 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
93.109.212.163
I2S can be tapped to give true 384k 32 bit output for ESS dac!
Not expensive either. Think I'll get one.
Follow Ups:
I remain curious about the discussion...
So this DAC can do 32/384. Cool.
Is there any music one can download that can benefit? It's hard enough getting 24/192 (mostly DVD-A rips for me).
Or is it interesting mainly because of the DIY potential?
I'm holding out for 64/768.
Since bigger numbers in high-end audio always sound better, this has to be the alpha and omega. I will just resample all my material to 64/768 on the fly and then I can use a capacitor for an anti-aliasing filter.
Wow, I say this all tongue in cheek. But, since the advent of the "race to 24/96" and "benefits of upsampling" some 30 years ago, there will always be those who believe that bigger IS better when it comes to bitdepth and sample rate - regardless of the resolution of the original material.
Me? I've upsampled to 24/96 and 24/88.2 and I'm not sure I'm not convinced it's worth doing *FOR ME* on my rig(s) with my ears.
I always conceed to the possibility of having excessive tin concentrations in my ears or insufficient resolving power in my speakers. Which is unlikely since they are phase corrected, time aligned and thus transient accurate - but hey, who can say... Just today I read that people over on the speaker asylum can hear the sonic degradation of bannana plugs. Myself, if I could hear such sonic degradation I would be hard-wiring from amplifier busses directly to the first component on the crossover boards - bypassing even a barrier strip if one existed on said crossovers.
But of course, the audio obsession is about nitty gritty, all of which is supposed to add up to transcend minutae status into something well worth doing...
Cheers,
Presto
'Me? I've upsampled to 24/96 and 24/88.2 and I'm not sure I'm not convinced it's worth doing *FOR ME* on my rig(s) with my ears.'
I am not surprised by your perception. Going from 44 to 88 smooth things out a bit but the improvement does not always grab you (execpt that there is less digititise). Going to 174 is a different matter and the improvement is much more obvious.
Use off line upsampling and discard preconceptions.
It's cheap enough to try setting it up to feed an ESS Buffalo dac directly to see how the two sound and perform. The unit can be fed from usb3 for headphone self powered use. The onboard dac does 24/192 only.
There is a facility they call Mulink which outputs via 4 wires at 32/384 which can be used to interface with a suitable dac.
There is some music on 2L.no but one can upsample 176.4/192k material for trial purposes.
Set up correctly, there should not be any issues with 192k DVDA. I rip them to wav first for pc playback.
http://exadevices.com/exaU2I/Overview.aspx...multichannel I2S
...comes fully isolated on the I2S side
...beside up to 32/384, even DSD is supported
...propriatary FPGA based data stream refresh/reclock
...requires propritary driver on W7people are raving about it.
Some trustworthy ( at least to me ) people consider it the best USB/I2S interface.
With a bit of PS tweaking it's supposed to perform extremely well.
Big disadvantage: W7 only. (Keep the fingers crossed that those guys survive - otherwise you'll get issues on driver maintenance)
Cheers
--------------------------------------------------------------------------------------------
::: Squeezebox Touch Toolbox 2.0 and more ::: by soundcheck
Edits: 10/08/11
Thanks, I know about it. But there is a waiting list and the Musiland costs only $150 including a 24 bit 192k dac.
It's going to be my first port of call to try.
Since a 44.1 KHz 16 bit CD album takes about 0.5 GB to store on a disc drive, that 384 KHz 32 bit album is about 16 times more memory, or 8 GB to store. That is a maximum of 125 albums for a 1 TB hard drive. As is the rule of mankind, give an inch and they take a mile.Believe it or not, most people in the world can't yet accommodate an 8 GByte download off HDtracks. With a 4 Mb/sec average download rate, that comes to approximately 4.5 hours. And most people will never now average that rate over 4.5 hours. But if HDtracks or elsewhere were to sell it on a memory stick (or flash drive or whatever) through the mail, maybe that would work.
-Kurt
Edits: 10/08/11 10/08/11
It's lossy and brought to you by the folks at Sandisk:
"That is a maximum of 125 albums for a 1 TB hard drive."
Let's say that these downloads cost $10.00 each, currently a low price for a 44/16 download. Purchasing these 125 albums is going to set you back $1250. I see that a 1 TB hard drive sells for under $65.00. So storage cost is not much of a problem.
Download time for 8 GB on my DSL service is 2 seconds per megabyte, which works out to 1.8 GB per hour. So it would be an easy overnight download, even assuming that the data wasn't losslessly compressed, which would reduce the size and time by about 40%. (FLAC can't compress 32 bit data, either integer or floating point, but Winzip 11 can.) Or if I were willing to pay $15 extra, I could up my download speed three times. That with compression would cut the time down to under one hour, entirely practical. I live in rural Vermont. If I lived in a major city more bandwidth would be readily available.
If you look at costs, you will see that Amazon would host the file on a server at a bandwidth cost of 0.12 per Gigabyte. This works out to about $1.00 per download if uncompressed, $0.60 if compressed. So you see that there aren't serious costs to the larger file size on the server end either. (These numbers keep going down. This is because communications technology is progressing rapidly. For example, Verizon recently announced that it was beginning the upgrade of its fiber optic backbone links to 100 Gbps technology.)
The costs of transmission and labor associated with manufacturing, loading, distributing memory sticks is much higher than these figures. For this reason, selling hi-res on memory sticks makes little economic sense, especially for low volume titles (such as these extreme hi-res would certainly be).
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
I'm picking up a 3TB Usb3.0 External HD for 140$ Canadian.
Storage is dirt cheap and just getting cheaper and better. Better being SSDs.
I recently downloaded "Low Country Blues" by Gregg Allman at HDTracks using
a Stelera HSPA+ wifi router at 21Mbps. Took all of ten minutes, 8 gigs yes.
I live on San Antonio Bay and must choose a 3G network provider as I can't
get DSL or other cheaper internet access.
Vista Ultimate 64 bit/e5300 Intel 45nm cpu/ASRock G41M-LE/Asus Xonar DS R 7.1/YamahaRX-V465 HT receiver/ Infinity RS1001 & Cambridge SoundWorks speakers/Yamaha YST-SW216 subwolf
Uhm... That's a pretty beast wireless connection. I thought at first you tried to say you downloaded 8gig in ten minutes over a plain 3g connection, was going to say that's impossible.. But a 4g Hspa network could cover that. Mind you service plans on those high speed wireless networks charge by the amount of Bandwidth you use. IE they sell you packages based on Megabytes you can download. My phone provider includes a 500megabyte package bundled with my normal service for about 10$ ( This is on a three year old contract, but prices are still high. )
So on top of paying for your album, you payed what I'm assuming is a sizable amount just for the bandwidth..?
I have a wired Cable connection topping out at 3.7mbps download speed. The connection comes with a monthly 450gig download limit I believe. Maybe even just 400gig. And it's not cheap.
*shrugs*
upsample yourslef to move brick wall filter as far away from 22k as posiible. Increasing bit depth also enhances performance of modern dacs significantly.
You won't do your veiwng at 16 bit on a monitor, or will you?
"upsample yourslef to move brick wall filter as far away from 22k as posiible. Increasing bit depth also enhances performance of modern dacs significantly."
Brick wall is on recording. There is no way to remove it, as the information is already gone. By using tailored filters and upsampling one can reduce ringing but this will come at the expensive of rolling off high frequencies. Preringing (if it's on the recording) can be eliminated by going to a linear phase filter but only if there is huge attenuation at the ringing frequency, i.e. more high frequency roll off.
Increasing the bit depth without resampling will do nothing, it does not and can not add any information, just adds a bunch of zeros to the end (32 bit integer) or a byte of exponent information (32 bit floating point). If one resamples, the filter operation does add extra bits but these don't represent any new information, just extra bits that need to be handled one way or other, or else the conversion will have lost information.
It should be clear that taking a 44/16 recording and processing it does not differ whether this is done in the computer or in the DAC. In both cases the same information is available. If doing this in the computer sounds better it is because (a) the audiophile has chosen a better algorithm than the DAC or (b) the processing done in the DAC in real time is somehow degrading playback. In both cases, the conclusion is that any differences can be attributed to faults in the DAC. There is no possible magic that will produce better sound than is in the original 44/16 file. If it were really possible to eliminate the problems with 44/16 audio by upsampling then there wouldn't be all the grousing about "fake hires" downloads on some sites.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
get out of the bits and software mind set and look at the performance of modern dacs fed with 16 bit and 24 bit data. Have a look at some valid test data.
