Some info for those interested 2 of 2
So what did I hear - real or imagined
I had the 8" silver in the system. I set up a series of cuts and went through them a couple of times to attempt to understand it well.
I plugged in cable 1. the standard usb cable.
My initial thought was - ****, this is good. Maybe I did imagine this. Now this is going to be a bit embarrassing, but man up and take the heat. I listened to all the cuts a couple of times thinking, damn this is very good.
I put back the 8" copper twisted pair and to my surprise, clearly better. The effect was not dramatic but clearly a fuller presentation. Like when a picture goes up in resolution. Same picture, just fuller, more detailed and clearer.
I listened to all the cuts. Then put in the same length silver cable. Now here was a difference worth talking about. Be clear, I don't maintain its just because its silver. I realize as an experiment I have changed 2 variables, silver vs copper and and twisted pair vs parallel signal lines. The impact of the 'silver' cable is more of a jump to a different level. Stuff that was previously equally detailed now just sounds real. No other way to put it. You could wax on about air and separation etc. Stuff is just real.
Important to note that the cable with the parallel silver lines did not work properly without a shield. bare and open in close proximity to AC cables and a lot of big transformers it picked up some stuff. The result bare is a wonderful music but with some static that sounds just like vinyl that needs to be cleaned. You could swear it was a turntable, static and all. I tried wrapping it in aluminum foil and the static vanished but so did the magic. Not to be daunted I made a 1" tube of foil and ran the usb cable through it with space between the signal lines and the foil. The magic was back and the static was gone. So clearly there is some stuff to play with hear to get it in any way robust. Matters not for me because all this stuff will be inside an enclosure so YMMV.
My final test was to see if a very short cable would work, and if it is any good. This was a bit of a tough test as it required opening the DAC, and leaning the ALIX into it without any attention to EMI (without touching 170v B+). The ALIX locked onto the WaveIO immediately and the music was very good. Actually I left it like this for a couple of hrs and quite enjoyed the presentation.
Now I can imagine there will be many cases where none of this is of any use. For me in my configuration it works well. The suggestion that impedance mismatch and violating the spec does not seem to result in perceived sound issues. Actually quite the opposite. I fully admit the limits of my technical analysis. I do understand that redbook playback in general expects to have data losses and it is designed to do its best to create sound with what it gets. Clearly improvements in CD transports to get more of the bits to the sound chip results in better sound. Could it be we have a similar effect with usb? Does usb have a checkbit. Is it constantly finding errors and re transmitting? One would never see that in a digital transmission. At the end you would always get a bit perfect file. What you don't see is variation in throughput speed based on how many retries of packets to get the message through. I don't know just one thought. I know I keep my buffers as small as possible as the sound gets better with smaller buffer size. Is it possible the short cable is in fact getting a correct packet through in one attempt more often and as a result fewer packets dropped and better sound in the end after the dac happily works with what it has? It does not seem to make sense in an async usb2 world, but then again changes in the pc should not make a difference but it certainly does.
Final pics are the beautifully engineered cable with its fancy foil shield and then of the ALIX taking a dive into the DAC to reach the tiny little cable. Engineering at its finest.
Everyone please have fun with this.
I am using the little shortty cable in my new music player and won't be taking this any further.
So what did I hear - real or imagined
I had the 8" silver in the system. I set up a series of cuts and went through them a couple of times to attempt to understand it well.
I plugged in cable 1. the standard usb cable.
My initial thought was - ****, this is good. Maybe I did imagine this. Now this is going to be a bit embarrassing, but man up and take the heat. I listened to all the cuts a couple of times thinking, damn this is very good.
I put back the 8" copper twisted pair and to my surprise, clearly better. The effect was not dramatic but clearly a fuller presentation. Like when a picture goes up in resolution. Same picture, just fuller, more detailed and clearer.
I listened to all the cuts. Then put in the same length silver cable. Now here was a difference worth talking about. Be clear, I don't maintain its just because its silver. I realize as an experiment I have changed 2 variables, silver vs copper and and twisted pair vs parallel signal lines. The impact of the 'silver' cable is more of a jump to a different level. Stuff that was previously equally detailed now just sounds real. No other way to put it. You could wax on about air and separation etc. Stuff is just real.
