|Block diagram of the USB 'PHYsical' chipset. UTMI stands for USB 2.0 Transceiver Macrocell Interface, a standard bus interface for the USB chipsets out there...|
One could see the "glass half full" when we run into imperfections in the audio system. For example, it was serendipitous that the Emotiva XSP-1 pre-amp had such a noise-sensitive unbalanced "HT Bypass" input that I was able to detect noise and measure it previously. If it were not for this issue, I probably would never have thought about using the Corning USB 3 optical cable or consider how to isolate the noise originating from the computer USB connection to the DAC simply because in my system using balanced interconnects throughout, I had never had a problem (subjectively audible or objectively verifiable).
As a quick recap, I was able to hear and measure the 8kHz "microframe packet noise" (125us) originating from the computer's USB port. The HTPC is based on the ASUS B85M-E/CSM motherboard running the Intel Pentium G3220 CPU. I use one of the USB 3 connectors although I have tried the USB 2 ones as well and they're just as noisy. So, step-by-step, I can show you how I brought that noise down:
0. Let's start just by showing what "silence" looks like in the system through the pre-amp's analogue out.If I don't have the USB cable connected to the DAC. Here's what the balanced output looks like from the XSP-1 preamp from 1kHz to 30kHz:
This gives us the "baseline" to start from. There is some inaudible noise in the system - realize that I have my Transporter and Onkyo receiver connected to this pre-amp at the same time so some of this low level stuff could be from these sources as well. Theoretically, if there's no noise at 8kHz (around where that cursor is, hard to "lock" on to 8000Hz as there is no tone there to snap the cursor on to), the noise floor is around -146dB. Ideally then, we'd like to see this when we connect the USB TEAC UD501 DAC to the computer.
I. Nasty 8kHz noise directly from computer to USB DAC:
HTPC --> 10' generic USB cable --> TEAC UD501 DAC --> 6' balanced analogue cables --> Emotiva XSP-1 preamp --> 6' balanced analogue cables --> E-MU USB0404 ADC --> measurement computer
To the left we can see the cursor is locked on 8kHz and the signal is -102.5dB. Note that this might seem low but in fact as it goes through the gain in my monoblocks, I can hear this annoying pitch (along with other electrical noises) as I demonstrated in the video previously.
The 8kHz primary frequency and resonant frequencies up at 16kHz, 24kHz can easily be seen. There's also some noise at 9kHz but below -120dB. Notice that I'm running this FFT out to 30kHz (above audible range just to make sure there's no obvious anomalies up there).
II. Let's put a USB hub between the USB cable and the DAC:Suppose now we put a USB hub in the path... Remember that in the previous post, I had to do this eventually anyhow because the Corning USB 3 optical cable does not supply power which was needed for my TEAC DAC to recognize the connection to the computer. So let's put in the same Certified Data USB 3 4-port unit that I used in my previous post.
HTPC --> 10' generic USB cable --> Certified Data USB 3 hub --> generic 3' USB cable --> TEAC UD501 DAC --> 6' balanced analogue cables --> Emotiva XSP-1 preamp --> 6' balanced analogue cables --> E-MU USB0404 ADC --> measurement computer
This hub is one of the early 4-port USB3 devices. I think I bought it more than 3 years ago (around 2012). As you can see, placing the hub in the data chain doesn't do much to the noise... It's down to -103.5dB which is just a 1dB improvement. Certainly the distortion does not sound any less annoying!
This shows us that a USB hub itself does not attenuate the conducted USB noise from the computer.
III. Let us now use the 33' Corning Optical USB 3 cable to connect to the computer, same hub as above:
HTPC --> Corning Optical USB 3 cable --> Certified Data USB3 hub --> generic 3' USB cable --> TEAC UD501 DAC --> 6' balanced analogue cables --> Emotiva XSP-1 preamp --> 6' balanced analogue cables --> E-MU USB0404 ADC --> measurement computer
There you go... Clearly the optical cable has done a very noticeable job in attenuating the USB noise from the computer! The 9kHz tone is below noise floor now. As I discussed previously, the optical connection provides significant electrical noise isolation from the computer's USB port. We see the 8kHz tone drop from -102.5dB to -118dB. A very significant difference of 15.5dB. This drop has now rendered the annoying high-pitched tone inaudible to me in the soundroom. Job pretty much done :-).
However, since curiosity is perhaps the most important trait to have in order to ascertain truth and the nature of things (and one which I try to instill in my children), the obvious question is: do other powered USB hubs, likely using other PHY chipsets result in lower 8kHz packet noise?
Well, since I have a few USB hubs at my disposal (and I was wanting to buy a new USB 3 hub anyway for work), let's measure a few in place of the Certified Data model I have used up to now and see if this 8kHz tone changes...
So for the following tests, the connections will go like this:
HTPC --> Corning Optical USB 3 cable --> [USB HUB BEING TESTED] --> generic 3' USB cable --> TEAC UD501 DAC --> 6' balanced analogue cables --> Emotiva XSP-1 preamp --> 6' balanced analogue cables --> E-MU USB0404 ADC --> measurement computerAs best I can, I will use the same USB cables between the hub and DAC (not that I believe USB cables make any difference unless faulty). Unless specified, the included wallwart power supply was used with each hub.
This 4-port USB 2 hub has been with me since 2006 at the office! It has worked flawlessly, is fast, never any issues. Good metal construction as well. Probably paid about $25 when it first came out.
Not bad, the 8kHz tone down to -119dB.
B. Cheap generic portable USB 2 hub:
Ewwww. Look what happened here! The 8kHz noise is up at -107.5dB which is horrible, almost as bad as straight off the computer USB port and negating most of the benefit from the optical USB cable. Furthermore, notice the noise floor elevation below 14kHz. This is audible as white noise. This is a bad, noisy hub!
