Post by thread:Displaying Mantissa and Exponent on an LCD

Currently I am using the floating point math routines in Microchip AN575 and converting the result to a 5 digit BCD number to display on a character LCD module. I would like to display the result in standard scientific notation. Does anyone know a relatively easy way to convert the PIC floating point base 2 format to a base 10 mantissa and exponent? Alan

> Currently I am using the floating point math routines in Microchip AN575 and > converting the result to a 5 digit BCD number to display on a character LCD > module. I would like to display the result in standard scientific notation. > Does anyone know a relatively easy way to convert the PIC floating point > base 2 format to a base 10 mantissa and exponent? I've never tried this, but the math shouldn't be too difficult: I'd create a routine that would take the float number , k, in the form (+-) f x 2^n, and convert f into 2^x (where x is a float). This can be accomplished by using a small lookup table and linear approximation. Add x and n, and the result leaves k in the form 2^x. Multiply by 0.30103 (log 2), to convert this to 10^x. Now, take the " f " portion of this float, and using a small lookup table and linear approx. again, convert it to the mantissa. The exponent portion is actually 1 shy of what it should be if you're using Microchip's float format. This should give you 2 significant figures reliably if you use a 16-point lookup table... I think I'm sure someone out there has a better idea, but I don't mind starting the thread :-)

At 14:11 04/05/99 -0700, Vogel, Alan wrote: >Currently I am using the floating point math routines in Microchip AN575 and >converting the result to a 5 digit BCD number to display on a character LCD >module. I would like to display the result in standard scientific notation. >Does anyone know a relatively easy way to convert the PIC floating point >base 2 format to a base 10 mantissa and exponent? i don't know what routines are provided in that app note to do what you want, but if you have a standard float representation, you usually have two binary (signed) numbers: the mantissa and the exponent. you just have to convert them separately to decimal/bcd, then you can display them in standard float notation, ie. "0.6444e-55". the "0.6444" part and the "-55" part are directly available (in binary) by extracting the appropriate bits from the bit field representing the number. just convert them. have a look at the way the routines store the float data, and you'll see what i mean. ge

Oops, I thought you meant any floating point number. For a 5-digit BCD, it would be as easy as shifting the decimal and counting how far you had to shift it... ---------- > From: Vogel, Alan <spam_OUTAVogelTakeThisOuTPACESETTER.COM> > To: .....PICLISTKILLspam@spam@MITVMA.MIT.EDU > Subject: Displaying Mantissa and Exponent on an LCD > Date: Monday, April 05, 1999 3:11 PM > > Currently I am using the floating point math routines in Microchip AN575 and > converting the result to a 5 digit BCD number to display on a character LCD > module. I would like to display the result in standard scientific notation. > Does anyone know a relatively easy way to convert the PIC floating point > base 2 format to a base 10 mantissa and exponent? > > Alan

Thanks for the reply. Unfortunately I think I would lose too much resolution after using 2 lookup tables with interpolation. I'm performing math on a 12-bit signed number and would like to preserve the 3 and 3/4 digit display capability (e.g., three digits to the right of the decimal point). > {Original Message removed}

Gerhard, Thanks. I was hoping there would be an easier way. After I convert the mantissa and exponent to decimal I still have to multiply them together (because the exponent is in the form of 2^n) and then extract the exponent by normalizing the result. Alan > {Original Message removed}

You're right, but my 5-digit number is derived from a floating point to integer conversion which doesn't always provide me with the resolution I need. I think maybe the fp to integer conversion I'm using should be replaced with what Gerhard suggested. Alan > {Original Message removed}

|Currently I am using the floating point math routines in Microchip AN575 and |converting the result to a 5 digit BCD number to display on a character LCD |module. I would like to display the result in standard scientific notation. |Does anyone know a relatively easy way to convert the PIC floating point |base 2 format to a base 10 mantissa and exponent? If speed is not critical, and if you don't mind having different routines for the >=1 case and the <1 case, I'd suggest that you do a variation of the BCD shift-and-add technique used elsewhere. It's not the most effecitent in terms of execution speed, especially with large negative or positive exponents, but what you do is simply mult- iply or divide the decimal representation by two the appropriate num- ber of times. If you're multiplying by two and you get a carry into the leftmost digit (your buffer should be one digit larger than what you're displaying) you should shift everything right one place and add one to the [displayed] exponent; if you're dividing and you get a zero in the leftmost place, shift left one place and subtract one from the exponent.