As for the playback brick wall filter, with the right dac such as a Buffalo, this, and the subsequent analog filter, can be pushed right out towards hf.
Fred, I'm about at the point where I am not going to waste any more of my time replying to your incorrect posts. I know there is no chance in hell that you will change your opinion. I make these posts only so that others won't be mislead. However, it's gotten to the point where most inmates have already figured out the situation, so there is little point in continuing.
It is no more possible to restore missing high frequencies from a 44/16 recording than it is to extract the sound of "The Lost Chord" in hi-res digital through the use of a Ouija board in together with the Edison cylinder.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
No one is claiming restoration of high freq.
Your posts was directed purely at bits status and not the bahviour of dacs, of which I have deduced that you are not prepared to accept ie that a modern dac performs much better fed with 24 bit data than 16 bit, whether these are zeros or ones.
As it happens I usually feed my DAC 24 bits, even when playing CDs without any upsampling. But I am willing to accept that the extra zero bits that you create are better than the extra zero bits that I create, if this makes you happy.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
It is not the bit value but the bit depth fed into a dac that affects particularly low level performance. 16 bit data carries high distortion at low levels, from 1 to 4% or thereabouts.
I am not surprised that you can't hear it with an integrated sound card.
"It is not the bit value but the bit depth fed into a dac that affects particularly low level performance."
Perhaps you should look at how DAC chips actually work when they are fed 16 bit data vs. 24 bit data. For starters, try the I2S specification:
"Serial data is transmitted in two’s complement with the MSB first.
The MSB is transmitted first because the transmitter and receiver
may have different word lengths. It isn’t necessary for the transmitter
to know how many bits the receiver can handle, nor does the
receiver need to know how many bits are being transmitted.
When the system word length is greater than the transmitter word
length, the word is truncated (least significant data bits are set to ‘0’)
for data transmission. If the receiver is sent more bits than its word
length, the bits after the LSB are ignored. On the other hand, if the
receiver is sent fewer bits than its word length, the missing bits are
set to zero internally. And so, the MSB has a fixed position, whereas
the position of the LSB depends on the word length."
"16 bit data carries high distortion at low levels, from 1 to 4% or thereabouts.
I am not surprised that you can't hear it with an integrated sound card. "
Please cut out the gratuitous insults.
It is very easy to hear the difference between different methods of creating 16 bit PCM data from higher resolution data. These methods represent different tradeoffs between background noise level, background noise spectrum, low level distortion and noise modulation. Each of these types of noise or distortion has a characteristic sound and is readily audible to the trained ear if music is played at a loud volume, particularly when quiet portions such as tail reverberations are turned up. The choice of dither algorithm (or lack thereof) is made in the production process and affects the sound of the resulting CD. There is no question or need of throwing bits away and no way of getting them back, either.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Your post typically talks about I2S, specs and downsampling which has nothing to do with what I'd been saying.Do this. Play a 16 bit 1 or 10 kHz file and use a good FFT to measure harmonics at -60 and -90dB ref 2V. Do that for a 24 bit file again. See what you find. Look also at the reconstituted -90 dB sine wave. The former looks like what?
The point about 16 bit and the digital filter is that you can only get good transient response at frequencies of the order of 1 kHz or less.
Edits: 10/09/11
This matter of 16 bits is rather elementary. One can generate a clean sine wave at 24 or 32 bits and take a spectrum and verify that it is clean. One can also see the noise floor. One can then convert this 24 or 32 bit file to 16 bit (at the same sample rate so as not to introduce other variables). There are various options that basically boil down to three strategies:
(1) truncation. This just takes the high order 16 bits and throws away the low order bits. If one does this one will see harmonic distortion in the output spectrum. The percentage distortion will be small with loud sine waves and huge with quiet sine waves. Sinewaves below about -96 dB will be grossly distorted, i.e. the sound will be totally gone. If one creates a sound track with the sound slowly dieing out one can hear this entire process, a relatively clean sound gradually decaying into an increasingly distorted sound and suddenly disappearing. Once one has heard this with a test tone and knows what it sounds like, one can move on to actual recordings where a similar effect occurs at the end of a track where the music gradually dies out as room reverberations dissipate.
