Migrate from PIC18F26K22 to PIC18F26Q10?

[Ttelmah]

If you look at the errata, it scans through the whole of memory.
It uses a count set to the memory size of the chip.
If any byte fails, you then have to stop, and wait for the chip to be powered
down/up again.
On a scale of 1 to 10 of a 'useable chip', this is about -30....
You could not use this in any application that requires even the slightest
degree of reliability. Has to be one of the worst faults I have ever seen
on a PIC.....

[Gabriel]

Well, at this point, I can't go back so FML.
I guess back in March/April I should have imagined it was too good to be true.
I had done my homework with the available information...sucks

My application is highly dependent on having those 5 UARTS.
GSM,WIFI,RS485,FTDI + 1 Free expansion port...

11k+ lines of code later... *Curses in multiple languages*
_________________
CCS PCM 5.078 & CCS PCH 5.093

[dluu13]

[Ttelmah]

Two questions:
Are you committed on the package?.
Do you have PCD?.

The older 67K40 has 5 UART's but is in a 64pin SMD package.
It's errata list is really minor.

Quite a lot of the DsPIC's have 6 UART's, but again usually only in larger
packages.

[Gabriel]

I do not have PCD and have 0 experience with 16bit MCUs. Can't imagine it's too difficult to switch though, cost of the compiler is not an issue.

Time is the enemy here.

Package change is not out of the question, i did intend to switch to SMD but after i had recouped my "Time Investment" on the current design.

I'm reading up on the 18F67K40 as per your suggestion.
Checking my stock of "cool boards i bought and never used", I see i have 3 of these boards with just that pic on it! https://www.mikroe.com/clicker-2-pic18fk

I'm going to finish my application on my current 18F47Q43 setup, and start redesigning my board for the 67K40...at least I'll have working code when the boards roll in and I'll get a chance to throw in that RTC i forgot and had to improvise with a timer. yeih!

This has been devastating.
_________________
CCS PCM 5.078 & CCS PCH 5.093

[AESPOSITO]

Q10 & Q43, (using 5093 & 5094)
any reported problems?
_________________
A Esposito, CEO
Avatar Engineering Corp.

[newguy]

Buy the PCD upgrade and don't look back. The only gotcha that might get you is that all variables are signed by default for PCD. You have to be very explicit with declarations - state whether they're signed or unsigned int8 int16, etc.

Just take care of the signed/unsigned issue and what you have written will "drop in" and compile just fine.

[Gabriel]

If i where to bump up to PCD ... what PIC with 5 uarts would you recommend? (Hold terrible errata inclusions please!)
If I'm already gonna have to redesign for the 18f67k40, might as well go all out and bump up to 16bit...

I have to admit i have not yet encountered the memory issue on the Q. Could it be power supply related? Maybe the fault is triggered by slow rails? And by some fluke my board is Just right?....bah... anyways.

I'm down for wicked 16bit pic recomendations... might as well do it.
_________________
CCS PCM 5.078 & CCS PCH 5.093

[newguy]

I haven't had a bad experience with any 16 bit PIC yet. But then again I always look at the errata first, then the data sheet (but only if the errata aren't too bad). I had a very quick look at the Microchip parametric search and there is a wide selection of PIC24FJ.... processors with 6 UARTs.

[dluu13]

There's the PIC24FJ128GA204 on the PIC Curiosity board provided with Microchip. It only provides 4 UARTS but it is a very convenient platform for playing with things and getting started on PIC24. It's not very expensive, it has a breakout header for the MCU, and a mikroBus header too which is very useful for quickly testing out various hardware.

[Ttelmah]

I'd suspect that the memory problem may only happen on some exotic
combination of power timing, or actual silicon source, that has not been yet
properly identified. They've obviously found some examples, but may not
have narrowed down what actually causes it....

On the PIC24's & DSPIC's, it can take a little time to get used to the simple
fact that the default 'integer', is a signed int16, and this is the default
alignment for memory. Ports etc., are also 16bit. The sheer performance
though is massively impressive. Remember it only takes two cycles of
the master oscillator to perform an operation, so even the chips that only
support 32MHz master clocks are giving you 16MIPS, while the ones that
run at clocks like 140MHz are giving 70MIPS. Some even go up over 100MIPS.
Also the basic instruction in most cases includes a hardware 16bit divide,
making the maths functions perform dozens of times faster than the best
PIC18.
One particularly 'nice' feature is PSV. With this a window can be created
in the RAM map, that points to the program memory, which allows things
like pointers to constants. A real improvement in convenience. It comes
at a 'cost' though, that stuff stored using this only uses 16bits from each
instruction word. However given the large ROM sizes that are available
this is not too bad.
Generally most of these chips do not have EEPROM. You can use a lump
of the program memory as a 'virtual EEPROM', but this costs code size
and performance. If you really need EEPROM, add an external chip.

[temtronic]

re: lots of UARTS...

Any chance you can use I2C<>UART or SPI<UART 'chips'.

I rememeber using DUARTS a zillion years ago to get extra UARTS. They even ame in a 'quadUART' configuration.
It might be an option that'd allow you to stick with your known GOOD PIC ( and programs...), 'just' add a daugther board to gain more UARTs. Reminds my of the good old PDP8 days where I have 9 or 11 serial interfaces.

Just trying to think of good, solid ways to get what you need.

[hmmpic]

I know this is a old thread, but it belong here i think.

Is there a way to know what "Revision ID" a chip have when ordering it?
PIC18F27Q10 have a "Revision ID" - "A4" & "B2", the "B2" is preferred according to it's Errata.

As I know it is not visible on the chip, only as a readout, when we have the chip running in a system... Then it is too late to know:-(

So how to know at order time what "Revision ID".

[newguy]

In my experience, I don't believe there is a way to specify a particular revision when ordering, unless (perhaps) you order direct from Microchip.

Years ago I had to track down an obsolete motor control IC, the Si9979. The caveat was that we knew that revision N742AB (don't ask me how I remember that) had a bug where the IC would "run away" (i.e. drive the motor 100%) no matter what the control input PWM actually was. An obsolete part vendor had stock, but of the N742AA. Against my recommendation, my employer at the time bought up the entire stock. Turns out that the 'AA revision had the same bug. Money wasted.

[Ttelmah]

You can order a revision from MicroChip, and from most of the major
suppliers. Mouser, and Digikey will both supply identified revisions if
asked.