Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Feeding filterless DAC with SOX (Read 28252 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Feeding filterless DAC with SOX

Already posted this else where on forum, reposted in more appropriate topic, this version varies slightly.
So. I've been researching the DAC in my spare time for several months after buying a product (and taking the research way too seriously as one does lol).
Familiarizing myself with the Digital Compact Disc as a more-than-adequate medium in the beginning, Nyquist, sampling theorem, the gradual evolution of DAC applications, ranging from the early 80s brickwall type, 16, 18, eventually 20bit 8x oversampling chips in the CD player, PCM vs PWM, the transition of playback over to personal computers in 90s among them being 16bit ISA interfaces and eventually 24 bit which immerses the market today.

After getting a feel of the direction of hifi, I concluded that external asynchronous DAC is ideal for playback.
But 16bit? hard to find still.
I find the average consumer being offered studio grade 24 bit designs complete with dsp chips which are designed for production. especially true for sound cards, 8 extra bits and a compromise in value when all one wants to do is playback, simply 44.1, 48 or even 96 usually, which only really requires 16bit chips.

So is it possible to feed a 16bit 44.1/48/96 (Filter-less) DAC with up-sampled 44.1 material (to 96) with Sox via foobar - or is 96 not enough to prevent artifacts?

Have software re-samplers replaced the need for hardware based re-samplers?

I understand the main reason up-sampling exists is because its required to make post-DAC (analogue) filtering easier which in turn prevents 44.1k entering analogue systems but what about analogue systems which are capable of handing 44.1 in the first place? These do exist.

How about
NOS DAC (with sinc filter +3db @ 22khz ofcourse) > valve pre amp > followed by amps hardware capable of reproducing 44.1khz. its only 44.1k I know of amps capable of flat response up to 100khz.

Another point.

Am I right in writing most PC sound cards with 24bit/192khz chips, have them so in parallel (each for L and R) so theoretically are actually 384khz DACs. Some DACs/Sound cards have chips for each channel, some dont!

To clarify what I mean, here is a 352.8khz (thats 44.1 x8) a true 8x DAC.
Musiland US 03 Dragon, capable of said rate across USB 3.0 in Async mode.
http://www.amazon.co.uk/Musiland-03US-Drag...l/dp/B00A2QL1CI

Thoughts?

Apologies too for long post, didnt want post seperate threads.

Feeding filterless DAC with SOX

Reply #1 –
Quote from: giro1991 on
I find the average consumer being offered studio grade 24 bit designs complete with dsp chips which are designed for production. especially true for sound cards, 8 extra bits and a compromise in value when all one wants to do is playback, simply 44.1, 48 or even 96 usually, which only really requires 16bit chips.


What's wrong with current oversampling DACs? For instance, my trusted EMU 0204 USB (€120) shows excellent performance all across the board, way, way beyond audibility. As far as the DAC alone is concerned, even my cheap Realtek ALC663 chip in my laptop (what, maybe $1 when sold in bulk?) shows excellent 16 bit performance.

Feeding filterless DAC with SOX

Reply #2 –
Any halfway decent DAC is going to be better than human hearing, so I wouldn't worry about what's "under the hood".    With modern electronics it's cheap & easy to build a good DAC.  The only real sound quality issue tends to be soundcards that sometimes pick-up electrical noise from inside the computer.    I'd look at "features".  i.e. Do you need S/PDIF, USB or HDMI?  Do you need Dolby or DTS decoding?  Do you need multichannel?  Do you want a nice looking cabinet/box?

Quote
So is it possible to feed a 16bit 44.1/48/96 (Filter-less) DAC with up-sampled 44.1 material (to 96) with Sox via foobar - or is 96 not enough to prevent artifacts?
You should never get audible artifacts with upsampling.  If you downsample too-low you'll loose high frequencies.  If you reduce the bit-depth too far, you'll get quantization noise.

