ISO REGEN launch thread! (product web page up; photos, etc.)

[JohnSwenson]

8 hours ago, greenleo said:

@John

 

I want to check if my understanding of Galvinic Isolation (GI) is correct.

 

1. Say, for a simple system a DAC + an Amp, both AC powered and chassis earthed, there ground (or ground plates) are basically connected through the signal cable.  Hence noises may flow between them.

 

2. If, say, the DAC is powered by battery, it will become floating and there will be no common earth between the two devices.  Hence noises between them may not flow to the other party.

 

3. GI effectively forbids noises flow between interconnecting devices.

 

4. If a system that flows from A to D (say PC to mR to IR to DAC), if B and C are both battery powered, then the noise from the ground plate of each of the devices will be isolated and cannot pass to other devices.

 

I understand IR does more for providing high SI.  I also understand that LPS-1 provides effectively clean DC and takes care of the AC leakage current.  What I really want to know is what GI is for and how it helps in SQ.  Recently I come across with the term "ground loops" and cannot figure out a coherent picture even after reading lots of web pages.  Hence my questions.

 

If John or other members would pm me on these, you're welcome. 

First off, everything I'm saying in this post has been covered in detail in other posts, I'm just not up to finding all of them right now.

 

I'm actually going to cover a bit more than your questions, I'll cover some things from a couple other recent posts.

 

My current thinking on USB interface is that there are at LEAST three things that can degrade the sound coming from a DAC. All of these work by increasing the jitter on the clock feeding the DAC chip(s). This can happen in two different ways: increasing the jitter of the local clock itself, or modifying how the DAC chip receives the clock, which effectively increases the jitter INSIDE the DAC chip. Both of these mechanisms are usually happening in most DACs.

 

This extra jitter gets to the clock and DAC chip through noise on the power and ground plane. Noise on the power plane can somewhat be dealt with by voltage regulators, but you can't do anything about noise on the ground plane except prevent it from happening in the first place.

 

So what is jitter? It is variations in timing of the clock edges. For example if a clock is running at a certain frequency, the time between edges is supposed to be an exact number. It never is EXACT. The time between edges varies a little bit from edge to edge. Sometimes a little more than what it should be and sometimes a little less than it should be. Note this is purely timing, the AMPLITUDE of the signal is not changing, it is just the timing that changes.

 

The audiophile world loves to characterize jitter with one number, such as 1 pico second (ps) of jitter. But that is a way over simplification. In any real system the differences from "perfect" are all over the place, one edge be just a little off, the next might be much larger etc. One of the most important methods to characterize this "difference from perfection" is called phase noise. It is amplitude of the variations vs frequency. For example if the variations from perfection were a sine wave, ie small positive variation, larger variation, even larger, then going back to small positive, then small negative, larger negative, back to small negative etc. This sinusoidal variation in timing would show up as a single vertical line in the phase noise plot.

 

Phase noise of real systems usually consist of random variations due to actual physical processes, these show up as a continuous, somewhat noisy curve, and variations due to specific non-random process, these show up as vertical lines on the plot (called "spurs"). A very common source is power supply bleed through which shows up as lines at 60Hz, 120Hz etc.

 

So now on to those three sources of ground plane noise in USB systems:

1)  PG (Power/Ground) noise generated by the USB receiver itself.

2) leakage loops.

3) clock bleed through from other clocks in the system.

 

#1: USB is a complex protocol, it takes a fair amount of circuitry to process the USB packets and extract the audio data. Every time a transistor in that circuitry switches it generates a little blip of noise on the PG planes. All of these transistors doing their thing generate a significant amount of noise on the planes.

 

On top of this the PHY (that stands for the physical interface) is connected to the USB signal, changes in the "signal Integrity" (SI) of the USB signal cause the PHY to change the amount of PG noise it generates. I have written a LOT about this so I won't go into great detail here.

 

#2. This is a completely different mechanism that is caused by leakage current from power supplies. All AC power supplies have some leakage from the AC line (the 60Hz in US etc) to the DC output of the supply. This current forms a loop, going from AC line, through supply, to interconnects, to another box, to supply of that box, and back to the AC line. This loop has nothing to do with the electronics powered by the supplies, it is purely a property of the supplies. I've gone into this in great detail in other posts.

