• AVR Freaks

Helpful ReplyHot!Is MCHP still developing the PIC32?

Page: << < ..6789 > Showing page 7 of 9
Author
DominusT
Super Member
  • Total Posts : 1424
  • Reward points : 0
  • Joined: 2005/07/22 08:31:18
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 09:25:58 (permalink)
0
Stefiff
DominusT Change and Trace Width/ Spacing to 4 mil.




Attached Image(s)

Stefiff
Senior Member
  • Total Posts : 99
  • Reward points : 0
  • Joined: 2012/07/15 15:26:29
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 09:44:13 (permalink)
0
My mistake, I did not explain correctly. 6mil / 0.4mm are usually good for a dual layer board. For 4+ layers should be reduced to 3 / 4mil - 0.25/0.3mm under BGA.
wdy
Starting Member
  • Total Posts : 34
  • Reward points : 0
  • Joined: 2016/10/05 06:32:30
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 11:13:28 (permalink)
0
hevatsOn the main topic of PIC32 being developed further, with our little insight & tiny insider knowledge, the MIPS based chips have been switched off to fade out slowly.

That'd be sad news.
 
hevatsNeither there is any real progress on RISC V.

That, too.
 
hevats.. only ARM in the near future.

And that as well.
NorthGuy
Super Member
  • Total Posts : 6230
  • Reward points : 0
  • Joined: 2014/02/23 14:23:23
  • Location: Northern Canada
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 12:23:46 (permalink)
4.5 (2)
Stefiff
Something for an example that is used for prototypes. - ALLPCB site.
Choose a size, for example 100x100mm. For example 10 pieces. Select a dual-layer board, and the standard options below. See what the price is.
Then select 4 (or more) layers. Choose the thickness of the track - 4mil. Hole - 0.25/0.3mm. You can try other options. See the price.
Do you feel the difference in price only on the blank board?



Now change the quantity to 1000 and see the price difference getting smaller.
 
Stefiff
My mistake, I did not explain correctly. 6mil / 0.4mm are usually good for a dual layer board. For 4+ layers should be reduced to 3 / 4mil - 0.25/0.3mm under BGA.

 
I fully routed PIC32MZ 169-pin 0.8 mm pitch BGA on 4-layer board with 5 mil traces and 0.25 mm holes. I left the ground plane completely intact, and I had only couple signal traces in the power layer. But, this would take much more time if I tried to squeeze everything on 2 layers. And it would take even longer if I used TQFP.
 
With 1 mm pitch BGA (very common in Xilinx FPGAs), you can make the whole design with 6 mil traces and 0.3 mm drills.
Stefiff
Senior Member
  • Total Posts : 99
  • Reward points : 0
  • Joined: 2012/07/15 15:26:29
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 13:20:58 (permalink) ☄ Helpfulby DominusT 2020/07/19 15:42:16
5 (1)
However, we comment Microchip ARM CPUs. We are talking about processors with many pins 250+. At least 6 layers and it depends. There are no 169 pins, and they are an exception. Take a look at the SIP versions for pins. All are 0.8 mm pitch.
I made my own board with PIC32MZ DAS 176pins, two-layer, 7 mil traces and 0.6mm holes. However, this is the exception rather than the rule.
 
As for 1000 pieces, well, good luck if you sell that much. For a month or two, for example. This is usually how new orders are made here.
I live in Europe, and many pieces are made somewhere locally. China is used only for prototypes.
NorthGuy
Super Member
  • Total Posts : 6230
  • Reward points : 0
  • Joined: 2014/02/23 14:23:23
  • Location: Northern Canada
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/19 17:16:02 (permalink)
0
Stefiff
We are talking about processors with many pins 250+. At least 6 layers and it depends.

 
The more pins you have the more layers you will need. Of course you need to fan out only the pins you actually use. It's easier to route 150 pins from 256-pin BGA than from 169-pin BGA, isn't it?
 
If you need 250+ pins, TQFP wouldn't solve the problem, but rather would make things worse. It would probably be 0.4mm pitch ... which would make it 30 x 30 mm.
 
<edit>I searched DigiKey for MCUs looking for "ATSAM", and the highest number of pins I found was 144 (except two obsolete out-of-stock 176-pin LQFP parts).
 
There are also ATSAM MPUs in 500-600 MHz range, but these would require DDR memory and such, so you don't just connect things with traces - you need to worry about SI and all that. This would be very hard on a 2-layer board, if at all possible. And they do need lots of pins - just 32-bit wide DDR would take 70+ pins.
 
