. They don't have to load boot stream from memory.

FPGAs with internal memory still have to boot from that memory to SRAM which actually controls logic gates.

The exception is Actel ProASIC3 and Igloo and their nano versions. Those are true flash FPGAs, flash cells are placed right next to logic gates just like CPLDs.

iCE40 can boot in a few ms, so unless you need absolutely zero boot time, iCE40 will work.
If you need a lot of registers, then CPLDs are not likely to work.  Xilinx makes some modest FPGAs with internal flash.  The small ones like XC3S50AN (under $15) has about 1500 FFs and macrocells.
Quote
Actually you are correct, I don't really need very fast boot time. I suppose I'm just looking to avoid an external EEPROM. My application is in a bit of an unwinnable position of preferring very low pin-out / small footprint (SPI + a handful of other IO) but needing large internal registers for counting
Boot time on the XC3S50AN is about 1/4 second.  If you have an LED hooked up to DONE/ you can just see it flash.

Jon
« Last Edit: April 09, 2018, 09:48:15 pm by jmelson »
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2946
  • Country: us
Re: Are CPLD's dead?
« Reply #19 on: April 09, 2018, 09:51:29 pm »
I'm toying with the idea about getting into CPLDs.... the prime reason being they are 5V tolerant... contemporary FPGAs seem less and less 5V tolerant.

[Either ATF1504AS-10AU44 or M4A5-64/32-10VNC48]
The only truly 5V tolerant part from Xilinx, I think, is the XC9500XL family.  Others can fake it with resistors.  All Xilinx parts have really weak ESD resistance, so any pin that goes to an external connector pin should be run through some voltage level translator.  Most of those are QUITE robust.

Jon
 
The following users thanked this post: NivagSwerdna

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1553
  • Country: wales
Re: Are CPLD's dead?
« Reply #20 on: April 09, 2018, 11:15:56 pm »
Haven't looked at using CPLDs for quite a while but when I've used them they've been a lot cheaper than FPGAs.
I would forget 5V tolerent parts, 5V is old school, translate down to 3V3. Xilinx 9500 series although 5V tolerent are more or less obsolete now and they run hot. If you need 5V tolerent then go for the 95xx series whilst they're still available.

Xilinx CoolRunner, based on a Phillips design, runs at 3V3 or less and is a good solution if you want low power. Some of the Altera MACH cplds don't have true CMOS output stages, they're totum pole so useless if you need a symmetrical output drive, read the data sheet. Can't comment on Lattice but they might fill the gap where Altera and Xilinx are probably less focussed on CPLDs these days.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf