Jump to content
IGNORED

Overall Isolation - network, USB, and power


Recommended Posts

A little more detail on what the Intona is doing.

 

It does isolate ground and power. The usb handshake works with detectors on both sides for this with separate paths through the isolation for these conditions on the bus so they can be recreated on the other side.

 

The 5V VBUS on the downstream side is provided by an isolated switching DC/DC converter. The VBUS from the upstream side drives a high frequency oscillator which feeds a transformer, the other side of the transformer drives diodes and a filter then into a regulator. The quality of the 5V coming out of this is decent but not spectacular.

 

The digital isolation system and the transformer in the power system DO block almost all leakage current. The transformer is a high frequency transformer so it does block most leakage components which are mostly lower in frequency. Some higher frequency components will pass through the transformer, but they are going to be much lower in level than the high frequency POWER signal which is what the transformer is designed to pass.

 

Putting a REGEN after the Intona cleans up the issues with the not so great signal integrity and noise on the 5V line, but can introduce an additional leakage loop through the REGEN supply unless you use a battery or LPS-1 to power the REGEN.

 

Going with a microRendu powered by an LPS-1 is a really good choice since it prevents leakage loops from happening without needing other equipment.

 

As far as what is going on with the ETHERNET side of things, I will be very interested in hearing reports from people that use an LPS-1 with a microRendu whether Ethernet isolators etc make any difference in this configuration. I still don't have any good theories as to why an Ethernet isolator makes any difference given that Ethernet already has transformers on both ends. (this is assuming you are not using shielded cables that actually connect ground from one box to another)

 

John S.

Link to comment
(Yes i understood that there is another thread about this but...)

 

@John : Daniel the intona CEO wrote on the thread about Usb isolators that the intona power regulators are LINEAR and not switching.

 

B.

 

 

Sent from my iPhone using Computer Audiophile

 

Tplink optical bridge/etalon streamer/audiogd nfb29/linn klout/athom sirrocco + athom rafale v38 hypex

 

The REGULATOR is linear, but the power isolating system is a transformer. Transformers don't work on DC, you have to take the DC from the upstream VBUS and convert it to AC in order to feed power through the transformer. On the downstream side the AC goes through diodes to convert to pulsating DC, then through some form of filter to try and smooth out the waveform, THEN into the linear regulator.

 

Everybody does the AC generating as simple switching, so it is a square wave going into the transformer. Thus the output of the transformer has a LOT of high frequency noise. The filter after the diodes has a lot to deal with and leaves some noise still there. Regulators are not perfect devices, most leak a fair amount of high frequency noise, so the output still has some high frequency noise going into the VBUS pin.

 

John S.

Link to comment
Hi

Square ac upstream (if it is square) will not even be square on the other side, and i don't know what kind of filter and linear regulator is behind... maybe Daniel could give us info on it...

 

What you described there is not what you called in your previous post "a switching regulator", that subjectively refers to somethething else, i mean a switching power regulator of a supply, that is supposed for everyone reading this thread to be noisy by itself.

 

 

 

 

Sent from my iPhone using Computer Audiophile

 

Tplink optical bridge/etalon streamer/audiogd nfb29/linn klout/athom sirrocco + athom rafale v38 hypex

 

I said it was an isolated switching DC/DC converter. I'm sorry if that gave the wrong impression. At the time I didn't have time to write out the full description of what that was.

 

Looking at the picture of the Intona board the transformer is driven by a CMOS 4030 which is a quad XOR gate, so definitely a square wave driving the transformer. There are two diodes that drive some form of electrolytic cap, I don't see any inductor in that area so I'm guessing it is a a simple single cap filter after the diodes.

 

The regulators are LD1117 which have a typical output noise of 100uV, the spec sheet does not have a graph of the PSRR vs frequency so it is hard to tell exactly how well it will reject the high frequency noise that gets past the capacitor. Other regulators which have similar specs do not have very good PSRR above 50KHz which is most likely where the converter is running at.

 

As I said before this is decent but not spectacular results. There are other isolating power converters out there which are much worse. Intona did a good job NOT going the easy route and using one of the off the shelf converters which are truly abysmal.

 

John S.

