In my other job I develop high-performance electric vehicle inverters, and I am itching to port over from a TI DSP chip to something that's actually nice to program....
What power and frequency do these run at ?
P2 should be able to interface to any number of isolated ADCs on the sense side.
If you needed dithered PWM, for finer control, a COG could be dedicated for that.
We have a 200 kW SiC inverter at 18kHz and a 2 MW IGBT inverter at 12 kHz, both at 600V.
I do have P1 assembly that outputs the commutation patterns and triggers external ADCs at just the right times in a dedicated cog. It needs around 10ns edge accuracy for good efficiency.
It turned out to not have sufficient math throughput so I gave up at the time.
We have a 200 kW SiC inverter at 18kHz and a 2 MW IGBT inverter at 12 kHz, both at 600V.
I do have P1 assembly that outputs the commutation patterns and triggers external ADCs at just the right times in a dedicated cog. It needs around 10ns edge accuracy for good efficiency.
It turned out to not have sufficient math throughput so I gave up at the time.
Wow, I had a feeling you might be talking about largish numbers ..!
How many bits are the ADC's ?
P2 should have 'sufficient math throughput' ? , and manage either external isolated ADCs or the internal ADC's may be 'good enough' for some channels.
I've recently used the P1 as a low-power EtherCAT slave with a LAN9252 at work and we've been pretty impressed with its flexibility, but it sounds like the P2 will completely blow it out of the water.
I'd be interested in getting up to speed on designing EtherCAT sensors. Any advise on difficulty of such an interface? Software and hardware. Also, if I can ask, what are you using for masters?
There is a driver here that can be adapted to the P1 pretty easily, though I did write a dedicated assembly SPI cog for better throughput. I believe we're seeing update rates close to 1kHz. It says Linux, but it fits in the Prop with a bit of trimming.
I won't say it's trivial, but most of the work is in understanding how EtherCAT works and optimizing a few sections of code for reasonably fast loop rates. https://github.com/OpenEtherCATsociety/SOES/blob/master/drivers/linux/lan9252/lan9252.c
We're running standard Linux on x86_64 with Preempt-RT patch for the master.
I'm still lurking, still developing with P1 and waiting for all the P2 tools etc to stabilise.
I don't seem to have as much time to experiment with things these days but still have a few plans I'd like to see come to fruition with Baggers that involve P2.
Hey guys, saw this thread pass by a while ago but had to startup laptop to reply.
I've been focusing on my music and learning how to use new gear I've acquired. Also trying to design a suitable case for the P2 Touchscreen prototype. A cardboard box just isn't going to cut it this time I already have some simple app ideas once I figure the last few details out. Haven't had much time to work on electronics lately but I'm getting excited for when I do have time because lots of ideas.
Oh yeah, I've been teaching the 9 year old some SPIN on the BOEbot. She's really excited about getting her own laptop for school, already asking if she can play with the Bot more.
It's quiet because we are all in shock at the realisation you are 'writing spin in 0x86 assembler'
If you want community support - you are better off to write in a language the community can understand and contribute to.
It's bad form to put out an 0x86 assembler tool and then expect someone else (Roy?) to convert it to c++ or something else others can understand before anything can happen.
Chip, do us all a favour, Stop right now on the x86 assembler, spend a few days ( what's a few days in 12 years ) to learn c++ or even better c# and put out at least the core tools ( PNUT, SPIN ) in a language that the community can actually contribute to.
I for one tried to get allocation addresses to support a debug and emulation tool I have developed for P1 and could not get what I needed from existing tools and could not readily modify OpenSIM to get what I needed either and so gave up and benched the entire project.
x86 assembly. Really?
The current path is almost complete using x86. I know it seems so retrograde, but that's where my momentum is. I could have had it running a few weeks ago, but have rethought a lot of the interpreter code which is in P2 assemlbly language. That is almost done and then I will go back to the x86 Spin2 compiler code and make the corresponding adjustments. After that, PNut will start compiling Spin2, not just pasm. I want to get it running before rewriting the complier in a high-level language, hopefully in Spin2, itself. A tiny bytecode interpreter could be written in C that would run it.
Chip, I know people give you a lot of Smile for using x86 assembler, but personally I think it's awesome. I was writing an operating system as a hobby project 13 years ago. I never finished it, but to this day x86 assembler is still my favorite programming language.
I've used the Propeller 1 for years and am really looking forward to the P2, I've been watching it's progress from the beginning and am super excited that the wait is almost over. I really wanted to pick up one of the P2-ES boards but unfortunately didn't have the funds available at the time.
