What is the magnitude of the numbers you are storing?
If your dataset is relatively small, the amount of overhead the TDMS file format adds for channel names and grouping is greater than the compression you get by storing the data as a binary file. Your data set isn't huge, but isn't all that small either.
However, you are saving doubles which would cost you 8 bytes per array element. With ASCII, you are storing the elements with about 6 or so characters per number since the default setting is 3 decimal places. You have a tab separator, 3 decimal places, the decimal point, and 1 character for each digit on the integer side of the decimal point.
Unfortunately, the VI you posted was missing some subVI's and typedef's and didn't reflect the image you posted. But cleaning it up and running it with a small 4x4 array with just a 3 in the lower right corner (all rest zeroes), I got 107 bytes for ASCII and 616 for TDMS. And the TDMS index file added another 456 bytes. So my test run showed both possibilities, very little data overall so that the extra info in the TDMS format was a greater percentage of the file. And relatively small numbers that took only 6 bytes to show and separate in ASCII as opposed to 8 bytes to store as a double in binary.
By looking at your file sizes, I say each value you are storing is costing about 2 more bytes with binary compared to an ASCII string with 3 decimal places. (147,160 array elements, ~258,000 difference in file size)
One disadvantage of ASCII is that the formatting to text will limit your precision to some amount that you define by the format string, but if you remain in binary, the value will be stored with all the precision/resolution that was present in its native datatype.