Important to note that the cable with the parallel silver lines did not work properly without a shield. bare and open in close proximity to AC cables and a lot of big transformers it picked up some stuff. The result bare is a wonderful music but with some static that sounds just like vinyl that needs to be cleaned. You could swear it was a turntable, static and all. I tried wrapping it in aluminum foil and the static vanished but so did the magic. Not to be daunted I made a 1" tube of foil and ran the usb cable through it with space between the signal lines and the foil. The magic was back and the static was gone. So clearly there is some stuff to play with hear to get it in any way robust. Matters not for me because all this stuff will be inside an enclosure so YMMV.
My final test was to see if a very short cable would work, and if it is any good. This was a bit of a tough test as it required opening the DAC, and leaning the ALIX into it without any attention to EMI (without touching 170v B+). The ALIX locked onto the WaveIO immediately and the music was very good. Actually I left it like this for a couple of hrs and quite enjoyed the presentation.
Now I can imagine there will be many cases where none of this is of any use. For me in my configuration it works well. The suggestion that impedance mismatch and violating the spec does not seem to result in perceived sound issues. Actually quite the opposite. I fully admit the limits of my technical analysis. I do understand that redbook playback in general expects to have data losses and it is designed to do its best to create sound with what it gets. Clearly improvements in CD transports to get more of the bits to the sound chip results in better sound. Could it be we have a similar effect with usb? Does usb have a checkbit. Is it constantly finding errors and re transmitting? One would never see that in a digital transmission. At the end you would always get a bit perfect file. What you don't see is variation in throughput speed based on how many retries of packets to get the message through. I don't know just one thought. I know I keep my buffers as small as possible as the sound gets better with smaller buffer size. Is it possible the short cable is in fact getting a correct packet through in one attempt more often and as a result fewer packets dropped and better sound in the end after the dac happily works with what it has? It does not seem to make sense in an async usb2 world, but then again changes in the pc should not make a difference but it certainly does.
Final pics are the beautifully engineered cable with its fancy foil shield and then of the ALIX taking a dive into the DAC to reach the tiny little cable. Engineering at its finest.
Everyone please have fun with this.
I am using the little shortty cable in my new music player and won't be taking this any further.Attachments
... pushing a lot of the "snake-oil" buttons!
Solid core copper, that Teflon word, silver wire, untwisted wires, then the '41A dac, 6n2p valve buffer, Lightspeed vol control, more silver, OTL amp, FR speakers, minimalist anything, and then 'different resistor sounds' to round it all off!!
You're going to get 'excommunicated, or something!!
Solid core copper, that Teflon word, silver wire, untwisted wires, then the '41A dac, 6n2p valve buffer, Lightspeed vol control, more silver, OTL amp, FR speakers, minimalist anything, and then 'different resistor sounds' to round it all off!!
You're going to get 'excommunicated, or something!!
The USB audio protocols (synchronous, adaptive and asynchronous) are all in the isochronous category and do not provide for retransmission (unlike the bulk protocol used for printers and hdd). There is some errors check but no possibility to resend data.
Interesting.. kind of supports my theory. Audio is moving a lot of data. If there are errors or omissions then the DAC just steams on and does it's best with what it gets even tho there may be missing or incorrect data. Same cable to a HDD however is always eventually bit perfect transmission. The variable wrt hdd tx errors would be speed of the file transfer depending on how many re-transmissions.
In my experience, USB cables affect the sound by means of common-mode noise. Not jitter, not bits being changed. Ferrite clamps around the cable will tend to increase the common-mode impedance and might well help. But the noise source is the computer's SMPSU where high voltages are swung and are coupled via low parasitic capacitances. Meaning the noise source is in effect a noise current source - in such circumstances its quite hard to make a significant increase in the common-mode impedance because to reduce the noise by say 6dB you need to double the impedance. When that starts out in the 100k - 1Mohm region, you've got your work cut out. Some CM chokes will have such impedance but only over a narrowish band near their self resonance. This suggests a possible solution with many dissimilar CM chokes in series tuned to target the switching frequency and the lower harmonics of same.
Hope this helps a little
<edit> Interesting Head-fi thread here where on at least one DAC, SQ is improved just by cutting the +5V wire on the USB. The thread starter is fairly savvy and is planning to experiment with CM chokes : http://www.head-fi.org/t/693798/ranking-of-17-dacs-and-dac-configurations.