So, what parts are you talking about?
 
post edited by NorthGuy - 2020/07/19 17:36:20
Stefiff
Senior Member
  • Total Posts : 99
  • Reward points : 0
  • Joined: 2012/07/15 15:26:29
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/20 01:39:42 (permalink)
5 (1)
I'm talking about processors at higher frequencies 400-600MHz.
 
The idea is to have processors with fewer pins.
 
When the chip has 250+ pins, it is clear that the only option is BGA.
 
I will try to explain it in another way.
/I'm talking about 64-144 pins everywhere/
 
1/ 64 pins.
PIC32MZ 2M Flash, 500k RAM. Very good processor.
Exactly the same processor, but to be at a higher frequency, let it be ARM.
I.e. the first option. A little RAM, but you need speed.
 
2/ Need more RAM.
Connecting to the DDR (any type) itself is quite complicated, and takes a lot of pins that are blocked for it.
Here comes the benefit of SIP, very good idea.
We have enough RAM, up to 256MB, which is already connected.
It remains to connect some flash / SQI, SD card/. Again, in most cases, we don't need many legs. And here it can easily be with 64 pins.
The only exception is a parallel NAND flash.
 
3/ More and more RAM.
250+ pins and BGA.
There is no other option here.
 
====
Here is an example for me.
PIC32MZ DAS 176pins, 32MB RAM.
What turns out in practice after testing each module in it.
*SD card. The processor is much slower than the card.
*TFT at 1024-600 is at the limit of its capabilities, there are no layers. Not enough speed.
*32MB is not enough for me, I need more.
*The processor is so slow that even sound can't decode without being 85% + busy for that. That is, it cannot be used. I tested on many codecs /MP3, OPUS, SPEEX/, for a decent quality 16 / 32kHz output.
*External WIZNET is used for LAN. Unloads the processor and easily reaches 1MB/s+ transfer, with virtually no use of the processor.
 
An SIP version with 128MB RAM and 500-600MHz would do a great job for me. It doesn't have to be with many pins, 144 are enough.
 
Especially for the number of pins, see the microchip website. At the top left, select the different options.
/SAM9, SAMA5, SIP/
https://www.microchip.com/ParamChartSearch/Chart.aspx?branchID=30077
hevats
Junior Member
  • Total Posts : 57
  • Reward points : 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/20 04:30:18 (permalink)
0
This may or may not be helpful as every project has its own design constrains, however based on the example you described we would go for embedded linux. Something like a beaglebone black (or even a RPi compute) with whatever extra makes the whole thing work. In our experience such a solution has multiple benefits, be it compilers, ease of updates, standardized libraries (and a whole lot of those for everything including SSL), maintenance etc. etc. Especially for anything that connects to the internet having linux would help a lot. Beaglebone is open hardware, one can redesign the board if they so wish (I believe that would only be suitable for large quantities). Like I said it may not suite all use cases, and would definitely need some additional hardware in most cases.
 
friesen
I think manufacturers make what people buy.

I couldn't agree more. For 1 and 2 above the most cost effective option probably would be a ST H7 series. The requirements you have mentioned would most likely make you a minority in the customer pool for any vendor. One possible way out might be to have standardized PCB modules that can be surface mounted or plugged to a base board (something like your own SIPs). This would not only help with production and testing, it would also make it possible to reuse a lot of code. Initially one can only make the PCBs for such modules and keep inventory, later depending on the project requirements you can populate the exact components required (e.g. skip the RAM for a particular design). Thats what we have been doing, it has made our lives a bit easier.
 
friesen
Super Member
  • Total Posts : 2149
  • Reward points : 0
  • Joined: 2008/05/08 05:23:35
  • Location: Indiana, USA
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/20 09:33:12 (permalink)
5 (1)
Embedded linux is such a beast.  What does it take, one person to deal with yocto or its equivalent?

Erik Friesen
NorthGuy
Super Member
  • Total Posts : 6230
  • Reward points : 0
  • Joined: 2014/02/23 14:23:23
  • Location: Northern Canada
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/20 09:49:15 (permalink)
3 (1)
Stefiff
An SIP version with 128MB RAM and 500-600MHz would do a great job for me. It doesn't have to be with many pins, 144 are enough.

 
Get the one with more pins and only use 144. Why is that a problem?
 
