Does a file with jitter keep the jitter when saved to a hard drive?

BlueFox · Jun 9, 2014

I have a question regarding jitter. Since jitter is being defined as a timing error as to when the leading or trailing edge of a digital signal occurs, then it would seem that as a byte with cable induced jitter is saved to a drive then that saved byte would now have the jitter.

For example, this diagram shows how the reconstructed byte has jitter. The new byte has bits that are not at the exact point they were when the byte originated. Since a digital music file is composed of numerous bytes, why doesn’t the jitter exist in the saved file since it obviously has jitter introduced? How can saving a file eliminate the timing errors introduced by the cable, or from other factors? It would seem that the system would just write the byte as it occurs to the drive. How would it know to correct the timing errors introduced into it?

The reason why I ask is some say jitter can be introduced by any link in the chain. Others are saying jitter is eliminated everytime the file is written either into RAM or onto the drive.

garylkoh · Jun 9, 2014

No, jitter is not kept when saved to a hard drive (SSD or spinning drive). That is not how digital works. Jitter only appears only when it is transported.

However, we have had experiments where jitter can be encoded on to a CD. This is because the pits and lands do not have well-defined edges, and the dimensions are not consistent due to errors in the writing mechanism. Nevertheless, this error was still well within the orange book spec, and theoretically not measurable when read. This is the basis of Yamaha's patent for their AMQR process.

amirm · Jun 9, 2014

As that data moves from one storage device to another it does pick up "jitter" on the way. But that jitter gets eliminated once the data is captured digitally and stored. The audio samples do not change. And we have no use for its timing.

The only time jitter comes into play is when the target is analog which is the case with DAC clock.

So no, there is no worry about jitter getting added to a file. Indeed you can do a binary comparison between the original and copied file and they will be identical.

elcorso · Jun 9, 2014

According to Jussi Laaka Signalyst’s HQPlayer music player creator, a digital engineer with full credentials, embedded jitter in musical tracks is possible, from the recording process.

If I take this as a true, I don't see why can't be copied to an SSD or HD.

To me, the interesting thing is how each digital player and / or DAC handle this jitter to be (almost) inaudible.

Discuss this can result in a long and tedious discussion, but the question is valid.

Some other discuss about two 'identical' files with identical checksums (then exact bit by bit) could bring a different SQ.

Roch

BlueFox · Jun 9, 2014

I understand the conventional feeling on this subject, but I can't grasp how timing errors are not preserved on a HD, SSD, or even ram. As long as the errors are within spec the CRC, or checksum, will be correct, but the music will not be 100% accurate.

My personal feeling in this area is that as digital music, especially high res, becomes even more popular we will start seeing new discoveries in this area. We might even find that a file downloaded from HD Tracks is slightly different from the file on their servers.

When it comes to music reproduction I learnt a few years ago there are no absolutes. Everything matters, as the saying goes

DonH50 · Jun 9, 2014

If you have already made up your mind you know the answer, why ask?

Timing errors made during the recording, at the ADC, are embedded in the digital data from that point forward.

As others have said, although jitter is technically added to the edges each time a transfer of data is involved, it would have to be huge and ERC would have to be exceeded for any bit errors to be generated. The digital bit stream can tolerate a LOT of jitter and so long as the files as bit-perfect (i.e. normal) nothing is changed. Jitter on the edges is not preserved in digital transfers; the edges are not used. Digital data is sensed in quadrature; the bits are sensed in the middle of the bit period, well away from any edges. Draw a line in the middle of the bits in your first image; that is where data is recovered. The next opportunity for audible changes is at the DAC where the clock is applied to create an analog signal. Clock jitter can cause errors, but asynchronous designs isolate the DAC clock and the audibility of jitter even when the clock is derived from the bit stream is a topic of considerable debate.

amirm · Jun 9, 2014

elcorso said:
According to Jussi Laaka Signalyst’s HQPlayer music player creator, a digital engineer with full credentials, embedded jitter in musical tracks is possible, from the recording process.

If you provide a link to what has been said, I can comment.

If I take this as a true, I don't see why can't be copied to an SSD or HD.

Digital storage is just that: digital. It cannot by definition store analog values. Jitter is an analog timing variation. It can't be stored in the digital storage device. There is no room to put it there.

As I explained, "jitter" does exist in digital domain when we transmit one bit from one place to another. A wire after all is an "analog" device. Assuming your digital systems are working the way they are supposed to, the data gets captured despite existence of jitter in that connection. Once there, we are back in digital domain and there is no place for jitter.

To me, the interesting thing is how each digital player and / or DAC handle this jitter to be (almost) inaudible.

They simply don't see it because it does not exist. When the media player fetches the data from hard disk, it gets PCM audio samples and that is it. It doesn't get timing with it. And if it has no timing, then there is no clock jitter. If the above author says there is jitter, then they should be able to show that in their player but I am confident they don't because there is no such data for them to display.

