Checksum of the used bytes, or of all of program memory including implied 0xff of unused bytes Both! For what I understand your usage to be, you could implement it in code inside your bootloader or main program. Designate an EEPROM or flash area where you will store the checksum and initialize it to 0xFFFF. Then in your bootloader calculate a checksum for whatever memory you want to include (you can calculate a checksum over both flash and eeprom, or even external memory) and if the previous value is FFFF store it. Each subsequent boot you recalculate and compare. Here is a small CRC-16 implementation:.

CRC Calculator (Javascript) CRC width RadioButton: CRC-8 CRC-16 CRC-32 CRC parametrization Predefined. Custom CRC detailed parameters Input reflected: Result reflected: Polynomial: Initial Value. (Cyclic Redundancy Check) calculation for an detailed discussion of CRC calculation.

Please Lee, I would really appreciate your comments on this as well as the other freaks Did you miss my suggestion of srec_cat? Have you looked at the manual? Oh and I only picked a big endian CRC16 as an example.

How to install sapd fr v2 550 trackside. It can do all sorts of checksums and CRCs. By default it will just checksum the data from input file but if you add a --fill on the command line you can instruct it to extend the data and virtually fill the remaining space with any value which would include 0xFF (the value in erased but unused flash bytes).

There's also --crop which you can use to limit the region that it operates on to some specific address range. In fact you should think of the srec_cat command line as almost a linear flow of programming operations. You first have it load a file into a buffer (it can read about 30+ different code formats). Then you list the operations to be performed on that buffer such as cropping it or expanding it or calculating a checksum and embedding it in there etc. It is VERY powerful. While one could easily use libbfd to write your own.hex file processing program to achieve similar the fact is that SRecord (the global name for the group of programs that includes srec_cat) does pretty much anything you can ever think of to.hex (.srec,.cosmac,.fairchild.) load images.

The only thing I ever found an 'issue' in srec_cat was working out the polynomial, starting value and augmentation of the CRC they actually use. I did work it out eventually and my decoder (at least of their --l-e-cr16) is built into this. David and Cliff, I don't think my knowledge is anywhere near enough for this stuff I think the most simple of all is the Modulo 256, which I think it simply adds all the bytes of memory and divides by 256 to give the remainder, so for an 8bit system it could only hold the checksum for 256 bytes of memory (assuming the memory is all 0xFF) Is this correct? An Atmega328P has 2048 bytes of data memory so the checksum for all that could theoretically be a maximum of 2048*0xFF/0xFF So it would require 11 bits am I right here? Did you miss my suggestion of srec_cat?

Have you looked at the manual? I had a look but it is like a foreign language to me, it just isn't easy reading at all If the OP doesn't mind I think an example would be best, on a really simple program. Is this correct? Well, yeah, if you only sum into an 8bit then the result clearly cannot be more than 0.255.

But if you are just 'adding up the bytes' why do you feel the need to squeeze it into 8bits? Why not a 16, 32, 64 bit sum? An Atmega328P has 2048 bytes of data memory so the checksum for all that could theoretically be a maximum of 2048*0xFF/0xFF So it would require 11 bits am I right here? 256 is 2^8 and 2048 is 2^11 so to sum them would take 8+11 = 19 bits. 2^19 is 524,288 the result of 256*2048. But who was talking about summing SRAM?

We're talking about checksums of flash here. It's usually used when you send a.hex file off to the factory. They program the code into a device then read it out again and get the checksum. If that matches what you told them when you sent the.hex file they know they have programmed the correct version of code successfully.

It just isn't easy reading at all I agree the way the manual is just made up of 'man' pages does not make it that easy to follow. In WinAVR there is a PDF version that pulls all the 'man' pages together than makes things easier to find. Cliff thanks BNow I am waiting for the OP to give me permission before I go hijacking another thread so to sum them would take 8+11 = 19 bits. Yes that would be the maximum value of the sum but for modulo 256 I thought we divide by 256 so the result would be a maximum of 11 bits, do I have this correct? We're talking about checksums of flash here.

Cohen, John Wiley and Sons, Inc. Economics by begg fischer and dornbusch pdf to jpg. Floyd Daniel I.

Yes I think so! I thought checksums were used to check if the flash had gone corrupt?, I can understand having a checksum to compare to the uploaded code but aren't checksums used to check the flash of a uC as it runs out in the field? Or have I got this totally wrong?