Fair enough - and as that is something you can't measure with an oscilloscope, frequency analyzer or any of the other usual audio measurement methods, you have to choose - do you want to find out what is objectively "better" or "good enough", or are you concerned about what is most emotionally pleasing to you. In most cases there is very little correlation between those two avenues.
...understanding, of course, that "emotionally pleasing" may (and usually does) have very little to do with actual sound. It's important not to conflate design and analytical listening (which has to do with the equipment and actual sound) with sitting back and enjoying music (which may or may not). The most satisfying emotional experiences I've had with music have largely involved absolutely terrible sound- but great musicians.
Sorry I have to disagree, just spent two weeks working on a PCB with DDR3 amongst other interfaces, and the engineers been happily simulating the design as I lay it out. The software used is this:
CADSATR PCB analysis and verification | Zuken
Its a real tool and its used for real high speed designs, and some of the waves you see are quite strange, and the tool is used to verify a design but also to try out different termination schemes.
Also don't forget it is a quick scenario made up od an 04 driver and three TL's. The rise time is about 30ns on the terminated one, its the way the x and y of the graph are shown that make it look so square.
Termination as well as absorbing the excess energy put into the line, visible as over shoot, also damps the reflections, thus reducing there effect on the transmitted signal.
Last edited:
Rather than trying to explain termination, and have to refer to documentation, a quick visual demo is easier and more understandable, that's all, but as shown it does help alleviate problems caused by mismatches.
It looks like you're running about 2.5MHz, which isn't very different than (say) spdif. If memory serves, you'll get a good square wave if the frequency response is flat for a decade above and a decade below the fundamental frequency (in this case, 250k-25M). So it seems to be completely reasonable data.
"As to the quality of a DAC or amp or whatever: suppose I have this very expensive, very well regarded amp, that sounds like heaven to me. Now I change the line cord and the amp suddenly sounds different! You can't maintain with dry eyes that this is a competently designed amp!
If I design a regulated power supply that is 'sensitive' to the upstream rectifier diodes or reservoir cap, I've done a bad job!
If a DAC suddenly sounds different with a differnt S/PDIF cable on it's input, it's a crap DAC.
Jan "
I've been selling, servicing, upgrading & designing audio dac units since they were first on the market, and I have YET to see a dac, no matter how well-designed, elaborately crafted or expensive, be even close to NOT affected by choice of digital cable. And I am talking about proper spec, manufactured cables, as well as off-spec, and EVERY TIME I've changed a cable, EVERY DAC changes sound. And NOBODY that I have ever known has contradicted that experience. Show me a dac unit that is immune to cable differences.
As for the other things you mention- Standard silicon rectifiers create ringing HF noise that is extremely difficult to completely do away with, which is why I have never yet found any piece of audio gear, no matter how elaborately filtered/regulated, that did not sound better with change to soft recovery rectifiers. Likewise, ac power noise, especially nasty when carried on the ground line in SE power systems like the USA's 120V power, can find ways around the best power supply circuits.
Point is, at least one major point, if you can't account for a difference in what you hear based on what you're measuring, you simply haven't figured out what you should be measuring or how to measure it.
If I design a regulated power supply that is 'sensitive' to the upstream rectifier diodes or reservoir cap, I've done a bad job!
If a DAC suddenly sounds different with a differnt S/PDIF cable on it's input, it's a crap DAC.
Jan "
I've been selling, servicing, upgrading & designing audio dac units since they were first on the market, and I have YET to see a dac, no matter how well-designed, elaborately crafted or expensive, be even close to NOT affected by choice of digital cable. And I am talking about proper spec, manufactured cables, as well as off-spec, and EVERY TIME I've changed a cable, EVERY DAC changes sound. And NOBODY that I have ever known has contradicted that experience. Show me a dac unit that is immune to cable differences.
As for the other things you mention- Standard silicon rectifiers create ringing HF noise that is extremely difficult to completely do away with, which is why I have never yet found any piece of audio gear, no matter how elaborately filtered/regulated, that did not sound better with change to soft recovery rectifiers. Likewise, ac power noise, especially nasty when carried on the ground line in SE power systems like the USA's 120V power, can find ways around the best power supply circuits.
Point is, at least one major point, if you can't account for a difference in what you hear based on what you're measuring, you simply haven't figured out what you should be measuring or how to measure it.
stephensank said:
Agree, my experience is the same.
Even cables with the same physical lenght, but different electrical signal lenght (different Vf) will sound different.
Depends. What are they waveforms of? How does the DAC output look like in each case?
And what property of the cable is the most important factor in that change in sound?
Indeed. But, conversely, when you find apparent changes in sound that don't correspond to traditional measurements, it is essential to make sure that the sonic differences actually correspond to real changes in the actual signal. What measures have you taken to eliminate expectation and perceptual bias when you have conducted your listening tests?
Some DDR2 memory waveforms at the bottom of the article; they look like that in real life, your DDR memory works.
DDR2 Signal Integrity
As long as the switching points are crossed data will be transmitted by either. The DAC should reclock the signals, so the slight change in mark space ratio should be sorted. A well designed DAC should be able to cater for both waveforms as you don't know the quality of the up stream transmitter or how good the cable is going to be.
From the waveforms shown there is no way of guessing the sound output...and bad data transmission is pretty obvious, as drop outs, not a change in the sound. jitter we have discussed and is down to the DAC and related circuitry to sort.
I would like to hear an explanation of how they could sound different?
Simple...
There are still a lot of DACs (the majority) which simply use the MCLK/BCK generated by a SPDIF receiver chip to clock the DAC directly. Those clocks are recovered using a PLL.
Of course when recovering a clock using a PLL, there is a compromise. If the PLL bandwidth is very low, source jitter will be attenuated, but lock is more difficult or takes longer, more FIFO buffers are necessary to absorb source rate drift, and local VCO jitter becomes a problem. So, the easiest solution is to use a rather wide PLL bandwidth.
If you look at the datasheet of CS8414, page 8, it shows the jitter attenuation of the internal PLL, which is zero for jitter/phase noise frequencies up to about 20 kHz. The chip adds its own jitter on top of that, but the main problem is that absolutely no clock cleaning is done...
Same datasheet mentions "200 ps rms" jitter (without any info on bandwidth or how it was measured) so it's not very relevant. It is probably the jitter added by the chip to whatever was in the source...
They also say "Most jitter introduced by the transmission line is high in frequency and will be strongly attenuated.", which is, incidentally, completely wrong. It is like saying "we filter out whatever is easy to filter, noone cares about the rest, right ?"
Better receivers like the Wolfson have a digital fractional PLL based on a local XO, and a FIFO, which allow smart locking algorithms and a low PLL corner frequency, hence good filtering. They provide relevant docs and even jitter measurements :
https://www.wolfsonmicro.com/docume...e_of_spdif_digital_interface_transceivers.pdf
There are other solutions : ASRC, roll your own PLL with a VCXO (complicated), ESS DACs...
IMO, SPDIF is obsolete, we have USB asynchronous now. Why bother ?
There is also the EMI problems. Most SPDIF inputs have a transformer... how do you isolate a coax with a transformer ? The right answer would be you don't... a coax whose shield is not connected to a metal box is a unhappy coax. I have a hunch that the interaction between transformer parasitics, the obligatory capacitor between coax shield ("isolated") and chassis ground, layout parasitics, etc, etc, and of course the cable itself could be an important part of the "different sound" of spdif cables...
Last edited: