how do I convert a hex to a floating point number

There is a great discussion of this topic on the main SSC site which includes Jeff Moden, Peter Larsson (whom we know as @Peso on this site) and Martin Wills. Here is the link: http://www.sqlservercentral.com/Forums/Topic356756-8-1.aspx. I just tried the version which Martin Wills posted on page 3 which was in turn a recoding of Peter's code. Here is the paste from the post referenced above:

/* Hats off to Peter Larsson and Martin Wills for this one */
CREATE FUNCTION dbo.fnBinaryReal2Real
(
    @BinaryFloat BINARY(4)
)   RETURNS REAL 
AS

BEGIN    

    DECLARE  @Mantissa REAL, 
    @Exponent SMALLINT, 
    @IntValue INT, 
    @Real2 REAL        

    SELECT 
        @Real2 = CAST(2.0 AS REAL), 
        @IntValue = CAST(@BinaryFloat AS INT),            
        @Exponent = (@IntValue & 0x7f800000) / 0x00800000,            
        @Mantissa = 1.0 + (@IntValue &  0x007FFFFF) * POWER(@Real2, -23)    

    RETURN  SIGN(@IntValue) * @Mantissa * POWER(@Real2, @Exponent - 127)

END
GO

Here is the test:

declare @input binary(4) = 0x43F8E354;
select dbo.fnBinaryReal2Real(@input) result;

result
-------------
497.776

Please note that the desired number 497.77600098 is not real (in both senses) because it has way too many digits for a real (a.k.a. float(24)) number to afford. However, if the accurate binary representation 0x43F8E354 of the single floating point number 497.776 is sufficient then the function above should help.

<!-- Begin Edit

Here is the link to Peter's post where he restated both binary to real and binary to float as inline scalar functions: http://www.sqlteam.com/forums/topic.asp?TOPIC_ID=81849. Scroll to the bottom of the post to see them.

End Edit -->

<!-- Begin Edit 2

I am glad that the links helped. However, I would like to comment on the function you use in SSIS. Please don't use it as is if you can because it leaks memory. Not much memory to speak of, but it still does which is never a good thing. Both MemoryStream and BinaryReader are disposable types but the code does not close and dispose any of them. Thus, the code will leak some memory and might eventually trash the server if the function is called frequently enough for a long enough period of time.

Usually the way any conversion works is consistent: it always converts as many bytes as needed. For example if you need to convert a byte array to REAL (System.Single in VB or float in C#) then only 4 bytes of the array will be read and converted. Conversion to FLOAT (System.Double in VB or double in C#) will take to read and then convert 8 bytes of the byte array input. Here is the function written in C# which you can use in place of the one you have (you can use online telerik code converter to convert it to VB from C#):

/// <summary>
/// Converts the byte array to a double precision floating point number.
/// These have a maximum number of significant digits limited to 15.
/// In T-SQL this type translates to float(53) or simply float by default.
/// </summary>
/// <param name="blob">System.Double (called double in C#) needs 8 bytes of
/// storage which means that the input should be limited to 8 bytes.</param>
/// <returns>Double precision floating point number</returns>
static double ReadDouble(byte[] blob)
{
    if (BitConverter.IsLittleEndian) Array.Reverse(blob);
    return BitConverter.ToDouble(blob, 0);
}

Similarly, if you need to convert a byte array to a single precision floating point number, you can feed the 4 bytes long byte array to a similar function like this:

/// <summary>
/// Converts the byte array to a single precision floating point number.
/// These have a maximum number of significant digits limited to 7.
/// In T-SQL this type translates to float(24) sometimes called real.
/// </summary>
/// <param name="blob">System.Single (called float in C#) needs 4 bytes of
/// storage which means that the input should be limited to 4 bytes.</param>
/// <returns>Single precision floating point number</returns>
static float ReadSingle(byte[] blob)
{
    if (BitConverter.IsLittleEndian) Array.Reverse(blob);
    return BitConverter.ToSingle(blob, 0);
}

As you can see, the code is pretty simple, and does not use any disposable types. Testing the latter using the bytes in your original question:

byte[] blob = new byte[] { 0x43, 0xF8, 0xE3, 0x54 };
float f = ReadSingle(blob);

Console.WriteLine(f);
Console.ReadLine();

will print 497.776 in the console window as expected.

End Edit 2 -->

Oleg

This would be a good place for you to start: Convert Hex to int.