I've resampled both ways between 44.1kHz (CD) and 48kHz (DVD), and I've never heard any difference, no matter what software I was using.  I used to worry about it because it's a theoretically lossy process (not bit-for-bit reversible).  But sometimes you don't have a choice, and it turns-out that there's no difference in the sound.

Quote
Have software re-samplers replaced the need for hardware based re-samplers?
I suppose it depends on why you are resampling.  If you are doing music production and you have  96kHz master files (or a variety of sample rates), you'd use software resampling if you want to make a CD.

I think true hardware resampling is rare.    On a computer, "hardware resampling" is usually done by the driver.

Quote
I understand the main reason up-sampling exists is because its required to make post-DAC (analogue) filtering easier which in turn prevents 44.1k entering analogue systems but what about analogue systems which are capable of handing 44.1 in the first place? These do exist.
I think you are confusing oversampling with resampling.

Quote
How about
NOS DAC (with sinc filter +3db @ 22khz ofcourse) > valve pre amp > followed by amps hardware capable of reproducing 44.1khz. its only 44.1k I know of amps capable of flat response up to 100khz.
So???  You can't hear that high, and the signal is filtered.  It's pretty easy to build a solid state amp/preamp that goes from DC (zero Hz) above 100MHz.  In fact if you simply do nothing to limit the frequency response, an op-amp circuit will typically go into the MHz range...    Tube preamps can go very high too.  It takes a bit more effort & cost to make a tube preamp that goes to DC.  Tube power amps have transformers, and that limits their frequency response on both ends.  (But with enough money, you can build a tube power amp that's flat from 20 - 20kHz)

Feeding filterless DAC with SOX

Reply #3 –
PC sound cards running at 192k are not really a single DAC running at 384 kHz or at least I have never seen that design used in audio (it is very common in other applications). In audio its not worth it because the DACs are so simple and cheap.

Instead they are nearly always 2x DACs @5-10MHz (so that aliasing is all but impossible).

Feeding filterless DAC with SOX

Reply #4 –
Quote from: giro1991 on
After getting a feel of the direction of hifi, I concluded that external asynchronous DAC is ideal for playback.


Asynchronous USB DACs are a solution looking for a problem. The problem that they would fix could be jitter, but audible jitter is a boogey man, not a real threat to sound quality.

Quote
But 16bit? hard to find still.


What are you talking about? Good 16 bit products literally grow on trees.

Quote
I find the average consumer being offered studio grade 24 bit designs complete with dsp chips which are designed for production. especially true for sound cards, 8 extra bits and a compromise in value when all one wants to do is playback, simply 44.1, 48 or even 96 usually, which only really requires 16bit chips.


Still plenty of 16 bit hardware around. My benchmark USB audio interface is the Behringer UCA 202. $30 or so all over the web. What can you hear wrong with it?

Quote
So is it possible to feed a 16bit 44.1/48/96 (Filter-less) DAC with up-sampled 44.1 material (to 96) with Sox via foobar - or is 96 not enough to prevent artifacts?


Why would someone do something that arcane?

Get back to me when you find something audibly wrong with a UCA 202 - with DBT in hand!

Quote
Have software re-samplers replaced the need for hardware based re-samplers?


Software resampling is certainly more perfectible, but why do any resampling? BTW this is not a slam on Sigma Delta converters which incorporate upsampling as a matter of their operational principles, but are just fine.


Quote
I understand the main reason up-sampling exists is because its required to make post-DAC (analogue) filtering easier which in turn prevents 44.1k entering analogue systems but what about analogue systems which are capable of handing 44.1 in the first place? These do exist.


The > 22.05 KHz images that reconstruction filtering removes sounds like @!$$.

Quote
How about
NOS DAC (with sinc filter +3db @ 22khz ofcourse) > valve pre amp > followed by amps hardware capable of reproducing 44.1khz. its only 44.1k I know of amps capable of flat response up to 100khz.


Why put boutique garbage like tubes in the signal path?

