High performance open-source 24bit USB DAC released | Hacker News Reader

High performance open-source 24bit USB DAC released

nwavguy.blogspot.co.nz

68 pointsdbcooper29 comments

This design is explicitly not open source. From the linked page:

NO SCHEMATIC: Because YoYoDyne is taking considerable financial risk to produce an unproven DAC in relatively high volume, it’s not reasonable to have it be open source (see the previous ODAC articles). Once he’s hopefully recovered his investment, we plan to re-visit the open source question.

However, there are several other open source DACs available with plenty of documentation. One such example: http://www.amb.org/audio/gamma2/

Curious about this. Monty (of xiph.org fame) claims[1] that the effective DNR of 16 bit audio is actually 120dB, not the 96dB you get by using the "standard" 6*(# of bits) formula, due to use of shaped dither (basically smarter quantization).

However, this DAC supports 24-bit audio, and claims a DNR of 111dB. He talks about guilt-free volume adjustment. That is, if you lower the volume in software, you're effectively decreasing your dynamic range because you're artificially limiting the range of values you can use. Makes sense. Is that the main reason you'd want to use a 24-bit DAC here? (Of course, this assumes you have 24-bit source material as well.)

Just really curious about this... I trust Monty's word on this stuff given his reputation and the amount of work he's done with audio. But the guy who designed this DAC clearly seems to know what he's talking about as well.

[1] http://people.xiph.org/~xiphmont/demo/neil-young.html

There are three separate things going on here: The standard "6 dB per bit" is the approximate signal-to-noise ratio with a flat dither. Then you can use a shaped dither to increase the dynamic range at some frequencies at the expense of others (the SNR actually decreases as the dither is louder overall, but quieter at frequencies the ear is most sensitive). Then in a real-world DAC there is analog noise getting in at various points in the circuit, so that is what limits the end result to 111 dB.

So yes, the main reason to use a 24-bit DAC here is for volume adjustment, so the lowest bits are well below the analog noise floor, not above it like they would be for a 16-bit DAC.

I think that one of the main points that Monty made, was that 16 bit is enough for storage, but you may want more bits when processing digital data. Digital volume control is transforming the data, and having some extra resolution to work with is helpful at that point. So in effect, they are both right here.

Now I know what I am going to do with my first slice of R-Pi http://www.raspberrypi.org/ " We now have in excess of 100,000 confirmed orders for the Raspberry Pi globally and can confirm that everyone who ordered before 18th April (i.e. today!) will definitely receive their Raspberry Pi before the end of June 2012, whatever your existing order confirmation says!"

Micro media-server?

absolutely, the audio performance of this usb powered DAC is knockout, hook this up to an amp and speakers or just a little headphone amp and you have networked reference quality audio for under a few hundred dollars that fits in a few square inches and sucks very little power

R-Pi is itself USB powered and then it should give power to the DAC. Would that deliver enough? I also wonder if iPhone/iPod can be connected to this DAC, deliver data via USB and power it?

If not, you could have a powered hub in between the rpi and the dac.

I used to get DAK catalogs back in the 80s, and this article reads very much like their product ads. "Here is what the audiophiles dont want you to know"

Here is some more background http://en.wikipedia.org/wiki/DAK_Industries http://ascii.textfiles.com/archives/1984

The designer has also released the designs for a headphone amplifier, and has manufacturers making it:

http://nwavguy.blogspot.co.nz/2011/08/o2-details.html#resour...

Would prefer a 1/4" headphone jack though. ;-)

Yeah I'd prefer one with bigjack outputs as well, or XLR. Like he says, balanced outputs aren't essential, but they definitely aren't snake oil: they are very useful at gigs.

Also, I'd like one with four channels please :)

It's not that hard to build a very good USB DAC since most of the work is done via hardware that's already pretty solid. The problem, like the article states, is that most of the parts are difficult to get and are very difficult to build by hand.

I would love this if this had big Japanese solid caps on it, and can do both native 44.1 and native 192khz output.

192kHz is unnecessary and actually harmful. See: http://people.xiph.org/~xiphmont/demo/neil-young.html

Yes, I've read that before. However, the argument fails because human hearing doesn't work that way and the DVD/post-CD industry already decided to use 48/96/192 instead of 44.1/88.2/176.4.

