There's no such thing as digital...

[richidoo]

http://www.audiostream.com/content/draft

An email conversation with Charles Hansen, Gordon Rankin, and Steve Silberman, hosted by audiostream.net, (Mike Lavorgna of Stereophile)

[rollo]

  That was a surprise to read. No ones and zeroes eh ! Analog square waves very interesting indeed Ollie.

charles

[JBNY]

I always find this discussions interesting. While I think you can say that with SPDIF you indeed have a bitstream, which is a sequence of bits being delivered at precise points in time, is very much like an analog signal and having any shift in the timing of the bitstream causes jitter.

But there the similarities end. Async USB or Firewire do not use a bitstream, but rather packets of data that are sent to the DAC when requested. This kind of communication is what networking devices have been doing for a long time, and networking jitter is one of those issues that was solved quite some time ago. Timing issues pretty much go away. As long as the data arrived uncorrupted and in time to fill the DACs input buffer (which should be trivial), the data is indeed "bit perfect".

Digital is a word coined by the audio industry, it's just data transfer now. no different than copying files or transmitting email. Packets is packets.

[richidoo]

Joe, with ASync USB, isn't it the USB receiver doing the asking for data, then sending a traditional stream to DA chip? Are there DA chips that can network? How does that work?

That SPDIF is really just an analog square wave explains why impedance makes such a difference. Ringing transients probably affects the sound too if they are extreme. 

[tmazz]

Joe, with ASync USB, isn't it the USB receiver doing the asking for data, then sending a traditional stream to DA chip? Are there DA chips that can network? How does that work?

That SPDIF is really just an analog square wave explains why impedance makes such a difference. Ringing transients probably affects the sound too if they are extreme. 

Rich let's keep in mind that as Dr. Fourier explained to us, complex waves (of which a square wave is one) are nothing more than the sum of a series of sine waves. Ringing in a square wave is most often caused by a restriction of bandwidth in the medium transmitting the wave. The bandwidth required to accurately transmit a square wave is actually many multiples of the fundamental frequency if the square wave itself to allow the passing of all the harmonics that actually do the "squaring" of the wave.

See http://en.wikipedia.org/wiki/Square_wave for an interesting article on square waves.
In particular look at the section called "examining the square wave" to find a nice animation that visually shows how the ringing in a square flattens out as the bandwidth of the signal us widened out to include more harmonics.

So any square wave, whether it is packetized or not, is fundamentally analog in nature and any transmission or signal processing that is not linear with respect to frequency can cause distortions that will cause the signal to depart from being a "clean" square wave.

[JBNY]

But the thing to keep in mind when you say that you still need  a "clean" square wave. When it is in packetized form, the pulses are sent make up a packet, think of it like a zip file. When put together it forms a little file, the packet. Inside the packet is data such as when it was sent, what the data contains, how many packets are in the group and what number it is, as well as CRC info. Regardless of any noise, or ringing etc, if the packet does not pass the CRC check it is considered corrupt and discarded and either re-transmitted by the bus or error corrected by the DAC. This is the main difference to the bit stream, the bits must arrive at a precise point in time. But the packets can arrive whenever, as long as when assemble in the right order they are in the buffer when the DAC needs it.

As to how the DAC processes the data from the input buffer, that is up to the DAC manufacturer. Mistakes could obviously be made there, and just because the the data made it from the transport to the DAC is no guarantee that is will sound great. So all DACs will not sound the same, it's just that one piece of the puzzle we will finally be able to worry about less.

But this idea that we have been hanging onto that noise on the line, for whatever reason, can cause square sign waves to be misinterpreted is gone, if the packet does not pass the CRC check it is flagged as corrupt. I'm not saying that it doesn't happen, I'm saying that now the unit will know when it happening, and can act accordingly.

[tmazz]

But the thing to keep in mind when you say that you still need  a "clean" square wave. When it is in packetized form, the pulses are sent make up a packet, think of it like a zip file. When put together it forms a little file, the packet. Inside the packet is data such as when it was sent, what the data contains, how many packets are in the group and what number it is, as well as CRC info. Regardless of any noise, or ringing etc, if the packet does not pass the CRC check it is considered corrupt and discarded and either re-transmitted by the bus or error corrected by the DAC. This is the main difference to the bit stream, the bits must arrive at a precise point in time. But the packets can arrive whenever, as long as when assemble in the right order they are in the buffer when the DAC needs it.

As to how the DAC processes the data from the input buffer, that is up to the DAC manufacturer. Mistakes could obviously be made there, and just because the the data made it from the transport to the DAC is no guarantee that is will sound great. So all DACs will not sound the same, it's just that one piece of the puzzle we will finally be able to worry about less.

But this idea that we have been hanging onto that noise on the line, for whatever reason, can cause square sign waves to be misinterpreted is gone, if the packet does not pass the CRC check it is flagged as corrupt. I'm not saying that it doesn't happen, I'm saying that now the unit will know when it happening, and can act accordingly.

I agree with all that you are saying, but identifying the bad packets is only half the issue. Once a packet is flagged as corrupt the unit needs to do something to correct for the bad data, whether that be packet retransmission or just some kind of interpolation. And pretty much, the less correction the DAC has to do, the better off we are. So anything that is done to insure cleaner transmission and less errors that need to be corrected will only help us have a more faithful final output signal (hence the improvements gotten from "better" digital and USB cables.) While it is good to know where the errors are, it is better to minimize the number of errors that exist in the first place.