Just had another thought for an experiment to test this - run your PC on a bank of SLAs. SMPSUs are happy with DC (its even better as far as they're concerned) so probably you'd only need about 12 12V SLAs in series. They're cheap enough.
http://www.head-fi.org/t/693798/ranking-of-17-dacs-and-dac-configurations
Hope this helps a little
<edit> Interesting Head-fi thread here where on at least one DAC, SQ is improved just by cutting the +5V wire on the USB. The thread starter is fairly savvy and is planning to experiment with CM chokes : http://www.head-fi.org/t/693798/ranking-of-17-dacs-and-dac-configurations.
Just had another thought for an experiment to test this - run your PC on a bank of SLAs. SMPSUs are happy with DC (its even better as far as they're concerned) so probably you'd only need about 12 12V SLAs in series. They're cheap enough.
http://www.head-fi.org/t/693798/ranking-of-17-dacs-and-dac-configurations
Last edited:
That doesn't even start to make any sense, really. Sorry.
What's the point ? The pc is full of switching regulators. Those are much more responsible for noise at the usb ports than the main power brick.
Do you really want to know or is this just a way of objecting to my statement? On the off-chance that you are genuinely curious here, try looking up the difference between common mode and normal mode noise.
How do you know? Got anything to support this last claim?
Short cables are less sensitive to impedance mismatch problems. In fact, short cables are less sensitive to almost all problems apart from plug/socket problems. This may be why you can get away with using bad cables.
The test appears to have been fully sighted, and combined with attempts to push 'audiophile' buttons (e.g. mention of resistors). To some this will appear to be an attempt to exhibit 'genuine audiophile credentials'; someone to be taken seriously in the 'ears' department. To others this will flag up the presence of strong expectation bias and/or placebo effect. Hence this test will tend to widen the divide rather than resolve it.
The test appears to have been fully sighted, and combined with attempts to push 'audiophile' buttons (e.g. mention of resistors). To some this will appear to be an attempt to exhibit 'genuine audiophile credentials'; someone to be taken seriously in the 'ears' department. To others this will flag up the presence of strong expectation bias and/or placebo effect. Hence this test will tend to widen the divide rather than resolve it.
Kind of both. My experience with recent laptops is that there isn't any significant audible difference for usb audio, in between running them on battery or on ac.
Thinking about it twice, the situation is different for desktops (much more reliance on the PSU and less on the stepdown switchers) and I might have been hasty in transferring reasoning from one case to the other.
A more honnest question: through which mechanisms do you suggest that common mode noise will affect usb audio ? Some (empirical audio for example) link common mode noise and jitter.
I explained (with hindsight, 'outlined' is a better choice of words) the mechanism in post #165 - there's a noise source inside an SMPSU which is referenced to the mains input and also to earth, depending on how filtering is implemented internally and where the dominant stray capacitances are. This couples to the laptop (say) via the 0VDC wire (normally has a ferrite around it to pass EMC tests for CE etc.). The loop gets completed via the USB cable, the DAC's mains transformer back to mains again.
There needn't be any audible effect if this noise current doesn't share common ground impedance with anything in the analog part of the DAC. However few designers are aware of the issue and many just 'bolt-on' a USB interface PCB to an existing design without really thinking through where the noise currents are going to flow. Perhaps the designers of the DAC you have knew what they were doing? Even if the DAC design is perfect though the amp it connects to might be sensitive to the noise currents - assuming the amp has an AC path back to mains.
There needn't be any audible effect if this noise current doesn't share common ground impedance with anything in the analog part of the DAC. However few designers are aware of the issue and many just 'bolt-on' a USB interface PCB to an existing design without really thinking through where the noise currents are going to flow. Perhaps the designers of the DAC you have knew what they were doing? Even if the DAC design is perfect though the amp it connects to might be sensitive to the noise currents - assuming the amp has an AC path back to mains.
Last edited:
Ok. I got the bit about how common mode noise appears in your earlier post. No disagreement about that. The question was more on how it would appear at the dac output. Considering smps run at around 50khz those days (at least those I have on hand), it's out of the directly audible range. So we still need how to explain how it affects sound. Degradation of the operating parameters of the usb receiver and dac ?
I'd love to think that the designers of my DAC knew what they were doing but I'm not too optimistic about that since I designed and built it myself. More seriously, there is an adum4160 in front of the usb receiver, a common mode filter at the 230Vac input and a groundloop breaker in between earth and gnd.