Quantization noise is a killer. Resampling audio is just a bad thing to do. Now, if it was a several mhz 1 bit DAC with a low pass filter, this wouldn't be much if an issue.

Why not resample in software before sending to the DAC? Benchmark Media claims that Windows 7's sample rate conversion is basically transparent (distortion below -140dBFS).[1]

[1] http://www.benchmarkmedia.com/wiki/index.php/Windows_7_Audio...

Because not everyone runs Windows, nor are they listening on something that isn't a portable music player in their pocket. Software or hardware accelerated conversion eats battery life.

> However, the argument fails because human hearing doesn't work that way and the DVD/post-CD industry already decided to use 48/96/192 instead of 44.1/88.2/176.4.

That's quite a strong statement to make in such an off-handed fashion, particularly in reply to such a (at least seemingly) well researched and often linked article. Furthermore, the citing the decision of the industry reeks a bit of appeal to authority.

Maybe you could back up your claims a bit better?

Go resample 48khz to 44.1khz using any resampling algorithm you choose. Go ahead. Thats all the proof I will ever need.

Alright, thanks for confirming my suspicion. You have no clue what you're talking about. Nobody is questioning that downsampling, particularly in that lower region, can degrade quality.

The author even explicitly states that 48khz and 24bit at least cannot hurt that much. He makes a point about whether that kind of "fidelity" is necessary, whether it produces superior listening and whether it can even have a damaging effect.

You claimed "Human hearing doesn't work that way" - well, then how does it work, in your mind? "Quantization noise is a killer" - The article clearly mentions that good dithering is the key here. "Resampling audio is just a bad thing to do" - Well, where does the article mention that resampling is great? Maybe we should better all go back to vinyl?

To quote the two paragraphs where resampling is even mentioned:

> Unfortunately, most samples are mastered to use the full digital range. Naive resampling can and often will clip occasionally.

> It is commonly assumed that resampling irreparably damages a signal; this isn't the case. Unless one makes an obvious mistake, such as causing clipping, the downsampled and then upsampled signal will be audibly indistinguishable from the original. This is the usual test used to establish that higher sampling rates are unneccessary.

Thats quite a long ad homonym attack you got there.

That word that you didn't even care to spell right: It does not mean what you think it means.

An ad hominem attack is "an attempt to negate the truth of a claim by pointing out a negative characteristic or belief of the person supporting it." I have not "pointed out a negative characteristic" of yours except my belief that you are wrong. Even more than that: You didn't really have much of a claim to start out with, just an opinion, which I challenged you to support with detailed claims and facts.

When you failed to deliver on that, I backed up my belief with arguments. That's how discussions work. Don't hide behind words when somebody is inviting you to state your case (remember my first reply to was asking you: "Maybe you could back up your claims a bit better?").

By now, however, you have already lost your case, so just accept defeat and try better next time.

I'm sorry, but what? All you're doing is trolling.

Please, find me a resampler that works for real world cases. You can point to URLs on the Internet or to research papers or to whatever you want; but when I use what the world considers the best available resamplers to deal with typical cases and the result audibly bad, then I'm not sure what you really want me to claim here.

Also, down voting me isn't helping your case. This is HN, you can't intimidate people here with cheap tricks.

I can't seem to find a link to the circuit design / PCB / firmware, anyone locate them?

That's good and all that, buy what about latency? That's the killer these days.

In the post he states that he'll be publishing measurements of latency in his next article on the ODAC.

In the comments section he writes:

"The total hardware latency of the ODAC is approximately 1.6 mS (it should be under 2 mS worst case). But that's just the hardware. You have to add whatever latency the driver (ASIO or native UAC1) requires for buffering, etc. I'll have more information in the next ODAC article. I haven't yet tried it with DAW software."

Pet peeve alert (as always in the interest of Educating The Public): that's "ms". No capital 'S', since that is a different SI unit, namely that of electrical conductance, the siemens (http://en.wikipedia.org/wiki/Electrical_conductance). Not sure which is more annoying, "mS" or "mSec". What's so wrong with "ms"?

And yeah, I realize the quoted text is at fault, not trying to assault the messenger in any way. :)

Yeah, I thought siemens at first too, then wondered why he'd gone for the wacky notation.

Reply on news.ycombinator.com