 

This is very different than traditional "ground loops" which are current loops through the safety grounds of the boxes. The leakage loop has nothing to do with the safety ground. So ways to deal with "ground loops" are frequently  ineffective with leakage loops. I wrote a post a while ago going into detail on the differences between between ground loops and leakage loops.

 

Switching power supplies which traditionally are used with computers usually have considerably high leakage current than other supplies, so the inclusion of computers into audio systems have elevated the importance of leakage loops.

 

In addition DACs are particularly susceptible to leakage loops, the leakage loop usually goes through the ground plane of the DAC PCB, directly increasing the jitter of the DAC clock.

 

Other devices such as pre-amps and power-amps can also be part of leakage loops, but their analog circuitry is less susceptible than DACs.

 

#3. In every digital stream the timing of the edges of the waveform are controlled by some clock. The clock signature of that clock (its phase noise) is directly imprinted on that data stream. When that stream goes through digital circuitry, it causes PG noise that is correlated to the clock signature of the generating clock. Remember that any time a transistor switches it creates a little blip on the PG planes, thus every edge from the data stream causes a little blip. The timing of those blips is exactly the same as the timing of the original clock.  Thus anything receiving that stream is going to be generating noise on its own PG planes that matches the signature from the originating clock.

 

This signature from the originating clock will be layered on top of the clock signature from any local clock. Thus even if a data stream is re-clocked by a local clock, the signature from the originating clock gets layered on top of the signature of the local clock. If a whole bunch of things are done right the intensity of the originating signature can be attenuated, but never eliminated.

 

The ISO REGEN is designed to address each of these three items.

#1, the ISO REGEN contains a hub chip which re-generates the USB stream from the source, the SI is extremely good. The result is that the USB PHY in the DAC does not have to work as hard and thus generates less noise on the PG planes in the DAC. I've written lots on this already.

 

#2, this is what the galvanic isolation is for. Again I have written a lot on this. Some notes on this, GI is necessary for breaking leakage loops, but not sufficient. For example almost all power supplies are galvanically isolated from the AC line, but are still the source of leakage loops.

 

The issue here is frequency. Galvanic isolation is only interested with DC, that the DC resistance is very high. But leakage current is AC, so you can have a circuit that blocks DC but passes low frequency AC (power line frequencies) and still have leakage loops go right through. So just because something says it is galvanically isolated does not mean it blocks leakage loops.

 

The isolation in the ISO REGEN is specifically designed to block leakage loops which go through USB cables.

 

#3, the local clock on the ISO REGEN has VERY low phase noise, but remember that the noise from the clock generating that USB signal going into the ISO REGEN still gets layered on top of the phase noise of the local clock. Now it DOES get attenuated, a lot of work went into designing it such that not very much of the signature from the USB source gets through, but some still does. The result is a very low phase noise USB signal that will provide very low degradation of the local clock in the DAC.

 

The ISO REGEN is designed to significantly attenuate all three mechanisms that cause degradation of sound quality in a DAC.

 

I hope that helps.

 

John S.

 

 

[JohnSwenson]

7 hours ago, BigAlMc said:

Hi @Superdad& @JohnSwenson

 

Apologies if this has been covered but I've been trying to track both this and the Listening experiences thread and didn't see it. 

 

What's your take on whether the isolation offered by the ISO Regen means that optical isolation is no longer needed or not. Theoretically do the two forms of isolation compliment each other or does the ISO Regen mean the optical isolation is no longer needed or less significant? 

 

By optical I mean a couple FMCs and a short run of Fibre optic cable in an ethernet run. 

 

Many Thanks, 

Alan 

Unfortunately the answer is quite complex.

 

Ethernet is already inherently galvanically isolated and will break leakage loops (the low frequency leakage current can't make it through Ethernet transformers). Thus the isolation in the ISO REGEN does not improve the overall isolation situation, as long as the Ethernet to USB device does not generate a new leakage loop. For example a microRendu powered by a LPS-1 does not generate a new leakage loop, but a mac mini (even if powered from a JS-2) does.

 

So the optical LAN does no give any better leakage current isolation than wired Ethernet. Any difference in sound is going to be due to clocking differences. THAT is much harder to analyze by just thinking about it. It depends on the clock generating the Ethernet packets feeding the wire in the first place, the clock in the wired->optical converter, the clock in the optical->wired converter. The power supply and board layout in each of these has a major impact.

