Here's why it's important.
Lets say you have a scanning mirror with a rotational speed of 3000RPM, this is 50RPS. If the mirror that's being spun has 6 sides, this then gives you 300 scan-lines per second. For a 48000Hz sample rate sound card, that gives you 160 samples (pixels) per scan line. An image with a 4:3 aspect ratio would then have dimensions 160x120, and to scan 120 lines (at 300 lines per second) would take 0.4 seconds. This is only slightly faster than 2FPS. That's quite slow, and as a result low frequencies must not get blocked out, or you will have a bad image.
I've approximated the frequency response of my current AC-coupled soundcard, with the following equation:
wave(n+1)-wave(n) + wave(n-1)*0.99
wave is the signal value
n is the current sample number
wave(n) is the current signal value
wave(n+1) is the next signal value
wave(n-1) is the previous signal value, and it is dependent on the output of the filter operation from the previous sample
This is a simple IIR filter, and it closely approximates the frequency response of my sound card, when the sample rate is 48000Hz.
To test how this would effect an image. I loaded a Windows7 sample image (tulips.jpg) into Gimp, converted it to grayscale, and reduced it in size to 160x120 (using bilinear interpolation).
I then exported the raw data (no header, just the raw 8-bit grayscale values) from Gimp. I loaded it into GoldWave audio editor software (which supports loading and saving raw audio data). I then used the above IIR filter equation in GoldWave's equation evaluator feature (which allows almost any equation you can think of to be used to process audio data, just by typing it into the text field and pressing the Ok button). I then saved the resulting filtered data back to raw data, and imported it back into Gimp. I then saved both the original and filtered copies of the image as PNG files from Gimp, and uploaded them to Imgur so I could embed them in this post.
Here's the original image.
For comparison, here's the filtered image.
The simulated AC-Coupling filter obviously messes up the image. Now you know why I will need a DC coupled soundcard before I can start my experiment at building a mechanical scanning infrared imager that interfaces to my computer via the Line Lin port on the soundcard.