Stefiff
Senior Member
  • Total Posts : 99
  • Reward points : 0
  • Joined: 2012/07/15 15:26:29
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/20 10:03:34 (permalink) ☄ Helpfulby DominusT 2020/07/20 12:33:26
5 (2)
BGA and the multilayer board. For these are my last few posts here. And without Linux of course. All descriptions of Microchip point there. Smile: Smile
post edited by Stefiff - 2020/07/20 10:17:06
hevats
Junior Member
  • Total Posts : 57
  • Reward points : 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/21 02:26:44 (permalink)
4 (1)
friesen
Embedded linux is such a beast.  What does it take, one person to deal with yocto or its equivalent?

Its not that bad, I believe most people on this post would get the hang of it within a day. Anyone who has a bit of command line experience would find it fairly trivial.
 
However I would advice beginners against yocto (or something like building your own kernel from source). Its far easier to start with getting a board that can run some form of mainline linux e.g. debian, most of the boards do. After that its just an exercise in reading the docs and writing in your favorite language. The advantage for someone from embedded world is that most of us have strong electronics knowledge and strong c / c++ abilities (I haven't ventured too much in assembly, happy with c). Compare this to others who for example write in python and hardly know anything about MCUs. Our code runs fabulously fast, and the boards resources seem enormous even for complex projects. Another killer weapon for us is hardware customization, which is almost always required.
 
To give you an example the beaglebone black (rev. c AM3359 based) has 2 CAN modules. However the transceivers are not implemented. Tons of people actually buy the so-called "CAN bus hat" for their projects. I believe it is because they come from the PC side trying to make electronics work. For people like us, going in the opposite direction, its easy to hook up 8 pin transceivers ourselves.
 
So to answer your question "how tough is it for one person to deal with embedded linux?" I would say my money is on the embedded engineer trying linux rather than a software engineer trying to solder a TQFP.
 
The beauty of what may be called a mixed solution, linux board + custom hardware, comes from faster POC, faster time to market, extreme ease of development, lots of code reuse, less reinvention of the wheel etc. (some people do like to reinvent the wheel, which is ok once in a while)
 
I can tell you one project that we did recently. It had a lot of sensors which where to talk to a cloud app, and a few actions going back from the cloud to the device. The POC was done in a day by just one guy. Later we had to add a speaker, a display, a webserver, SSH access, encrypted remote firmware upgrades (actually just device side application code, not the OS itself) and a few more functions. That took us another couple of weeks. I remember a PIC32MX based solution we did few years ago, to get the ethernet to work properly took more time than that.
 
I do not know of any large company where their devices connect to the internet not using linux (or VxWorks or QNX, but mostly linux) I would strongly recommend giving embedded linux a try. The cost of experimentation is not very high, you just need to buy a board, usually cheaper than "our kind" of dev boards. GCC is free and so is almost everything else that you'll need to get started. I can help with whatever little we have learned so far.
friesen
Super Member
  • Total Posts : 2149
  • Reward points : 0
  • Joined: 2008/05/08 05:23:35
  • Location: Indiana, USA
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/21 02:38:55 (permalink)
4 (1)
About Linux:
1. How to update firmware?
2. How to make the thing reliably boot from SD card post power loss.

In my experience this pretty much precludes off the shelf distribution. Also, one can't really rely on BBB hardware to be available or the same.

I did one project using the ts4900 from technologic, but ended up moving it to the pic32mz dah after some years.

Erik Friesen
hevats
Junior Member
  • Total Posts : 57
  • Reward points : 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/21 03:02:53 (permalink)
0
friesen
About Linux:
1. How to update firmware?

In most cases "firmware" would be your application which is just an executable stored somewhere, made to run at startup by a startup script. So updating firmware would simply mean replacing this file with a new one. There could be multiple ways to do this, SD cards, USB drives, serial connection... over the network is the easiest if you have network access. The update to the kernel is possible as well using some of the above methods, we haven't encountered a problem where we had to do this yet. 
 
On the point of firmware updates one very interesting difference is that you may have multiple executable running different hardware functions. E.g. a standalone application that takes user input & does something and another that logs & posts to cloud. You could update one of these without affecting the other, without any downtime! 
 
friesen
2. How to make the thing reliably boot from SD card post power loss.

