EEVblog Electronics Community Forum
Electronics => FPGA => Topic started by: jayk on October 24, 2024, 02:46:25 pm
-
Looks like TI is looking to get into the low-end of the programmable logic market:
https://www.ti.com/logic-voltage-translation/configurable-programmable-logic/programmable-logic-devices/overview.html?keyMatch=tpld&tisearch=universal_search (https://www.ti.com/logic-voltage-translation/configurable-programmable-logic/programmable-logic-devices/overview.html?keyMatch=tpld&tisearch=universal_search)
I can't find details on programming these in the docs... maybe some low pin-count proprietary programming interface. Looks like these are OTP-only, which is a bit of a bummer (reprogrammable flash-based like Lattice would have been nice).
-
Yes OTP, but there's more. No idea exactly what they mean but this is in the datasheet:
System designers can create circuits and configure the macro-cells, I/O pins, and interconnections by temporarily emulating the non-volatile memory or by permanently programming the one-time programmable (OTP) through InterConnect Studio.
-
Seems to be more in line with Renesas GreenPAK devices. Hopefully TI ones have more coherent architecture.
-
With the TPLD1201 you may create an "intelligent 555", provided you would add three 5k resistors.. ;D
Btw., we miss today such small CPLDs like the XC9536/72 (fast and 5V). The Lattice ones are slow and 3.3V only.. This TI's are 5V, slow as well, and low LUT count (10 sounds too low to me).
-
With the TPLD1201 you may create an "intelligent 555", provided you would add three 5k resistors.. ;D
It's rather the other way around. Those mixed signal functions look interesting for power management applications and that's probably what they are meant for.
-
I would love to have small devices which can be configured for different logical functions, as they are often required for various auxiliary functions, and right now one is forced to stock a whole bunch of different SKUs which perform different logical functions. If such device would exist, it would allow stocking just a single SKU and configure it for required function on the go. And for volume production you can always order pre-programmed parts from manufacturer.
-
Basically yet another much too tiny OTP part you don't really need.
GAL is well estabilished, can be programmed with Tl866
Step up, XC9536/72, then EPM240/570/1270
Cyclone1/2/4 if not enough, or some old Spartan3 or Gowin
Everything for a few $, complete boards $8-30 or so.
Go cheap with garbage parts for hobby projects usually is a bad choice, go big or go home, no one cares about a few extra $$ with prototyping/hobby stuff.
-
All the things you mentions are digital. This device has a analog peripherals. And depending on the details of the design they may be very useful. It all depends on details.
OTP is not a huge problem, since devices are simple and you should be reasonably be able to make them configured for your application with minimal number of iterations.
This targets the same market as Renesas devices. And more options is always better.
-
Basically yet another much too tiny OTP part you don't really need.
GAL is well estabilished, can be programmed with Tl866
Step up, XC9536/72, then EPM240/570/1270
Cyclone1/2/4 if not enough, or some old Spartan3 or Gowin
That stuff is too ancient and too large. I want something like that in ~3x3 mm package, not ancient DIP, which will take as much space on a board as some modern FPGA/CPU...
-
OTP is not a huge problem, since devices are simple and you should be reasonably be able to make them configured for your application with minimal number of iterations.
Yep, and TI also promises software which would allow full simulation of part behavior. Historically TI has been pretty good in that department, so assuming simulator will be sufficiently true-to-life, it will be great. Especially for what I'm looking for - which is essentially a 3-6 pin device (+power pins) which can be programmed to perform some logical function. These would be very useful as there are many places where one would need some kind of logical gate/LUT, but placing CPLD would be overkill.
-
Basically yet another much too tiny OTP part you don't really need.
GAL is well estabilished, can be programmed with Tl866
Nobody uses those obsolete parts in modern designs. There's really a gap in the market between simple logic ICs and too expensive, too large FPGA/CPLD.
-
With the TPLD1201 you may create an "intelligent 555", provided you would add three 5k resistors.. ;D
It's rather the other way around. Those mixed signal functions look interesting for power management applications and that's probably what they are meant for.
I see several applications for this. Pushbutton controllers, simple clock generators, sequencers, writing configuration registers for different chips. It also has analogue functions, very simple ADC with a few levels. The datasheet doesn't fully describe what can be done with this, the software is much more interesting.
-
Basically yet another much too tiny OTP part you don't really need.
I think you mean "I don't need". If others didn't need them TI wouldn't have started the project. They don't invest on a whim. The functionality AND the packaging its presented in will have been carefully worked out with potential volume customers.
-
5V supply - that's a pleasant surprise these days
-
Skimming through the data I'm struggling to get excited - too simple & pretty slow ( 8MHz max ext clock). Internal osc a bit rough at +/-5%.