I'd love to think that the designers of my DAC knew what they were doing but I'm not too optimistic about that since I designed and built it myself.
More seriously, there is an adum4160 in front of the usb receiver, a common mode filter at the 230Vac input and a groundloop breaker in between earth and gnd.
Ah I see, thanks for clarifying. I'm not totally sure on this but this is my best guess at present. The noise currents are high frequencies - because that's what best gets through the trafo interwinding capacitance. As you say, 50kHz and up. Its out of the audio range but it gets back into the audio as noise by means of the electronics' non-linearities - IMD in other words. Opamps with low-noise input stages are particularly prone to IMD from ultrasonic hash on their inputs and power rails. Subjectively all this extra noise ruins the dynamics.
I'd love to think that the designers of my DAC knew what they were doing but I'm not too optimistic about that since I designed and built it myself.
That sounds like a good start - the ADI USB isolator will present only single digit pF capacitance (guessing as didn't find this spec in the DS) so be a major barrier to CM currents. Then are you using a common 0V between the analog and digital parts, and if so where do they join? They really need to join together at the place where the CM currents exit - the connection to the trafo secondary.
It's getting a bit OT maybe...
A bit complicated to explain without drawing the thing. I'll try.
The adum4160 is external (a hifimediy thingy, also providing isolated power). Inside the dac box, you have a pcm2707 on its pcb, powered from USB. From that PCB only exit I2S lines (including gnd), to the "dac pcb". That pcb has src4192 and pcm1798 onboard. Single gndplane, onboard regs (one for clock, one for src4192 and digital section of pcm1798, one for the analog section). The + and - signals at the outputs of the pcm1798 then goes to the "I/V pcb". It also has it own regs onboard. No direct gnd connection in between the two pcbs.
The "dac pcb" and the "I/V pcb" are powered by two different xformers, followed by the unregulated supplies (+6V on one hand, +/15V on the other). These supplies are linked at their final caps. That's also where the gndloop breaker is tied.
Probably plenty to correct but that's the arrangement that resulted in the least audible noise (meaning none using headphones and a 5X gain headphones amp).
A bit complicated to explain without drawing the thing. I'll try.
The adum4160 is external (a hifimediy thingy, also providing isolated power). Inside the dac box, you have a pcm2707 on its pcb, powered from USB. From that PCB only exit I2S lines (including gnd), to the "dac pcb". That pcb has src4192 and pcm1798 onboard. Single gndplane, onboard regs (one for clock, one for src4192 and digital section of pcm1798, one for the analog section). The + and - signals at the outputs of the pcm1798 then goes to the "I/V pcb". It also has it own regs onboard. No direct gnd connection in between the two pcbs.
The "dac pcb" and the "I/V pcb" are powered by two different xformers, followed by the unregulated supplies (+6V on one hand, +/15V on the other). These supplies are linked at their final caps. That's also where the gndloop breaker is tied.
Probably plenty to correct but that's the arrangement that resulted in the least audible noise (meaning none using headphones and a 5X gain headphones amp).
I have the same Dacmagic too. Rojo earlier in this thread suggested only the D+ and D- twisted pair is required. No +5V and ground connections between notebook and DAC. You kept the ground connection and worked.
My experience is that it won't work without the ground connection. I tried:
1. only D+ and D- connected. Didn't work.
2. Then added ground connection. Worked.
3. To re-confirm, I disconnected the ground connection. Didn't work.
4. Re-connect ground, worked again.
Analog_SA was probably correct when he said: "Or if it does work it is only thanks to the common mains ground. Snip that and there will be silence."
I use a toroidal transformer to provide the 12VAC to the Dacmagic instead of the supplied AC transformer, but failed to see the reason why it worked for Rojo and not me when ground connection snipped.
Hi Crom
No ABX at my place. Invited some good people, but no real interest in this test and some understandable reluctance to venture into this snake pit.
I continue to enjoy the fruits of my labor but pretty much decided to keep my findings to myself.
Next revision of music server will take usb to somewhere minimum possible length. I am thinking next step after that will be to eliminate USB all together. Watching with some interest the musings of the twisted pair folk playing with rasberry pi with direct to I2S.
http://www.diyaudio.com/forums/twis...-open-embedded-audio-applicance-new-post.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.