Link to comment
  • 2 weeks later...

Hi All,

unfortunately I haven't had much time to participate in this thread, but this is related to a LOT of what I have been doing for the last several years, some of which have made it into products you all know about.

 

I have a little bit of time right now so I'll go into a broad overview of what I have found in my wanderings through digital audio.

 

First USB.

 

There are at least three issues with USB that need to be addressed, it is rare for all three to be dealt with well in the same system, making it possible to do this is what I have spending a lot of time on.

 

The three things are:

 

timing

signal integrity

leakage loops

 

Timing should be familiar to everybody on this thread. Its is the age old issue of where the clock that feeds the DAC comes from. Early implementations were terrible, having to sync a local oscillator to the USB signal. Asynchronous mode fixed that, allowed a local oscillator in the DAC to be in charge. Unfortunately MANY implementations did this very poorly so the advantage of async was not realized fully. Fortunately things are MUCH better today, with many manufacturers finally realizing how to do this.

 

Everyone thought this would BE IT, the ultimate interface. Unfortunately it was not so, there were OTHER things involved here.

 

The first I started looking at was signal integrity of the USB bus. Exactly how this makes it through to sound coming out of the DAC is not fully known at this point. BUT it does look like increasing the signal integrity at the DAC USB chip decreases small noise voltages in the ground plane of the DAC. Devices such as the REGEN are designed to increase the signal integrity going into the DAC USB chip.

 

Some manufacturers have added isolation techniques after the USB chip, trying to block the ground plane noises from getting to the clock and DAC chips. Again unfortunately many of these implementations have not been done very well. But even the ones that ARE done well still do not seem to completely get rid of cable and source dependencies. There still has to be something else.

 

Enter leakage loops.

 

I don't have time here to go into detail on leakage loops, I have written extensively about this in recent posts. A few important points. This is caused by leakage from the AC line through capacitances in the power supply to the DC output of the supply. They have NOTHING to do with the traditional "ground loops" that happen through the safety ground (third pin) of the AC wiring, it is something different. Earthing (ground rods etc) have nothing to do with this. Leakage loops always go through at least two power supplies: AC line, PS A, DC output to circuit A, interconnect to other box, DC of circuit B to PS B back to AC line.

 

Leakage loops depend on the type and implementation of both power supplies. Leakage current is line frequency (50/60Hz) plus harmonics, plus any crud from the PS. It is primarily low frequency AC, but higher frequencies can exist. The problem from the leakage loop is the voltage noise generated when the leakage current flows through a non-zero impedance, such as the shield on an interconnect between boxes (analog or digital).

 

ALL AC mains connected power supplies have leakage currents, SMPS seem to have considerably more leakage current than most LPS. These leakage loops have always been there in audio systems (except the very earliest ones that ran off batteries!), but computers in audio have exacerbated the issue considerably due to most computer systems running exclusively from SMPS. Thus many of the ills of digital audio have actually come from the leakage current of the SMPS rather than the "noise" of the digital systems themselves.

 

There are two ways to deal with the noise: block the loop or decrease the impedance of the loop. Blocking the loop means blocking the flow of the leakage current anywhere in the loop. There are several ways to do this, a battery supplying a component prevents leakage current from an AC powered supply. Breaking the ground of an interconnect with say an optical interface breaks the loop. But beware of this, the "receiving" end of the optical interface requires power, if that comes from an AC line connected PS, you are adding in another leakage loop. It may be better than the original, but it will still be there.

 

The LPS-1 and the Intona deal with these two sides of loop breaking, the LPS-1 at the PS side and the Intona at the digital interconnect side.

 

The OTHER way to deal with leakage loop noise is by decreasing the impedance of the loop. One interesting way to do this is on the AC mains side of things. Remember this IS part of the loop. Many audiophile have special "power conditioning" devices designed to "lower noise on the AC line". Unfortunately the implementation of these devices can greatly INCRESE impedance between devices plugged into them! They do decrease the noise between neutral and hot, but increase the impedance between devices, thus increasing the noise generated by leakage loops. My experience has been that in many cases the leakage loop noise increase wreaks more havoc than any advantage from decreased neutral to hot noise.

