Thanks for reading I have achieved a nice increase in sound quality using a usb dac connected to a PC.
There is a 5 volt cable, in the usb cable supplying the dac.
I removed the 5 volt usb cable and replaced it with an external low noise power supply, so that the dac is not being powered by the PC anymore.
The PC power supply can be noisy, hard drives, optical drives, cpu, can inject noise into the power supply
Now that the dac is being powered by a cleaner external supply it sounds better, more detail coming through, smoother sound.
There are 4 cables inside the usb cable. 1) 5 volts 2) data positive 3) data negative 4) ground. I have snipped the 5 volt cable from out of the usb cable, but i have not snipped the ground cable out from the usb.
I have connected the negative from the power supply to the usb ground cable.
Is this correct? I thought that perhaps the ground cable in the usb cable may be needed by the PC to do some thing or other, so best not to be removed
What do you think. Should i remove the usb ground cable as well
And just one other thing, i want to put a thermistor in my pc to limit inrush current. What impedance should the thermistor have when cold, for use with british 240volts mains. I was told that the popular cl-60 thermistor is only 10 ohms and will probably be too low.
Thanks
There is a 5 volt cable, in the usb cable supplying the dac.
I removed the 5 volt usb cable and replaced it with an external low noise power supply, so that the dac is not being powered by the PC anymore.
The PC power supply can be noisy, hard drives, optical drives, cpu, can inject noise into the power supply
Now that the dac is being powered by a cleaner external supply it sounds better, more detail coming through, smoother sound.
There are 4 cables inside the usb cable. 1) 5 volts 2) data positive 3) data negative 4) ground. I have snipped the 5 volt cable from out of the usb cable, but i have not snipped the ground cable out from the usb.
I have connected the negative from the power supply to the usb ground cable.
Is this correct? I thought that perhaps the ground cable in the usb cable may be needed by the PC to do some thing or other, so best not to be removed
What do you think. Should i remove the usb ground cable as well
And just one other thing, i want to put a thermistor in my pc to limit inrush current. What impedance should the thermistor have when cold, for use with british 240volts mains. I was told that the popular cl-60 thermistor is only 10 ohms and will probably be too low.
Thanks
Last edited:
Don't remove the ground. It is needed for the handshake otherwise your DAC wont be recognised by the PC.
Here is the mod you have done so far with the explanation.
https://www.tnt-audio.com/clinica/221_diy_usb_e.html
cheers, Bob.
Here is the mod you have done so far with the explanation.
https://www.tnt-audio.com/clinica/221_diy_usb_e.html
cheers, Bob.
Its debatable if the sound will have more detail.
The signal is digital and so the detail is always the same.
You might get a noise improvement but I doubt if its much.
My PC USB DAC is dead silent.
The signal is digital and so the detail is always the same.
You might get a noise improvement but I doubt if its much.
My PC USB DAC is dead silent.
Depending on the setup you can sometimes end up with a ground loop that runs through the PC via the USB cable to the DAC and then into the analog part of the system. PC power supplies are not optimized for analog performance, so this can result in an elevated noise floor and such. A galvanic isolator or an optical link between the PC and the DAC solves it.
Tom
Tom
What if your model is too simplistic? What about possible veiled sound from computer RFI/EMI on the power rails? Page 141 of the attached describes some audible effects which IME can arise from EMI/RFI when using USB 5v. Basically, the problem appears to be due to resulting intermodulation of audio with RF noise. Again IME, isolated power can do a lot to clean up SQ of some devices, such as, for example, Amanero USB boards. Probably USB powered dacs too.
Last edited:
Thanks! What do you think of using a powered usb hub? Do the hubs run off the small psu that comes with them, or does they still use the usb power in some way?
IME powered USB hubs vary with the particular design. Sometimes they can remain powered by the computer if their dedicated power supply is disconnected. In that case there may be less voltage at end devices. Some devices may continue to function and others not.
In any case, powered hubs usually rely on SMPS that are not particularly clean relative to what can matter for analog circuitry. Experimentally, what works best and most reliably is to isolate dacs from computers using techniques like isolated power, galvanic isolation, and in some cases reclocking. Not every computer/hub/SMPS/DAC/etc., is the same, so symptoms may vary from one situation to the next.
Also IME, its not uncommon that some noise problems are not so easy to measure, although sometimes evaluating FFTs in non-standard ways can be used to show correlation between (possibly nonquantitative) measurements and audible changes in sound. Obviously care needs to be taken not to fool one's self by so-called 'sighted' listening. The latter can create some expectation that biases perception. Its even possible for such bias to be introduced by looking at an FFT before listening since how a measurement looks, or otherwise being aware of a measurement result, can have some influence on expectation. It may help some to keep in mind that not every problem shows up very well on PSS FFTs, as ESS correctly pointed out.
In any case, powered hubs usually rely on SMPS that are not particularly clean relative to what can matter for analog circuitry. Experimentally, what works best and most reliably is to isolate dacs from computers using techniques like isolated power, galvanic isolation, and in some cases reclocking. Not every computer/hub/SMPS/DAC/etc., is the same, so symptoms may vary from one situation to the next.
Also IME, its not uncommon that some noise problems are not so easy to measure, although sometimes evaluating FFTs in non-standard ways can be used to show correlation between (possibly nonquantitative) measurements and audible changes in sound. Obviously care needs to be taken not to fool one's self by so-called 'sighted' listening. The latter can create some expectation that biases perception. Its even possible for such bias to be introduced by looking at an FFT before listening since how a measurement looks, or otherwise being aware of a measurement result, can have some influence on expectation. It may help some to keep in mind that not every problem shows up very well on PSS FFTs, as ESS correctly pointed out.
Last edited:
Do you have a reason to use inrush control for your PC? I wouldn’t want to upset the switching power supply myself, and have found that they always last longer when left on anyways.
As for the 5v isolated supply, that has always been a welcome change for me as well.
As for the 5v isolated supply, that has always been a welcome change for me as well.
Yeah? What about it? If the RF crap gets onto the analog circuits I can see how it could affect the sound quality. As long as the crap stays on the digital side and doesn't cause the PLL in the SPDIF/I2S receiver to lose lock I don't see a problem.
Um. No. Page 141 describes the perceived degradation in sound quality from RFI/EMI. The statements made on Page 141 are made without any reference to a controlled experiment. So all you have is anecdotal evidence and/or Whitlock's opinion. Furthermore, there is no mention of USB on Page 141.
I do agree that EMI/RFI can enter through a USB cable. I stated as much above. I also agree that EMI/RFI can degrade the sound quality. But I disagree that Page 141 puts those two together.
Intermodulation or straight-up rectification. I was working for National Semiconductor when they launched the LMV85x series of EMI hardened opamps. It was a pretty interesting design review. Among other things they placed a GSM cellphone next to a circuit that had a plain opamp and the LMV851. They then called the cellphone. The resulting output voltage of the circuit is shown in Fig. 2 of this app note: https://www.ti.com/lit/wp/snoa817/snoa817.pdf
AN-1698 goes into a bit more detail: https://www.ti.com/lit/an/snoa497b/snoa497b.pdf
National used to have an engineering brief that showed the setup but it looks like TI took it down in favour of the more boilerplate documentation.
Tom
USB board clock oscillators are analog. IME clock oscillators can be surprisingly sensitive to power supply noise and or nonlinearity, including from simple things like using ferrite beads and or X7R bypass caps. Some of the effects can be so audible as to be considered obvious. When possible effects are not clearly obvious I call in some trusted listeners and use a process that reduces the number test subjects and the number of trials to much less than would be needed if using more conventional methods. Its a separate subject in itself that probably justifies its own thread if there were sufficient interest. Since such methods are outside of the area of expertise of most EEs, it can take some time and effort to get up to speed. In the past I have offered to advise individual members about how to go about setting up such as system for themselves if they would like to contact me by PM. Nobody ever did, so it appears that nobody really cares.
Just as very brief observation, Sean Olive has described the benefits of using trained listeners for his work evaluating speakers and headphones. I find similar results for audible differences beyond that of transducers.
Regarding veiled and or grainy sound, the above methods are also used to evaluate any perceptual changes of that nature. We have observed the effects described by Whitlock, as well as other effects.
On the subject of intermodulation or rectification, of course it can be either. A single PN diode can be used as a small signal modulator/demodulator. Large signals produce rectification.
Last edited:
The Fox and the Grapes is one of Aesop's fables,[1] numbered 15 in the Perry Index.[2] The narration is concise and subsequent retellings have often been equally so. The story concerns a fox that tries to eat grapes from a vine but cannot reach them. Rather than admit defeat, he states they are undesirable. The expression "sour grapes" originated from this fable.[3]
https://en.wikipedia.org/wiki/The_F...erns a fox,grapes" originated from this fable.
https://en.wikipedia.org/wiki/The_F...erns a fox,grapes" originated from this fable.
USB can be fussy about long cables.
I have a Kodak camera with 18 inches of USB lead.
I use a PICKIT 3 emulator and that sometimes drops out on a 3 meter cable requiring disconnect and reconnect.
Try to keep USB cables short as possible.
I have an el cheapo Chinese USB DAC on my PC and with no sound it is dead silent.
So no problems at my end.
My USB DAC drives speaker direct at a bout 1 watt.
If you start adding amplifiers etc to the DAC you can get in the realms of grounding problems causing noise.
I have a Kodak camera with 18 inches of USB lead.
I use a PICKIT 3 emulator and that sometimes drops out on a 3 meter cable requiring disconnect and reconnect.
Try to keep USB cables short as possible.
I have an el cheapo Chinese USB DAC on my PC and with no sound it is dead silent.
So no problems at my end.
My USB DAC drives speaker direct at a bout 1 watt.
If you start adding amplifiers etc to the DAC you can get in the realms of grounding problems causing noise.
Have you ever measured a difference in DAC performance as result of any of this? I agree that oscillators need clean supplies if you want low clock jitter, but ultra-low clock jitter may not matter for the DAC.
Many of the better DACs have clock jitter cleaners built in so they're really quite tolerant of clock jitter. The fs clock oscillators in audio are about marketing.
Tom
When jitter is random, J-test is not useful. For close-in phase noise it is possible to take a look at dac output spectral line noise skirts on a very high resolution FFT.
However it can be harder to disentangle phase noise from Vref amplitude noise which also affects the skirts. There is difference in how they affect skirt shape which is discussed in the attached document in pages 3-15, with the noise skirt difference illustrated on page 15. (forum post example at: https://www.diyaudio.com/community/threads/phase-noise-in-ds-dacs.387862/post-7063038 )
One issue is that the visualization non-quantitative in nature, much like looking at some differences in scope waveforms.
The other type of phase noise we might want to measure is called far-out phase noise. It may show up as a signal correlated shift in the average noise floor of an FFT. Because its energy is distributed across many bins, its significance may be underestimated. The test involves measuring noise floor with and without a high level test signal. Again though, there are potential complications when sigma-delta dacs are involved. IN that case what appears as noise on an FFT can vary with audio signal level in various ways. in some cases noise variation can appear as a periodically repeating pattern as dac output voltage is stepped across the output voltage range.
Also worth mentioning is that some of the literature treats jitter and close-in phase noise as two different things, since can affect signal detection in different ways. In that case jitter refers to the time domain effect that corresponds to far-out phase noise in the frequency domain.
Of course, merely showing a measurement result of something that may be not so easy to measure very well with commonly available instrumentation, says nothing about audibility. Correlation of audibility with measurements, and or with experimental circuit perturbations to circuits, requires listening tests that are devised to effectively counter the effects of perceptual bias. These days it seems that most testing of that nature is done for internal or proprietary use, rather than for academic publication. That's especially true now that that a lot public source grant funding for academic research has dried up. The cost in time and effort for publication quality research is beyond reach of most companies, and even it were done to that level there would no point in giving it away for free to competitors. Thus, we never see a lot of modern research in the public domain. I could go on and on about how difficult it can be to convince skeptics that old research and related old models are wrong enough that they need to be superseded. One case that readily comes to mind was where a multitude of research publications over a period of 20-years ending up being needed to change an existing and dogmatic academic behavioral model. Thus no interest on my part in trying to convince skeptics by that mechanism. I have invited and continue to invite people here to listen to the same system at same time. So far, no skeptic has responded to my offer. IMHO there is nothing more convincing that hearing something for yourself, including some blind testing if you want. Of course, there will always be people who haven't learned out to listen for audible effects in reproduction systems. One person who figured out the right way to learn describes what he found in his own words at:
https://www.diyaudio.com/community/threads/the-black-hole.349926/post-7310119
Getting back to what we can hear at this location using trained, expert listeners, yes, the effect of jitter/close-in phase noise can be quite obvious. For one particular example, it can be obvious with certain oscillators when their power passed though a typical ferrite bead (presumably related to hysteresis noise).