8kHz noise down at -118.5 dB. Not bad.
8kHz noise at -119.3 dB. Excellent.
Summary and Conclusion:Okay, let's just cut to the chase.
1. I've shown again that the Corning USB 3 optical cable paired with a powered hub is capable of removing much of the nasty USB noise originating from the computer's motherboard USB connection. The vast majority of the improvement I saw came from the use of this fibreoptic cable.
2. Powered hubs are not created equal. Here's the summary of the improvement over a direct connection to the computer USB port (sorted from worst to best in terms of lowering of the 8kHz packet noise):
Optical USB + Cheap USB 2 hub bought in China = -5.2 dBAs you can see, the cheap USB 2 hub I bought in China was atrocious and noisy! Not only did the 8kHz tone attenuate by only a small amount, but it even caused audible broadband elevation of the noise floor below 14kHz (see FFT above). Otherwise, the others were all relatively close; essential +/- 1dB difference. Note that I did measure each hub a couple of times using 2 different ports and the relative attenuation remained essentially the same. There are obviously too many hubs out there to measure each one but this sampling did not show any pattern as far as I can tell other than the best performing being the two Kensington products. One was an old USB 2, 4 port model and the other being one of their latest 7-port USB 3 hubs.
Optical USB + Certified Data 4-port USB 3 = -15.5 dB
Optical USB + Logiix 7-port USB 2 = -16.1 dB
Optical USB + Kensington PocketHub SE USB 2 = -16.7 dB
Optical USB + Kensington UH7000C 7-port USB 3 = -16.8 dB
Other than the very inexpensive hub bought inside China, the others were essentially playing in the same ballpark.
3. The only lingering question left for me... How does this level of noise attenuation - especially that 8kHz packet noise reduction compare to commercial audiophile products. I posted a few weeks ago after the Vancouver Audio Show about the forthcoming inexpensive (~US$50) AudioQuest JitterBug. As I said, the quote I heard was "reduces noise by 6dB, and jitter by 8%". Hmmm, in the context of what I'm showing here, 6dB doesn't seem like much. Furthermore, I have yet to see anyone demonstrate the effect of jitter with modern asynchronous USB DACs in the analogue output so would question the benefit of this "8%" jitter reduction on the sound quality even if objectively demonstrable in the USB digital signal.
Then for US$175 we have the UpTone Audio USB Regen. Basically, we have a single port hub... But custom built to reduce the packet noise and "ground plane" noise as well as a better, lower noise power supply. I'd be very curious whether this device works in the same fashion as what I've been describing and measuring here. The Corning Optical USB 3 cable + Kensington USB 2 hub costs about US$130 before taxes and you can likely get free shipping off Amazon these days. Nonetheless the price differential is small. However the optical cable and USB hub can be repurposed for other uses in the future. Lots of words and descriptions on the UpTone website along with claims that $0.40 worth of resistors "increased its musical performance by 40-50% - most especially the bass!" presumably using some kind of Squeezebox Touch setup and significant differences in sound between a 'nasty' 6" cable and 'smooth' Supra USB cable. Right, so that's how audio design is done... I see they also looked at the Corning optical cable in that post and makes various claims about audible sonic changes. Strangely no measurements to confirm changes since this is primarily a noise reduction device, isn't it? Needless to say, I would love to get my hands on this USB Regen device (one night would be enough to make some observations :-).
There are of course other products out there like the iFi iPurifier. Notice in that web page they have a little graph showing EMI noise as 34dB from the iPurifier vs. 39dB from the computer USB port. 5dB noise reduction? Similar to the JitterBug; I can see this as realistic.
Okay, to end off... Let me remind everyone one more time that what I'm measuring with the 8kHz noise here is because of the Emotiva XSP-1's sensitivity to noise through the "Home Theater Bypass" input. The 8kHz tone is NOT something I have ever heard / measured coming out of a reputable DAC's analogue outputs! Nor is it something that's found in Stereophile USB DAC measurements. If it were not for this noise sensitivity, I would not even have bothered with looking at ways to attenuate the computer USB noise. Nonetheless, I think this is an interesting real-life demonstration of the noise pollution that can come out of the computer's USB port and a solution that works reasonably well. I did not bother with jitter tests this time as I have never seen the Dunn J-Test change in any substantial fashion with the use of a hub with an asynchronous DAC (see the measurements last time). Think you have a jitter issue? Save up the cash and buy a better asynchronous USB DAC - forget cables and tweak products IMO.
Well, they're at it again... I mean LH Labs and this time it's the Geek Out V2 (US$250+shipping) on crowd-sourcing.
About a year back I actually pre-ordered the first Geek Out but later cancelled as it was taking too long; this time word is that they might be ahead of schedule. Well, I'm a sucker for little DACs and since I don't have an ESS Sabre DAC to play with, I figure this might be fun to have at not much cost. This new model also has a few new features like balanced output and 2 headphone amp power settings (100mW & 1000mW), the 100mW setting should be good to keep the little DAC cool. Looks like it can handle up to DSD128 as well.
Now, if any of you have an interest in getting one, feel free to CLICK HERE and order using my referral code... Apparently it's one of those deals where they'll give me an upgrade or something if a few folks buy. Seriously, no pressure, whatever comes gets measured :-) !
Remember: Don't forget the Digital Filters Test is ongoing! We're at 35 responses at the 1/2 way mark. Thanks to everyone who has experienced for themselves and entered their results. Would love to get to 50 by the end of this month!
Well, it's getting sunny and hot on the West Coast; time to go play with the kids. Hope you're all enjoying the music as we get into June...