 

Thus my recommendation is to get rid of the expensive filters and instead use a simple power strip that has nothing between outlets other than thick wire. EVERYTHING goes into this strip. This will dramatically decrease leakage loop noise simply by decreasing the impedance on the AC mains side of the loop.

 

So in order to get the best from USB you need an async implementation done right, good signal integrity at the DAC and some means of blocking the leakage currents from the computer's power supply from getting into the rest of the system. It takes all three to really get the best from USB.

 

Note there was nothing in here about "the noise from all that digital stuff going over the wire to the DAC". My research seems to show that this doesn't happen very much, the leakage current from the power supply is FAR more important to great sound.

 

Now that we have covered USB, lets briefly touch on Ethernet. It seems like the same things affect SQ. You still need the main timing control to be in the DAC, signal integrity matters and leakage currents are very important.

 

Fortunately Ethernet is inherently galvanically isolated if you are using UTP. BUT if you are using STP (shielded) with connectors that connect the shield you are right back in the leakage current mess.

 

At this point in time I am not sure what the passive isolators are doing. They are high frequency transformers which usually have almost no response at the frequencies usually associated with leakage currents. They might be changing signal integrity in a good way, I have never looked at one on a scope so I don't know. One possibility is that the power supplies frequently supplied with network equipment (routers, switches) are some of the worst I have ever seen with some pretty strong high frequency components to their leakage current. These isolators may have better rejection at these frequencies than the ones commonly used in equipment.

 

The Optical system completely block any leakage current of course, but you can get some back from the power supply feeding the downstream side of the optical system. The signal integrity of the downstream side is also very important and is also usually dependent on the quality of the power feeding it. An LPS-1 will do well for both purposes on the downstream side of the optical system.

 

BTW the speed of signaling on gigabit Ethernet is much less than high speed USB. Both 100 megabit and 1 gigabit use the SAME rate of 125 mega symbols per second. High speed USB is 480 mega symbols per second. HUH? 100 and 1000 are the same? 1000 uses four pairs at the same time, 100 uses just one pair, AND 100 uses 3 voltage levels (high low and in the middle) but 1000 uses 5 voltage levels. The result is one bit per symbol vs two. So bandwidth of transformers needs to be exactly the same for 100Mb and gigabit, just a lot more of them for gigabit.

 

I guess that is enough for now, and I didn't even talk about vibration isolation.

 

John S.

Link to comment
  • 3 months later...
If the intona reclocks and reproduces the signal on the downstream as stated on the website is the usb signal or data reclocked on the upstream path from intona back to computer? if so if the dac is controlling the timing of packets of information in its communication to the computer how then is the intona not affecting this process again on the upstream if it also reclocks. the intona i had sounded nice but is it changing the sound signature of the dac?

 

You need to understand how asynchronous USB mode works. First some background on USB audio:

 

In high speed mode packets are sent at 8KHz, period. Different sample rates work by sending differing number of samples per packet. In order to get a sample rate that is not an integer multiple of 8KHz, different packets have different number of samples. For example you might have 10 packets of 12 samples per packet then have one packet of 13 samples, then back to 12 sample packets etc. The computer assumes the packet rate is exactly 8KHz and produced a sequence of packet sizes to get the correct average sample rate assuming the 8KHz packet rate. This is done for both adaptive and asynchronous.

 

In asynchronous mode the data at the DAC side goes into a buffer and is clocked out by the local clock. If the buffer starts getting too full or too empty (because the local clock is running slightly slower or faster than it should be) it sends a packet back to the computer telling it to speed up or slow down. The computer does this by slightly changing the sample rate (say 44.104 KHz), which changes the sequence of packet sizes, which changes the overall average data rate.

 

Thus no actual clocks are sent from the DAC to the computer, the computer is free running sending the data, but the DAC tells it to speed up or slow down every so often so the average data rate matches the local clock.

 

The Intona DOES reclock the feedback packets coming from the DAC to the computer, but that doesn't matter, the precise timing of the bits on the bus does not mater, it is the DATA in the packet that matters (the speed up or slow down). The asynchronous protocol still works fine with the reclocking.

 

I hope that makes sense.

 

John S.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...