It's quiet because we are all in shock at the realisation you are 'writing spin in 0x86 assembler'
If you want community support - you are better off to write in a language the community can understand and contribute to.
It's bad form to put out an 0x86 assembler tool and then expect someone else (Roy?) to convert it to c++ or something else others can understand before anything can happen.
Chip, do us all a favour, Stop right now on the x86 assembler, spend a few days ( what's a few days in 12 years ) to learn c++ or even better c# and put out at least the core tools ( PNUT, SPIN ) in a language that the community can actually contribute to.
I for one tried to get allocation addresses to support a debug and emulation tool I have developed for P1 and could not get what I needed from existing tools and could not readily modify OpenSIM to get what I needed either and so gave up and benched the entire project.
x86 assembly. Really?
The current path is almost complete using x86. I know it seems so retrograde, but that's where my momentum is. I could have had it running a few weeks ago, but have rethought a lot of the interpreter code which is in P2 assemlbly language. That is almost done and then I will go back to the x86 Spin2 compiler code and make the corresponding adjustments. After that, PNut will start compiling Spin2, not just pasm. I want to get it running before rewriting the complier in a high-level language, hopefully in Spin2, itself. A tiny bytecode interpreter could be written in C that would run it.
Chip, I know people give you a lot of Smile for using x86 assembler, but personally I think it's awesome. I was writing an operating system as a hobby project 13 years ago. I never finished it, but to this day x86 assembler is still my favorite programming language.
I've used the Propeller 1 for years and am really looking forward to the P2, I've been watching it's progress from the beginning and am super excited that the wait is almost over. I really wanted to pick up one of the P2-ES boards but unfortunately didn't have the funds available at the time.
KaylaMarie, thanks for the support! The hardest thing I face is feeling like what I'm working on is relevant, because so many aspects of my work are dated in terms of language, awareness, and so on. I know, though, that people need stuff that actually works if they are going to succeed in making anything that works beyond the next wave of "upgrades". I feel like a lot of modern approaches are akin to building on a sandy foundation, with "code" squirreled away in graphical menus and IDE's being 1GB installs. How does anything even get to be 1GB?. Maybe nobody knows. Simple is good. Hopefully, this next chip works as planned and we get you one.
KaylaMarie, thanks for the support! The hardest thing I face is feeling like what I'm working on is relevant, because so many aspects of my work are dated in terms of language, awareness, and so on. I know, though, that people need stuff that actually works if they are going to succeed in making anything that works beyond the next wave of "upgrades". I feel like a lot of modern approaches are akin to building on a sandy foundation, with "code" squirreled away in graphical menus and IDE's being 1GB installs. How does anything even get to be 1GB?. Maybe nobody knows. Simple is good. Hopefully, this next chip works as planned and we get you one.
Chip, just do whatever works. Let others follow with whatever is the current "fad"!
Well if you are writing something for yourself, you can use whatever tool you wish.
But if you expect that tool to be used by others, it needs to be in something people commonly write with, or else you are expecting someone else has to duplicate your work. x86 is not even a native instruction set anymore.
KaylaMarie, thanks for the support! The hardest thing I face is feeling like what I'm working on is relevant, because so many aspects of my work are dated in terms of language, awareness, and so on. I know, though, that people need stuff that actually works if they are going to succeed in making anything that works beyond the next wave of "upgrades". I feel like a lot of modern approaches are akin to building on a sandy foundation, with "code" squirreled away in graphical menus and IDE's being 1GB installs. How does anything even get to be 1GB?. Maybe nobody knows. Simple is good. Hopefully, this next chip works as planned and we get you one.
Chip, just do whatever works. Let others follow with whatever is the current "fad"!
Easily seconded. Proceed with full faith and confidence Chip. You won't miss. It will be fine. A lot of people want what you would create, and all of this is part of "your way" and it matters.
There will be lots of options for people over time.
KaylaMarie, thanks for the support! The hardest thing I face is feeling like what I'm working on is relevant, because so many aspects of my work are dated in terms of language, awareness, and so on. I know, though, that people need stuff that actually works if they are going to succeed in making anything that works beyond the next wave of "upgrades". I feel like a lot of modern approaches are akin to building on a sandy foundation, with "code" squirreled away in graphical menus and IDE's being 1GB installs. How does anything even get to be 1GB?. Maybe nobody knows. Simple is good. Hopefully, this next chip works as planned and we get you one.
Chip, just do whatever works. Let others follow with whatever is the current "fad"!
Easily seconded. Proceed with full faith and confidence Chip. You won't miss. It will be fine. A lot of people want what you would create, and all of this is part of "your way" and it matters.
