SERIAL BUG or PIC Problem???????????????????

[cfernandez]

Hi,

I have a problem with my project in 18F4620 20Mhz, I need to obtain data from UART (C6,C7) to 115k and the fgetc NOT WORK. I try with FORCE_SW and this clausule work in 57600 but not in 115k. I remove the directive #ZERO_RAM and fgetc work in 115k without clausule FORCE_SW.

The other problem with solution is that I can not comunicate with my serial device (C1,C2) to 115k, the solution is put the clausule SAMPLE_EARLY in the #USE definition. This is a bug from compiler or a bug from Microchip 4620?

Somebody know which is the problem with ZERO_RAM clausule.

I hope that this information help to you in found the solution.

Thank you very much!

Best Regards,

[cfernandez]

CONTINUE....

I send a simple program for test the fgetc

This LST not work fgetc (BAD)

[ckielstra]

Some small background info on RS232:
The sample clock is only allowed to have an frequency error of about 3%. A data byte consists of 10 bits (8 data bits, a start bit and a stop bit), 10 times 3% error = 30% total error. With bad quality lines the error margin will drop to 2% or below. Maxim has a good application note on this, too bad I can't find the link now.

This 3% error margin is combined for both the sender and transmitter!

In your 2nd example code with the hardware UART the baudrate is set to 20MHz/4( 0x2A + 1) = 116.279 baud, a positive error of almost 1%. If your transmitter has an error of -2% or more you will receive garbage.
Many PCs are known to have large frequency offsets....

Do you have a scope? If yes, you can check the timing of the other device.
Or just quick, try the effect of reducing the baudrate in your 2nd example to 113k5.

A good thing with the hardware UART is to add the 'ERRORS' directive which will clear the error bits in the UART for you (otherwise the UART will stop receiving).

Edit: Removed an invalid remark about software UARTs.

[cfernandez]

Thank you very much!! for explain me this. But the problem not is the baud rate, is the ZERO_RAM.

I send this to CCS support and they tell me:

CCS Support wrote:
>
> This message is a reply to CCS e-mail id: 5B2624
>
> There is a problem with #ZERO_RAM on this part. Do not use it until the next release.

The next release is the Friday.

What you do think?

Best Regards,

[ckielstra]

Comparing your two examples I don't see how the #zero_ram is affecting the hardware UART. Both generated assembly listings are equal, except for the RAM clearing part. The RAM clearing part only writes to RAM, not to any of the UART related file registers as far as I can see.
I still think you can fix your problem by configuring the next lower baudrate of 113k5. Especially the fact that the software UART started working when you enabled early sampling is a good hint in this direction.

[cfernandez]

I have too many problems with my 18F4620, I dont know if all problems are bugs or the microchip part have some problem. The point is that the problem exist!, if a case for the RIPLEY's

Other strange problem is that my board to 20Mhz can comunicate by the UART to 115kb excellent, but I change the crystal to 25Mhz and to 115Kb I see garbage!!. Now I am crazy!!, I want to found some people that develop on 18F4620 and CCS and ask if have some problem.

Sorry by my english.

Thank you very much!.

[ckielstra]

[cfernandez]

Dear ckielstra,

Thank you very much!, for your information.

I have a question, Why in the table 2-2 Page 24 say 25Mhz????. and where you see that other people have the same problem?


Best Regards,

[rwyoung]

[ckielstra]

[rwyoung]

[cfernandez]

I make a ticket in Microchip support and this is your answer:

** PLEASE DO NOT RESPOND TO THIS AUTOMATED MESSAGE VIA E-MAIL **
** TO RESPOND SUBMIT ADDITIONAL COMMENTS TO THIS TICKET ON SUPPORT.MICROCHIP.COM **

Below is a response from Microchip Engineering Support Team for your submitted ticket #: 1-13744
A description of the problem and our proposed resolution is contained below.

Area: PICmicro Products
Product Group: PIC18FXX20
Product: PIC18F4620

Problem Description:
I have a some question:

Question 1)
We buy 30 parts 18F4620 in DigiKey, and the idea are use this with 25Mhz and 115Kb, but we have problem with the UART for this speed, we change the crystal to 20Mhz and the UART in 115Kb work OK.

I see that in the table 2-2 of 18F4620 (39626b.pdf) say 25Mhz, but now I see that in the table 26-6 say that the max. speed in HS is 20Mhz.

Can you explain me this difference.

Question 2)
Exist some bug in the 18F4620 parts??

Question 3)
How to idenfy if my parts are OK or are lots with bugs?

Question 4)
I try with 10Mhz PLL and I have garbage in the comunication, This feature work fine?


I wait your answer for make a new order to DigiKey for 100 parts more.

Thank you very much!

Bert Regards,



Problem Resolution:
For a 25Mhz oscillator the error in baud rate will be just over 13%, which is greater than the tlerances in the rs-232 spec.
Dropping the oscilator frequency to 20 mhz will improve the baud rate error to under 9.6% which may be close enough.

There is an error in the datasheet where HS oscillator mode is rated to 25 Mhz. There is curently documentation rectifying this in the process of being published.

There are several errata items in teh PIC18F4620 these are covered in the errata document.
http://ww1.microchip.com/downloads/en/DeviceDoc/80224a.pdf
The errata applies to the currently available devices.

You should be able to clock the device at 40 mhz with an external 10 mhz cyrstal and the x4 PLL. You may have to check your baud rate and configuration of the USART.


If the resolution provided does not solve your problem, you may respond back to the support team through the web interface at support.microchip.com. Telephone support is also available Monday-Thursday between the hours of 7:30am and 4:00pm MST and on Friday between 10:00am and 4:00pm. You may contact the Support department by calling (480) 792-7627. You will be prompted to enter the Ticket ID shown in the subject line of this message for verification. Enter the Ticket ID which is 1-13744 substituting the '-' with the '*' key. Press # on your telephone when finished.

Please do not respond to this message via e-mail.


Microchip Engineering Support

[cfernandez]

Where found the exactly documentation for know why 115kb in 25Mhz have more errors that 115kb in 20Mhz.

I remenber see this doc in microchip, but now I not found. Somebody know where is this?

Thanks,

[Chas]

I find it most curious that when I go to Microchip's website and search for the PIC18F4620 that there are no errata sheets in the search results. There are for most other processors. If you check the document in the above post by cfernandez, it is 10 pages long!!! I wonder what other dirty little secrets, oops I mean errata sheets, are lurking in the weeds???

[treitmey]

16F877A,.. oops that doesn't go to 25Mhz