A PIC16F15xxx with its CLCs isn't far off in terms of programmable logic functionality.
-
Love the mixed signal-ness of it, there's plenty of applications for these if the price is right.
-
I've requested access to their config tool - curious to see what it's capable of.
-
Skimming through the data I'm struggling to get excited - too simple & pretty slow ( 8MHz max ext clock). Internal osc a bit rough at +/-5%.
A PIC16F15xxx with its CLCs isn't far off in terms of programmable logic functionality.
And there's the new ones with tiny PLDs
https://ww1.microchip.com/downloads/aemDocuments/documents/MCU08/ProductDocuments/DataSheets/PIC16F13145-Family-Microcontroller-Data-Sheet-DS40002519.pdf
You don't even need the MCC bullshit to use it (that IS a pleasant surprise), they are providing the design tool as standalone (albeit in a webpage... just give me the damn JAR or PY or whatever you're using, god.) with means to import/export a design, and outputting an array with the bytecode that you shove into the appropriate registers
-
Yep. Obviously, there's also quite a bit you can do with the RP2040 (and now RP2350) PIO.
-
Got access to the tool - and it looks like TPLD801 is exactly what I'm looking for. It can implement up to 10 4-input LUTs, 4 D flip flops, 4 counters and some other things. If they will actually be $0.19@1ku as TI currently says, that's going to be a great deal.
TPLD1202 is more advanced, it's got a lot more available blocks, among them are frequency detectors, multi-channel analog comparators, PWM, edge detectors, and even I2C and SPI receiver interfaces.
-
With the TPLD1201 you may create an "intelligent 555", provided you would add three 5k resistors.. ;D
Btw., we miss today such small CPLDs like the XC9536/72 (fast and 5V). The Lattice ones are slow and 3.3V only.. This TI's are 5V, slow as well, and low LUT count (10 sounds too low to me).
Infineon/Cypress PSoC 3/5 are macrocell-based, up to 192 MCs, can be powered from 0.5V (solar cell) to 5.5V with a single inductor for integrated boost converter. The UDB architecture of PSoC 3/5 originated from Quantum 38K / Delta 39K (can be inferred from its characteristic "carry" product term), ultimately dated back to Cypress MAX340, a.k.a. Altera MAX 5000 series. Also, logic compiler PSoC Creator "Warp" has existed after 1996, at least.
-
Basically yet another much too tiny OTP part you don't really need.
..If others didn't need them TI wouldn't have started the project. They don't invest on a whim. The functionality AND the packaging its presented in will have been carefully worked out with potential volume customers.
Not always, imho.. This parts are tailored either for a specific customer and thrown in the general public afterwards, or simply knitted with hot needles.. Frankly I doubt people here on eevblog would ever approve that product for taping after a vote..
-
Considering today's tiny chip geometries, I'm surprised it's worth doing chips with such low logic count - AIUI the die size is limited by the bond pads, so within that constraint, logic in teh middle comes almost for free ( apart from testing)
-
I would love to have small devices which can be configured for different logical functions, as they are often required for various auxiliary functions, and right now one is forced to stock a whole bunch of different SKUs which perform different logical functions. If such device would exist, it would allow stocking just a single SKU and configure it for required function on the go. And for volume production you can always order pre-programmed parts from manufacturer.
I have been hearing this for almost 30 years already. 'TTL logic' will dissapear :-DD . But the reason to stick with discrete logic is the vast amount of second sources. On top of that TI has a poor track record of keeping up with demand in times of crisis. TI's parts where among the hardest to procure during the credit crunch and Covid pandemic.
-
I would love to have small devices which can be configured for different logical functions, as they are often required for various auxiliary functions, and right now one is forced to stock a whole bunch of different SKUs which perform different logical functions. If such device would exist, it would allow stocking just a single SKU and configure it for required function on the go. And for volume production you can always order pre-programmed parts from manufacturer.
I have been hearing this for almost 30 years already. 'TTL logic' will dissapear :-DD . But the reason to stick with discrete logic is the vast amount of second sources. On top of that TI has a poor track record of keeping up with demand in times of crisis. TI's parts where among the hardest to procure during the credit crunch and Covid pandemic.
They got scalped due to the TI store. We could always buy enough parts, but they came from China from not authorized distributors you never heard of. And 10x the price. They wouldn't have this issue if they would have stayed with arrow and avnet.
-
Considering today's tiny chip geometries, I'm surprised it's worth doing chips with such low logic count - AIUI the die size is limited by the bond pads, so within that constraint, logic in teh middle comes almost for free ( apart from testing)
You noticed yourself that these things are 5V tolerant, so they aren't in a fine geometry. The I/O ring can be less of a limitation than it used to be. A staggered I/O ring is run of the mill these days, and can really push up the I/O count on a small die.