What's wrong with my benchmark $30 DAC?  Where are your DBT results showing that it has audible artifacts?

Quote
Am I right in writing most PC sound cards with 24bit/192khz chips, have them so in parallel (each for L and R) so theoretically are actually 384khz DACs. Some DACs/Sound cards have chips for each channel, some dont!


Your big mistake all though this post  seems to me to be that you are way to focused on means and seem to have no reliable clue about results. It sounds to me like you have read way too many golden ear pontifications and not done nearly enough reliable listening tests.

Feeding filterless DAC with SOX

Reply #5 –
Quote from: Arnold B. Krueger on
why do any resampling?


For mixing two or more streams with different sampling rates together (i.e. normal PC audio, e.g. 44.1kHz and 48kHz, two very common sampling rates).
 

Feeding filterless DAC with SOX

Reply #7 –
Quote from: Arnold B. Krueger on
Of course, but this discussion was about a single stream, no?


To be honest, I have no idea what the discussion (or the OP) is about.
Seems like the OP is talking about problems that don't really exist (anymore?).

Feeding filterless DAC with SOX

Reply #10 –
http://www.astellnkern.com/eng/htm/ak120/ak120_feature.asp

Quote
From input to output, the design of the dual Wolfson WM8740 DAC implemented in the Astell&Kern AK120 independently separates the left and right audio channels, creating a true Dual-Mono audio output system.
When the analog audio signal passes through the dual Wolfson WM8740 DAC, not only does the AK120 use the Ultralow Distortion and Noise Technology, but executes the optimal digital noise reduction technology combined into the analog amp. The dual DAC built-in to the AK120 forms a dual-monaural audio ecosystem that brings wider soundstage, channel separation, and an overall richer sound.

Feeding filterless DAC with SOX

Reply #11 –
Quote from: Arnold B. Krueger on
Of course, but this discussion was about a single stream, no?

Yes only single stream.

Quote from: phofman on
Which soundcard uses a separate DAC chip for each channel?

Onkyo 90PCI

Seems like the OP is talking about problems that don't really exist (anymore?).[/quote]

Apologies but why answer at all if you're unsure?
I spent a considerable amount of time making the OP as clear as possible. Including the ~8 months research prior to me posting (lol).

Though OP is really about why there is a lack of such products, here are two close contenders for what I am seeking.
Can you feed either of these with (foobar>sox) 96k, or is 96 not enough to remove imaging?
Pro-Ject BOX USB 'FL'
or
firestone fubar30MKIII

Feeding filterless DAC with SOX

Reply #12 –
Quote from: giro1991 on
Quote from: phofman on
Which soundcard uses a separate DAC chip for each channel?

Onkyo 90PCI

Separate channels for greater separation. Many products do this. DAC chips usually can process Lout and Rout, but sometime manufacturers choose to put 2 chips in.


Theres actually only 1 DAC chip on that board.  You're probably looking at the buffering opamps, not the DAC.

Quote from: giro1991 on
I'd however like a basic 16 bit only DAC to reproduce as close to the original. Many products which state 16bit are '16/24bit' are normally 24bit chips capable of 16bit. I want purely 16bit.


To be clear, any 24 bit DAC is also a 16 bit DAC.  I think the idea you have is that you have 16 bit samples so you don't want to zero pad them to 24 bit, but this operation is lossless.  There is no advantage to what you are thinking.  You'll get the best performance by picking the device with the highest SNR and lowest distortion. 

Quote from: giro1991 on
Quote from: DVDdoug on
Quote
So is it possible to feed a 16bit 44.1/48/96 (Filter-less) DAC with up-sampled 44.1 material (to 96) with Sox via foobar - or is 96 not enough to prevent artifacts?

You should never get audible artifacts with upsampling.


Let me re-iterate. Referring to sampling theory, is a jump from 44.1k to 96k using sox suffice to prevent imaging? The reason I'm asking this is because 16bit-only chips can do 96k max.