Regarding a sort the point of diminishing returns for low jitter/phase noise in dacs, we have tried Andrea Mori SOA-measured Driscoll oscillators with SC-cut crystals as compared to Crystek 957 clock modules which I have babied in every way which has been found here to make any difference. This includes careful attention to layout, power, bypass, and output buffering, and the babying of output buffers to the same extent, with the sole exception of buffering their outputs. The result with a custom AK4499 dac, running in DSD volume bypass mode, was that additional phase noise performance of the SOA clocks was audible, but deemed only slightly so by the listening panel. (It may help to bear in mind that we use very low distortion ESL panel speakers, or else planar headphones, either of which result in much less transducer distortion masking of upstream electronics as compared to box speakers and or to more commonly used headphones. This is one of the reasons I invite people here to listen, because of system transparency relative to DUT audible characteristics.)
However it can be harder to disentangle phase noise from Vref amplitude noise which also affects the skirts. There is difference in how they affect skirt shape which is discussed in the attached document in pages 3-15, with the noise skirt difference illustrated on page 15. (forum post example at: https://www.diyaudio.com/community/threads/phase-noise-in-ds-dacs.387862/post-7063038 )
One issue is that the visualization non-quantitative in nature, much like looking at some differences in scope waveforms.
The other type of phase noise we might want to measure is called far-out phase noise. It may show up as a signal correlated shift in the average noise floor of an FFT. Because its energy is distributed across many bins, its significance may be underestimated. The test involves measuring noise floor with and without a high level test signal. Again though, there are potential complications when sigma-delta dacs are involved. IN that case what appears as noise on an FFT can vary with audio signal level in various ways. in some cases noise variation can appear as a periodically repeating pattern as dac output voltage is stepped across the output voltage range.
Also worth mentioning is that some of the literature treats jitter and close-in phase noise as two different things, since can affect signal detection in different ways. In that case jitter refers to the time domain effect that corresponds to far-out phase noise in the frequency domain.
Of course, merely showing a measurement result of something that may be not so easy to measure very well with commonly available instrumentation, says nothing about audibility. Correlation of audibility with measurements, and or with experimental circuit perturbations to circuits, requires listening tests that are devised to effectively counter the effects of perceptual bias. These days it seems that most testing of that nature is done for internal or proprietary use, rather than for academic publication. That's especially true now that that a lot public source grant funding for academic research has dried up. The cost in time and effort for publication quality research is beyond reach of most companies, and even it were done to that level there would no point in giving it away for free to competitors. Thus, we never see a lot of modern research in the public domain. I could go on and on about how difficult it can be to convince skeptics that old research and related old models are wrong enough that they need to be superseded. One case that readily comes to mind was where a multitude of research publications over a period of 20-years ending up being needed to change an existing and dogmatic academic behavioral model. Thus no interest on my part in trying to convince skeptics by that mechanism. I have invited and continue to invite people here to listen to the same system at same time. So far, no skeptic has responded to my offer. IMHO there is nothing more convincing that hearing something for yourself, including some blind testing if you want. Of course, there will always be people who haven't learned out to listen for audible effects in reproduction systems. One person who figured out the right way to learn describes what he found in his own words at:
https://www.diyaudio.com/community/threads/the-black-hole.349926/post-7310119
Getting back to what we can hear at this location using trained, expert listeners, yes, the effect of jitter/close-in phase noise can be quite obvious. For one particular example, it can be obvious with certain oscillators when their power passed though a typical ferrite bead (presumably related to hysteresis noise).
