Need help with python sign extension

**Wimpinator** · 06-17-2008, 10:23 AM

I am working on a project that has a computer (32 bit x86 machine) reading data (numbers) over the serial port from a 16 bit micro controller. My software on the computer is written in python.

Here is the problem. A 16 bit microcontroller is represents all numbers with only 16 bits. When I read the data with the 32 bit computer it misses the sign bit like this:

What the microcontroller sent: 1000010001000000 (MSB is set so the # is negative)
What the computer puts into the memory: 00000000000000001000010001000000 (MSB is not set so the number is wrong).

I'm a hardware guys so for the life of me I can't figure this one out. I was thinking to AND the data with 0x8000 to look for the MSB and if it is set then do an inversion (~number + 1) but no dice so far. Anyone have any ideas?

**RandomGuy** · 06-17-2008, 07:19 PM

You don't want to invert or take the negative of your number, just set the upper 16 bits appropriately. A C compiler would handle this for you if you used the correct types, I'm not sure about python.

Code:

if (data & 0x8000)
   data |= 0xFFFF0000;

that's c syntax, btw

Steamers · 06-17-2008, 09:05 PM

i'm lost o.0

**Dick Blonov** · 06-17-2008, 10:11 PM

Shift your number 16 bits to the left, do your math, then shift it back 16 bits to the right when you're done ?

Can you post the piece of code that processes that number ?

DB

**Wimpinator** · 06-18-2008, 10:59 AM

Well I think I figured it out. If the read number is negative then I invert it. I then AND it with 0xffff to chop off the extra ones (greater than 16 bits). Inverting one more time seems to give me the correct number. I was also misunderstanding how they were representing negatives. They are using twos complement.

The data is read over the serial port and I load it into a list for easy manipulation. I look at the MSB of the high byte and if it is set then I know that the number is negative. That is when I do the inversion stuff. Here is a code snippet for reference.

Code:

if (lst[4] & 128)  :
	Xaccel = Xaccel * -1
	Xaccel = (Xaccel & 0xffff) * -1

Well anyway it works now. Thanks for the advice guys.

**Dick Blonov** · 06-18-2008, 09:50 PM

Cool!

(that reminds me why I love assembler, one instruction: cwd, 2 bytes of code, done

)

DB

06-17-2008, 10:23 AM	#1 (permalink)
Wimpinator Join Date: Jan 2006 Location: United states, TN Age: 36 Posts: 2,982	Need help with python sign extension I am working on a project that has a computer (32 bit x86 machine) reading data (numbers) over the serial port from a 16 bit micro controller. My software on the computer is written in python. Here is the problem. A 16 bit microcontroller is represents all numbers with only 16 bits. When I read the data with the 32 bit computer it misses the sign bit like this: What the microcontroller sent: 1000010001000000 (MSB is set so the # is negative) What the computer puts into the memory: 00000000000000001000010001000000 (MSB is not set so the number is wrong). I'm a hardware guys so for the life of me I can't figure this one out. I was thinking to AND the data with 0x8000 to look for the MSB and if it is set then do an inversion (~number + 1) but no dice so far. Anyone have any ideas? __________________ \|TG-6th\|Wimpinator BF2 Rules and SOP's\|Teamspeak information\|TG on IRC

06-17-2008, 07:19 PM	#2 (permalink)
RandomGuy Join Date: May 2003 Location: K-W, Ont. Age: 27 Posts: 1,740	Re: Need help with python sign extension You don't want to invert or take the negative of your number, just set the upper 16 bits appropriately. A C compiler would handle this for you if you used the correct types, I'm not sure about python. Code: if (data & 0x8000) data \|= 0xFFFF0000; that's c syntax, btw

06-17-2008, 09:05 PM	#3 (permalink)
Steamers Join Date: May 2008 Posts: 158	Re: Need help with python sign extension i'm lost o.0

06-17-2008, 10:11 PM	#4 (permalink)
Dick Blonov Join Date: Feb 2006 Location: In a Vortex! Age: 51 Posts: 3,398	Re: Need help with python sign extension Shift your number 16 bits to the left, do your math, then shift it back 16 bits to the right when you're done ? Can you post the piece of code that processes that number ? DB __________________ \|TG-6th\|Blonov «Any situation that involves the line "I was ran over by a dumptruck..." can't be good.» eGoatBoy BattleField2 SOPs \| Teamspeak \| Server Rules and SOPs \| The 6th Devil's Brigade Last edited by Dick Blonov; 06-17-2008 at 11:10 PM.

06-18-2008, 10:59 AM	#5 (permalink)
Wimpinator Join Date: Jan 2006 Location: United states, TN Age: 36 Posts: 2,982	Re: Need help with python sign extension Well I think I figured it out. If the read number is negative then I invert it. I then AND it with 0xffff to chop off the extra ones (greater than 16 bits). Inverting one more time seems to give me the correct number. I was also misunderstanding how they were representing negatives. They are using twos complement. The data is read over the serial port and I load it into a list for easy manipulation. I look at the MSB of the high byte and if it is set then I know that the number is negative. That is when I do the inversion stuff. Here is a code snippet for reference. Code: if (lst[4] & 128) : Xaccel = Xaccel * -1 Xaccel = (Xaccel & 0xffff) * -1 Well anyway it works now. Thanks for the advice guys. __________________ \|TG-6th\|Wimpinator BF2 Rules and SOP's\|Teamspeak information\|TG on IRC

06-18-2008, 09:50 PM	#6 (permalink)
Dick Blonov Join Date: Feb 2006 Location: In a Vortex! Age: 51 Posts: 3,398	Re: Need help with python sign extension Cool! (that reminds me why I love assembler, one instruction: cwd, 2 bytes of code, done ) DB __________________ \|TG-6th\|Blonov «Any situation that involves the line "I was ran over by a dumptruck..." can't be good.» eGoatBoy BattleField2 SOPs \| Teamspeak \| Server Rules and SOPs \| The 6th Devil's Brigade

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)

Sponsored links

Sponsored links