To be honest, I don't really understand what you are asking.  Could you explain in more detail what it is you are trying to do?

Feeding filterless DAC with SOX

Reply #13 –
Sorry, see last part of my second post / reply and the products I'm referring to.

Feeding filterless DAC with SOX

Reply #14 –
Quote from: saratoga on
I'm assuming he means two separate physical DACs located on the same packaging, not two discrete packages.


Well, any stereo/multichannel DAC are multiple physical DACs located on the same packaging. :-)

Feeding filterless DAC with SOX

Reply #16 –
Quote from: giro1991 on
Sorry, see last part of my second post / reply and the products I'm referring to.


So basically, you have these two obsolete DACs which do not oversample and consequently will have problems with aliasing and non-uniform frequency response and want to know how high you will have to oversample in software to eliminate the problem?

The answer depends on the design of the two devices.  You'd have to get them, measure what the actual frequency response of each was, and then design an oversampling/equalization filter that compensated for it. 

My general advice in this situation would be to not bother.  Quality DACs are not expensive.  You will save a lot of time and get better results if you simply start with a product that meets your requirements.

Feeding filterless DAC with SOX

Reply #17 –
Quote from: giro1991 on
Separate channels for greater separation.


My EMU 0204 USB has -86dB of stereo crosstalk on the line-out, and -71dB on the headphone out with my Denon AH-D2000 headphones. That's quite enough! I tried to ABX a much higher value (-52dB, with the Sansa Clip+) with normal music, as well as with music with large stereo separation (Rock from the 60's), and for the life of me, I couldn't. When I saw my measurements of the Clip+, I really thought that would be problematic, but as often is the case, a simple ABX test set me straight.

Quote from: giro1991 on
I'd however like a basic 16 bit only DAC to reproduce as close to the original. Many products which state 16bit are '16/24bit' are normally 24bit chips capable of 16bit. I want purely 16bit.


Why?

Quote from: giro1991 on
I wish only for playback, as in, the sound a dedicated/standalone audiophile CD player unit would produce.


Again, my EMU does that, complete with a full 2 Volts on the line-out, on par with standalone CD players. Whatever problems you are referring to, they've been solved. I haven't used an actual CD player in over a decade.

Feeding filterless DAC with SOX

Reply #18 –
Just to clarify, there are a lot of misconceptions, misunderstandings, myths and raw ignorance about sound quality in audiophile circles. If you want to learn about audio, stick around, ask questions, and stay far away from audiophile sites like Head-Fi, Computer Audiophile, Steve Hoffman forums, etc…

Once you start to (somewhat) understand how (digital) audio works, and what actually matters, you'll be able to make an informed decision when purchasing audio gear. And once you realise that your (not-so-expensive) gear is so good, so far beyond audibility, you'll get peace of mind, daily aurgasms and be forever content without having spent your entire salary (or more) on some silly, awfully overpriced gear marketed by incompetent charlatans who prey on naive, uneducated customers looking for the audio nirvana.

That is not to say that finding a perfect, affordable product is easy, but they exist. Once you've learned a bit about digital audio and know what to look for, they're easier to find.

Edit: NwAvGuy's blog is a good place to start. He's the designer of the Objective 2 (O2) headphone amplifier and Objective DAC (ODAC), in an exercise to prove to audiophiles that excellent gear can be affordable, if well designed, even when it's not mass produced. Those products may not be strictly perfect in every way, but they're damn close. There are other excellent products that don't cost an arm and a leg (and much less than the O2/ODAC), you just have to find them, knowing what matters.

Feeding filterless DAC with SOX

Reply #19 –
Quote from: skamp on
Quote from: giro1991 on
Separate channels for greater separation.


My EMU 0204 USB has -86dB of stereo crosstalk on the line-out, and -71dB on the headphone out with my Denon AH-D2000 headphones. That's quite enough! I tried to ABX a much higher value (-52dB, with the Sansa Clip+) with normal music, as well as with music with large stereo separation (Rock from the 60's), and for the life of me, I couldn't. When I saw my measurements of the Clip+, I really thought that would be problematic, but as often is the case, a simple ABX test set me straight.



