I'm a Design Fellow Emeritus for Analog Devices and for a number of years I designed retiming PLL and DLL loops for various interfaces etc. including USB. The 1s and zeros that you refer to are generally caused by current changes into a fixed load at the receiver that matches the characteristic impedance of the "transmission line" that is the USB cable in order to minimize reflections.The data is also sent in packets retimed at the receiver. This retiming occurs "perfectly" as long as the transitions from 1 to 0 and back occur in a given period of time- a so called retiming window- the purpose of which is to eliminate the effect of the finite bandwidth and jitter of the interconnect/transmitting system. The receiver adds its own imperfections to the reconstructed data stream, but as long as the cable, transmitter and receiver comply to the standards the resulting data stream will have a one to one correspondence to the source. In addition, the data is checked upon reception using a CRC and if the packet is corrupted it will be dropped and the host can be asked to resend it. So, obviously, the system is not a dumb one and in reality- spec compliant cables CANNOT matter in an analog sense.
USB2.0 is also able to transmit 480Mb/s so audio bit streams are well below the max.
I've heard audio streams where the USB data is actually corrupted and packets are being lost.
The result is not subtle...
|
I'm not using technical jargon to obscure the issues. Quite the contrary.
I'm happy to explain, however I'm also trying to avoid being condescending (and imprecise) in avoiding its use. Sometimes terms with precise meanings are necessary to add clarity to an issue, and sometimes those terms require a great deal of explanation that would just cause people to turn off. Better to try and explain as requested...
|
Perhaps you view it that way. I do not. CD readback is imperfect because there are mechanical and optical imperfections in the recording and playback processes that result in timing issues and there also is stochastic jitter. These processes are intrinsically more complex than the rather simple USB interface impairments.
The attached CD data recovery circuitry is tasked with the restoration of the near ideal bit sequence that then is applied to the DAC.
There's plenty of room for the introduction of timing errors/jitter in that process even if the result is ultimately bit perfect.
Ideally the DAC will retime the input data to restore a pristine, low jitter, bit perfect, data sequence for conversion to analog.
|
Again, the fact that the input sequence is not perfectly timed is not relevant providing the data recovery circuitry/algorithms are able to cope with the imperfections. The use of data corrections (CRCs etc) can ensure bit perfect output sequence results. The use of quality retiming can ensure extremely low jitter prior to the conversion (or indeed during the conversion) to analog.
For high rate data recovery PLLs with jitter in the 100fs region or lower are de rigeur. For audio the best PLLs/DLLs used are sub ps. The theory concerning the eventual SINAD of the data at the output and the resulting impairments is quite clear. There's a reason that digital recording/playback of audio has improved by so much in the last three decades.
|
One other thing that needs clarification- the concepts of bit perfect and low jitter are NOT homomorphic.
Bit perfect simply means that the recovered output data sequence is the same as the original input sequence- which in turn means that the recovered output changes in state are within the correct timing windows.
It tells you nothing about the stochastic and deterministic components of the jitter other than the combination of both in any given timing window does not cause the data to fall outside the recovery region.
|
The arguments are dissimilar.
This is not about whether a digital format produces a perfect reproduction of an analog source or otherwise.
This is about whether the USB data transfer protocol is bit perfect as long as the interface is spec compliant- including the transmitter and receiver.
The answer to the latter question is yes, it is.
The answer to the former question is far more complex.
There is some possibility that the USB can introduce noise through the ground connection. The USB receiver can be designed to eliminate the possibility of ground loops. However, sometimes the problem does occur.
By the way, to put this in context, my system is tri-amped with a sub woofer and a pair of slightly modified ML Montis, The electrostatic panels are driven by one of the following- Rogue M180s, Hypex Ncore 400s or a Benchmark AHB2, depending on my whim. The non electrostatic units are all driven by DSP equalised class D amps.
The preamp is a modified Hovland HP100 or a home made phono amp driving a digital ADC/DAC- the RME ADI-2 PRO FS. My cables are all home made (except for the USB ;-))
My cartridge is a Miyajima Madake. I also have many 15IPS 1/4" 2 track tapes- safety masters or similar- that I play on a Otari MX50.
I have digitized many LPs/tapes and I play them back with the RME device using a USB connection.
|
Incidentally, the same kinds of circuits/techniques are used to recover data for high speed data links such as is used on the backbone of the internet. The jitter requirements for those applications are orders of magnitude lower than the needs of audio and although the noise bandwidths are different the issues remain the same, particularly as modulation of the phase and/or modulation of the amplitude is used to increase the data capacity. |
A sufficiently corrupted data stream cannot be reconstructed at the receiving end, but the corrupted packets can be identified and dropped.
This is quite audible and unmistakable. It does not result in the loss of ambience or differences in top hat shimmer.
|
Download your digital media onto a computer. The ripping process includes CRC checks for bit perfection. That way you can be certain that the bits out are what was intended to be recorded on the original medium and any discrepancies will be flagged.
This also applies for any lossless compression schemes such as FLAC.
After that, well you still need to have quality retiming etc.
|
Stochastic is random, such as the addition of noise to a voltage level.
Deterministic is due to the interference of the pulses (they're usually called symbols) with each other due to the fact that the bandwidth of the interconnect is not infinite and as a result the edges get "smeared". Because the edges are at variable times relative to one another the amount of interference varies from each bit of data to the next. This causes the point in time at which the data is detected to also shift from bit to bit causing jitter relative to the average. These two effects combined (plus other second order effects) are responsible for deterioration in the sound if they are not corrected. If the shift is too large the transition will not be correctly detected at all and the bit is either lost or misplaced and a data error occurs and is flagged by the CRC.
|
Agreed. My point is that the loss of a packet results in a discontinuity in the signal. In my experience such an event is audibly catastrophic and not in the least subtle and does not require audiophilese to describe.
USB interfaces will report packet loss, but I'm not sure that audio gear is able to enlighten the listener when such events occur.
I would imagine that computer based playback software would be able to report it, but I don't know that for a fact.
|
Yes, many CDs are just dross, while a few are quite magical.
This applies also to high res recordings, but as always caveat emptor.
I know of one download that on my system sounds noticeably superior to the live performance of the same piece with the same two musicians but performed in a different hall than the recording and with the violinist playing a different violin- a, as he put it, "fine copy of my strad" and the pianist playing a Steinway model A rather than the recorded model D.
This is not just my opinion, other attendees agree.
The playback was over USB using an RME ADI-2 PRO FS, and using a non audiophile USB cable with an active extender.
|