Regarding a sort the point of diminishing returns for low jitter/phase noise in dacs, we have tried Andrea Mori SOA-measured Driscoll oscillators with SC-cut crystals as compared to Crystek 957 clock modules which I have babied in every way which has been found here to make any difference. This includes careful attention to layout, power, bypass, and output buffering, and the babying of output buffers to the same extent, with the sole exception of buffering their outputs. The result with a custom AK4499 dac, running in DSD volume bypass mode, was that additional phase noise performance of the SOA clocks was audible, but deemed only slightly so by the listening panel. (It may help to bear in mind that we use very low distortion ESL panel speakers, or else planar headphones, either of which result in much less transducer distortion masking of upstream electronics as compared to box speakers and or to more commonly used headphones. This is one of the reasons I invite people here to listen, because of system transparency relative to DUT audible characteristics.)
Last edited:
Thanks! I aciddentally damaged the thermistor in the pc psu, and there are no markings on it so cannot tell what would be a good replacement. I'm thinking about the cl-60 that i heard pass labs use, but it only has an impedance of ten ohms with four amps current handling, am not sure of this will be enough impedance to have the desired soft-start effect for british ac 240 volts. What do you think, will 10 ohms be enough
With respect. My dac is silent as well, absolutly no noise hiss crackling coming through the speakers, but when i hooked up the external psu and cut off the pc's 5v supply through the usb cable, the sound quality definitely mproved. I swapped it around a few times and the modded cable set-up sounds so much better. My brother agrees as well. This is real life not just a point of view or opinion
It could be the pc's noise is preventing the dac from acheiving its full potential
For a quick cheap mod and a big improvement in SQ it is definitely worth a try if anyone wants to corroborate these findings
Before diving into the deep end you could first explain how you compare the oscillator with/without ferrite bead or X7R? No A/B is possible so this is purely subjective (i.e. figment of imagination).
And since you often like to cite Sean Olive this is a good reading:
http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html
A couple of quick comments, then will have to return later. First, I probably should have mentioned that a claim to the effect that "dac jitter is inaudible" according to some literature would appear to result at least in part from a misreading of that literature. In particular, a 'threshold of audibility' is not a hard limit below which no human can hear. Rather such thresholds are estimates of an average value for a population. IOW, it is expected that 50% of the population should be able to hear below the threshold. To put it yet another way, a threshold of audibility is an estimate of the center of a bell curve.
Secondly, on the matter of subjective versus objective, we have been through this many times. Measurements of human auditory discrimination are considered to be objective, not subjective. An example almost nobody would claim was subjective would thresholds of audibility. In such cases the sensor is biological, yet the measurement is objective. Not clear why some people appear stubbornly resistant to understanding that.
EDIT: Regarding A/B discrimination testing, the purpose of such tests is to keep the time between comparisons short relative to average auditory working memory duration. Humans also have some long term memory capability. When A/B differences are plainly obvious to a particular test subject, the time between listening to A and then listening to B can be rather long. As long as one or more days can be feasible if sounds have sufficiently been committed to memory. For example, it doesn't take a quick A/B comparison to reliably discriminate the sound of an oboe from that of a violin.
Secondly, on the matter of subjective versus objective, we have been through this many times. Measurements of human auditory discrimination are considered to be objective, not subjective. An example almost nobody would claim was subjective would thresholds of audibility. In such cases the sensor is biological, yet the measurement is objective. Not clear why some people appear stubbornly resistant to understanding that.
EDIT: Regarding A/B discrimination testing, the purpose of such tests is to keep the time between comparisons short relative to average auditory working memory duration. Humans also have some long term memory capability. When A/B differences are plainly obvious to a particular test subject, the time between listening to A and then listening to B can be rather long. As long as one or more days can be feasible if sounds have sufficiently been committed to memory. For example, it doesn't take a quick A/B comparison to reliably discriminate the sound of an oboe from that of a violin.
Last edited:
It is not only about subjective versus objective. Auditory sensory memory has a short span (maybe up to 1 minute) so if proper A/B comparison is not possible the possible difference observed is purely a figment of imagination. Not clear why some people appear stubbornly resistant to understanding that.