The fact audiophiles talk about “DAC”s is really telling. Transparent digital-to-analogue conversion is a solved problem. Any computer or smartphone worth a damn has a DAC whose output is indistinguishable to the human ear from anything “better”.

The truth is that DAC is not the problem… everything else in the analogue audio chain is. Amplifiers are messy analogue devices. Speakers and headphones are incredibly messy analogue devices. Power supplies and power conditioners are messy analogue devices. And noise is not down to any one component, but is a whole-system design problem. A particularly cool thing about power supplies is that they often create noise that will be picked up by other devices on the same circuit.

Of course, when people are buying a “DAC” they are really buying a box of some kind that also includes an amplifier, but this naming choice surely contributes to people paying attention to the wrong specs.

I agree that you can have cheap and great DACs nowadays that far exceeds what a consumer needs but some manufacturers still insists in putting less than ideal ones, or fumbling their implementation, in all sorts of device. I've got a computer with a fairly recent motherboard that I'm using for an htpc and the audio output is much noisier and produced a less punchy bass than the now 20 year old (!) X-fi sound card I've been using for... 20 years. Too bad, I wanted to give it a respite but I guess It's gonna keep trucking for at least another decade.

I then proceeded to investigate that by using REW to produce measurements of each output and the findings confirmed my hearing, the motherboard audio was outputting 5db less, plus a noise floor 15 db higher than the x-fi, resulting in 20db of extra noise (When you compensate with 5db extra at the amp). The resulting frequency response also revealed a much more aggressive high pass filter, I was actually getting another -3db@30hz compared to the x-fi confirming where the lack of "punchiness" was coming from. And then the cherry on the top was that all the extra channeles (surrdound/sub/center) had an even more aggressive highpass filter, they cheapened out on the cheapening out, probably assuming you are not gonna use more than stereo.

Here is the graph, red is x-fi outs and green is motherboard outs: https://i.imgur.com/dxoLXJO.png

Obviously this is just my anecdote, but I suspect this is very widespread.

EDIT: Oh and I haven't even talked about the line-ins! It's pretty much unusable in ALL motherboard audio I've cared to try. There is some insane noise gating and AGC going on to try and mask the fact that they use some really low bit depth ADCs with terrible dynamic range. Meanwhile I can capture pristine audio from my n64 into the x-fi no problem.

Most of your noise floor is not going to be coming from the DAC, and your frequency response difference is probably caused by a difference in impedance, which is a fun topic because it's not like there's a “correct” impedance and you may just be more fond of the effect a particular impedance pairing gets you.

AFAIK impedance mismatch should only matter in relation to speakers and amps or headphones and dacs/pre-amps. It shouldn't matter when doing line level output to an amp or even a line level loopback (dac to adc). The fact it has such behavior in a consistent manner leads me to believe it is just out of spec or badly implemented, even if it is impedance matching after all.

Regarding noise floor, the "DAC" (really, the audio source as a whole) did matter in this case. The volume is being set at the amp to as loud as I need because the volume is being controlled at the source. In the xfi setup I couldn't hear any hiss, noise or EMI from the pc itself, while on the onboard audio it bothered me pretty much instantly with a constant hiss and EMI buzzing even from moving the mouse. I do want to note again that it's likely the manufacturer implementation at fault rather than purely the DAC fault, regardless, the user is the one being harmed with a subpar product.

I do my setup this way because I have a pile of DSPs running in that PC, including an equal loudness contour compensated volume "dial" (bass/treble gets boosted as volume goes down essentially), so controlling the volume at the source is a must.

this is because nobody who talks of DACS means just the DAC circuit, it means the whole unit.

like where you talk about your phone you actually mean pocket computer, but it's too long to say.

True, but plenty chase specs like sample rate (beyond 96 kHz) and bit depth (beyond 24) that have no practical benefit beyond the DAC circuit.

You're right. In serious audio world these numbers are kind of like saying "64-bit CPU". Other stuff like power is much more important. Someone could make big bucks advertising that if there was a catchy way to do it

> 100-1000x lower than any transducer

This seems like a lot but it's only 20-30db lower than whatever reference they're using.

This is the spec that really matters: THD+N: 0.0003% which is roughly -110 dB. It's very good and completely inaudible but not exceptional these days.

THD+N is irrelevant for the issue the author is describing, through. You need to spec PSRR (power supply rejection ratio). Many individual ICs do not spec this, and pretty much every system does not spec this.

Indeed. I was only commenting on the 100-1000x claim.