We achieve this by running everything in RAM. Actually we do not use SD cards as well, we found them to be quite unreliable. For us everything is flashed to emmc, then loaded to RAM at boot. Its pretty much the same way a home router works. In case of power loss you only lose RAM variables, which is similar in non linux solutions. Critical runtime data can be saved to battery backed custom hardware if so desired. Even further, the whole solution can have a small battery which would enable "proper shutdown"... we are still waiting for orders from some space exploration company to test this one :P
 
post edited by hevats - 2020/07/21 03:06:35
jdeguire
Super Member
  • Total Posts : 583
  • Reward points : 0
  • Joined: 2012/01/13 07:48:44
  • Location: United States
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/21 08:45:21 (permalink)
0
I wonder if something like IncludeOS would be worth looking into for the Cortex-A5 devices.  I don't know how far they are on an Arm port and the GitHub repo's last update was six months ago, so maybe it's abandoned anyway.
 
https://www.includeos.org/
NorthGuy
Super Member
  • Total Posts : 6230
  • Reward points : 0
  • Joined: 2014/02/23 14:23:23
  • Location: Northern Canada
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/21 18:13:19 (permalink)
5 (3)
Stefiff
BGA and the multilayer board. For these are my last few posts here.



You keep repeating that like this will make it true. If you have 0.8mm pitch 256-pin BGA, you can fan out the outer two layers on a single layer and you'll get 116 pins, using exactly the same layout as if you used 0.4mm pitch 116-pin TQFP. In addition to that, you can fan out 84 pins on the second layer. That's 200 pins you can use on a 2-layer board. This is way more than you said you needed.  Moreover, it takes less space than 200-pin TQFP would. That's the reality. 
post edited by NorthGuy - 2020/07/21 18:16:25
EngSam
New Member
  • Total Posts : 18
  • Reward points : 0
  • Joined: 2019/04/17 09:28:54
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/22 06:19:36 (permalink)
0
my 2 Cents on BGA:
leave the routing and layer count and " via in pad" aside since i donot think a processor running at 200mhz+ could manage EMC in 2 layer ( See henry ott book: EM compatibility ) . However I also hate BGAs since (to my knowledge) are harder to solder reliably. Which if not done properly could lead to pre-mature failures due to solder joint cracks.  
 
questions on Embedded Linux:
Do you use beagle bone or RSpi in your products directly ?
I see these board similar to arduino in terms using them in an industrial product. I have heard how unreliable they are especially running the code from SD card and they suffer from Random REBOOT ing etc.... maybe if they make a more rugged version with higher temperature range and compliance.
 
 
post edited by EngSam - 2020/07/22 06:20:57
Stefiff
Senior Member
  • Total Posts : 99
  • Reward points : 0
  • Joined: 2012/07/15 15:26:29
  • Location: 0
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/22 12:33:17 (permalink)
5 (1)
When the first SOM1 appeared, there was a presentation of Microchip, along with other products. It had just gone on sale, and I had set out to buy it with the kit for experimenting. Linux turned me down. And I was very impressed by the words of the man from Microchip. He said this in front of the whole hall, literally the following.
We know that it is difficult to use BGA, and that is why we try to make your work easier. It is not just to draw and have the circuit board made. We have personally reworked this board several times. We had it made in several different companies. Just because they will make your board does not mean that it will work.
We had to repeat the whole process several times until we got a working board that met all the requirements. The easiest way is to give you the files and you can order it somewhere in China. But that doesn't mean it will work. Not everyone can make a working board for you, even in China.
 
Personally, I don't think I can compare myself to the resources that Microchip has, so I can count on faster success if I construct it.
That's why I have no desire to make a board with XXX+ pins BGA, running at 500MHz.
friesen
Super Member
  • Total Posts : 2149
  • Reward points : 0
  • Joined: 2008/05/08 05:23:35
  • Location: Indiana, USA
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/22 12:43:07 (permalink)
0
I have some doubts that the ATSAM SOM is designed the way it should be, but rather cost optimized.  Look at the way the bypass caps connect to the core, overall the SOM hardly follows their own design recommendations.

Erik Friesen
dlindbergh
Super Member
  • Total Posts : 483
  • Reward points : 0
  • Joined: 2006/09/11 18:17:11
  • Location: Boston USA
  • Status: offline
Re: Is MCHP still developing the PIC32? 2020/07/22 12:48:03 (permalink)
4.5 (2)
This has become quite a thread.
 
I wonder if we're going to see any new PIC32 parts?  ;-)
Page: << < ..6789 > Showing page 7 of 9
Jump to:
© 2020 APG vNext Commercial Version 4.5