(2) TPDF dither. This adds a special random noise pattern to the original 24 or 32 bit data and then rounds the resulting signal to 16 bits. Doing this adds noise to the file (e.g. takes the broadband noise floor from roughly -96 dBfs to roughly -90 dBfs). However, when the tests above are repeated the harmonic distortion will be gone. Through more sophisticated measurements one can show that the distortion components are not just masked by the added noise, but are actually gone. This can also be easily heard by repeating the listening tests mentioned above. For example, one will hear the diminishing tone gradually die out without gaining any distortion. In addition, instead of the tone abruptly disappearing the tone will remain clearly audible at very low levels, perhaps as far down as -115 dBfs depending on volume, quiet in the listening room, etc. Similarly the tail reverberations at the end of a recording will dissipate naturally rather than suddenly disappearing (assumes the recording was competently made and these were not simply chopped in the production process).
(3) Noise Shaped Dither. There are a number of proprietary dither algorithms that attempt to do a better job than TPDF. To the extent that they succeed, they achieve lower (audible) noise by reducing the approximately 7 dB noise penalty added by TPDF dither. Because it is mathematically impossible to remove noise without adding bits, all these do is to shift the noise to frequencies where it is assumed to be less audible. Generally, these methods improve sound, but not always. They are such that they can only be used once. If noise shaping is applied to a recording that has already had noise shaping then the results are likely to be audibly inferior, sometimes the problems are gross, not merely subtle.
All the necessary test signals, spectrum analysis, etc., are readily available and built in to any audio editor program, including free ones and ones which come with free trial downloads. To hear these effects requires nothing more than a 24 bit sound card and a decent amp/speakers or headphones. At least in the past, however, some of these editor programs (including those used to produce commercial recordings) did not do a very good job of implementing either TPDF or noise shaped dither. So this is a matter of caveat emptor . The conversions in iZotope RX2 and Soundforge 10c are satisfactory. If one doesn't want to play with an audio editor and generate various test files there are test disks that one can purchase. In either case, it is easier to hear these effects by turning up the volume. If one does this, it is wise to take care.
I don't believe the transient response of a filter is affected by the bit depth output, it's a function of the design. However, a 16 bit file ends up at least 24 bits after being filtered and there will be unnecessary noise/distortion added if this is then chopped before it is sent to the DAC. This provides an argument for using a 24 bit DAC with 16 bit recordings, and a 32 bit DAC for 24 bit recordings, but at some point one runs into cost vs. diminishing returns. Good digital filters involve the use of 48 bit fixed point arithmetic or 64 bit floating point arithmetic in the internal calculations. This is not terribly expensive to do these days, either in special purpose chips, FPGAs or software.
If one engineers digital recordings it is absolutely essential to be familiar with the above concepts and to recognize audible distortion caused by various ways of misusing one's tools. In addition, some understanding of theory may be useful, particularly if one is devising experiments to evaluate digital audio components and software. Ideally, one should use a combination of measurements and listening tests.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
something else altogether which has no practical bearing on how dacs react to signals. I suggested that you did a simple test on your dac. Instead you talk about dither and whose functions and effects are well known. Conducting experiments with iZotope is, pure and simple, a software 'thought' experiment, but not one in Einstein's footstep..
It is obvious that there is no point in me spending further time on the subject; your posts are, I am afraid, just hot air.
Are you are saying that there is a difference in sound quality between sending a 16 bit file to your DAC vs. sending the same 16 bit file in 24 bit format with the low order 8 bits all zero?
Are you further saying that these audible differences are significant, in that they affect the enjoyment of music?
Are you further saying these differences are measurable as well as audible?
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
have a read of John Atkinson's test methods. What I suggested you do is simple enough and you will be able to see for yourself and learn something about dacs on top of you knowledge of bits of software. Do not dither the test signals and you can compare the reults with dithered ones of any description.