-
Yep, and TI also promises software which would allow full simulation of part behavior. Historically TI has been pretty good in that department, so assuming simulator will be sufficiently true-to-life, it will be great. Especially for what I'm looking for - which is essentially a 3-6 pin device (+power pins) which can be programmed to perform some logical function. These would be very useful as there are many places where one would need some kind of logical gate/LUT, but placing CPLD would be overkill.
A word of caution. TI is big and has different divisions doing different things, so such expectations may not be met if their PLD team is completely new.
In the past, with TI, I have learned that if you do not use TI devices in the exact way they describe or the way their engineers have intended things to be used, even though the spec of their component says it can be done, you may be debugging in the dark until you decode the occasional mis-documentation about their core functionality.
-
In the past, with TI, I have learned that if you do not use TI devices in the exact way they describe or the way their engineers have intended things to be used, even though the spec of their component says it can be done, you may be debugging in the dark until you decode the occasional mis-documentation about their core functionality.
I worked with their parts extensively before COVID (when they have became unobtanium), and I don't remember ever having any issues with their parts not working like they should. Infact I know about a few cases when they shouldn't work according to spec, but they actually do in practice.
-
I have been hearing this for almost 30 years already. 'TTL logic' will dissapear :-DD
I want them to solve my specific problem, and they do just that according to what I see so far.
On top of that TI has a poor track record of keeping up with demand in times of crisis. TI's parts where among the hardest to procure during the credit crunch and Covid pandemic.
That problem can be easily solved by maintaining a sufficient stock of those parts. And given how cheap they are claimed to be, I see no problem in stocking enough of those parts for several years worth of my needs.
-
Meh. Looks like GreenPaks, only worse.
GreenPaks rule.
If one is about to bring competition, after all this time one would expect to see something original and better.
-
Nonsense. These are in differenet worlds.
Try making somethign that ALSO workd as a voltage shifter from anywhere from 1.8V to 5.5V
.
GreenPaks can utilize dual power supply and each one has its I/O ad bank.
GreenPaks also have plenty of analog functions. Does your GAL have enalog comparator or MOSFET switch ?
Can your GAL be either programmed or just configured (== setting the circuits just by wrtitng SRAM configuration registers, without burning on-board PROM) ?
Can you reconfigure it on the fly while working through I2C interface ?
-
Is there a way to write some source code to configure these chips, instead of using their GUI tool though?
-
GreenPaks
Is made by Renesas, therefore a lot of us is not going to touch it with a 10 feet pole.
I would tell you why, but first let's sign an NDA, and you have to promise a million dollar in business revenue with me.
-
Is there a way to write some source code to configure these chips, instead of using their GUI tool though?
As far as I can tell, no.
-
Meh. Looks like GreenPaks, only worse.
GreenPaks rule.
If one is about to bring competition, after all this time one would expect to see something original and better.
Interesting chips. I've never seen them before. From my observation, only two versions are programmable in-system via I2C. The rest (the interesting ones) seem to only be programmable by Renesas. The documentation says you must send an email and wait 6 weeks for samples :palm:. Strange design choice. I would use them if they were easily programmable. I wonder how TI parts are programmed. If it's in-system, preferably 1-2 pins with a cheap dongle, then I'm in.
That's odd. Mouser has a lot of stock SLG46140V, but if they can't be programmed, why bother? Am I missing something?
-
The rest (the interesting ones) seem to only be programmable by Renesas.
Not exactly:
The SLG46140 is a user programmable device with One-Time-Programmable (OTP) memory elements that are able to construct
combinatorial logic elements. Three of the I/O Pins provide a connection for the bit patterns into the OTP on board memory. A
programming development kit allows the user the ability to create initial devices. Once the design is finalized, the programming
code (.gpx file) is forwarded to Renesas Electronics Corporation to integrate into a production process.
So you can program it (not 'in system' thought) by special programmer from Renesas
And programmability by Renesas can be considered as additional service, not as mandatory process.
-
So you can program it (not 'in system' thought) by special programmer from Renesas
And programmability by Renesas can be considered as additional service, not as mandatory process.
Yes, I just realized that. There is a kit called "GreenPAK Advanced Development Platform" SLG4DVKADV or SLG4DVKINTRO to program them. Edit: There is also a much cheaper SLG4DVKLITE board. There are also adapters with sockets for the chips or breakout boards with goldpins. Now it makes sense.
It looks to be the same on TI parts. There is a USB "programmer" TPLD-PROGRAM and TPLD EVM board with a socket for the device itself. Programming requires 8V.