Some other discuss about two 'identical' files with identical checksums (then exact bit by bit) could bring a different SQ.

Roch

The reason two identical files *may* sound different is not because each has different jitter stored with it. It has to do with the activity of the system changing and that somehow bleeding into the digital output from the computer. This is possible. But right now is domain of conjecture and not anything to back it. Let's say this can happen. Its cause is not because the data has jitter with it in the storage device but because we are trying to convert the digital data to analog and problem manifests itself.

dallasjustice · Jun 9, 2014

Peter found someone who at 3:48 on the video claims to change the molecules on a cd after file downloaded onto it and then upload to a server and the end result will be better digital than original file. It's funny.
http://www.whatsbestforum.com/showthread.php?14968-IPC-Devices-another-entry-into-the-unknown

BlueFox · Jun 10, 2014

DonH50 said:
If you have already made up your mind you know the answer, why ask?

If you are talking to me I am trying to find out if there is any new research into the subject. At this point, my instinct is that the science is not yet settled with digital audio. Also, sometimes we simply repeat something so often that is becomes a 'fact', even if it is incorrect.

So, one final question based on the diagram in post 1. If the top wave form is the accurate music file, and the second wave form is what is reconstructed from the signal on the Ethernet cable after arriving at the computer then which one is stored on the hard drive. The top one, or the second one with the extended plus signal and reduced minus signal?

elcorso · Jun 10, 2014

Amir,

As I said before, "discuss this can result in a long and tedious discussion". Then I prefer not to fully reply.

The link to Jussi Laaka (Miska in Computer Audiophile) belongs to another forum, precisely CA, and I don't want to mix different forums.

The only thing I can tell you, and learned from experience is, digital theory is this, a theory, that may change with the time as every theory, or maybe would never change, who knows? "Bits are bits" had being discussed ad nauseam, but every day a wise digital player and / or DAC designer improves the way they reach our ears, in analogue, of course.

Maybe this is why so many music listeners remains in full analogue and don't foul his souls in ADC / DAC conversions.

Roch

rbbert · Jun 10, 2014

elcorso said:
...Maybe this is why so many music listeners remains in full analogue and don't foul his souls in ADC / DAC conversions.

Roch

So they only have to deal with record warps, turntable speed and isolation, cartridge VTA and SRA, tonearm tracking and phono preamp issues...

FrantzM · Jun 10, 2014

rbbert said:
So they only have to deal with record warps, turntable speed and isolation, cartridge VTA and SRA, tonearm tracking and phono preamp issues...

BlueFox · Jun 10, 2014

BlueFox said:
So, one final question based on the diagram in post 1. If the top wave form is the accurate music file, and the second wave form is what is reconstructed from the signal on the Ethernet cable after arriving at the computer then which one is stored on the hard drive. The top one, or the second one with the extended plus signal and reduced minus signal?

In doing more research on this subject, I found this from IBM.

"The data flows into a cache where it is encoded using special mathematically derived formulae, ensuring that any subsequent errors caused by noise can be detected and corrected."
http://www.research.ibm.com/research/gmr/basics.html

"Subsequent" implies the jitter error is preserved.

On the other hand, if the width of a bit is predetermined by a clock then I can understand how a clock would transition a 1 to a 0 even while the original bit is still a 1. Same thing with a 0 going to a 1 even while still 0. This would eliminate timing errors in the example. I am still Googling for that level of detail. Maybe I need to buy a text book.

Ken Newton · Jun 10, 2014

Bud,

All three waveforems you have shown convey exactly the same binary data or bits. That data is conveyed by the flat tops and flat bottoms of the signal, not by the precise positioning of the edges. So long as the waveform top and bottom levels are correctly arbitrated, then any data transfer will contain zero errors, which means that it will be a PERFECT copy, no matter how many times it undergoes digital to digital transfer. Yes, there is an theorectical Bit-Error-Rate for any digital data transfer system, but it will be extremely low in digital audio applications.

Jitter, then, only matters in two cases. One case is, if the jitter is gross enough to provoke incorrect arbitration of the binary levels, as DonH50 has said. While this is a real concern for long-haul telecom links, as a practical matter, it is not an real concern in a properly functioning digital audio system. The second case is, jitter affecting the A/D or D/A domain conversion instants. This second area is very much of real concern in digital audio, because humans have proven sensitive to the unique distortions produced thereby. In short, jitter is not an concern in the transfer of binary data files in a digital audio system. Jitter is only an concern at the point of A/D to D/A domain conversion.

I suspect, that what causes some of the confusion is the well deserved bad press about the ubiquitous S/PDIF link between transports and DAC boxes. The S/PDIF link mixes the clock signal with the data signal, which then have to be seperated. At the DAC box, the jitter added by the link interface will not affect the accuracy of the recovered binary audio data, but it will affect the stability of the recovered clock signal. This clock signal synchronizes the D/A conversion instants to the recieved data, which then opens a window through which jitter can modulate the conversion instants.

elcorso · Jun 10, 2014

rbbert said:
So they only have to deal with record warps, turntable speed and isolation, cartridge VTA and SRA, tonearm tracking and phono preamp issues...