 

There are so many variables that it is almost impossible to determine whether a particular configuration with optical is going to sound better than a particular system without optical.

 

My personal opinion is that the optical system is just adding complexity that is probably not necessary. You will probably get better results by getting a very low phase clock on the circuit that generated the wired Ethernet in the first place (computer, switch etc). Of course each of those can be affected by the clock generating the data to THEM.

 

It gets essentially impossible to figure any of this out in advance. A particular optical configuration MIGHT attenuate the clock signature from an upstream clock, or it might make it worse. Impossible to tell without measuring.

 

Unfortunately nobody has any test equipment to measure this. I'm working on designing my own, but this will take quite some time to get that working. So for now it is purely trial and error.

 

John S.

[JohnSwenson]

2 hours ago, Quadman said:

Alex, thanks.  I will remove the AQ Jitterbug tonight.  That custom homemade 6" usb cable with very high quality pure silver data lines and separately shielded voltage lines has been in my system for over 1 year, with the regen and no issue ever occurred with it.  In fact when I compared it to the USBPC adaptor you sent with IR I definitely preferred my custom cable, music was more alive (not bright)  the usbpc seemed to lose dynamics and life compared to my cable, kind of like what @swl3600 reported on the other thread.

 

As to the IR and the T+A the T+A does some pretty good isolation of the USB and digital signal from analog signal, I think they use the same chip as you at least in separating digital from analog.  The regen or isoregen does not make a huge change with the T+A as it did say with my Chinese dac the Gustard x20 or my NAD M51 that was a OMG moment when the regen was inserted.  That said the IR does offer some additional improvement over just straight USB connection to the T+A.  Images focus a bit better, separation is noticeably better and bass tightened up just a touch, subtle but improvements.  Now I need to listen more with the usbpc vs. my cable once the disconnect issue is sorted out.  Last session was when the IR would not connect when I inserted the usbpc, I pulled it (IR) and inserted the amber regen and music played.  Switching with regular regen does not cause this behavior.

The ISO REGEN and USB REGEN handle VBUS differently. The USB REGEN has the downstream VBUS always on so even if the hub chip is in suspend mode the VBUS will still be applied to the DAC.

 

The ISO REGEN uses a VBUS switch controlled by the hub chip, this was necessary to make proper connections when using the USB isolator in the path. But that also means that the VBUS gets turned off when the hub  is in suspend mode. The solution is to prevent the computer from suspending the hub chip. I know how to do this for windows, but not for other OSs.

 

For Windows open up the device manager, click on the "Universal Serial Bus controllers". This brings up a list which should contain the line:

USB2.0 MTT Hub

This will be the ISO REGEN.

 

Right click on the line, click on Properties, this brings up a properties window. Click on the "Power Management" tab, make sure the line "Allow the computer to turn off this device to save power" is un-checked. (if it is checked click on the check to un-check). Click on OK.

 

This should prevent the computer from suspending the hub when it has been inactive for a long time, thus keeping the connection to the DAC alive.

 

Let us know how this goes.

 

Thanks,

 

John S.

 

 

[JohnSwenson]

1 hour ago, Quadman said:

 

John, Thank you for the explanation.  I run windows 10 pro 64 bit, creators update.  I just fired up the PC, before doing so installed the USPCB adaptor.  Went into device manager clicked on Universal serial bus controllers and the image below is what I see The name USB2.0 MTT Hub does not appear.  I believe the IR is the top "Generic USB hub", I checked the pwr mgnt tab and the box you refer to was unchecked.  I did check all listings and only 2 others had a pwr mgnt tab, the bottom two (usb root hub) I unchecked the box next to "Allow the computer to turn off this device to save power".  Any reason why device manager does not show the IR as USB2.0 MTT Hub?

 

 

 

Hmm, on Windows 7 it shows up as USB2.0 MTT Hub. You should be able to leave the device manager open and plug and unplug the ISO REGEN and see which line changes.

 

If it is unchecked and you still have the disconnection with long term non-use then we will have to do some more detective work. 

 

John S.

[JohnSwenson]

9 hours ago, Ciukas said:

 

Are the supercapacitor batteries a problem at airport securities etc? I know the max limit is 99wh, but just making sure.