There will be lots of options for people over time.
Thanks, Doug. I keep thinking the interpreter is done when I realize some cool improvement is possible. I'm at the point, again, where I can get back to the compiler.
Chip,
Is spin2 and the compiler in a form that we could start testing?
While I’m in the UK (in HK on the way - isn’t the internet wonderful) I would like to try some spin2 to get my P1 OS running on P2. Mostly it is in spin and a little PASM too. The PASM is mainly the SD access which is mostly done in the P2 ROM - just needs write support which is a simple command as all the underlying SPI R/W is already there
Then we have to get the 386 compiler working in P2ASM
Comments
We have a 200 kW SiC inverter at 18kHz and a 2 MW IGBT inverter at 12 kHz, both at 600V.
I do have P1 assembly that outputs the commutation patterns and triggers external ADCs at just the right times in a dedicated cog. It needs around 10ns edge accuracy for good efficiency.
It turned out to not have sufficient math throughput so I gave up at the time.
How many bits are the ADC's ?
P2 should have 'sufficient math throughput' ? , and manage either external isolated ADCs or the internal ADC's may be 'good enough' for some channels.
12-bit ADCs are sufficient for motor control, though more is always nice. P2 appears more than capable of handling the number crunching.
As far as EtherCAT, I'd suggest this guy
https://www.microchip.com/wwwproducts/en/LAN9252
There is a driver here that can be adapted to the P1 pretty easily, though I did write a dedicated assembly SPI cog for better throughput. I believe we're seeing update rates close to 1kHz. It says Linux, but it fits in the Prop with a bit of trimming.
I won't say it's trivial, but most of the work is in understanding how EtherCAT works and optimizing a few sections of code for reasonably fast loop rates.
https://github.com/OpenEtherCATsociety/SOES/blob/master/drivers/linux/lan9252/lan9252.c
We're running standard Linux on x86_64 with Preempt-RT patch for the master.
I've never patched my kernel before. Something new to learn. I suppose best not do it to my main workstation.
I don't seem to have as much time to experiment with things these days but still have a few plans I'd like to see come to fruition with Baggers that involve P2.
I've been focusing on my music and learning how to use new gear I've acquired. Also trying to design a suitable case for the P2 Touchscreen prototype. A cardboard box just isn't going to cut it this time
Oh yeah, I've been teaching the 9 year old some SPIN on the BOEbot. She's really excited about getting her own laptop for school, already asking if she can play with the Bot more.
Chip, I know people give you a lot of Smile for using x86 assembler, but personally I think it's awesome. I was writing an operating system as a hobby project 13 years ago. I never finished it, but to this day x86 assembler is still my favorite programming language.
I've used the Propeller 1 for years and am really looking forward to the P2, I've been watching it's progress from the beginning and am super excited that the wait is almost over. I really wanted to pick up one of the P2-ES boards but unfortunately didn't have the funds available at the time.
KaylaMarie, thanks for the support! The hardest thing I face is feeling like what I'm working on is relevant, because so many aspects of my work are dated in terms of language, awareness, and so on. I know, though, that people need stuff that actually works if they are going to succeed in making anything that works beyond the next wave of "upgrades". I feel like a lot of modern approaches are akin to building on a sandy foundation, with "code" squirreled away in graphical menus and IDE's being 1GB installs. How does anything even get to be 1GB?. Maybe nobody knows. Simple is good. Hopefully, this next chip works as planned and we get you one.
Chip, just do whatever works. Let others follow with whatever is the current "fad"!
Well if you are writing something for yourself, you can use whatever tool you wish.
But if you expect that tool to be used by others, it needs to be in something people commonly write with, or else you are expecting someone else has to duplicate your work. x86 is not even a native instruction set anymore.
Easily seconded. Proceed with full faith and confidence Chip. You won't miss. It will be fine. A lot of people want what you would create, and all of this is part of "your way" and it matters.
There will be lots of options for people over time.
No word from them, yet. Should hear something any day. They're committed to August 1st, but they are usually early.
Thanks, Doug. I keep thinking the interpreter is done when I realize some cool improvement is possible. I'm at the point, again, where I can get back to the compiler.
Is spin2 and the compiler in a form that we could start testing?
While I’m in the UK (in HK on the way - isn’t the internet wonderful) I would like to try some spin2 to get my P1 OS running on P2. Mostly it is in spin and a little PASM too. The PASM is mainly the SD access which is mostly done in the P2 ROM - just needs write support which is a simple command as all the underlying SPI R/W is already there
Then we have to get the 386 compiler working in P2ASM