Agreed.

The audio world has been carried down the road to dual mono snake oil many times before. As usual the weak link in the chain is the irrefutable  fact that in the end, both channels are listened to by the same brain. A secondary link is bone transmission through the skull that provides a separation-robbing path between the l & R ear mechanisms. Then there is the slight matter that a live recording puts both microphones in the same room, and often within inches of each other or even touching each other.

Probably the best way to test the dual mono issue is to see how much separation loss can be added to a dual-mono recording (such as early Beatles stereo tracks) before it is noticed in an ABX test.  My recollection is that the effect is frequency dependent, but once you get to 30 or 40 dB separation, further increases in separation are not perceptible.

Feeding filterless DAC with SOX

Reply #20 –
giro1991:  You should probably explain what it is you want to do.  Drive speakers?  Headphones?  What type?  If you just say that and how much you want to spend someone can probably recommend what model of device would be appropriate for your goals.

Feeding filterless DAC with SOX

Reply #21 –
Thank-you for replies, I see what you both mean about the industry as a whole.

But what do we think about SOX? is Sox different from that of upsamplers found on/in hardware?
If sox were to up-sample 44,100 to 96,000, is the technique different (newer) than hardware up-sampling (which may or may not be software too). Will artifacts still be present if up-sampled to 96k only?

Regarding this image, is the sweep faster / more efficient than a 'typical over/re/up-sampler'.

Apologies for being stubborn here about this but i'm curious about sox.

Feeding filterless DAC with SOX

Reply #22 –
Quote from: giro1991 on
But what do we think about SOX? is Sox different from that of upsamplers found on/in hardware?


Yes, it has very little in common with typical upsampling used on sound cards, which is actually just zero stuffing + lowpass filtering.

Quote from: giro1991 on
If sox were to up-sample 44,100 to 96,000, is the technique different (newer) than hardware up-sampling (which may or may not be software too).


This isn't a conversion that is even possible on a typical sound card, so the question doesn't really make sense.  Instead, if the sound card actually supports 44.1k, its directly zero stuffed to a couple MHz and then played back.  If it doesn't support 44.1k, its upsampled by the CPU in software using whatever the driver specifies (probably 48k for cheap devices) and than that is zero stuffed to a couple MHz.

Quote from: giro1991 on
Will artifacts still be present if up-sampled to 96k only?


I don't understand the question.

Quote from: giro1991 on
Apologies for being stubborn here about this but i'm curious about sox.


Its a good resampler, but the answer hasn't changed from the thread you posted on this earlier in the year:  upsampling only makes sense if there is some specific problem with your hardware and a specific sampling rate.  If you hardware works fine, there is no advantage. 

Since you haven't explained what it is you're using or what you want to do, its impossible for anyone to give you a more specific answer.

Feeding filterless DAC with SOX

Reply #23 –
Ok. All the NOS DACs i mention feature 16bit chips capable of 44.1, 48, 96k only.

I'm interested in whether feeding a NOS DAC with 96K straight from Sox (which is re-sampling 44.1 to 96k) would be satisfactory.

Sampling theory states you have to up-sample x amount of times to eliminate imaging. But is SoX a different approach to sampling all together? seeing as its software based.


Thank you for you patience either way

Feeding filterless DAC with SOX

Reply #24 –
Finally a question that makes a little bit of sense.

Setting aside that it makes no sense to use a NOS DAC at all, if the DAC has an extremely good (read expensive) analog reconstruction filter for 44.1 ksps then its performance could be almost as good as that of an average, inexpensive oversampling DAC with its trivial analog filter.

At 48 ksps the analog filter wouldn't need to be quite as good, and at 96 ksps both DACs could use the same inexpensive analog filter.