Ha, ha, ha. It's like to find the perfect wife: At this stage of my life I guess I would go for jitterless digital one, perhaps because it is cheaper to correct some jitter that full aesthetic reconstruction of an 'analogue woman'.

Roch

elcorso · Jun 10, 2014

BlueFox said:
In doing more research on this subject, I found this from IBM.

"The data flows into a cache where it is encoded using special mathematically derived formulae, ensuring that any subsequent errors caused by noise can be detected and corrected."
http://www.research.ibm.com/research/gmr/basics.html

"Subsequent" implies the jitter error is preserved.

On the other hand, if the width of a bit is predetermined by a clock then I can understand how a clock would transition a 1 to a 0 even while the original bit is still a 1. Same thing with a 0 going to a 1 even while still 0. This would eliminate timing errors in the example. I am still Googling for that level of detail. Maybe I need to buy a text book.

Bud, here is a nice reading about jitter & jitter types:

http://class.ee.iastate.edu/mmina/ee418/Notes/Introduction_to_Jitter.pdf

But please do not suffer too much with this as I did ... Now only guided by my ears (never from others ears) and leave this work to music player & DAC developers, plus some implementation on the music server I use, in my case a highly tuned Mac Mini (suggestions from those developers help a lot).

Roch

DonH50 · Jun 10, 2014

BlueFox said:
If you are talking to me I am trying to find out if there is any new research into the subject. At this point, my instinct is that the science is not yet settled with digital audio. Also, sometimes we simply repeat something so often that is becomes a 'fact', even if it is incorrect.

So, one final question based on the diagram in post 1. If the top wave form is the accurate music file, and the second wave form is what is reconstructed from the signal on the Ethernet cable after arriving at the computer then which one is stored on the hard drive. The top one, or the second one with the extended plus signal and reduced minus signal?

As I said, draw a line in the middle of each bit period and that is what gets stored. It will be the same for both. The data is not sensed at the edges. Jitter on the edges does not matter until it corrupts virtually the entire bit period, and even then error correction will fix it. Unless the link is so bad it generates bit errors, but in that case you have all kinds of other things going wrong. If serial links were that sensitive to edge jitter your computer would never work.

Where jitter matters* is when the signal is used as a clock for an ADC or DAC where the edges matter. In between it is digital signals detected in the middle of the bit period, not on the edges.

* Until it is bad enough to cause the link to fail... That is not what we are talking about here.

DonH50 · Jun 10, 2014

BlueFox said:
In doing more research on this subject, I found this from IBM.

"The data flows into a cache where it is encoded using special mathematically derived formulae, ensuring that any subsequent errors caused by noise can be detected and corrected."
http://www.research.ibm.com/research/gmr/basics.html

"Subsequent" implies the jitter error is preserved.

On the other hand, if the width of a bit is predetermined by a clock then I can understand how a clock would transition a 1 to a 0 even while the original bit is still a 1. Same thing with a 0 going to a 1 even while still 0. This would eliminate timing errors in the example. I am still Googling for that level of detail. Maybe I need to buy a text book.

You are mixing up several things. The errors in a disk drive are when the actual bits are wrong, and there is all sorts of error checking and handshaking that take of that. The disk drive electronics deal with analog signals that do indeed look just like noise. Massive analog and digital processing is used to ensure you get the right data off your disc drive (usually). At the read head the signal is nothing like the bit stream coming out of the drive and on to the rest of the PC, let alone S/PDIF or USB signals. You cannot equate that article with jitter causing noise and/or distortion at the output of your audio DAC.

This would be way easier with a white board.

rbbert · Jun 10, 2014

As far as a computer is concerned, two "bit-identical" (or just identical) files are the same and will always behave the same way when that data is put into a program; if you are transmitting it to other hardware (like a DAC), other factors may come into play affecting what that hardware does with the data.

FrantzM · Jun 10, 2014

The confusion stems from the waveforms that have come to represent digital in the mind of the public. I am not sure it will disappear anytime soon.

Does a file with jitter keep the jitter when saved to a hard drive?

Member Sponsor

Attachments

WBF Technical Expert (Speakers & Audio Equipment)

Banned

VIP/Donor

Member Sponsor

Member Sponsor & WBF Technical Expert

Banned

Member Sponsor

Member Sponsor

VIP/Donor

Well-Known Member

Member Sponsor & WBF Founding Member

Member Sponsor

Well-Known Member

VIP/Donor

VIP/Donor

Member Sponsor & WBF Technical Expert

Member Sponsor & WBF Technical Expert

Well-Known Member

Member Sponsor & WBF Founding Member

Similar threads