 

When the feeder supply is unplugged or turned off the supercaps are specifically discharged. After a a few seconds there is almost no charge left in them. There is about 70 mv  across the whole string after this. This is an extremely small amount of charge left in the caps. That comes out to about 0.09 watt seconds. There are 3600 watt seconds in a watt hour, so that is 0.000025 watt hour residual charge. I think it is safe to say that is well below 99 watt hour.

 

John S.

[JohnSwenson]

15 hours ago, Em2016 said:

Hi @JohnSwenson

 

If you get a spare moment can you share (at a high level) how these isolators like the Silanna chip used in the ISO REGEN actually work, i.e. how do they isolate the USB data lines? 

 

And is the USB ground of the ISO REGEN isolated by the Silanna chip or is that a different isolation method?

 

Asking these questions in another way, how does stuff actually get from one side of 'the moat' to the other side and how does this then block most of the (low impedance) leakage?

 

Cheers!

 

The reason you use an isolator is that you want to isolate the ground. The ground is where the leakage current travels so it is the critical path to isolate. The problem is that without the ground connection normal signals will not pass, since they use the ground for return current. This is where the digital isolator comes into play, it transfers digital signals across a ground break.

 

There are many types of digital isolators, the most famous is the optocoupler, (LED pointing at photostransistor), there are many others: transformers, capacitively coupled, radio transmitters and my favorite, GMR (Giant MagnetoResistive).

 

There are two things that make high speed USB difficult:

1): high speed -- HS USB runs at 480 Megabit per second (Mbps), most digital isolators do not run anywhere near that fast. Some do, but they are not easy to come by and are very expensive.

 

2): HS USB is a bidirectional bus, all of the digital isolators are uni-directional, you need two, one for each direction. The hard part is that the isolators need to know which end is driving the bus, the isolator or what is at the other end. There is no wire in the bus that specifies which direction the data is going in at any given time. The ONLY way to do this is a full blown USB protocol engine that follows the commands on the bus and can thus figure out which direction the bus is going at any given time.

 

#2 is not easy, none of the big chip companies has a chip that does this, anybody that wants to to isolate HS mode has to do this themselves, very few companies have been successful at this.

 

Things get more complicated for capacitive and transformer coupled isolators. Neither of these transfers a steady state, they only transfer a CHANGE in signal. This means more complex encoding so they can handle the state of a signal not just that it changed. This limits the data rate they can handle.

 

I don't know for sure what the Sillana uses, but I think it is capacitively coupled.

 

That is the overview, if you want more detail, let me know.

 

John S.

[JohnSwenson]

9 hours ago, Em2016 said:

 

Hi John I will zoom in on your favourite there (GMR), if you don't mind.

 

Is there still capacitive coupling involved? Or is GMR as good as optical isolation (absolute zero leakage currents getting across)?

 

Have you played around with GMR isolation, especially with your new custom made leakage currents test gear?

 

GMR is magnetic, but is different than a transformer. A transformer passes change in a magnetic field, thus it cannot properly pass a steady high or low, just something changing. Thus all transformer based isolators have to send a more complicated series of changes to define a steady state signal.

 

GMR works differently, it has a coil that generates a magnetic field, and magnetically sensitive resistors. They change resistance with differing magnetic fields applied. Thus they inherently handle steady state signals as well as changing ones. They can run pretty fast, but are still fairly expensive.

 

The common optocouplers are fairly large and can't go very fast. The LPS-1 uses optocouplers all over the place to get signals between power domains. They work well for this because the signals are slow (32KHz) and the couplers are dirt cheap (40 cents per channel). GMRs are much faster (150MHz) but are much more expensive ($3 per channel). I use GMRs when I am isolating I2S signals.

 

Both optocouplers and GMRs have very low cpacitance, hence very good leakage isolation.

 

The problem with the ISO REGEN was not the isolation technology per se, but that for some reason the chip didn't work when there was a very rapidly changing voltage between the two sides. It turns out this can easily happen with older style SMPS supplies, particularly those used for residential networking equipment. Since most of you guys are using this with computer audio where residential networks are used, this happens a lot.  The only way to deal with this is to put a filter across the "moat" which slows down the edges, but also lets high impedance leakage through. In REALLY bad cases that isn't even enough which is what the switch is for.

 

John S.