[Werd]

Well i know what they are getting but dac outputs have been described as quasi digital. This is due to the imperfect sine waves. As you get higher bit depth the analogue output becomes more sine. Filters are used to smooth analogue after the dac before output to compensate for lower bit depth.

[JBNY]

So anything that is done to insure cleaner transmission and less errors that need to be corrected will only help us have a more faithful final output signal (hence the improvements gotten from "better" digital and USB cables.) While it is good to know where the errors are, it is better to minimize the number of errors that exist in the first place.

So we are pretty much in agreement, but the part I still think that I don't agree with is the USB cable part. If the cable is up to specs, then the cable should not be causing issues at all. All I am trying to get across is the computer with async USB or firewire is a different animal than SPDIF. As soon as the tech is a little more more mature, I think that is will be very difficult to tell the difference from a $5000 audio PC and a $75 one. The PC as a transport will become irrelevant and the differences will mostly be at the DAC implementation level. YMO

[Werd]

^^^^^

I am hoping the PC transport will be less of an issue in the future generations but i doubt we are any where near that. I don't consider normal data transmission the same as - real-time transmission into a dac that sits infront of an audio system amplified into a pair of speakers.

IOWs we hear the transmission amplified. For eg. I can hear a noticeable drop in noise when i put my WD (usb powered) 1.5gb HD on my Symposiun ultra platform opposed to just laying on top of my BDP player. I do not know what the micro effects on the data might be but i can attest to the drop in noise for sure. That is just moving a small portable HD onto a different sitting environment.

USB cabling for sure when the HD is plugged in. In fact a bad usb can make  an adaptor plugged  HD unlistenable due to all the treble clang.
But I am hopeful like yourself that the next mass usb and HD generations are audio friendly.

[Carlman]

I had to re-read the article again.. He's not saying there are no 1's and 0's, he's saying that is how a difference in voltage is represented.  I have always wondered about this because how can all the different sounds really just be 1's and 0's?  It's all relative! That's how... that fundamental phrase really explains a lot.

We over simplify the digital to analog process to market it and argue about it...

I found this statement on AQ's site while researching their little Dragonfly DAC:

Remember that digital audio is stored on computers and delivered to DragonFly as streams of 1’s and 0’s.

Really? They're doing what they're saying isn't the case... Stupid marketing...

But then they go on to say:

Making beautiful music out of 1’s and 0’s isn’t a case of simply getting all the music data from point A to Point B. Maintaining subtle digital timing relationships is crucial in order to be able to reconstruct the analog waveform that we hear as dialog or music.

More consistent with what was in the article... and somewhat redeeming.

I have since decided the HRT Music Streamer may be a better option for me.. sounds a little better and costs a little less... win-win.

[tmazz]

I have since decided the HRT Music Streamer may be a better option for me.. sounds a little better and costs a little less... win-win.

I have always liked HRT products. A very goo bang for the buck company. 

[bhobba]

Yea - read it.

All true.

Digital is not 1 and 0's it's a changing voltage from which digital is extracted.  For DAC's that also includes timing errors.  This is why USB cables, for instance, can, and do, sound different.

Thanks
Bill

[JBNY]

Timing errors do not happen like that with USB though. the 1's and 0's are part of a packet. if the voltage is off slightly and the USB receiver misinterprets the 0 as coming in slightly later it does not matter as long as the rest of the 0's and 1's come over in the proper order. 

People keep thinking that the 1's and 0's come over and are immediately translated into voltage, so that if the timing of one of the 0's is off by a few milliseconds the voltage to the DAC is also off and you get jitter. That is how is works with a bitstream, but not USB.

With USB you get a packet. A string of 0's and 1's like so 01100010001000111100010011000100. The USB receiver receives all of them. Once it does it reads it as a packet, think of it like an envelope, and it is considered one entity, and inside it are the various contents of the packet. One part is the CRC check, that tells it if the packet is the same as what left from the other end. Once that is verified the rest of the contents are 100% correct and read as well. So then the packet tells what part in the sequence it is, say 45 of 890 packets coming through. Then the voltage number and then the timing of the voltage. All that info is then held in the buffer and handed out to the DAC when needed. (This is a very simplified explanation)

If the CRC fails then the packet is discarded (really up the manufactured what they do). But this idea that voltage timing is changing the sound on USB is not correct, it will only change the sound if you have so much noise on the line that the packets are continually getting corrupted on the way to the USB receiver, at that point the DAC tries to error correct for the missing data. But if that is happening something is very wrong, as the USB protocol is very robust and has lots of ways to try and avoid that from happening.

[bhobba]

Timing errors do not happen like that with USB though. the 1's and 0's are part of a packet. if the voltage is off slightly and the USB receiver misinterprets the 0 as coming in slightly later it does not matter as long as the rest of the 0's and 1's come over in the proper order.

In saying its timing errors and you need to think of it as an analogue signal that is obviously an over simplification.

What is going on is quite complicated.  For example one of the simplest things you can do with a USB cable to make it sound better is to have a separate power cable or even better remove it entirely - but some DAC's may not work if you do that.  The reason is junk from the power supply contaminates the data and via the power wire makes its way into the usb converter and/or DAC where it can cause problems.  You will find posts by genuine experts like Steve Nugent (he is a digital engineer of impeccable qualifications who holds many key patents in the design of the Pentium since he was a team leader in the group that designed it) explaining exactly whats going on and it is very subtle and complex.  Even now, if you go over to his Empirical Audio forum on Audio Circles you will find he is discovering new stuff all the time.

For guys like us all we have to do is listen - not get caught up in technicalities.  At the end of the day that even applies to guys like Steve Nugent.

Thanks
Bill