|
Home
/ FAQ
/ News Classifieds / Events |
Audio Asylum Thread Printer |
Get a view of an entire thread on one page |
85.140.94.233
Wanted to share this with you and ask for any comments or recommendations. I have a standalone DAC with PCM 2902 inside. The DAC is used via SPDIF from my CD player and TOSLINK from my DVD player. I intended to use it via USB in from my laptop (fed from a separate power line), but was concerned about 120 mV of AC between the grounds – clearly not good. In addition, my stereo system is more than 5 meters away from my laptop. So, I came up with this solution.
A USB hub at the end of a 5m USB cable, then 0,5 meter USB cable to DAC. The hub uses a modern NEC 720114 chip. I made it run in self-powered mode off a battery. Still, the 720114 chip needs Vbus from the PC for device recognition (on a separate Vbusm pin), but the USB bus is not used to power the hub (there are diodes installed for this purpose). Then, PCM 2902 in the DAC again needs Vbus for recognition (this time it is the power from the battery via the USB hub, which has some sort of DC line filtering via ferrites and caps), although all of its critical areas are powered internally within the DAC.
This is what I did: first of all, I separated the shield ground lines between the upstream and downstream USB port (had to saw a little across one side of the USB hub PCB). Second, I tried disconnecting the USB ground line running from the PC (by unsoldering the respective pin on the IN socket in the hub). This resulted in a system message stating that USB hub not being recognized properly. To solve this AND provide some sort of noise isolation between PC – hub – DAC, I soldered an inductor (Bourns RLB 100mH) between the USB bus ground line and the hub system ground. Now, my multimeter does not show any AC between the devices. So far, things are working, but it is too early to make any observations re the sound. Also, I do not have a scope at my disposal, so can’t make any measurements as to groundline noise reduction.
This might prove an effective way to provide some sort of isolation between PC and hi-fi system, definitely more accessible than $500 cables. However, maybe Gordon Rankin, audioengr, John Swenson or other interested parties can provide some input on the following questions:
1. In your opinion, is a 100mH inductor an adequate solution? What could be better for filtering the most typical PC noise on the USB ground line? (again, don’t have a scope, so can’t tell what are the offending frequencies).
2. It is my impression that the NEC 720114 chip actually buffers and “reclocks” USB packets, thus potentially helping reduce the resulting jitter, especially on a link over 5 m long. What do you think?
The biggest problem with USB is the USB clock. Devices like the Trend UD10 "reclock" the datastream in a clean external environment.
Soundcards also reclock like the UD10, so toslink off a sound card is another option that will have better timing and also provides a non-conductive cable to the DAC.
A soundcard clock, while always better than USB, must also perform in the harsh environment of the PC. So it's not ideal.
The proper solution is a DAC that reclocks all input. Most do not. Transports of any kind don't matter with a good reclocking solution.
Since we are into the subject, what about removing pin 1 ( power) from the cable at the PC end. Since there is some noise in the PS wouldn't it be beneficial to not have it running in the cable under the shield next to our data line?
Bob
www.PlateauLight.com
AFAIK, most of USB chips (at least the ones involved here) need both the differential data pair, AND the power lines for device identification, even if running in self-powered mode.
I think it is Gordon who uses a different TI USB chip at the DAC end that needs just the differential data pair. I guess your suggestion might work in this case.
Actually even the 2706/7 can run fine without the actual +5 volt wire on the cable in most circumstances.
You HAVE to have the ground wire though. You still need a voltage reference (ie ground) for the differential pair. As I said before putting chokes etc in line with the ground is a bad thing to do as well. Some designs do that to try and meet EMI specs, but all it can do is mess up the quality of the data transmission.
John S.
Very interesting post, I've actually been thinking along these lines for awhile.First off, inductor in the GROUND line is very bad, almost certainly going to make things worse.
A few general thoughts on ground loops:
first notice the word LOOP, in order to have a ground loop you must have a loop, a single ground wire does not a loop make, in order to have a ground loop you have to have at least two ground paths. So if the ONLY electrical connection between the laptop and the DAC is the USB cable, you cannot have a ground loop and you are fine. So if your laptop is running off batteries and there is no other connection to anything else, just the USB cable you can't have a ground loop.The problem occurs if you have the laptop plugged into the wall, and the ground of the USB cable is connected to either the neutral or the safety ground, then presto, ground loop. Laptops actually have an advantage here because a lot of the power supplies do not connect to the safety ground and are isolated from the neutral. If yours has a three pin plug then things get a bit more complicated. Use the meter and measure the resistance from the ground on the USB port to the ground pin on the supply. If the resistance is high you are fine, if its low then you are going to have a ground loop through the power line.
For long distances and or ground isolation I like the optics cable, its not cheap but really does work well, BUT as Gordon said you need to get rid of the supply that comes with it and make your own high quality supply. Its not hard, the thing barely uses any current so a big honking supply is NOT needed.
On the reclocking and hubs, this is an interesting subject that I have been looking into recently. It turns out that all hubs that I could find DO reclock, but they use a PLL to generate the clock. They use either a 6 or 12MHz crystal and a PLL based frequency multiplier on chip to generate a 48MHz signal used to read the input and clock the output. Since the power supplies are rerely done to "audiophile" standards, and they usually get driven off VBUS anyway, the feed to these PLLS is usually noisy as all get out so the clock jignal has very bad jitter.
I searched and searched and found a hub chip that allows you to bypass the on chip PLL and feed in an external 48MHz clock. So it looks like its possible to build a decent reclocking hub using a really good PS and a very low jitter 48MHz clock. With good board layout it should be possible to make a hub whoose output jitter is at least an order of magnitude less than any other USB connection. The only problem is I forgot to write down the chip number! (you'd think I'd remember my memory is now the proverbial steel sieve) If I can ever find this again I'll post it.
John S.
I found the chip! Its the TI TUSB2046B, digikey has them.
John,
Jitter as it applies to audio is the clock associated with the data. Since the 48MHZ clock is merly the USB clock and the relaying of data it has nothing to do with the quality of audio since there is no real "clock" associated with the audio data.
In adaptive mode the dac will still be goverened by the computer not the hub.
But a well designed hub could isolate the power from the PC which might be a good thing.
Thanks
Gordon
J. Gordon Rankin
Hi Gordon,
I did a lot of experimenting and testing the recovered mclk with a 2706 and found that the jitter on the bus itself has a significant effect on the mclk jitter. Since the hub chips do retransmit the data based on the 48MHz clock, lowering jitter will improve things at leat for a 270x receiver. I have not done these tests with a 1020 so I'm not sure how it will respond to this.
John S.
John,
Basically the way the PCM270x works is that it links the Spact to the SOF frame. Now since the SOF frame has no acknowlegments to it. There is also a VSOF (virtual) timer in the PCM part in case the real SOF frame is missed. The problem is the VSOF timer will always fire after the expected SOF. Which actually throws off the Spact unit, because now the expected SOF time differential is lost.
I have seen on allot of ocasions especially with my crappy HUSH pc that SOF frames are the first to corrupt for some reason and maybe this will lead to what you are saying.
Or maybe this has nothing to do with Jitter on the USB link at all.
I did some really in depth testing with ASYNC mode running 1KHz sine wave from the MAC (Faber Acoustics Signal Suite 2) to the emulator running on the TAS1020 board. In a 24 hour period running at 44.1/16 I did not recieve any packet errors.
This of course would have the same effect on any Adaptive part not only the PCM270x series.
Only packet errors would effect ASYNC mode.
Thanks
Gordon
J. Gordon Rankin
![]()
I thought it might be curious to see this hub design (prior to my modification). It seems at least some sort of consideration went into designing the power supply. If you read the NEC 720114 datasheet, it appears that Vbus can be used just for "handshake", and the rest of the chip is powered by the +5 V supply through the EH11A regulator. In my case, it runs off a rechargeable battery. I might eventually remove D2 for the extra assurance. Thus, I did not see it relevant to install the inductor on the Vbus line. The way I have it now is in hope it alleviates some hi-frequency currents between my hi-fi and my laptop via USB ground. By the way, ALL of my power supplies are TWO pin, for better or for worse.
The fact that there is a crystal on this board may indicate that it is probably using a PLL internally to reclock. The datasheet is not much help. One thing that you could try is to replace the crystal with a low-jitter oscillator.
Steve N.
Thank you for your comments. Another reason I have a feeling this chip might reclock is the way it simultaneously handles USB 1.1 and 2.0 "streams" ("split transaction", "transaction translation", etc.). Do not pretend to have an in-depth understanding of USB protocol, but my guess is that it might involve reclocking.
So far I find an audible improvement when powering the hub from an external battery. Will have to listen more. I will be also trying to remove the D2 diode to disconnect for sure the downstream USB bus power from the PC (in this case the only connection to the PC power bus is via an RC circuit to the Vbusm pin - second from the borrom on the left side of the chip - used ONLY for monitoring whether there is a computer attached).
Will also try other things to bring the picture to a resonable ideal - only the differential USB pair running to the USB chip, and power and ground lines isolated (or filtered) as much as possible.
Then might try the oscillator route...
The inductor may help in your situation, but it really does not provide isolation, only attenuation of high-frequencies. Listening tests required. As for the hub "reclocking", it is certainly possible, but doubtful. It is more likely that it buffers and clocks the data through a flip-flop. This is a far-cry from useful reclocking. The jitter will still be there and get worse.Isolation can be achieved using an RF or inductive coupling device. These are active devices. One could also use a pulse transformer, but the USB interface may need DC to operate properly. These will break the ground connection between the computer and the audio system - galvanic isolation.
I chose to isolate further down the chain, in the I2S interface or the S/PDIF output interface.
Steve N.
Thanks for your comment, Steve. Could you just elaborate a little on this "flip-flop" part?
Flip-flop is a fundamental element of synchronous logic. It stores the state of a single bit. If the clock is recovered from the USB stream, then this clock can be used to clock the data bits through the flip-flop and then the stream can be re-encoded and re-transmitted. Not sure if this is what the device you have is doing, but possible. When you reclock this way, it can improve the edges by speeding them up, but it will not improve the timing. The jitter still gets worse as you put the stream through more stages of buffering or clocking.
Steve N.
Pry,
You would be better off with an opticis cable and a better power supply.
First using a 5M cable into a hub is not a good idea. A better one would be too use a a 3m then hub then 2m. But even then it's not a good idea and the ground imbalance between the computer and the dac is going to have real issues.
You need to have the computer, dac and stereo on the same ground potential and to do that you usually need to be on the same circuit.
The opticis is optically isolated. The power supply they send is crap but there are other solutions for that.
Thanks
Gordon
J. Gordon Rankin
Opticis is very effective and I like it a lot. Please power it with battery. If you are using some battery pack with universal voltage output, please make sure that it is linear regulated DC output. I read one post here with negative feedback using Opticis. He uses a universal voltage battery pack. I suspect the DC-DC converter is switch mode converter for high energy efficiency which is very noisy.