Fred,
Please help me out. I asked you three questions that could have been answered "yes" or "no". Instead you browbeat me with a suggestion that I do not understand. I have no way of considering your suggestion until I understand it. I can't even tell if I have already done similar tests and if/how these might be modified.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
If you simply carry out the tests I outlined, which are not difficult, you will understand clearly what I said.
Answering your questions the way you put them will simply lead to more circumlocution from you and more comments that have no relevance. This is quite clear from the above exchanges.
By the way, I would hope those who do not understand the technical aspects of digital audio appreciate the extra effort that you put into your posts.
The USB is only supported on PC, should have expected it to with USB3 capability.
Hopefully Mac drivers and other outputs like TRS or XLR may be released by
George
It's available from China on fleaBay :-)~160 each.
Would love to grab one and try the USB3 but not sure if I can justify yet another toy :-). Also WHERE'S THE MUSIC at such sampling rates???!
Edits: 10/07/11
How to you plan to cable the I2S and terminate it?
Steve N.
Haven't thought that far, evaluate first. Thet also have a mysterious 384k Mulink output board for some of their soundcards. AT their prices, I am also going to try out.
The problem with Musiland is clear literature. I also find your site descriptions condusing.
To make this exercise worthwhile it going to be desirable to import the clock that's in the I2S DAC to the USB/I2S converter, i.e. the clock will have to go from the DAC to the Musiland and the data from the Musiland to the DAC. Depending on the clock rates, this may not be straightforward, either in terms of modifying the Musiland or in terms of getting the DAC data synchronized to the DAC clock due to transmission time on the cable.
If the clock signal goes in the opposite direction then any jitter on this line is going to make it into the DAC circuitry. (How much this matters to the output is debatable as the SABRE chip has a good ASRC capability. I don't know if ASRC is enabled in fmak's DAC.)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
I read that the Japanese and Koreans are already doing it; no details yet
My point is that at 384kHz SR word-clock, the master clock frequency is 98.304MHz. Good luck getting this down a cable with any signal integrity.I am able to get 192KHz SR master clock, which is 49.152MHz down a cable cleanly, but I have a lot of experience with transmission-lines.
Steve N.
Edits: 10/07/11 10/07/11
They make a card called Mulink, which provides a 4 wire 32/384 output. There is no information on it but at, the price, I'll probably try one.
"My point is that at 384kHz SR word-clock, the master clock frequency is 98.304MHz"
I don't see the need for such a high clock rate. The total audio data rate is 384000 x 32 x 2, roughly 25 million bits per second. This figure would apply to three wires (clock, word select, data). If four wires are used (clock, word select, left, right) the rate would be 12.5 million bits per second. This is not a high data rate.
Why are you suggesting a 256x multiple at this sample rate? Most of the DAC chips I've looked at use lower multiples when running at the higher sampling rates.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
do it all the time.
http://hifiduino.wordpress.com/2011/08/01/musiland-monitor-03-us/
Click the link
Cut-Throat
.
With Firefox, triple click on the text that is the URL and then right click and select the menu item "Open Link", "Open Link in new Tab" or "Open Link in new Window" as desired.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
There is zero extra work for a poster to make the referenced site an actual link vs pasting it in the comment box. By pasting in the Optional Link URL: section instead, readers can just click on the link (one step, not a threellion) to get to the referenced site.
[Yes, I know. Kinda like printed cheese, n'est-ce pas?]
Different posters have different priorities. Some don't give a damn if their saving one mouse movement results in 1000 :-) readers having extra work. Other posters are clueless or careless, rather than selfish or thoughtless. If one previews one's posts and checks all the links as considerate posters do, then one is saving oneself time as well.
I use these little niceties as clues when evaluating inmates as to their attitude and competence. Other clues include the care in editing, typos, quality of written English, etc... I cut newbies and non-native English speakers slack.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
"Some don't give a damn if their saving one mouse movement results in 1000 :-) readers having extra work."
is Fred clueless, careless, selfish or thoughtless because he can't paste a string of characters onto a form?
To be honest, the great majority of the time I don't visit a referenced site if it's posted as a string of characters instead of a link.
Thanks Fred,
Looks to be discontinued, at least at Pacific Valve.
Rick
It's new and just out, try eBay (UK)
Post a Followup:
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
