The problem with sound cards.

I stumbled on an article the other day that piqued my curiosity, but I forget what it was about. There were a few comments at the end, and before I knew it, I was reading a debate about the best sounding PC motherboard for gaming. My pick would be the Gigabyte GA-Z170X Gaming G1 board.

[Actually, I have no real opinion, nor am I a gamer. The PCM1794 caught my attention, and it’s inclusion in a motherboard inspired this post.]

Here’s something to think about: Why wouldn’t this board make a great network DAC?

After all, this is the same chip we use in our Virtuoso DACs.

So by DAC chip spec alone, we’ve got a winner. But there’s so much more to it than the D to A processor.

What jumps out at me from my earlier experience with trying to tame PCs is noise and ground. Not “earth” ground, but the reference ground against which signals are measured. A PC motherboard is a hellish place, and finding a quiet stable ground amidst the plethora of switching regs is nearly impossible.

“But hey,” you’re thinking, “this article is about sound cards, not on-board audio.”

The same applies for them, living in the PC’s power pool.

But there’s something that’s equally bad, and that’s the PCI bus. (whether or not it’s the Express line)

Did you ever play the telephone game when you were a kid? When you whisper a message to the person next to you, and they do the same? When you get to the end of the line, what comes out isn’t exactly what went in.

When you translate signals, you run the risk of generating errors, both as noise and as timing issues. PCI is one of those universal buses (Peripheral Component Interconnect, to be precise) that requires a translation (or two) to send audio data in, and get some other flavor of audio data out.

In the early days, we developed a daughter board that connected to a Juli@ card to pull an I2S signal. It worked well and sounded nice. But it wasn’t until we developed a player with a couple less signal translations that we could begin to hear the troubles with PCI.

I mean, if you can grab a music signal from an isolated Ethernet connection, clock the bits locally to generate an I2S signal then send that (through high speed isolators, of course) to your DAC chip, why wouldn’t you? In this case, I don’t know if “less is more”, but it is certainly better.