• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

TPA - USB Transport

Status
Not open for further replies.
Member
Joined 2007
Paid Member
There is good information. I only object to the absolute tone in the first article.
Agreed. The commentary on 'the golden ear' is naive. If the parameters of bandwidth and dynamic range were the only important features of hearing, mine would be normal! But many nights happily spent with noisy woodworking tools have slightly damaged things, resulting in a heightened sensitivity to interference and noise. Not the ideal way to develop a 'golden ear'. :p

To your point about greater bit depth and sample rates - whether they matter theoretically or not, why do so few redbook recordings equal the SQ of HD files? I do have some - precious few... Two well-mastered recordings I use for reference are Ray Charles with guests Gladys Knight and Van Morrison. The 96/24 versions flat out sound better than the same masters at 44.1/16. I can't control every detail of why or why not - just give me what works with my room and 'lectronix!

Cheers!
 
I do agree to a certain extent that higher bit rate could be better, leave more headroom. I have plenty of recordings from HDTracks, Linn, etc... Unfortunately, many people buy HD music only to realize that it sounds the same as 16/44.1. There are many reason for this, including equipment not able to render such recordings, master was recorded in much lower bitrate and then resampled (seen many of those and it is a shame), etc...

If you take good material like the TBM sounds in CD 16/44.1 you would have a hard time believing it is a CD. There is so many parameters involve in the process that many times 24/192 and 24/96 is just a seller's propaganda and this is too bad since we only find out when we bought the album and it is too late. Go on the Elusive web site and you can find some amazing CD transfers from FIM and many more.

My comment was aimed at 384 and 768 samples which I could not see any use for... I have some 384 recordings from 2L and they sound exactly the same as their 24/192 counterparts... And I do have some really good gears to play them.

Yes, the article can be taken with a grain of salt since we don't know what is the author's position on digital audio. Is he biased?

Do
 
The article on Xiph.org is written by Monty who worked with Ogg Vorbis which is like MP3. Seems the article is written for non-audiophile non-professional reader or computer users. I don't even consider MP3 or anything that uses lossy compression listenable. I don't own a portable MP3 player device.

Since the Xiphg.org article came many instigators reference the article just to stir up debates on every forum. I would absolutely believe those like Rob Watts who have done extensive research for over 20 years and make highly regarded products. Or Daniel Wiess too. Even thought their products seem overpriced, they still sound good.

And nothing wrong with longer word lengths. Makes it easier to get great dynamic range and low level detail without compromise. And as Barrows wrote, storage is dirt cheap these days.

" Unfortunately, many people buy HD music only to realize that it sounds the same as 16/44.1. There are many reason for this, including equipment not able to render such recordings"

I can hear the difference on my notebook computer and it isn't anything special. I've emailed people music clips who are not audiophiles, and every single person could hear the difference between 192 and 44.1 versions of the same recording. I think a lot of regular people who don't think there's a difference haven't listened. And some couldn't be bothered even if there is a difference.

The biggest problem I see with HD material is lousy production. The record companies are ruining their own market by releasing upsampled or other crap of reduced quality.

As far as 768? I saw the impulse response comparisons between various sample rates which means there's some difference. So why not include it in a USB to I2S interface? DSD64,128,256 why not put it in there for experimental reasons. There's plenty of bandwidth available. FPGA chips can do everything now and there is processing power and storage are abound these days.
 
Last edited:
Impulse tests...

I have seen some of these, thay are often used to demonstrate the superiority of DSD to lower rate PCM. The tests I have seen pretty much show that by the time one gets to 24/176.4, there is no need to go much further.
Additionally, the tests I have seen appear to be based on a single sample impulse, at full scale. I very much doubt that there is any musical event which produces a full scale impulse in one sample of time, even at 44.1, right? I speculate that any actual musical event's real impulse could be fully represented even by 44.1, as I just find it very hard to believe that any real world impulse reaches full scale that quickly. Given an impulse lasting a few samples, all these rates would look equivalent, right?
 
Impulse_response_comparison_zpsd1e40551.jpg


Source: Signal processing for Direct Stream Digital - A tutorial for digital Sigma Delta modulation and 1-bit digital audio processing - authors Derk Reefman and Erwin Janssen (comments and proof reading S. P Lipshitz, J. Vanderkooy, J. D. Reiss and H. ten Pierick).

I think this article is also interesting, although I believe most of you here have already read it.

Cheers.
 
Additionally, the tests I have seen appear to be based on a single sample impulse, at full scale. I very much doubt that there is any musical event which produces a full scale impulse in one sample of time, even at 44.1, right? I speculate that any actual musical event's real impulse could be fully represented even by 44.1, as I just find it very hard to believe that any real world impulse reaches full scale that quickly. Given an impulse lasting a few samples, all these rates would look equivalent, right?

I am trying to understand that impulse response also.

Imagine several instruments with fundamental tones and harmonics. It may seem uncongest with a Fourier frequency plot. Now plot all that in a amplitude vs. time graphic. Pretty agressive peaks and dips.

At each sample the hardware need to switch to a different power level. It seems easy with only one fundamental tone. It is not the with several tones. I also believe that, when playing all that tones at the same time, the hardware does not reach full scale, because hardware may not be able to reproduce the transients between samples (how many current and voltage it can deliver per unit of time), which may lead to harmonics that were not in the input signal (intermodulation).

I believe the same reasoning may be used to understand square waves, isn't it?
 
Sorry for polluting the thread. My post was not comprehensible. :eek:

I wanted to say:

I also believe that, when playing all that tones at the same time, the hardware does not reach full scale.

But the DAC hardware may not be able to reproduce some transients (it depends on how many current and voltage it can deliver per unit of time).

This may lead to harmonics that were not in the input signal (intermodulation).

There is already an slew rate specification for sigma delta DAC's. So I think at least in the audible range the hardware is able to reproduce the transients, even with complex tones.

I will keep my mouth shut...
 
I know that some knowledge is secret, but is there a thread to discuss the ADC and DAC technologies, integrated circuit topologies, algorithms, filters etc.?

I just wanted to post this presentation made by Bob Adams from Analog Devices, which is also really interesting: http://www.cscamm.umd.edu/programs/ocq05/adams/adams_ocq05.pdf

There are several frequency domain graphs comparing one bit and multi-bit sigma-delta DAC's, thermal encoding, scrambling etc.

If I use all foil resistors in a ladder DAC physically coupled to a "stabilization thermal oven" is it possible to dispense thermometer encoder and scrambler?

I know Sabre must do that better within its integrated circuit, but I am trying to understand what are the pros and cons of doing it discrete (for instance, having the discrete DAC in the voltage domain).

I would like also to understand why some people prefer a DSD 1bit, 128x file for distribution (there must be a reason why Korg says it is a future proof master recording file).

Anybody knows someone who explains why it is "easier" or "less harming" to convert 1 bit, 128x (5.6MHz) sample rate distribution file to a 6 bit, equivalent sample rate ready to conversion stream than convert a 16 bit or 24 multi-bit with 1x (48KHz), 2x (96KHz), 4x (192Khz) or 8x (384KHz) sample rate distribution file to a 6 bit, higher oversampled rate ready to conversion stream ? I mean an explanation without complex equations.

This could also explain what would be an waist of space to use the raw output of a multi-bit sigma-delta ADC, since the 1 bit, 128x is a smaller file that may already solve the major problems. Then I would give up on the "raw rant" I have done in other threads... :eek:

I know that I am usually trying to understand things (digital filter theory, mathing resistors etc.) that one would simple does not get it without an engineering degree, but perhaps I learn something from posting and reading here... :usd:
 
Status
Not open for further replies.